AMERICAN NATURALIST Volume 190, Number S1 | August 2017 Convergence, Natural History, and Big Questions in Biology A symposium organized by Anurag Agrawal Toward a Predictive Framework for Convergent Evolution: Integrating Natural

History, Genetic Mechanisms, and Consequences for the Diversity of Life Anurag A. Agrawal, pp. S1–S12

Pattern and Process in the Comparative Study of Convergent Evolution D. Luke Mahler, Marjorie G. Weber, Catherine E. Wagner, Travis Ingram, pp. S13–S28

Convergently Evolved Toxic Secondary Metabolites in Plants Drive the Parallel Molecular Evolution of Insect Resistance Georg Petschenka, Vera Wagschal, Michael von Tschirnhaus, Alexander Donath, Susanne Dobler, pp. S29–S43

Convergent Phenotypic Evolution despite Contrasting Demographic Histories in the Fauna of White Sands Erica Bree Rosenblum, Christine E. Parent, Eveline T. Diepeveen, Clay Noss, Ke Bi, pp. S44–S56

Convergence and Divergence in a Long-Term Experiment with Bacteria Richard E. Lenski, pp. S57–S68

Evolutionary Scenarios and Primate Natural History Harry W. Greene, pp. S69–S86

Convergence, Consilience, and the Evolution of Temperate Deciduous Forests Erika J. Edwards, David S. Chatelet, Bo-Chang Chen, Jin Yao Ong, Shuichiro Tagane, Hironobu Kanemitsu, Kazuki Tagawa, Kentaro Teramoto, Brian Park, Kuo-Fang Chung, Jer-Ming Hu, Tetsukazu Yahara, Michael J. Donoghue

pp. S87–S104

Geographical Variation in Community Divergence: Insights from Tropical Forest Monodominance by Ectomycorrhizal Trees Tadashi Fukami, Mifuyu Nakajima, Claire Fortunel, Paul V. A. Fine, Christopher Baraloto, Sabrina E. Russo, Kabir G. Peay, pp. S105–S122

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Toward a Predictive Framework for Convergent Evolution:

Integrating Natural History, Genetic Mechanisms,

and Consequences for the Diversity of Life*

Anurag A. Agrawal†

Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, New York 14853; and Department of Entomology, CornellUniversity, Ithaca, New York 14853

abstract: A charm of biology as a scientific discipline is the diversityof life. Although this diversity can make laws of biology challenging todiscover, several repeated patterns and general principles govern evolu-tionary diversification. Convergent evolution, the independent evolu-tion of similar phenotypes, has been at the heart of one approach to un-derstand generality in the evolutionary process. Yet understanding whenand why organismal traits and strategies repeatedly evolve has been acentral challenge. These issues were the focus of the American Societyof Naturalists Vice Presidential Symposium in 2016 and are the subjectof this collection of articles. Although naturalists have long made in-ferences about convergent evolution and its importance, there has beenconfusion in the interpretation of the pattern of convergence. Doesconvergence primarily indicate adaptation or constraint? How oftenshould convergence be expected? Are there general principles thatwould allow us to predict where and when and by what mechanismsconvergent evolution should occur? What role does natural historyplay in advancing our understanding of general evolutionary princi-ples? In this introductory article, I address these questions, review sev-eral generalizations about convergent evolution that have emerged overthe past 15 years, and present a framework for advancing the study andinterpretation of convergence. Perhaps the most important emergingconclusion is that the genetic mechanisms of convergent evolutionare phylogenetically conserved; that is, more closely related species tendto share the same genetic basis of traits, even when independentlyevolved. Finally, I highlight how the articles in this special issue furtherdevelop concepts, methodologies, and case studies at the frontier of ourunderstanding of the causes and consequences of convergent evolution.

Keywords: adaptation, comparative biology, constraint, evolutionaryecology, phylogenetic ecology, plant-insect interactions.

Introduction

The search for convergent evolution and its causes is oneway to make sense of the wonderfully bewildering biolog-

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† E-mail: aa337@cornell.edu.

Am. Nat. 2017. Vol. 190, pp. S1–S12. q 2017 by The University of Chicago.0003-0147/2017/190S1-57341$15.00. All rights reserved.DOI: 10.1086/692111

This content downloaded from 132All use subject to University of Chicago Press Term

ical diversity on our planet. The study of convergence ispart of a program to discover repeated patterns and generalprinciples that govern evolutionary diversification. Indeed,with the advent of non-model-omics, the study of conver-gent evolution is enjoying a new surge of interest, and thissymposium and special issue of the American Naturalistbrings together a superb group to address classic and novelquestions on the topic.Here I define convergence as the independent evolution

of similar phenotypes. As has been noted many times, theintrigue of convergence was not lost on one of our greatestnaturalists, Charles Darwin, when he identified traits suchas luminescent organs in seemingly distantly related in-sects and pollen packages in distantly related plants. Dar-win (1859, p. 193) wrote: “although the general appearanceand function of the organ may be the same . . . some funda-mental difference can generally be detected . . . Natural se-lection . . . has sometimes modified in very nearly the samemanner two parts in two organic beings, which owe but lit-tle of their structure in common to inheritance from thesame ancestor.” Darwin recognized that convergent traitsare not necessarily identical in all respects and that their evo-lution was largely independent, although not completely so,given that all organisms ultimately share a common ancestor.Most convergent evolution falls under the umbrella of

what may be considered “constrained adaptation”—evolu-tion that is limited by the strength of natural selection, geneticarchitecture, and fitness costs and benefits, all of which nar-row the number of possible evolutionary outcomes. Althoughconvergence is often interpreted as evidence of both adapta-tion and constraint (sagaciously reviewed by Losos [2011]),the two processes are intertwined and thus can be difficultto separate. In this context, constraint has been broadly de-fined as “restrictions or limitations on the course or outcomeof evolution” (Arnold 1992) or, more generally, the “unequalprobability of outcomes in evolution” (Schwenk 1994/1995).Issues of constraint will be important inmydiscussion of con-vergent evolution, but because of the varied historical use and

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S2 The American Naturalist

abuse of the term (Futuyma 2010), I will frequently use theterm “bias” to simply refer to the pattern of unequal probabil-ity of outcome in evolution (table 1).

As Losos (2011) pointed out, repeated trait-environmentassociations are suggestive of convergent adaptation, andmeasures of natural selection and trait function can eluci-date the adaptive nature of convergent traits (box 1). None-theless, such measures do not address the extent to whichadaptation may be constrained. Many functional traits canbe under selection and yet may be limited in how they canevolve. As will be discussed later in this article, a pattern ofphylogenetic bias in both the degree of convergence andthe underlying mechanistic basis of convergent traits isreshaping our view of constrained adaptation. The challengeof studying convergence in our era is thus interpreting pat-terns in natural history on phylogenies, with repeated ordivergent genetic mechanisms helping to elucidate generalissues in evolution. This was the focus of the American So-ciety of Naturalists Vice Presidential Symposium in 2016.

In this introductory article, I aim to (1) introduce generalissues and concepts in the study of convergence and in par-ticular link the study of convergence to its roots in naturalhistory, (2) provide a framework for addressing modernquestions in convergence and summarize the state of thefield in terms of progress and open questions, and (3) high-light themes addressed in the subsequent articles in thisspecial issue. The current growth in studies of convergencehas been fueled by increasingly sophisticated modern ana-lytical tools of comparative biology, the ability to pinpointgenetic mechanisms of convergence using molecular biol-ogy, and the expansion of observational (as opposed to ex-perimental) science in the past few decades. While perhapsless common, experimental approaches and novel statisti-cal methods in the study of convergence are helping to ex-plain patterns with process (Lenski 2017; Mahler et al.2017).We are beginning to study the cellular andmolecularmechanisms leading to convergence in distantly related or-

This content downloaded from 132All use subject to University of Chicago Press Term

ganisms subject to the same selection, a critical approach tounderstanding phylogenetic biases in evolution (Petsch-enka et al. 2017; Rosenblum et al. 2017).Convergence has also been at the center of thinking

about major ecological patterns and their causes. For exam-ple, Edwards et al. (2017) seek to understand the causes ofdeciduousness in woody plants, a highly convergent traitthat defines several global biomes. In some cases, it has beenhypothesized that convergence occurs at the communitylevel, driving greater similarity of species within a commu-nity than would be predicted from their phylogenetic relat-edness (Schluter 1986). In other cases, dissimilarity evolvesdue to the convergent evolution of community membersinto different niches (Gillespie 2004; Losos 2009). Identify-ing traits that may be involved in community assembly arecritical, as such traits are the link between evolutionary his-tory and ecological outcomes. Together, these approachesrepresent an exciting frontier in the study of convergentevolution whereby we are able to study the causes and con-sequences of repeated evolutionary change integrating fromgenes to communities.

Natural History: The Roots of the Studyof Convergent Evolution

Natural history concerns the description of organisms (in-cluding their scientific name), their traits (any phenotypeor description of their genome), location and distribution,and interactions with the biotic and abiotic environment.Although natural history serves as a foundation for biology,there continues to be controversy about the role that naturalhistory should play in advancing knowledge in modern biol-ogy (Greene 2005). Yet clearly “reciprocal relationships amongthe growth of robust theory, experimentation, and accuratenatural history” (Greene 2005, p. 23) are essential for ad-vancement in our understanding of evolutionary biology. Nat-ural history continues to be an important source of inspira-

Table 1: Glossary of terms relevant to this article

Adaptive phenotype

A trait with a current function maintained by natural selection Constraint Restrictions or limitations on the course or outcome of evolution Biased convergence Convergent evolution where the trait consistently has the same mechanistic basis Contingency Chance events that shape evolutionary trajectories Convergent adaptation Repeated independent evolution of similar phenotypes by the same agent of natural selection Convergent evolution Repeated independent evolution of similar phenotypes Environment Abiotic conditions and biotic interactions Mechanism The causal basis of a particular trait, usually several fine-grained traits that underlie a higher-level

(coarse-grained) trait

Phylogenetically conserved

convergence

Convergent evolution where the trait consistently has the same mechanistic basis within a clade

Unbiased convergence

Convergent evolution where the trait has distinct mechanistic bases, even within a clade

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergence and Natural History S3

tion, is critical for knowing and conserving what we have onthe planet, and, most germane tothis article, is central to theconceptual advancement of evolutionary biology in thebroadest sense. Using his classic approach of natural historyand phylogenetic thinking, Greene (2017) addresses conver-gence in animal behavior, with implications for understand-ing our human selves as animals.

Natural history in the context of comparative biology hashistorically been a critical hypothesis generator, which hasoften yielded general patterns and sometimes pushed the en-velope of theory—especially in our understanding of conver-gent evolution. Two examples from my own corner of biol-ogy, plant-herbivore interactions, illustrate this point.

By the early 1960s, the birth of chemical ecology as adiscipline yielded possible explanations for chemical di-versity in plants and hypothesized consequences for majorpatterns in biological diversification (Fraenkel 1959; Ehr-lich and Raven 1964). Key hypotheses about the functionof plant chemistry were informed by careful natural his-tory observations on distantly related groups of organisms(Fraenkel 1959). In a classic study on the “raison d’êtreof catnip,” Thomas Eisner (1964) reasoned that the chem-ical components (including the terpene nepetalactone)causing catnip’s effects, feline euphoria, were a defenseof plants against herbivorous insects. His logic followednot from the incidental effects on cats but from the factthat nearly identical compounds were produced conver-gently by several insects, were insecticidal, and in somecases were ejected by the animals in response to risk of pre-dation. Eisner was an observer who poked around and fol-lowed his nose. The ejection of nepetalactone by molestedbeetles suggested an explanation for the function of thiscompound in plants.

Since then, many other defensive compounds have beenfound to be produced by both animals and plants, includingalkaloids and cyanides. The genetic basis of such convergentevolution in defenses continues to be revealed (Jensen et al.2011; Denoeud et al. 2014), and the hypothesis generationhas not stopped with identifying the functional or mechanis-tic basis of these defensive traits. Conceptual developmentsstarting in the 1970s and advancing with phylogenetic think-ing since 2000 have radically improved our understanding ofwhich plant traits repeatedly evolve together to produce mul-tivariate strategies of defense (Feeny 1976; Kursar and Coley2003; Agrawal and Fishbein 2006; Fine et al. 2006; Mooneyet al. 2010; Johnson et al. 2014; Mason et al. 2016). In eachof these studies, either closely related species (many of whichinhabit different habitats) or coexisting species (which span abroad swath of phylogenetic diversity) have been studied andarrayed along multiple axes of growth and defense. A cluster-ing of phenotypes has suggested repeatedly evolving syn-dromes that can be tested for their ecological effects and asso-ciations with particular environments.

This content downloaded from 132All use subject to University of Chicago Press Term

Another, perhaps more grandiose academic pursuit em-ploying convergence and natural history involved the searchfor general patterns or rules that might govern nature. AlfredRussel Wallace (and other contemporaries, including J. W.Slater and E. Haase) posed hypotheses about bright colora-tion, toxicity, and the acquisition of noxious substancesfrom the host plants of butterfly caterpillars. Each of thesenaturalists, through their travels, observations, and records,noted a set of associations that lead to a hypothesis: brightlycolored lepidopteran larvae obtain toxins from their hostplants and use them in defense, typically against vertebratepredators. Subsequently, in the 1950s therewas a race to dem-onstrate such sequestration of toxins by caterpillars, which,once successful, gave rise to the search for general rules of se-questration, aposematism, and mimicry (Reichstein et al.1968). Themonarch butterfly in particular was the first spe-cies shown to sequester toxins from its host plant and be-came a model to address such general rules (Agrawal 2017).In a series of studies led by Miriam Rothschild between

1967 and 1973, the association between aposematism andsequestration was cemented by studies of several distantlyrelated insect groups that feed on related host plants (vonEuw et al. 1967, 1971; Rothschild et al. 1970, 1973), ulti-mately leading her to study 23 aposematic insect speciesfrom six insect orders that sequestered the same toxic com-pounds from their host plants in the Apocynaceae. Suchfindings opened the door to more rigorous statistical anal-yses, conceptual development, and theory on defense, se-questration, aposematism, and mimicry (Ruxton et al. 2004).Although the frequent association of sequestration, apose-matism, and mimicry now seems obvious, it was the re-ciprocal interaction between natural history observations,studies of convergence, hypothesis testing, and theory de-velopment that led to the paradigm. Contemporary workaddresses the genetic basis of convergent associations suchas specialization and sequestration. In at least some cases,a diverse set of distantly related organisms utilizes the samemechanism to achieve adaptation to the same environ-mental challenge (Dobler et al. 2012; Bramer et al. 2015;Petschenka and Agrawal 2015).

Mechanisms of Convergence: A New Roadto Assessing the Role of Constraint?

Traits are hierarchical in nature, from complex phenotypes(sometimes referred to as coarse-grained traits) down tothe products of gene expression (i.e., fine-grained traits,beginning with messenger RNA). Higher-level traits in thehierarchy are typically dependent on several traits at lowerlevels (Conner and Hartl 2004), and the interpretation ofconvergent evolution has different meanings at these differ-ent hierarchical scales (Currie 2013). At the highest scale (orcoarse grain), an ecological outcome of several traits may

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Box 1: A key for categorizing, interpreting, and advancing the study of convergent evolution

1. Has the trait repeatedly evolved?(yes)—the trait exhibits convergent evolution; go to 2 to address why.(no)—the trait is not convergent, and it may be difficult to assess the general importance of the trait except from logic (e.g., pho-

tosynthesis is clearly important) or functional studies. The lack of repeated evolution could potentially be addressed experimentally tounderstand constraints.

2. Is there a pattern of repeated trait-environment association?(yes)—the pattern of convergent evolution is likely due to natural selection; go to 2b (to understand the agents of selection) or 3 (to

address biases in convergence).(no)—the trait appears to have evolved in different environmental contexts (not a convergent adaptation), suggesting alternate func-

tions. Go to 2b, which may help refine your hypothesized selective environment if you are convinced that the trait might be a conver-gent adaptation.

2b. Do functional analyses or other evidence suggest that the convergently evolved trait is the target of repeated natural selectionby the same agent?(yes)—the pattern of convergent adaptation is suggested; go to 3 to address selective mechanisms underlying convergence.(no)—the trait may be correlated with traits under selection but may not be directly subject to natural selection (i.e., pleiotropy

or hitchhiking).

3. Is the mechanistic basis of the convergently evolved trait identical (or nearly so) in the independent origins of the trait?(yes)—biased convergence is suggested: there were very few possibilities for the way in which this trait could evolve; there is a single optimal

solution or alternative solutions either are not as beneficial or occur with some delay and thus are less frequently realized. Go to 3b.(no)—multiple mechanisms of convergence, suggesting that natural selection has favored particular phenotypic space, and the out-

come was achieved in distinct ways. Go to 4.

3b. Ask question 3 again if you can go down a mechanistic level. For example, if anthocyanins repeatedly evolved as a sunscreen, didthe same genes evolve? Or if the same gene was involved in convergent evolution, was it the same specific site mutation?

4. Are the alternate mechanisms of convergent evolution phylogenetically conserved?(yes)—phylogenetically conserved convergence: history (i.e., phylogenetic signal) and selection are both important, and there is some pre-

dictability in what traits evolve by which mechanism in different lineages. Continue to 4b to understand why.(no)—unbiased convergence: selection has resulted in a limited set of phenotypic solutions across lineages, but the means to which we

get there have been mechanistically diverse even within lineages.

4b.What is the basis of the phylogenetically conserved convergence in a trait?Does a particular phenotype, gene, genetic architecture(including duplication), or extrinsic factor (such as ecology or life-history strategy) predispose a lineage to evolve the same mech-anistic basis for a convergent trait? Answering this question allows us to synthesize the joint role constraint and selection in adaptation.

S4 The American Naturalist

be the trait (or extended phenotype) that exhibits conver-gent evolution. For example, flight has evolved at least fourtimes, in insects, dinosaurs, birds, and mammals. Addition-ally, the highly similar ecological niches employed by placen-tal versus marsupial mammals on different continents alsoexemplifies convergent evolution of coarse-grained traits. Ifwe zoom in, however, the physiological, morphological, andbehavioral mechanisms that underlie convergent niches inthese radiations may or may not be the same (Luo 2007).

Another such coarse-grained trait is the phenomenon ofindirect defense, where protection of an organism from en-emies is achieved through the attraction of, or patrolling by,animal bodyguards. The specific traits and conditions un-derlying indirect defense may be manifold, and these occuracross diverse species and systems (e.g., ant-plants [Heil2008], phloem-sucking hemipteran bugs [ant-tending;Styrsky and Eubanks 2007], leaf-chewing caterpillars [alsoant-tending; Pierce et al. 2002], and even some fish andother vertebrates [Poulin and Grutter 1996]). The conver-

This content downloaded from 132All use subject to University of Chicago Press Term

gent evolution of such indirect defense as a high-level traitinvolves many distinct fine-grained traits, from housingstructures and food rewards to volatile attractants, traps,and undiscovered mechanisms. Given that indirect defenseis composed of multiple traits, we can begin to move downin scale to examine specific traits that generate indirect de-fense. In plants, a mechanism of indirect defense (one steplower in the hierarchy) may be extrafloral nectaries. We mightnext examine the chemical composition of extrafloral nectarsecreted or, more mechanistically, the genes that code for thenectary and its chemical constituents. These finer-grainedtraits are mechanisms that generate coarse-grained traits.Understanding underlying mechanisms is key to deter-

mining the causes of convergent evolution. For example,if convergent coarse-grained traits have distinct mecha-nisms, especially within a lineage, we may conclude thatthe evolution of those traits was less constrained. In otherwords, even within the backdrop of conserved genetic ar-chitecture, traits, and ecology common to a clade, the same

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergence and Natural History S5

evolutionary outcome was achieved by distinct means.Conversely, if mechanisms of convergence are themselvesphylogenetically conserved (the mechanisms are sharedamong close relatives that have independently evolved thetrait), then convergence likely reflects constraint. I empha-size that I am using the term “constraint” here to simply in-dicate bias in the outcome of evolution, here driven byshared traits in a lineage (table 1; box 1).

Phylogenetic Bias in the Mechanisms of Convergence

Several examples illustrate phylogenetic bias in the mecha-nisms of convergent traits. C4 photosynthesis in plants evolvedover 60 times, and several different physiological means havebeen employed as carbon dioxide–concentrating mechanismsin C4 plants (Sage et al. 2011). Recent evidence demonstratesthe convergent recruitment of particular genes within plantlineages, but different lineages employ different genes in con-trolling this photosynthetic pathway (Christin et al. 2015).Also consider red coloration in flowers, a highly convergenttrait that is underpinned by distinct genetic mechanisms.Evolutionary transitions to red flowers have repeatedly evolvedin several lineages, but the prevalence of different mech-anisms of red pigment production differs among lineages(Ng and Smith 2016). In addition to these two plant exam-ples, resistance of animals (insects, amphibians, reptiles, andmammals) to toxic cardiac glycosides is highly convergent,and although many of the specific genetic changes are con-served across lineages, there is again some phylogenetic bias(Price et al. 1990; Croyle et al. 1997; Dobler et al. 2012; Ujvariet al. 2015).

The mechanisms by which vertebrates adapt to high-elevation environments, typically involving the evolution ofaltered oxygenation properties of hemoglobin, have beenwidely studied (Storz and Moriyama 2008) and represent an-other case of phylogenetically conserved mechanisms in con-vergent evolution. In a remarkable recent study, Natarajanet al. (2016) studied 28 phylogenetic pairs of bird species andfound that the oxygen affinity of hemoglobin in highlandbird species was consistently higher than in closely relatedlowland species. Although the altered genes involved wereconsistent, the specific amino acid sites of substitution sub-stantially varied. Thus, evolution was convergent at the levelof individual genes but not within genes. Within humming-birds, which had multiple origins of the highland habit, thegenetic substitutions were nearly identical. Moreover, Natara-jan et al. (2016) demonstrated that those genetic substitu-tions were functional only in the hummingbird’s genetic back-ground, not in that of more basal species. Thus, phylogeneticbias in the mechanisms of convergent evolution may be drivenby the genetic background of a clade.

Phylogenetic bias in the mechanisms of convergence ap-pears to be a general result, one that should be quantita-

This content downloaded from 132All use subject to University of Chicago Press Term

tively evaluated in the coming decades. Understanding thisbias and its causes (such as a constraining effect of conservedgenetic architecture) will lead to a greater understandingof convergence across scales. A potential exception to thepattern of phylogenetic bias in the mechanisms of conver-gence may be loss of functions, which can presumably occurthrough many mechanisms (i.e., mutations) that have thesame ultimate consequence (Manceau et al. 2010; Smithet al. 2013). Thus, when the mechanisms of convergenceare phylogenetically conserved (i.e., a given mechanism re-peatedly evolves within a lineage but distinct mechanismsevolve between lineages), a biasing effect of genetic architec-ture, life history, or some other attribute of the lineage issuggested.Biases in convergent evolution can also be studied ex-

perimentally. Comparing what has evolved in the naturalworld with experimental populations has the potential toreveal the extent of constraint (Weinreich et al. 2006; Stern2013; van Ditmarsch et al. 2013). For example, both muta-genesis screens for particular phenotypes and experimen-tal evolution studies can reveal a greater number of possi-bilities than exist in nature, while their natural counterpartstypically reveal biased outcomes. A next step will be to re-veal why certain possible mechanisms are not realized. Thebiasing effect of genetic architecture is the tip of the iceberg(Natarajan et al. 2016).In some cases, advantageous mutations to groups of genes

involved in adaptation may be quite limited, even in highlycontrolled and benign laboratory conditions. In these cases,convergence may simply be the product of constrained pos-sibilities. However, in mutagenesis studies of microbes, plants,and animals (mostly insects), screens have typically revealedmany more potentially functional mutations than those thatare realized through the natural evolutionary process (reviewedin Stern 2013). For example, in the case of molecular adap-tations of animals to toxic cardiac glycosides (produced byplants and some animal prey), there are relatively few muta-tions that have repeatedly evolved (across many orders ofinsects and in some vertebrates) in the target site (e.g., theubiquitous animal enzyme or the sodium-potassium ATPase;Dobler et al. 2012, 2015; Ujvari et al. 2015). Nonetheless,animal cell mutagenesis studies reveal a greater possibility ofpotentially functional mutations in sodium-potassiumATPase,though this may not be the case when examined in the samegenetic background (Price et al. 1990; Croyle et al. 1997).Work is just now beginning where uncommon or unrealized(but functional) mutations are being introduced to orga-nisms (or cell lines) to address why such mutations are nottypically found in nature overall or at least in particular lin-eages. Presumably, the uncommon or unrealized mutationsdo not exist because of negative fitness consequences dueto either pleiotropic effects on some other function or epi-static effects given the genetic background.

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S6 The American Naturalist

Contingency, Convergence, and Parallel Evolution:New Takes on Classic Questions

Contingency and convergence are sometimes thought toplay mutually exclusive roles in evolutionary biology. InGould’s influential book, Wonderful Life (1990), he madethe case for the importance of contingent or chance eventsthat have had a profound effect on the diversification of life.Yet despite being a hero in Gould’s story, Simon Conway-Morris subsequently took issue with several of Gould’s theses.In fact, in two subsequent books, Conway-Morris providedmany examples of convergent evolution that he argued pro-vided evidence against Gould’s conclusion of the impor-tance of contingency (Conway-Morris 2003, 2015).Althoughit might be an attractive proposition that contingency andconvergence are alternatives, is that truly the case? No. Con-vergence can be found at many levels, and yet contingentevents also shape patterns of life on Earth. The polarizationof the contingency and convergence debate has taken holdconceptually but is largely a false dichotomy.

Gould’s contingency was conceptualized at a deep scaleof chance events that shape the future (e.g., extinction ofmajor lineages such as the dinosaurs). If certain traits evolvedonly once and those lineages were subject to a contingentcatastrophe, then clearly contingency rules. But the impor-tance of such contingent events is not under debate. Rather,what is at issue is whether convergence overrides the impor-tance of these chance events, because ultimately organismaltraits would return to the convergent state. In all likelihood,the truth lies somewhere in between. While chance eventscan reset or change the course of evolutionary history, manytraits have and will continue to exhibit convergent evolution.Contingent events create evolutionary opportunities for con-vergent and divergent evolution. The geographic isolationof large groups of placental and marsupial mammals allowedfor remarkable convergence as these groups diversified. Al-ternatively, certain other contingencies, such as mass extinc-tion events, have surely shaped the diversity of life on Earththat we see today.

Experimental evolution has allowed a reconciliation ofissues and deep insights into the debate over parallel, con-vergent, and contingent evolution (Elena and Lenski 2003).For example, when a single clone is used to found replicatepopulations, one can quantify whether parallel or conver-gent evolution occurs (Arendt and Reznick 2008). Follow-ing Stayton (2015), here I refer to parallel evolution as twolineages starting with the same character state and ultimatelyevolving into a different but shared character state. Manytraits in Lenski’s replicated long-term evolution experiment(LTEE; Lenski 2017) have repeatedly evolved (typically inparallel; e.g., Meyer et al. 2010; Lenski et al. 2015). Alterna-tively, other traits have evolved only in single populations(Blount et al. 2008). Experimental evolution approaches have

This content downloaded from 132All use subject to University of Chicago Press Term

allowed us to ask, When there is a lack of convergence, doesthis indicate that contingency dominates? In the LTEE, onlyone out of 12 replicate populations evolved the ability to utilizecitrate as a carbon source, and this emerged after 30,000 gene-rations. Interestingly, the citrate-utilizing line shows repeat-able evolution of citrate use after 20,000 generations. Thishas been tested by restarting the experiment at different timepoints using frozen ancestors. The genetic architecture waspotentiated by specific mutations after 20,000 generations,and thesemutations serve to repeatedly favor subsequentmu-tations that allow for the use of citrate (Blount et al. 2012)—but in only one out of 12 populations that started from thesame clone. In an observational and phylogenetic context,similar patterns have been observed, where a starting char-acter state has a strong influence on the subsequent macro-evolutionary trajectory taken (Smith et al. 2013).Various authors have proposed that the repeated evolu-

tion of traits may be potentiated by past events. For exam-ple, an original trait may have a low probability of evolving(i.e., dependent on contingencies, as in the case of citrateuse described above), yet what follows is highly repeatable.Interestingly, there is no agreed-upon terminology for suchtraits that then cause a bias, predisposition, or inherency(Conway-Morris 2003) toward a particular evolutionary out-come. Traits that enable (Donoghue 2005), channel (Gould2002), and potentiate (Blount et al. 2008) or that are pre-cursors (Marazzi et al. 2012) to subsequent change have allbeen suggested. The extent of convergence and shared mech-anisms of the traits that convergently evolve may frequentlydepend on such phenomena.In experimentally tractable systems, if one has a hypoth-

esis for what causes the predisposition, this can be addressedby introducing specific changes while controlling for otheraspects of genetic architecture. In some cases, a particulargene may cause the predisposition (i.e., epistasis), and inother cases it may be a constellation of genetic factors (in-cluding gene duplication; Riehle et al. 2001; Stern 2013).Even the ecology or life history of a group may cause theseevents, and such predispositions may well be evident in phy-logenies (Marazzi et al. 2012). In either case, a somewhatinfrequent event may predispose a lineage to convergentlyevolve subsequent adaptations. As discussed above and inbox 1, the same mechanism may repeatedly (convergently)evolve within lineages, but different mechanisms may beconvergently employed in other lineages, all toward the sameend of adapting to a particular selective agent. The examplesof phylogenetically conserved convergences discussed above(C4 photosynthesis, red flower color in plants, animal adap-tation to high elevation, and animal resistance to cardiacglycosides; Dobler et al. 2012; Christin et al. 2015; Ujvari et al.2015;Natarajan et al. 2016;Ng and Smith 2016) are outstand-ing candidates to study the joint role of biases and selectionin adaptation.

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergence and Natural History S7

What Is the Expected Level of Convergence?

Convergent evolution is inherently intriguing because at somelevel it seems unexpected, especially when convergent traitshave the same genetic underpinnings. Nonetheless, wherepossible, the extent of convergence should be contextual-ized against a null model or hypothesis of what might beexpected, both by chance and owing to selective processes.Stayton has argued for the importance of genetic drift in de-veloping null models for the extent of convergence (Stayton2008). Below I consider ecological opportunity, genetic archi-tecture, loss versus gain of functions, and genes of small ver-sus large effect when generating expectations for the levelof convergent evolution.

In a qualitative sense, the likelihood of convergence maybe driven by ecological opportunity, which may be basedon what resources are in excess and can be easily employed.Viewed through this lens, it is not surprising that plantshave evolved extrafloral nectaries many times (Weber andKeeler 2013). Sugar is often not limiting to plants, pre-dacious ants often live nearby, and plant genomes sharemany genes. Similarly, as discussed above, many phloem-sucking hemipteran bugs have evolved indirect defense byants who drink their sugary honeydew excretions (Styr-sky and Eubanks 2007). Their honeydew is a waste prod-uct, but nonetheless, there are closely related Hemipterathat are ant tended or not, and the evolution of tending typ-ically involves specific morphological traits as well aschanges in the amount or composition of honeydew (Völklet al. 1999; Fischer et al. 2002; Shingleton and Stern 2003;Shingleton et al. 2005). Thus, the Hemiptera apparentlyhave an ecological resource in excess,which,with some mod-ification, may be highly sought after by an interaction part-ner.

In contrast, one group of leaf-chewing insects, the Ly-caenid butterflies (and some close relatives) have evolveda nectar-producing gland and recruit ants as defensive body-guards as well (Pierce et al. 2002). Within the Lepidoptera thismay well be the only case of the evolution of defense by ants,and to my knowledge it has not been reported for beetlesor flies. Some gall wasp lineages have repeatedly evolved theability to induce nectar production on their galls (producedby their host plants), and they gain protection from patrol-ling ants (Inouye and Agrawal 2004; Nicholls et al. 2016). Asfar as I know, no vertebrate produces nectar to reward tend-ing ants. Thus, among animals, the high ecological oppor-tunity for the evolution of defense by ants appears to havewidely spurred ant-hemipteran mutualism (and the traitsthat support it), whereas what caused the limited (single?)evolution of the same strategy in Lepidoptera is more diffi-cult to know. Perhaps it was a chance event or circumstancein the history of the Lycaenids. The challenge for us now isto somehow quantitatively address the expectation for such

This content downloaded from 132All use subject to University of Chicago Press Term

convergences, perhaps depending on both extrinsic attributessuch as ecological opportunity and intrinsic attributes suchas genetic architecture.Given the prevalence of epistasis, gene duplication, and

effects of genetic background on the fitness advantages ofparticular mutations (e.g., Riehle et al. 2001; Weinreich et al.2006; Tenaillon et al. 2012; Kryazhimskiy et al. 2014), thenull expectation of convergence should be informed by knowl-edge of such genetic effects. Given a particular genetic back-ground, a new mutation may be more compatible or func-tional and result in having higher fitness than without thatspecific background. As discussed above, we are at just thebeginning of being able to understand biases in the extentof genetic convergence, but given that they exist, we shouldwork toward a predictive framework that would include nullexpectations based on genetic architecture.It has been suggested that loss of function mutations

may on average be less mechanistically convergent thangain of functions (Arendt and Reznick 2008; Manceau et al.2010; Smith et al. 2013). It is certainly the case that con-vergent loss of pigmentation, which often occurs throughmutations in the same gene, occurs by distinct mutationsat different sites (Protas et al. 2006). The rampant conver-gent loss of the gas (swim) bladder among teleost fishes alsooccurs by many distinct mechanisms; even losses within asingle species, wild-caught zebra fishes, occurred by over20 distinct mechanisms (McCune and Carlson 2004). In asense, loss of function is analogous to a highly polygenictrait, since genetic modifications at many locations willresult in the same phenotypic outcome.Are convergent phenotypic traits that are controlled by

many genes also less likely to have a commonmechanistic ba-sis? The genetic basis of body size is highly polygenic, and yetgeographical clines (e.g., across latitude) in phenotype arehighly repeatable across species and convergently evolve innative and introduced populations (e.g., see Lomolino 2005for a review on vertebrates). Nonetheless, the genetic basisof such convergent phenotypic clines appears to occur by di-vergent mechanisms (reviewed for Drosophila in Gilchristand Partridge 1999; Huey et al. 2000; Arendt and Reznick2008). A polygenic basis for adaptation to high temperaturemay also underlie convergence in experimental studies ofEscherichia coli. For example, experimental adaptation to hightemperature in over 100 replicate lines revealed that anygiven pair of lineages shared relatively few nonsynony-mous point mutations (2.6%), but at a higher hierarchicallevel, modified genes and operons were much more likelyto be shared (120%; Tenaillon et al. 2012). In the adaptationof fitness in 65 closely related yeast genotypes, Kryazhimskiyet al. (2014) demonstrated that although fitness evolvedin highly repeatable trajectories, the specific mutations werehighly variable and dependent on epistasis based on past fit-ness gains. In contrast to these studies on polygenic trait evo-

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S8 The American Naturalist

lution, in cases of resistance to particular toxins or diseaseswith highly specific target sites (with genes of major effect),much stronger convergence at the molecular level has beenexpected and observed (ffrench-Constant et al. 1998; Ash-field et al. 2004; Dobler et al. 2012; Farhat et al. 2013; Yanget al. 2013; Brodie and Brodie 2015; Ujvari et al. 2015).

Beyond Mechanism: Evolutionary and EcologicalConsequences of Convergence

Thus far, this article has focused on interpreting patternsof convergent evolution and integrating studies to under-stand the causes of convergence. As discussed above, theevolution of particular genes or traits may predispose alineage to converge on a particular solution in responseto natural selection. In addition, when convergence occurs(at whatever mechanistic level), are there predictable out-comes for the evolution and ecology of a lineage?

Perhaps the greatest interest among evolutionary biolo-gists in consequences of convergence has come in the formof searching for macroevolutionary key innovations, thosetraits that evolve and allow organisms to interact with theenvironment in new ways and increase the net lineage di-versification rate (Hunter 1998). Although there are nowsophisticated analytical tools to detect shifts in diversifica-tion rates on phylogenies, even if they occur only once(Rabosky 2014), single occurrences have little ability toprovide generality in terms of how traits impact diversifi-cation. Even though we have little understanding of whyparticular traits may facilitate speciation (or retard extinc-tion; Heard and Hauser 1995; Futuyma and Agrawal 2009),recent work on plant-insect interactions suggests that suchtraits exist (Fine et al. 2004; Kaminski et al. 2010; Foristeret al. 2011).

Indeed, our best cases of traits acting as key innovations,especially from traits that have repeatedly evolved, comefrom the defensive traits of plants: latex (Farrell et al. 1991)and extrafloral nectaries (Weber and Agrawal 2014). In bothcases, the traits have convergently evolved in numerousplant families, the traits are associated with defense againstherbivores, and lineages that have evolved these traits havehigher diversification rates than closely related lineageslacking the traits. Convergent evolution of these defensivetraits has had profound and predictable consequences forplant diversification. Remarkably, on the coevolutionaryflip side, herbivory as a trait in insects is also an iconic caseof a convergent trait (feeding strategy) that is widely asso-ciated with elevated diversification rates (Mitter et al.1988; Wiens et al. 2015). Future work will certainly identifyother convergent key innovations, but more importantly,we must address how and why the traits impact speciationor extinction.

This content downloaded from 132All use subject to University of Chicago Press Term

From an ecological perspective, there is tremendous in-terest in understanding the species and traits that may causepredictable outcomes in community assembly and compo-sition. Such processes may occur over long periods of time(which include evolutionary change) or over shorter pe-riods of time (e.g., following disturbance or during suc-cession). Do particular species and their traits, once presentin a community, fundamentally change the course of as-sembly? In the classical cases of adaptive radiations inbounded communities (i.e., lakes and islands), there is of-ten an association between convergently evolved animalecomorphs and assembly of a fauna of related species(Losos et al. 1998; Gillespie 2004; Turner 2007). Nonethe-less, cause and effect between the convergent traits and as-sembly of the community can be difficult to discern. Theconsideration of traits that modulate positive and negativespecies interactions is an especially important frontier, asboth can impact community structure, though in differentways. Fukami et al. (2017) take on the issue of convergencein plant mycorrhizal associations and how these may haveshaped predictable outcomes of community structure indiverse tropical forests.In a general sense, convergent traits may strongly impact

the process of community assembly through a few differentprocesses. Onemetric of community assembly is the phyloge-netic structure of a community, defined as a nonrandom pat-tern of evolutionary relatedness among species (Kraft et al.2007). In the simplest case where habitat filtering is critical,convergence in traits among species will result in a patternof even (sometimes called over-dispersed) communities, thosewhere species are less closely related than would be expectedby chance (Cavender-Bares et al. 2004). Conversely, if traitsare phylogenetically conserved (e.g., tolerance of some stress),habitat filtering will result in phylogenetically clustered com-munities. However, when traits are evolutionarily conservedand yet there is selection for dissimilarity within a commu-nity (because of competition or other negative species inter-actions among close relatives), the outcome will result inan even community. Finally, for the case where some habitatfiltering occurs as well as where species interactions are im-portant and species’ traits are convergent (likely the mosttypical scenario), the community outcome is unclear (Kraftet al. 2007). Thus, quantitatively parsing out the impact ofconvergent evolution on community structure is a critical fron-tier in understanding the role of deep evolutionary processesin community assembly.

Looking Back and Looking Forward

There are two related issues for why convergence providessuch an important and compelling approach to biology, es-pecially when describing phenomena or taking a natural

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergence and Natural History S9

history approach. As pointed out by Felsenstein (1985), in-dependent evolution provides statistical power in compar-ative biology. Correcting for phylogeny, though sometimeshaving a pejorative connotation, is simply aboutmaking ro-bust statistical arguments about the association betweentwo phenomena.Whether two traits show correlated evolu-tion, a trait repeatedly evolves in a particular environment,or the repeated evolution of a trait is consistently associatedwith increased diversification, convergence in all casesprovides power for stronger inference than if the associa-tion occurred only once. This is not to downplay the impor-tance of contingent events that occurred only once; indeed,these may be some of the most critical events in evolution-ary history. Nonetheless, to gain insight into whether thereare general rules in biology, there is no substitute for evolu-tionary replication. Evolutionary replication allows us tocome closer to understanding the cause of a particular asso-ciation because, if repeated in independent lineages, the ge-netic background, life history, and ecology of the distinctlineages are likely to be different, and yet the associationstill stands.

The current renaissance in descriptive biology has grownfrom exploring the natural histories of the genomes of manyorganisms (e.g., Parker et al. 2013), engaging in new forms ofdiscovery, and comparative analyses. There has been tremen-dous growth in the analysis of ecological gradients and under-standing the natural pattern of species traits, distributions,and interactions along latitude, altitude, and other gradients.Part of this revolution has come from a renewed interest intrait-based ecology, the availability of phylogenetic informa-tion, and climatic databases. Given this resurgence, I con-clude with four hopeful messages. First, let’s embrace this in-terest, as natural history is an important basis of inspiration,discovery, and the conservation of species. Second, the use ofconvergence can provide rigor and replication to addresssome of the biggest questions, ranging from understandingconstraints to diversification. Third, there is tremendous po-tential to understand the underlying drivers of convergenceby addressing the extent to which the mechanisms of conver-gent traits are biased by the phylogenetic lineage in whichthey evolve (box 1). Last, the combined use of comparativebiology and experimentation (ranging from mutagenesisscreens to reciprocal transplant experiments) will be criticalin advancing biology (Weber and Agrawal 2012).

Acknowledgments

Funding and stimulation to think about this topic was pro-vided by the John Templeton Foundation and especiallyP. Wason. Funding for this symposium and the speakerswas provided by the American Society of Naturalists. Fordiscussion, I am grateful to L. Arcila Hernandez, K. Böröczky,J. Elias, H. Greene, A. Hastings, K. Holmes, H. Inamine,

This content downloaded from 132All use subject to University of Chicago Press Term

P. Jones, M. Kirkpatrick, R. Lenski, A. McCune, K. Niklas,R. Petipas, B. Reed, and S. Scheiner. This article was improvedby comments from J. Bronstein, J. Conner, J. Losos, J. Maron,A. McCune, J. Thompson, and M. Weber.

Literature Cited

Agrawal, A. A. 2017. Monarchs and milkweed. Princeton UniversityPress, Princeton, NJ.

Agrawal, A. A., and M. Fishbein. 2006. Plant defense syndromes.Ecology 87:S132–S149.

Arendt, J., and D. Reznick. 2008. Convergence and parallelismreconsidered: what have we learned about the genetics of adapta-tion? Trends in Ecology and Evolution 23:26–32.

Arnold, S. J. 1992. Constraints on phenotypic evolution. AmericanNaturalist 140(suppl.):S85–S107.

Ashfield, T., L. E. Ong, K. Nobuta, C. M. Schneider, and R. W. Innes.2004. Convergent evolution of disease resistance gene specificityin two flowering plant families. Plant Cell 16:309–318.

Blount, Z. D., J. E. Barrick, C. J. Davidson, and R. E. Lenski. 2012.Genomic analysis of a key innovation in an experimental Esche-richia coli population. Nature 489:513–518.

Blount, Z. D., C. Z. Borland, and R. E. Lenski. 2008. Historical con-tingency and the evolution of a key innovation in an experimentalpopulation of Escherichia coli. Proceedings of the National Acad-emy of Sciences of the USA 105:7899–7906.

Bramer, C., S. Dobler, J. Deckert, M. Stemmer, and G. Petschenka.2015. Na1/K1-ATPase resistance and cardenolide sequestration:basal adaptations to host plant toxins in the milkweed bugs(Hemiptera: Lygaeidae: Lygaeinae). Proceedings of the Royal Soci-ety B 282:20142346.

Brodie, E. D., and E. D. Brodie. 2015. Predictably convergent evolu-tion of sodium channels in the arms race between predators andprey. Brain, Behavior, and Evolution 86:48–57.

Cavender-Bares, J., D. D. Ackerly, D. A. Baum, and F. A. Bazzaz.2004. Phylogenetic overdispersion in Floridian oak communities.American Naturalist 163:823–843.

Christin, P.-A., M. Arakaki, C. P. Osborne, and E. J. Edwards. 2015.Genetic enablers underlying the clustered evolutionary origins ofC4 photosynthesis in angiosperms. Molecular Biology and Evolu-tion 32: 846–858.

Conner, J. K., and D. L. Hartl. 2004. A primer of ecological genetics.Sinauer, Sunderland, MA.

Conway-Morris, S. 2003. Life’s solution: inevitable humans in a lonelyuniverse. Cambridge University Press, Cambridge.

———. 2015. The runes of evolution: how the universe became self-aware. Templeton, West Conshohocken, PA.

Croyle, M. L., A. L. Woo, and J. B. Lingrel. 1997. Extensive randommutagenesis analysis of the Na1/K1-ATPase a subunit identifiesknown and previously unidentified amino acid residues that alterouabain sensitivity implications for ouabain binding. EuropeanJournal of Biochemistry 248:488–495.

Currie, A. 2013. Convergence as evidence. British Journal for thePhilosophy of Science 64:763–786.

Darwin, C. 1859. On the origin of species by means of natural selec-tion, or the preservation of favoured races in the struggle for life.Murray, London.

Denoeud, F., L. Carretero-Paulet, A. Dereeper, G. Droc, R. Guyot,M. Pietrella, C. Zheng, et al. 2014. The coffee genome provides in-

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S10 The American Naturalist

sight into the convergent evolution of caffeine biosynthesis. Sci-ence 345:1181–1184.

Dobler, S., S. Dalla, V. Wagschal, and A. A. Agrawal. 2012.Community-wide convergent evolution in insect adaptation to toxiccardenolides by substitutions in the Na,K-ATPase. Proceedings ofthe National Academy of Sciences of the USA 109:13040–13045.

Dobler, S., G. Petschenka, V. Wagschal, and L. Flacht. 2015. Conver-gent adaptive evolution—how insects master the challenge of car-diac glycoside-containing host plants. Entomologia Experimentaliset Applicata 157:30–39.

Donoghue, M. J. 2005. Key innovations, convergence, and success:macroevolutionary lessons from plant phylogeny. Paleobiology 31:77–93.

Edwards, E. J., D. S. Chatelet, B.-C. Chen, J. Y. Ong, S. Tagane, H.Kanemitsu, K. Tagawa, et al. 2017. Convergence, consilience, andthe evolution of temperate deciduous forests. American Naturalist190(suppl.):S87–S104.

Ehrlich, P. R., and P. H. Raven. 1964. Butterflies and plants: a studyin coevolution. Evolution 18:586–608.

Eisner, T. 1964. Catnip: its raison d’être. Science 146:1318–1320.Elena, S. F., and R. E. Lenski. 2003. Evolution experiments withmicroorganisms: the dynamics and genetic bases of adaptation.Nature Reviews Genetics 4:457–469.

Farhat, M. R., B. J. Shapiro, K. J. Kieser, R. Sultana, K. R. Jacobson,T. C. Victor, R. M. Warren, et al. 2013. Genomic analysis identi-fies targets of convergent positive selection in drug-resistant My-cobacterium tuberculosis. Nature Genetics 45:1183–1189.

Farrell, B. D., D. E. Dussourd, and C. Mitter. 1991. Escalation ofplant defense: do latex and resin canals spur plant diversification?American Naturalist 138:881–900.

Feeny, P. P. 1976. Plant apparency and chemical defense. Pages 1–40in J. W. Wallace and R. L. Mansell, eds. Biochemical interactionbetween plants and insects. Plenum, New York.

Felsenstein, J. 1985. Phylogenies and the comparative method. Ameri-can Naturalist 125:1–15.

ffrench-Constant, R. H., B. Pittendrigh, A. Vaughan, and N. An-thony. 1998. Why are there so few resistance-associated mutationsin insecticide target genes? Philosophical Transactions of theRoyal Society B 353:1685–1693.

Fine, P. V. A., I. Mesones, and P. D. Coley. 2004. Herbivores pro-mote habitat specialization by trees in amazonian forests. Science305:663–665.

Fine, P. V. A., Z. J. Miller, I. Mesones, S. Irazuzta, H. M. Appel,M. H. H. Stevens, I. Saaksjarvi, et al. 2006. The growth-defensetrade-off and habitat specialization by plants in Amazonianforests. Ecology 87:S150–S162.

Fischer, M. K., W. Völkl, R. Schopf, and K. H. Hoffmann. 2002. Age-specific patterns in honeydew production and honeydew compo-sition in the aphid Metopeurum fuscoviride: implications for ant-attendance. Journal of Insect Physiology 48:319–326.

Forister, M. L., Z. Gompert, C. C. Nice, G. W. Forister, and J. A.Fordyce. 2011. Ant association facilitates the evolution of dietbreadth in a lycaenid butterfly. Proceedings of the Royal SocietyB 278:1539–1547.

Fraenkel, G. 1959. The raison d’etre of secondary plant substances.Science 129:1466–1470.

Fukami, T., M. Nakajima, C. Fortunel, P. V. A. Fine, C. Baraloto,S. E. Russo, and K. G. Peay. 2017. Geographical variation in com-munity divergence: insights from tropical forest monodominanceby ectomycorrhizal trees. American Naturalist 190(suppl.):S105–S122.

This content downloaded from 132All use subject to University of Chicago Press Term

Futuyma, D. J. 2010. Evolutionary constraint and ecological conse-quences. Evolution 64:1865–1884.

Futuyma, D. J., and A. A. Agrawal. 2009. Macroevolution and the bi-ological diversity of plants and herbivores. Proceedings of the Na-tional Academy of Sciences of the USA 106:18054–18061.

Gilchrist, A. S., and L. Partridge. 1999. A comparison of the geneticbasis of wing size divergence in three parallel body size clines ofDrosophila melanogaster. Genetics 153:1775–1787.

Gillespie, R. 2004. Community assembly through adaptive radiationin Hawaiian spiders. Science 303:356–359.

Gould, S. J. 1990. Wonderful life: the Burgess Shale and the nature ofhistory. Norton, New York.

———. 2002. The structure of evolutionary theory. Harvard Uni-versity Press, Cambridge, MA.

Greene, H. W. 2005. Organisms in nature as a central focus for bi-ology. Trends in Ecology and Evolution 20:23–27.

———. 2017. Evolutionary scenarios and primate natural history.American Naturalist 190(suppl.):S69–S86.

Heard, S. B., and D. L. Hauser. 1995. Key evolutionary innovationsand their ecological mechanisms. Historical Biology 10:151–173.

Heil, M. 2008. Indirect defence via tritrophic interactions. New Phy-tologist 178:41–61.

Huey, R. B., G. W. Gilchrist, M. L. Carlson, D. Berrigan, and L. S.Serra. 2000. Rapid evolution of a geographic cline in size in an in-troduced fly. Science 287:308–309.

Hunter, J. P. 1998. Key innovations and the ecology of macroevolu-tion. Trends in Ecology and Evolution 13:31–36.

Inouye, B. D., and A. A. Agrawal. 2004. Ant mutualists alter thecomposition and attack rate of the parasitoid community forthe gall wasp Disholcaspis eldoradensis (Cynipidae). EcologicalEntomology 29:692–696.

Jensen, N. B., M. Zagrobelny, K. Hjernø, C. E. Olsen, J. Houghton-Larsen, J. Borch, B. L. Møller, et al. 2011. Convergent evolution inbiosynthesis of cyanogenic defence compounds in plants andinsects. Nature Communications 2:273.

Johnson, M. T., A. R. Ives, J. Ahern, and J. P. Salminen. 2014. Mac-roevolution of plant defenses against herbivores in the eveningprimroses. New Phytologist 203:267–279.

Kaminski, L. A., A. V. Freitas, and P. S. Oliveira. 2010. Interactionbetween mutualisms: ant-tended butterflies exploit enemy-freespace provided by ant-treehopper associations. American Natural-ist 176:322–334.

Kraft, N. J., W. K. Cornwell, C. O. Webb, and D. D. Ackerly. 2007.Trait evolution, community assembly, and the phylogenetic struc-ture of ecological communities. American Naturalist 170:271–283.

Kryazhimskiy, S., D. P. Rice, E. R. Jerison, and M. M. Desai. 2014.Global epistasis makes adaptation predictable despite sequence-level stochasticity. Science 344:1519–1522.

Kursar, T. A., and P. D. Coley. 2003. Convergence in defense syn-dromes of young leaves in tropical rainforests. Biochemical Sys-tematics and Ecology 31:929–949.

Lenski, R. E. 2017. Convergence and divergence in a long-termexperiment with bacteria. American Naturalist 190(suppl.):S57–S68.

Lenski, R. E., M. J. Wiser, N. Ribeck, Z. D. Blount, J. R. Nahum, J. J.Morris, L. Zaman, et al. 2015. Sustained fitness gains and variabil-ity in fitness trajectories in the long-term evolution experimentwith Escherichia coli. Proceedings of the Royal Society B 282:20152292.

Lomolino, M. V. 2005. Body size evolution in insular vertebrates: gen-erality of the island rule. Journal of Biogeography 32:1683–1699.

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergence and Natural History S11

Losos, J. 2009. Lizards in an evolutionary tree: ecology and adaptiveradiation of anoles. University of California Press, Oakland.

Losos, J. B. 2011. Convergence, adaptation, and constraint. Evolution65:1827–1840.

Losos, J. B., T. R. Jackman, A. Larson, K. de Queiroz, and L. Rodriguez-Schettino. 1998. Contingency and determinism in replicated adaptiveradiations of island lizards. Science 279:2115–2118.

Luo, Z.-X. 2007. Transformation and diversification in early mam-mal evolution. Nature 450:1011–1019.

Mahler, D. L., M. G. Weber, C. E. Wagner, and T. Ingram. 2017. Pat-tern and process in the comparative study of convergent evolu-tion. American Naturalist 190(suppl.):S13–S28.

Manceau, M., V. S. Domingues, C. R. Linnen, E. B. Rosenblum, andH. E. Hoekstra. 2010. Convergence in pigmentation at multiplelevels: mutations, genes and function. Philosophical Transactionsof the Royal Society B 365:2439–2450.

Marazzi, B., C. Ané, M. F. Simon, A. Delgado-Salinas, M. Luckow,and M. J. Sanderson. 2012. Locating evolutionary precursors ona phylogenetic tree. Evolution 66:3918–3930.

Mason, C. M., A. W. Bowsher, B. L. Crowell, R. M. Celoy, C. J. Tsai,and L. A. Donovan. 2016. Macroevolution of leaf defenses andsecondary metabolites across the genus Helianthus. New Phy-tologist 209:1720–1733.

McCune, A. R., and R. L. Carlson. 2004. Twenty ways to lose yourbladder: common natural mutants in zebrafish and widespreadconvergence of swim bladder loss among teleost fishes. Evolutionand Development 6:246–259.

Meyer, J. R., A. A. Agrawal, R. T. Quick, D. T. Dobias, D. Schneider,and R. E. Lenski. 2010. Parallel changes in host resistance to viralinfection during 45,000 generations of relaxed selection. Evolu-tion 64:3024–3034.

Mitter, C., B. Farrell, and B. Wiegmann. 1988. The phylogeneticstudy of adaptive zones: has phytophagy promoted insect diversi-fication? American Naturalist 132:107–128.

Mooney, K. A., R. Halitschke, A. Kessler, and A. A. Agrawal. 2010.Evolutionary trade-offs in plants mediate the strength of trophiccascades. Science 327:1642–1644.

Natarajan, C., F. G. Hoffmann, R. E. Weber, A. Fago, C. C. Witt, andJ. F. Storz. 2016. Predictable convergence in hemoglobin functionhas unpredictable molecular underpinnings. Science 354:336–339.

Ng, J., and S. D. Smith. 2016. Widespread flower color convergence inSolanaceae via alternate biochemical pathways. New Phytologist209:407–417.

Nicholls, J. A., G. Melika, G. N. Stone, and M. A. McPeek. 2016.Sweet tetra-trophic interactions: multiple evolution of nectar se-cretion, a defensive extended phenotype in Cynipid gall wasps.American Naturalist 189:67–77.

Parker, J., G. Tsagkogeorga, J. A. Cotton, Y. Liu, P. Provero, E. Stupka,and S. J. Rossiter. 2013. Genome-wide signatures of convergentevolution in echolocating mammals. Nature 502:228–231.

Petschenka, G., and A. A. Agrawal. 2015. Milkweed butterfly resis-tance to plant toxins is linked to sequestration, not coping witha toxic diet. Proceedings of the Royal Society B 282:20151865.

Petschenka, G., V. Wagschal, M. von Tschirnhaus, A. Donath, andS. Dobler. 2017. Convergently evolved toxic secondary metabolitesin plants drive the parallel molecular evolution of insect resis-tance. American Naturalist 190(suppl.):S29–S43.

Pierce, N. E., M. F. Braby, A. Heath, D. J. Lohman, J. Mathew, D. B.Rand, and M. A. Travassos. 2002. The ecology and evolution ofant association in the Lycaenidae (Lepidoptera). Annual Reviewof Entomology 47:733–771.

This content downloaded from 132All use subject to University of Chicago Press Term

Poulin, R., and A. S. Grutter. 1996. Cleaning symbioses: proximateand adaptive explanations. Bioscience 46:512–517.

Price, E., D. Rice, and J. Lingrel. 1990. Structure-function studies ofNa,K-ATPase: site-directed mutagenesis of the border residuesfrom the H1-H2 extracellular domain of the a subunit. Journalof Biological Chemistry 265:6638–6641.

Protas, M. E., C. Hersey, D. Kochanek, Y. Zhou, H. Wilkens, W. R.Jeffery, L. I. Zon, et al. 2006. Genetic analysis of cavefish revealsmolecular convergence in the evolution of albinism. Nature Ge-netics 38:107–111.

Rabosky, D. L. 2014. Automatic detection of key innovations, rate shifts,and diversity-dependence on phylogenetic trees. PLoSONE 9:e89543.

Reichstein, T., J. Voneuw, J. A. Parsons, and M. Rothschild. 1968.Heart poisons in monarch butterfly. Science 161:861–866.

Riehle, M. M., A. F. Bennett, and A. D. Long. 2001. Genetic architec-ture of thermal adaptation in Escherichia coli. Proceedings of theNational Academy of Sciences of the USA 98:525–530.

Rosenblum, E. B., C. E. Parent, E. T. Diepeveen, and K. Bi. 2017.Convergent phenotypic evolution despite contrasting demo-graphic histories in the fauna of White Sands. American Natural-ist 190(suppl.):S44–S56.

Rothschild, M., J. von Euw, and T. Reichstein. 1970. Cardiac glyco-sides in oleander aphid, Aphis nerii. Journal of Insect Physiology16:1141–1145.

———. 1973. Cardiac glycosides (heart poisons) in the polka-dotmoth Syntomeida epilais Walk. (Ctenuchidae: Lep.) with some ob-servations on the toxic qualities of Amata (p Syntomis) phegea(L.). Proceedings of the Royal Society B 183:227–247.

Ruxton, G. D., T. N. Sherratt, and M. P. Speed. 2004. Avoiding attack:the evolutionary ecology of crypsis, warning signals and mimicry.Oxford University Press, Oxford.

Sage, R. F., P.-A. Christin, and E. J. Edwards. 2011. The C4 plant lineagesof planet Earth. Journal of Experimental Botany 62:3155–3169.

Schluter, D. 1986. Tests for similarity and convergence of finch com-munities. Ecology 67:1073–1085.

Schwenk, K. 1994/1995. A utilitarian approach to evolutionary con-straint. Zoology 98:251–262.

Shingleton, A. W., and D. L. Stern. 2003. Molecular phylogenetic ev-idence for multiple gains or losses of ant mutualism within theaphid genus Chaitophorus. Molecular Phylogenetics and Evolu-tion 26:26–35.

Shingleton, A. W., D. L. Stern, and W. A. Foster. 2005. The origin ofa mutualism: a morphological trait promoting the evolution ofant-aphid mutualisms. Evolution 59:921–926.

Smith, S. D., S. Wang, and M. D. Rausher. 2013. Functional evolu-tion of an anthocyanin pathway enzyme during a flower colortransition. Molecular Biology and Evolution 30:602–612.

Stayton, C. T. 2008. Is convergence surprising? an examination ofthe frequency of convergence in simulated datasets. Journal ofTheoretical Biology 252:1–14.

———. 2015. What does convergent evolution mean? the interpre-tation of convergence and its implications in the search for limitsto evolution. Interface Focus 5:20150039.

Stern, D. L. 2013. The genetic causes of convergent evolution. NatureReviews Genetics 14:751–764.

Storz, J. F., and H. Moriyama. 2008. Mechanisms of hemoglobin ad-aptation to high altitude hypoxia. High Altitude Medicine and Bi-ology 9:148–157.

Styrsky, J. D., and M. D. Eubanks. 2007. Ecological consequences ofinteractions between ants and honeydew-producing insects. Pro-ceedings of the Royal Society B 274:151–164.

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S12 The American Naturalist

Tenaillon, O., A. Rodríguez-Verdugo, R. L. Gaut, P. McDonald, A. F.Bennett, A. D. Long, and B. S. Gaut. 2012. The molecular diversityof adaptive convergence. Science 335:457–461.

Turner, G. F. 2007. Adaptive radiation of cichlid fish. Current Biol-ogy 17:R827–R831.

Ujvari, B., N. R. Casewell, K. Sunagar, K. Arbuckle, W. Wüster, N. Lo,D. O’Meally, et al. 2015.Widespread convergence in toxin resistanceby predictable molecular evolution. Proceedings of the NationalAcademy of Sciences of the USA 112:11911–11916.

van Ditmarsch, D., K. E. Boyle, H. Sakhtah, J. E. Oyler, C. D. Nadell,É. Déziel, L. E. Dietrich, et al. 2013. Convergent evolution of hy-perswarming leads to impaired biofilm formation in pathogenicbacteria. Cell Reports 4:697–708.

Völkl, W., J. Woodring, M. Fischer, M. W. Lorenz, and K. H.Hoffmann. 1999. Ant-aphid mutualisms: the impact of honeydewproduction and honeydew sugar composition on ant preferences.Oecologia 118:483–491.

von Euw, J., L. Fishelson, J. A. Parsons, T. Reichstein, and M.Rothschild. 1967. Cardenolides (heart poisons) in a grasshopperfeeding on milkweeds. Nature 214:35–39.

von Euw, J., T. Reichstein, and M. Rothschild. 1971. Heart poisons(cardiac glycosides) in the lygaeid bugs Caenocoris nerii andSpilostethus pandurus. Insect Biochemistry 1:373–384.

“Phases of Kangaroo Life. At Peace.” From “General Notes

This content downloaded from 132All use subject to University of Chicago Press Term

Weber, M. G., and A. A. Agrawal. 2012. Phylogeny, ecology, and thecoupling of comparative and experimental approaches. Trends inEcology and Evolution 27:394–403.

———. 2014. Defense mutualisms enhance plant diversification.Proceedings of the National Academy of Sciences of the USA111:16442–16447.

Weber, M. G., and K. H. Keeler. 2013. The phylogenetic distributionof extrafloral nectaries in plants. Annals of Botany 111:1251–1261.

Weinreich, D. M., N. F. Delaney, M. A. DePristo, and D. L. Hartl.2006. Darwinian evolution can follow only very few mutationalpaths to fitter proteins. Science 312:111–114.

Wiens, J. J., R. T. Lapoint, and N. K. Whiteman. 2015. Herbivory in-creases diversification across insect clades. Nature Communications6:8370.

Yang, S., J. Li, X. Zhang, Q. Zhang, J. Huang, J.-Q. Chen, D. L. Hartl,et al. 2013. Rapidly evolving R genes in diverse grass species con-fer resistance to rice blast disease. Proceedings of the NationalAcademy of Sciences of the USA 110:18572–18577.

Editor: Judith L. Bronstein

: Zoölogy” (The American Naturalist, 1884, 12:1271–1279).

.236.027.111 on July 22, 2017 05:12:23 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Pattern and Process in the Comparative Study

of Convergent Evolution*

D. Luke Mahler,1,† Marjorie G. Weber,2 Catherine E. Wagner,3 and Travis Ingram4

1. Department of Ecology and Evolutionary Biology, University of Toronto, Toronto, Ontario M5S 3B2, Canada; 2. Department of PlantBiology, Michigan State University, East Lansing, Michigan 48824; 3. Biodiversity Institute and Department of Botany, University ofWyoming, Laramie, Wyoming 82071; 4. Department of Zoology, University of Otago, Dunedin, Otago 9016, New Zealand

abstract: Understanding processes that have shaped broad-scalebiodiversity patterns is a fundamental goal in evolutionary biology.The development of phylogenetic comparative methods has yieldeda tool kit for analyzing contemporary patterns by explicitly modelingprocesses of change in the past, providing neontologists tools for ask-ing questions previously accessible only for select taxa via the fossil rec-ord or laboratory experimentation. The comparative approach, how-ever, differs operationally from alternative approaches to studyingconvergence in that, for studies of only extant species, convergencemust be inferred using evolutionary process models rather than beingdirectly measured. As a result, investigation of evolutionary patternand process cannot be decoupled in comparative studies of conver-gence, even though such a decoupling could in theory guard againstadaptationist bias. Assumptions about evolutionary process underly-ing comparative tools can shape the inference of convergent pattern insometimes profound ways and can color interpretation of such pat-terns. We discuss these issues and other limitations common to mostphylogenetic comparative approaches and suggest ways that they canbe avoided in practice. We conclude by promoting a multipronged ap-proach to studying convergence that integrates comparative methodswith complementary tests of evolutionarymechanisms and includes eco-logical and biogeographical perspectives. Carefully employed, the com-parative method remains a powerful tool for enriching our understand-ing of convergence inmacroevolution, especially for investigation ofwhyconvergence occurs in some settings but not others.

Keywords: convergence, phylogenetic comparative methods, adap-tive radiation, evolutionary process, adaptation.

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† Corresponding author; e-mail: luke.mahler@utoronto.ca.ORCIDs: Mahler, http://orcid.org/0000-0001-6483-3667; Weber, http://orcid

.org/0000-0001-8629-6284; Wagner, http://orcid.org/0000-0001-8585-6120; In-gram, http://orcid.org/0000-0003-0709-5260.

Am. Nat. 2017. Vol. 190, pp. S13–S28. q 2017 by The University of Chicago.0003-0147/2017/190S1-57359$15.00. All rights reserved.DOI: 10.1086/692648

This content downloaded from 132All use subject to University of Chicago Press Term

Introduction

The phenomenon of phenotypic convergence plays a fun-damental role in the study of organic evolution. Althoughconvergence itself is not necessarily indicative of any par-ticular evolutionary process (Losos 2011a; Speed and Ar-buckle 2016), the repeated appearance of similar forms indisparate lineages stands in apparent contrast to the ex-pected pattern of divergence over time during evolutionand thus demands an explanation (Wake et al. 2011). Con-vergent evolution has been attributed to a great diversity ofcauses, at times being invoked as evidence for the impor-tance of multiple and sometimes opposing evolutionaryprocesses. As such, confusion persists around the connec-tion between patterns of convergence, mechanisms of evo-lution, andmodeled processes in comparative methods (seebox 1 for definitions of these terms).Our aim is to clarify these concepts and show how they

relate to common assumptions in the comparative studyof convergence, recommending best practices and advocat-ing integrative ways to link convergent patterns with mech-anistic hypotheses. We begin by reviewing the comparativestudy of phenotypic convergence in continuously valuedtraits and the factors that make it uniquely challenging tostudy deductively. We then discuss the relationship betweenpattern and process in the comparative study of convergence,giving special attention to the facts that (1) all comparativetools for studying convergence make inductive inferencesabout convergence and thus assume an underlying modelof evolution and (2) despite the assumption of amodeled pro-cess, there can often be a many-to-one mapping of real evo-lutionary processes to modeled processes. Finally, we arguethat because these limitations can hinder interpretation, in-tegration of comparative phylogenetic models of conver-gence with other forms of inference is needed to increaseconfidence in links between pattern and process. We sug-gest several research directions to improve future prospectsfor gaining meaningful insights about evolutionary pro-

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Box 1: Definitions

Convergence

We define convergent phenotypic evolution (or “convergence”) as the pattern of evolution in which species intwo independently evolving lineages become phenotypically similar. Importantly, such a pattern is defined as con-vergence regardless of the evolutionary process that gave rise to it (Stayton 2015a). In this article, we will discussconvergent patterns in which lineages evolve greater phenotypic similarity than was exhibited by their ancestors.We do not discount convergence definitions that include patterns in which two or more lineages independentlyevolve similar traits even when their ancestors were also similar (i.e., parallelism; Arendt and Reznick 2008a,2008b; Leander 2008; Scotland 2011; Wake et al. 2011; Rosenblum et al. 2014); nonetheless, for clarity and becausemost comparative convergence tools are specialized for the study of the evolution of greater similarity among de-scendants than ancestors, we do not discuss parallelism in this article. As any statement about convergence requiresinformation about ancestral phenotypes, our discussion of the role of evolutionary process models in inferring an-cestral phenotypes using phylogenetic comparative methods (see the main text) should be relevant regardless of thespecific definition of convergence assumed.

Evolutionary Process/Evolutionary Mechanism

Evolutionary processes, which we use interchangeably with the term “evolutionary mechanisms,” refer to specificunderlying agents of evolutionary change, such as genetic drift or natural selection, that give rise to observedpatterns of organismal diversity (Eldredge and Cracraft 1980; Chapleau et al. 1988). A fundamental goal in evolu-tionary biology is to test hypotheses about the evolutionary mechanisms that shape patterns of biological diversitythrough time.

Evolutionary Process Model

In this article, the term “evolutionary process model” refers to a model of the evolutionary process assumed by aphylogenetic comparative method. Most such methods assume phenomenological evolutionary models, whichmeans that they are thought to represent macroevolutionary expectations arising from a given evolutionary processbut do not explicitly model the mechanism of evolutionary change itself. For this reason, some evolutionary processmodels may be consistent with more than one mechanism of evolutionary change (see the main text).

Ecological Mechanism

We refer to “ecological mechanism” as an ecological interaction that is an agent of natural selection on an or-ganism. Ecological mechanisms describe how an organism interacts with its environment or with other species, andinclude competition, mutualism, and abiotic tolerances, among others. There has been relatively little investigationof how different ecological mechanisms are expected to shape patterns of convergent evolution.

S14 The American Naturalist

cesses from the comparative study of convergent patterns,including extended model development, pairing compara-tive study with other approaches for studying the evolution-ary process, and better incorporating fossil information.

The Potential Causes of Macroevolutionary Convergence

The investigation of convergence plays an important role inevolutionary study. To many, instances of convergence are

This content downloaded from 132All use subject to University of Chicago Press Term

interesting as evidence for adaptation or for the importanceof ecology in the evolution of phenotypic diversity. For ex-ample, the evolution of pale dorsal coloration in numerousvertebrates and invertebrates inhabiting White Sands Na-tional Monument indisputably reflects adaptation for in-creased crypsis (Rosenblum et al. 2010), and the repeated evo-lution of cold tolerance in distantly related conifers is clearlyattributable to adaptation to similar environmental condi-tions (Yeaman et al. 2016). Large-scale convergence betweenwhole faunas occurring in similar ecological communities

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Pattern, Process, and Convergence S15

and environments suggests that evolution can in some cir-cumstances exhibit a surprising degree of predictability andthat ecological factors can repeatedly and predictably shapemacroevolutionary diversification (Nevo 1979; ConwayMor-ris 2010; Mahler et al. 2013; Esquerré and Keogh 2016; Moenet al. 2016).

Alternatively, convergence has been taken as evidence forconstraints on the production of variation (Haldane 1932;Maynard Smith et al. 1985; Schluter 1996), manifested atone or more hierarchical levels of biological organization(Wake and Larson 1987; Wake et al. 2011). Such constraintscan arise from biased mutation, pleiotropic gene networks,structural limitations, or limits on phenotypic variation im-posed by ontogenetically nested developmental sequences(Gould 1980; Alberch 1982; Oster and Alberch 1982; Ar-nold 1992; McCune and Carlson 2004; Brakefield 2011;Streisfeld and Rausher 2011; Stern 2013). For example,Bright et al. (2016) concluded that the vastmajority of cranio-facial variation in predatory birds was attributable to con-served patterns of allometry and covariation between traits(phenotypic integration), with only a small amount of resid-ual variation explained by feeding ecology. Convergenceresulting from such factors suggests that genetic or develop-mental constraints play an important and perhaps dominantrole in shaping the evolution of phenotypic diversity (Gould1980, 2002; Alberch 1982; Wake and Larson 1987).

Finally, some degree of evolutionary convergencemay bean expected outcome of chance, especially for traits with lowdimensionality (Wagner 2000; Stayton 2008). These possibil-ities are not mutually exclusive, of course, and convergencedue to chance may be most likely in scenarios in whichconstraints on the generation of variation restrict evolu-tionary outcomes to a small set of phenotypes with compa-rable fitness (Losos 2011a; Spor et al. 2014). Likewise, someauthors have uncovered phylogenetic patterns suggestingthat certain convergent adaptations are hierarchically con-strained by the presence or absence of preadaptations (Ma-razzi et al. 2012; Beaulieu et al. 2013). For example, the con-vergent evolution of arboreal adaptations in oribatid mitesappears contingent on the prior evolution of sexual repro-duction and strong sclerotization (Maraun et al. 2009).

Challenges to Measuring and StudyingConvergent Evolution

Despite its importance in evolutionary inquiry, conver-gence is difficult to directly identify, and once identified,it is difficult to ascribe to underlying mechanisms with cer-tainty (Maynard Smith et al. 1985). A principal challenge instudying convergence is that it is rarely possible to study us-ing deductive inference due to the difficulty of observingboth ancestor and descendant phenotypes. Empirically, thisis generally possible only in exceptional circumstances, such

This content downloaded from 132All use subject to University of Chicago Press Term

as when the fossil record preserves unambiguous ancestor-descendant sequences (e.g., Bell 1987; McCune 1987), whensource populations that gave rise to convergent daughterpopulations are still extant (e.g., Hoekstra et al. 2006; Ro-senblum et al. 2010), and when convergence is particularlyrapid (e.g., Pascoal et al. 2014), including in laboratory ex-periments on organisms with very short generation times(e.g., Meyer et al. 2012; Spor et al. 2014). Studies documentingconvergence in this way represent some of the best and mostcherished evidence for convergent evolution, but they are un-common and are limited in what they can tell us about theevolution of convergent patterns across the tree of life.Convergence has a long history of study despite the lim-

ited opportunity for deductive inference, but aside fromexceptional cases such as those described above, much his-torical study of convergence has been qualitative in nature.Most of the canonical examples of convergence describedin introductory biology textbooks, such as the streamlinedprofiles of fast-moving pelagic vertebrates or the winter pel-age of Arctic foxes and snowshoe hares, are so visually strik-ing and occur in such distant relatives that there can be littlequestion about their convergent origins. What was histori-cally lacking was a cohesive quantitative framework for in-ferring the trajectories of convergence; the lack of such astructure imposed formidable limits on the study of theultimate drivers of convergent evolution. This frameworkemerged with the union of the comparative method with anexplicitly phylogenetic perspective in the 1980s and 1990s(Felsenstein 1985; Harvey and Pagel 1991).

Phylogenetic Approaches to Studying Convergence

The advent of phylogenetic comparative approaches to study-ing trait evolution expanded the scope of convergence studies,making it possible to test quantitative hypotheses about con-vergent evolution in any group with sufficient phylogeneticand phenotypic information. This development effectivelyopened the quantitative study of convergence, previouslylimited to exceptional cases, to the entire tree of life. Whenviewed within a phylogenetic comparative framework, re-peated convergence of any kind provides researchers with adegree of statistical replication rarely afforded to studentsof the explicitly historical science of evolution. Regardlessof the question of interest, the repeated evolution of similarphenotypes in disparate lineages provides independent rep-licates in a grand, unplanned evolutionary experiment. Theability to repeatedly query putative cause and evolutionary ef-fect allows investigators to overcome the risks of “just so” sto-rytelling (Gould and Lewontin 1979) in the study of adapta-tion, structural constraint, and even chance (Maynard Smithet al. 1985; Harvey and Pagel 1991; Losos 2011a).The development of tools for investigating convergence

accelerated especially rapidly during the last decade (Speed

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S16 The American Naturalist

and Arbuckle 2016). New inductive tools provide a wide va-riety of methods for quantifying convergence—includingits occurrence, frequency, extent, and historical trajectory.Such tools come in a variety of forms, the details of whichare described in depth elsewhere (see Mahler and Ingram2014; Stayton 2015a; Arbuckle and Speed 2016; Speedand Arbuckle 2016 for more in-depth descriptions). Most,however, can be classed into one of three categories: (1) sta-tistical indices, which measure an expected emergent fea-ture of convergent phenotypic evolution on a phylogeny;(2) ancestor reconstruction methods, in which ancestralphenotypes are estimated under some assumed evolutionarymodel and then used to identify and quantify convergencepatterns; and (3) model-fitting approaches, in which evolu-tionary models explicitly incorporating processes expectedto cause convergence are parameterized and compared tomodels in which any convergence occurs by chance.

All existing comparative methods for studying conver-gence have particular limitations and weaknesses, as we willdiscuss below. More importantly, however, is that all of thesetools make inductive inferences about convergence, an ap-proach that has practical consequences for the design and in-terpretation of comparative studies. Regardless of themethodused, the results must be interpreted in light of an assumedmodel of the evolutionary process; as a result, it is not possibleto decouple the quantification of convergent pattern from thestudy of evolutionary process using comparative data, as hasbeen recently recommended (Stayton 2015a). This issue iscompounded by the fact that many widely used evolutionaryprocess models may plausibly represent multiple underlyingevolutionary mechanisms—a potential many-to-one map-ping of mechanism to model. At the heart of these issues isthe complex relationship between pattern and process incomparative biology.

Pattern and Process

An important property of any evolutionary phenomenonis the extent to which it represents a pattern versus a pro-cess (box 1). For example, the term “adaptation” is mean-ingfully defined as both a process (e.g., adaptation occurswhen a population evolves greater fitness via natural selec-tion) and a pattern (e.g., an adaptation is a trait that in-creases an individual’s fitness compared to individualswithout that trait; Gould and Vrba 1982; Futuyma 2005).In the case of evolutionary convergence, we feel that in al-most all scenarios convergence is best defined as a pattern(Stayton 2015a). Convergence is less meaningful as a pro-cess, because convergent evolution is nearly always an emer-gent outcome of evolutionary processes operating indepen-dently in multiple lineages rather than any intrinsicallyconvergent processes. For example, short-limbed twig spe-cialist anoles on different Antillean islands are similar be-

This content downloaded from 132All use subject to University of Chicago Press Term

cause they have adapted to similar arboreal substrates(Williams 1983; Losos 2009), not because they have been se-lected to resemble one another. There are a few exceptionswhere species are directly selected to converge with one an-other, such as mimicry complexes (Endler 1981) or charac-ter convergence driven by competition for nonsubstitutableresources (MacArthur and Levins 1967; Abrams 1987; Foxand Vasseur 2008), but otherwise the evolutionary processesresponsible for increasing similarity in a converging lineageare blind to the phenotype of the lineage to which it is con-verging. As standard evolutionary theory can explain theprocesses by which independent lineages evolve to be moresimilar, there is no need for a special theory of convergence(Speed and Arbuckle 2016), and convergence is thus bestdefined as a pattern.Although we agree with several recent reviews that con-

vergence should be defined as a pattern, we argue that whenusing the comparative approach, convergence must bestudied with the evolutionary process in mind. We there-fore reject arguments that comparative analyses of conver-gence should proceed in a two-step manner—first testingfor convergent pattern, and then investigating potentialevolutionary processes responsible for this pattern (Stayton2015a, 2015b; Speed and Arbuckle 2016). In a recent re-view, Stayton (2015a) made a case for this two-step ap-proach, specifically criticizing the use of process-based com-parative tools for identifying convergence. Stayton’s concernis defensible—in applying an evolutionary model to com-parative data (e.g., multiple-optimum Ornstein-Uhlenbeck[OU]; Butler and King 2004), the investigator assumes thatthe process being modeled is an appropriate representationof trait evolution in lineages of interest. In the absence ofappropriate model comparison, this could bias an investi-gation toward a particular evolutionary explanation forconvergence. The risk of bias is arguably greatest for adap-tive explanations (sensu Gould and Lewontin 1979), andStayton marshaled evidence for adaptationist bias in con-vergence definitions provided in many prominent biologytexts (Stayton 2015a). To safeguard against adaptationistbiases, Stayton recommended that comparative studies firstemploy process-neutral statistical tools to identify andmea-sure macroevolutionary convergence and then, patterns inhand, test alternative hypotheses about the evolutionary pro-cesses that may have given rise to these patterns.While the goal to investigate convergence without mak-

ing assumptions about process is a worthy one, in phylo-genetic comparative biology it is impossible to study evo-lutionary pattern and process independently (Pagel andHarvey 1989; Harvey and Pagel 1991; Freckleton et al.2011; Hunt 2012). The purpose of phylogenetic compara-tive methods is to study patterns that inherently arise as aconsequence of evolutionary processes, both to under-stand how history has shaped these patterns and to infer

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Pattern, Process, and Convergence S17

the processes associated with this history. As referencedabove, phylogenetic comparative biology is largely an in-ductive science, due to the usual lack of direct observa-tional evidence of historical processes and patterns. Thecomparative method provides a framework for testing hy-potheses about these phenomena, but an essential compo-nent of this framework is the assumption of evolutionaryprocess models.

Evolutionary Process Models Underlying Testsof Convergent Pattern

In the case of convergence, the assumption of an evolu-tionary process model is required at some step (often im-plicitly) by all available comparative tools. The most widelyassumedmodel is Brownianmotion, which is used tomodeltrait evolution in a variety of contexts. Brownianmotion is avery simple model that represents the expectations of con-tinuous phenotypic evolution under neutral genetic drift(Lande 1976; Felsenstein 1988). Like most models used inphylogenetic comparative methods, Brownian motion is aphenomenological model of the evolution of mean species-level characters, reflecting macroevolutionary expectationsbut not incorporating microevolutionary mechanisms. It isoften used to represent a hypothesis of neutral evolution(especially as a null model), but some have argued for itsutility in representing other evolutionary mechanisms suchas fluctuating directional selection or adaptive radiation ona dynamic adaptive landscape (although with importantcaveats; Felsenstein 1988; O’Meara et al. 2006). In studiesof adaptation, a popular generalization of Brownianmotionis the Ornstein-Uhlenbeckmodel (Hansen 1997; Butler andKing 2004; O’Meara and Beaulieu 2014). The OU modelincludes a Brownian drift term as well as a parameter de-scribing the strength of attraction to some optimum value.Extensions allow different lineages to be attracted to differ-ent optima, whichmay be interpreted as peaks on an adaptivelandscape. Unlike Brownian motion, specific OUmodels canmodel processes consistent with deterministic evolutionaryconvergence, though it is important to note that both are evo-lutionary process models (box 1). Recently developed Lévyprocess models represent an alternative generalization ofBrownian motion in which a Brownian drift process is punc-tuated by large, instantaneous shifts in trait value (i.e., evolu-tionary jumps; Eastman et al. 2013; Landis et al. 2013). Lévyprocess models lack parameters specifically expected to pro-duce convergence but can be used to test hypotheses aboutthe frequency of convergent jumps in groups for which thephenotypic similarity of certain species has already been es-tablished (Eastman et al. 2013).

Although all comparative methods for studying conver-gence employ evolutionary models in one fashion or an-other, the role and potential impact of the model vary across

This content downloaded from 132All use subject to University of Chicago Press Term

approaches. Index methods implicitly use process models togenerate a frame of reference with which to compare puta-tively convergent evolutionary patterns. For example, theWheatsheaf index (Arbuckle et al. 2014) uses pairwise phy-logenetic distances as a yardstick for evaluating and thenweighting observed pairwise trait differences, to be con-trasted with the correspondence between pairwise phy-logenetic and trait differences expected under Brownianmotion. Index tools have been described as “process-neutral”or “process-free” (Stayton 2015a; Speed and Arbuckle 2016),which is accurate in the sense that convergent patterns are notassumed to have evolved under any given evolutionarymech-anism. However, these measures are useful only in referenceto expectations under a particular evolutionary processmodel,which is sometimes unspecified but most often Brownianmotion. Furthermore, because no underlying historical pro-cess model is applied to the data themselves, these tools canbe limited by an inability to distinguish convergence fromother causes of evolutionary similarity between distant rel-atives, such as a simple lack of divergence (Stayton 2015a;Speed and Arbuckle 2016). We suggest that additional in-sights may be gained by comparing statistical indices to dis-tributions simulated under alternative models of the under-lying process (sensu Slater and Pennell 2014).Ancestral state reconstruction (ASR) methods critically

rely on an assumed model for quantification of both thefrequency and the strength of convergence. Although sev-eral kinds of ASR methods have been used to assess con-vergence, they all model a historical trajectory of evolutionunder an assumed model (almost always Brownian mo-tion) and then analyze estimated ancestral phenotypes todetect or quantify convergence (reviewed in Stayton 2015a,2015b; Arbuckle and Speed 2016; Speed and Arbuckle2016). ASR methods have been used to study convergencesince the dawn of the comparative methods era (e.g., Dono-ghue 1989; Brooks and McLennan 1991; Losos 1992) buthave gained renewed popularity with the recent developmentof phylomorphospace tools for visualizing evolutionary tra-jectories and quantifying the frequency and strength of con-vergence (Sidlauskas 2008; Stayton 2011, 2015b). Stayton(2015b) and Speed and Arbuckle (2016) classified availableASR methods as process-free on the grounds that they donot assume that convergent evolutionary patterns were theresult of adaptivemechanisms. These approaches are not trulyprocess-free, however—they rely on parameter estimatesfrom a model that assumes the observed patterns evolvedunder a process consistent with Brownian motion, suchas genetic drift (as well as some, but certainly not all, alter-native evolutionary mechanisms; Felsenstein 1988; Hansenand Martins 1996; O’Meara et al. 2006). ASR methods canemploy alternative macroevolutionary process models, in-cluding models with explicitly adaptive processes, so longas it is possible to reconstruct ancestral states under such

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S18 The American Naturalist

models (e.g., Elliot and Mooers 2014; Uyeda and Harmon2014). However, this is rarely done, perhaps because toolsfor carrying out ASR have not kept pace with the rapid de-velopment of methods for fitting alternative models of traitevolution.

Naturally, evolutionary process models play a promi-nent role in methods for studying convergence that are ex-plicitly based on model fitting. Such tools take one of twoapproaches. In the first, an investigator parameterizes anevolutionary process model in a way that explicitly incorpo-rates hypothesized convergence events, fits the model todata, and then compares the fit to that of an alternativemodellacking convergence. This is most commonly done usingmultiple-optimum OU models to represent the evolutionof a clade on an adaptive landscape, with occasional peakshifts in which a lineage escapes the influence of its histor-ical adaptive peak and is attracted to another (Hansen 1997;Butler and King 2004; Bartoszek et al. 2012; Beaulieu et al.2012; O’Meara and Beaulieu 2014). Because it is straightfor-ward to design OU models that permit independent line-ages to evolve toward a shared adaptive peak, they providea natural framework for the investigation of adaptive con-vergence. Generalized OU models permit a great deal offlexibility in parameterization, including multiple attractionstrengths or rates of Brownian drift in addition to multipleoptima (Beaulieu et al. 2012) andmultivariate OU (Bartoszeket al. 2012), although highly complex OUmodels can sufferfrom parameter identifiability issues (Ho and Ané 2014;O’Meara and Beaulieu 2014; Cressler et al. 2015; Cooperet al. 2016b; Khabbazian et al. 2016). Most methods usingOU models require a prior hypothesis for the phylogeneticplacement of adaptive peak shifts, which limits their utilityin estimating the frequency of convergence and precludestests for convergence in clades where putatively convergenttaxa have not been identified. These limitations can beavoided by using a second type of modeling approach inwhich the number or rate of evolutionary peak shifts is es-timated as a model parameter. To achieve this, some suchtools simply extend the multiple-peak OUmodeling frame-work by automating the evaluation of candidate peak shiftconfigurations (IngramandMahler 2013; Uyeda andHarmon2014; Khabbazian et al. 2016), while alternative methodsassume a Lévy process model of punctuated evolution(Eastman et al. 2013; Landis et al. 2013). Both techniquesallow assessment of the frequency of evolutionary shifts,including convergence events (Eastman et al. 2013; Ingramand Mahler 2013).

Why Treating Comparative Approaches as Process-FreeCan Lead to Problems in the Study of Convergence

For each of the comparative approaches to studying con-vergence outlined above, the resulting inferences about

This content downloaded from 132All use subject to University of Chicago Press Term

convergent evolution critically depend on the assumedmodel of the underlying evolutionary process. Quantita-tive measures of convergent pattern from a given methodmay differ markedly if an alternative model of the evolution-ary process is assumed. This can be particularly problem-atic for Brownian motion–based ASR methods described asprocess-free which may overestimate or underestimate thefrequency of convergence in groups evolving on rugged adap-tive landscapes and yield inaccurate estimates of the strengthor extent of convergence in any circumstance in which Brown-ian motion is a poor model of the true evolutionary process.To illustrate, we consider a clade diversifying on an adaptivelandscape in which several subclades have undergone peakshifts to much larger phenotype values but without conver-gence to the same optima (fig. 1). The reconstruction of an-cestral states assuming Brownian motion imposes an aver-aging effect on ancestral phenotype estimates that is mostpronounced at the root of the tree. This reconstruction sub-stitutes the true pattern of iterated divergence from small tolarge phenotype values with a pattern in which at least fourmajor lineages converge from intermediate to small values.Several ASR-based convergence metrics suggest substantialconvergence in this clade (fig. 1A). In contrast, if the true(i.e., generating) Ornstein-Uhlenbeck model is instead as-sumed when carrying out ASR, the same pattern-based met-rics accurately capture the lack of true convergence (fig. 1B).This is a particularly striking example but we suspect not anunrepresentative one, due to the well-documented tendencyof Brownian motion–based ASR methods to infer increas-ingly intermediate ancestral states for deeper nodes (Sch-luter et al. 1997; Oakley and Cunningham 2000). Similarproblems can be anticipated any time ASR methods are usedto investigate a group for which Brownian motion is not areasonably good representation of the evolutionary process,and model misspecification can likewise lead to failure toidentify convergence events or grossly inaccurate estimates ofthe strength or extent of convergence.The issue we discuss here is shared across phylogenetic

comparative biology. Hunt (2012) made similar pointsabout the measurement of evolutionary rate, showing thatmeasures based on process models were more meaningfulacross evolutionary timescales than traditional interval-based (and process-free) rate measures. However, theserate estimates were accurate only if the investigator as-sumed the correct model of evolution, due to the complexand model-specific relationship between evolution’s tempo(i.e., change over time) and mode (the process underlyingthis change). Due to the inseparability of tempo and mode,Hunt argued that the two must be considered in concert instudies of the evolutionary rate. A similar considerationapplies to comparative studies of lineage diversificationrates among clades, which are more accurately estimatedusing methods that assume an underlying diversification

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Pattern, Process, and Convergence S19

model than with process-free methods that simply controlfor elapsed time (Rabosky 2012). These examples reflect ageneral truth—pattern and process are inseparable in thestudy of phylogenetic comparative data, and it is not possi-ble to make inferences about evolutionary patterns withoutassuming something about the evolutionary process (PagelandHarvey 1989; Freckleton et al. 2011; Hunt 2012). Not allcomparative inferences are equal, of course, and somemay bemore robust than others to violations of model assumptions.However, convergence metrics that explicitly incorporatemodel-based reconstruction of ancestral phenotypes arelikely to be particularly sensitive to violations of assumptionsabout the underlying evolutionary model (Oakley and Cun-ningham 2000).

Given the need to assume an evolutionary process tostudy convergence, how can one avoid the potential foradaptationist bias? The potential for such bias is irrefut-able (Hansen 2014; Stayton 2015a, 2015b), but this is a

This content downloaded from 132All use subject to University of Chicago Press Term

risk that can be managed in part through conscientiousconsideration of alternative evolutionary models duringanalysis and careful interpretation of results (see box 2for discussion of best practices). The fit of adaptive modelsshould always be compared to nonadaptive alternatives,and interpretation should favor results obtained underthe best-performing model or, in the absence of a singlebest model, results obtained under all plausible candidatemodels. However, applying these best practices can onlytake one so far in avoiding adaptationist pitfalls, due tothe issue that evolutionary process models may effectivelymodel more than one evolutionary mechanism.

Many-to-One Mapping of Real EvolutionaryProcesses to Modeled Processes

Most available phylogenetic comparative methods employphenomenological models of the evolutionary process that

ba

j

gi

lk

m

pon

ec

rq

t

s

fd

h

0.00 0.25 0.50 0.75 1.00

01

23

45

time

phen

otyp

e

0.00 0.25 0.50 0.75 1.00

time

ba

j

gi

lk

m

pon

ec

rq

t

s

fd

h

A BC2 = 1.12 C2 = 0.10C5 = 6 C5 = 0

Figure 1: Models assumed when reconstructing ancestral states can dramatically affect inferences about the frequency and strength of con-vergent evolution. Here, data for 20 species were simulated using a known phylogeny under a three-optimum Ornstein-Uhlenbeck (OU)process with strong selection and no convergent evolution (a p 4; j2 p 1; v p 0, 3, 5; total tree length p 1; colors represent correspondenceto phenotypic optima). We reconstructed ancestral states assuming a standard Brownian motion model (A) and a three-optimum OU model(B) that closely represents the true evolutionary process under which the data were generated (the number of optima and phylogenetic loca-tions of shifts between optima were known a priori; all other parameters were estimated). We then used ancestral state reconstruction–basedcomparative methods from Stayton (2015a) to estimate the frequency (C5) and magnitude (C2) of convergent evolution for the set of specieswith small phenotypes (species a–j; several related measures of the magnitude of convergence yield similar results but are not shown). C5 talliesthe number of independent lineages that cross into the phenotype space occupied by the focal set of species (here this space is simply defined asthe range of extant phenotype values for this set). C2 indicates the phenotypic distance closed by evolution for this set of species and is cal-culated as the average value of (Dmax 2 Dtips) for all pairs of species in the focal set, where Dmax is the maximum phenotypic difference between apair of species since their divergence and Dtips is the phenotypic difference between the same pair of species in the present. Note that assumingBrownian motion in this example (A) leads to an overestimate of the frequency and strength of convergence events in this subset of species(from intermediate to small phenotype values). If we assume the true OU model (B ), we correctly infer no convergences (C5 p 0) and twodivergences from small to large phenotype. Because assuming a different evolutionary model can fundamentally alter comparative inference aboutconvergent evolution, we contend that there is no such thing as a process-free comparative measure of convergence.

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Box 2: Best Practices

The comparative method has important limitations that must be taken into account in any study (Losos 2011b;Maddison and FitzJohn 2015; Rabosky and Goldberg 2015; Cooper et al. 2016a), including studies that focus onconvergence. Here we recommend several considerations that we think are essential to making strong inferencesabout historical patterns and processes of convergence.

Study Design

The comparative approach investigates patterns that result from uncontrolled natural processes rather than ex-perimental manipulation (Freckleton et al. 2011). Nonetheless, the considerations that guide experimental designequally apply to comparative analyses.Statistical replication is necessary for addressing most questions about convergence, although the relevant form

and degree of replication can depend on the question being asked. For example, if one simply wishes to test whethertwo taxa have in fact converged, this can be tested with a simple model comparison or ancestral state reconstruction(ASR)–based test, although the scope of inference will be limited. Other study objectives will require relatively diverseclades to have any statistical power. For example, testing whether a shift to a new habitat is consistently associatedwith convergence will require a system containing numerous habitat shifts. It will almost never be possible to deter-mine the cause of single convergence events using comparative methods alone because of the inability to rule out thepossibility that an observed correlation is spurious (Gould and Lewontin 1979; Maddison and FitzJohn 2015).Study design should also involve consideration of phylogenetic scale. Many evolutionary models can be useful at

some scales but inadequate at others (Estes andArnold 2007; Hunt 2012). For example, larger clades aremore likely tohave been shaped by a more heterogeneous mixture of evolutionary processes (Beaulieu et al. 2013). At the verysmallest phylogenetic scales, it may not be possible to distinguish complex evolutionary models from simpleralternatives (Boettiger et al. 2012), even if the former better reflect reality.

Model Comparison

Alternative evolutionary process models can differ profoundly in how they reconstruct historical patterns (fig. 1),making it essential to compare models in any phylogenetic study of convergence. While model comparison hasbecome routine in many areas of comparative biology (Posada and Crandall 1998; Harmon et al. 2010; Morlon2014; Pennell et al. 2015), it is often neglected in phylogenetic studies of trait convergence. This may be becausecommon tools for measuring convergence assume Brownian motion evolution; investigators interested in assumingalternative models must customize existing tools to do so. This is especially true for ASR methods (but see Elliotand Mooers 2014; Uyeda and Harmon 2014). Because ancestral state estimates can differ so profoundly under al-ternative evolutionary models, we suspect that a greater appreciation for the importance of model comparisonmight result from the development of more flexible ASR tools.Although model comparison is essential for studying convergence, we caution against discussing results from

alternative models on equal footing when some models clearly outperform others. Comparative studies commonlyreport and interpret the results of several alternative methods, with results from different methods regarded ascomplementary. This is to be encouraged when methods are internally consistent with one another but can be mis-leading when they assume different models of evolution, especially if these differences lead to meaningful differ-ences in the quantification of convergent evolution. For example, if Brownian motion is found to yield a muchworse fit to data than a multiple-peak Ornstein-Uhlenbeck (OU) model (such as in the toy example in fig. 1),the use of comparative methods that assume Brownian motion, such as most ASR and index methods, does notmeaningfully contribute to our understanding of convergence and may even undermine it. Care should be takenthat results that rely on alternative evolutionary process models are themselves regarded as fundamentally distinct(and potentially incompatible) rather than complementary per se.

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:28 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Model Parameters and Model Adequacy

The parameters of fitted evolutionary models can be richly informative with respect to convergence, and many ofthe interesting features of evolutionary convergence that have inspired pattern-based tools may be effectively cap-tured by the parameters of evolutionary models. For example, the strength of attraction in an OU model can beinterpreted as a rate of adaptation in lineages that converge on a shared adaptive peak and may represent the bal-ance between historical constraint and adaptation (Hansen 1997, 2012; Beaulieu et al. 2012; Collar et al. 2014).Inspection of model parameters can also be used to identify when models provide a poor fit to data. For example,

multi-optimum OU models frequently return estimates for some optima that fall outside the observed range ofspecies trait data. This may reflect ongoing adaptation toward an extreme phenotype (Hansen 1997) or a mismatchbetween model assumptions and reality (Mahler and Ingram 2014). Some data sets cannot be fit well by multi-optimum OU models, highlighting the importance of testing whether a model can adequately reproduce keypatterns in the data rather than simply assessing which model from a set of candidates fits best (Pennell et al. 2015).Simulation-based approaches can help ensure robust inference in virtually any scenario, including empirical conditionsfor which model performance may yet be unknown (Boettiger et al. 2012; Mahler et al. 2013; Elliot and Mooers 2014;Slater and Pennell 2014; Pennell et al. 2015; Clarke et al. 2017).

Pattern, Process, and Convergence S21

may plausibly represent more than one kind of evolution-ary mechanism—that is, a many-to-one mapping of trueevolutionary mechanism to modeled macroevolutionaryprocess (Hansen and Martins 1996; O’Meara et al. 2006;Revell et al. 2008; Hansen 2012; Pennell 2014). For exam-ple, a single-peak OU model may represent adaptive evo-lution in a clade that has already reached a phenotypic op-timum (Hansen 1997), or it could represent evolutionarystasis due to a constraint on the production of variation(e.g., Harmon et al. 2010). Many-to-one mapping is possi-ble for a diversity of evolutionary process models, fromBrownian motion to early burst and saltational macroevo-lutionary models (Freckleton and Harvey 2006; O’Mearaet al. 2006; Mahler et al. 2010; Venditti et al. 2011; Pennellet al. 2014). Although many such models were introducedwith specific microevolutionary mechanisms in mind, theydescribe variation at a comparatively coarse macroevolu-tionary scale and contain no direct link to such fine-scalemechanisms. The lack of mechanistic detail in these modelspresents another formidable limitation to the interpretationof macroevolutionary convergence. The investigator mustconsider such possibilities in any analysis of convergence—as we will argue below, such considerations can be aided byinvestigation of the study group using complementary ap-proaches that allow more direct tests of mechanism and byconsidering the natural history of the organisms and theirenvironments.

Integrative Approaches to the Study ofMacroevolutionary Convergence

A complete understanding of any evolutionary phenome-non requires knowledge of both the detailed mechanismsby which evolutionary change occurs (i.e., the how) andthe circumstances that ultimately bring about such change

This content downloaded from 132All use subject to University of Chicago Press Term

or shape its course (i.e., the why). Replicated convergenceprovides a powerful framework for both avenues of inquiry.The power of this replication has been harnessed in combina-tion with high-throughput sequencing technologies in thelast decade to greatly increase our understanding of themo-lecular mechanisms behind convergent phenotypic change(Elmer and Meyer 2011; Stern 2013; Rosenblum et al. 2014).By comparison, somewhat less progress has beenmade in un-derstanding the ecological and phylogenetic context in whichconvergence occurs.The phylogenetic comparative method can be especially

useful for addressing questions about causes of convergentevolution because it is well suited for investigation at thelarge spatial and temporal scales at which such factorsshape the evolution of biodiversity. In many cases, though,comparative models on their own may fail to distinguishamong alternative hypotheses about the causes of conver-gent evolution, even when carefully applied. This is an in-trinsic feature of the comparative approach that results fromthemany-to-onemapping of process to pattern inmacroevo-lution (Hansen and Martins 1996; Pennell 2014). Thus, thecomparative approach will often be much more powerfulwhen integrated with complementary avenues of investiga-tion. Here we discuss ways in which comparative inferencesabout the processes underlying convergent evolution can bestrengthened by the incorporation of (1) more diverse causalmechanisms, (2) biogeography, and (3) fossil data into com-parative approaches to studying convergence.

Incorporating a Greater Diversity of Causal Mechanismsinto Evolutionary Process Models

A key goal in the study of evolution is to understand howmicroevolutionary mechanisms shape macroevolution, butelucidating this link has proved challenging (Uyeda et al.

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S22 The American Naturalist

2011; Rosindell et al. 2015). The comparative study of con-vergence can help to evaluate this relationship to the extentthat we can make meaningful connections between poten-tial causative factors and convergent pattern. Candidatefactors that may result in macroevolutionary convergenceinclude developmental mode (e.g., Wake 1982), genome ar-chitecture (e.g., Stern 2013), changes in climate (e.g., Yea-man et al. 2016), mutualistic interactions that involve pheno-type matching (e.g., Hoyal Cuthill and Charleston 2015),repeated antagonistic interactions (e.g., Siepielski andBenk-man 2007), and competition for nonsubstitutable resources(e.g., Abrams 1987; Scheffer and van Nes 2006). However,despite ongoing research interest in these areas, we stilllack answers to basic questions such as, Do evolutionaryshifts in reproductive system change the likelihood thatclades will exhibit convergence? Is convergence due to abi-otic selection more or less common than convergence dueto biotic selection? and How likely is convergence to occur,persist, or break down under antagonistic or mutualisticselection?

One source of improvement may come from renewedattention to model development. As the scope of phyloge-netic comparative methods has grown in recent years,models that focus on constrained or bounded evolutionhave received somewhat less attention than those inspiredby explicitly adaptive mechanisms. New work in this areacould help to diversify the scope of comparative inquiry(e.g., Boucher and Démery 2016). Future developmentsin the field should also work to clarify what kinds of mac-roevolutionary patterns we expect to arise from differentecological processes. Recently developed comparativemodelsbased on simple species interactions provide a welcomefirst step in this direction (Yoder and Nuismer 2010;Pennell and Harmon 2013; Nuismer and Harmon 2015;Drury et al. 2016; Clarke et al. 2017). Although thesemethods do not explicitly model convergence, the ecolog-ical processes underlying them may result in convergentphenotypes. In addition, recent years have seen the rapiddevelopment of a more sophisticated theory of species co-existence and community ecology (Chesson 2000; Hubbell2001; Pennell and Harmon 2013; Vellend 2016), and fu-ture modeling efforts would do well to identify a set ofexpected evolutionary outcomes that reflects current eco-logical thinking. Combining modern ecological theory withthe phylogenetic replication made possible in comparativestudies of convergence will make for a powerful approachto studying the macroevolutionary signature of ecologicalmechanisms. We speculate that a principal roadblock tothe development of more diverse and detailed macroevolu-tionary models has been the difficulty (or impossibility) ofrepresenting such models as closed-form likelihood expres-sions. Simulation-based approaches (including approxi-mate Bayesian computation methods) may be required to

This content downloaded from 132All use subject to University of Chicago Press Term

fully diversify the models available in our macroevolution-ary tool kit (e.g., Elliot and Mooers 2014; Clarke et al. 2017).Even the development of improved models may not

overcome the issue of many-to-one mapping of mecha-nism to evolutionary process model. However, hypothesesand models are not the same thing, and the utility of thecomparative method depends on the ability of the investi-gator to use natural history knowledge and comparativetools together to craft specific mechanism-inspired hy-potheses that can be tested using comparative data. In ad-dition to improvements associated with integrating morediverse causal mechanisms into comparative methods,the study of convergence will be aided by designing studiesthat creatively use ecological experiments to test for specificecological mechanisms acting to produce patterns observedat the clade level (Weber and Agrawal 2012). Ecological hy-potheses about the drivers of convergence generally containpredictions about the relationship between trait similarity,abundance, and the relative fitness of species in a givencommunity (table 1). For example, in instances where con-vergence is hypothesized to result from selection for Mül-lerian mimicry (i.e., selection for greater phenotype match-ing among aposematic species), experiments manipulatingtrait similarity in contemporary communities can be pairedwith phylogenetic tests of convergence (in relation to timingof sympatry). The prediction in this case is that (1) the re-duction of phenotypic similarity decreases species fitnessby increasing predation and (2) convergence occurred whenspecies were in sympatry, not before. This integrative frame-work can be applied to antagonistic hypotheses as well andrepresents a powerful approach to testing adaptive hypoth-eses about the ecological drivers of convergence.

Integrating Biogeographic Approaches into ComparativeStudies of Convergence

Integrating an understanding of species and clade bioge-ography can greatly enhance attempts to link ecological pro-cess to macroevolutionary pattern. Evolving lineages can di-rectly interact only when they co-occur, and accounting forco-occurrence patterns will be essential if we are to distin-guish the influence of species interactions on convergencefrom alternative factors. Furthermore, biogeographic per-spectives can disentangle the influence of species range overlapfrom abiotic factors such as climate or soil type. The biogeo-graphic approach is ripe for application to convergence gen-erally, and in table 1 we outline several ways in which phylo-genetic comparative methods may be combined with such anapproach to augment their resolving power when investigat-ing the ecological and evolutionary processes underlying con-vergent patterns.The incorporation of a biogeographic perspective has

yielded new insights in recent studies of replicated adap-

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Pattern, Process, and Convergence S23

tive radiations in which entire well-structured ecologicalguilds have evolved convergently (Schluter 2000; Mahlerand Ingram 2014). The existence of numerous replicatedadaptive radiations suggests an important and determinis-tic role for interspecific competition and subsequent char-acter displacement as a cause of convergence into the sameset of niches (e.g., Frédérich et al. 2013; Grundler andRabosky 2014; Esquerré and Keogh 2016; Moen et al. 2016).An alternative possibility, however, is that such convergenceresults more from biomechanical trade-offs involved in spe-cializing on particular resources than from competitive in-teractions per se and that such specialists may have emerged

This content downloaded from 132All use subject to University of Chicago Press Term

via mechanisms other than interspecific competition. Here,information about range overlap can be informative. If in-terspecific competition played a role in the replicated evo-lution of ecological specialists, we would expect the conver-gent species to occur allopatrically and to have evolved onlya single time in a given region, but we would have no suchexpectation if competition were not important in this diver-gence. This pattern is observed in replicated Greater Antil-lean Anolis radiations and in concert with experimentalstudies of both competition and character displacement(Pacala and Roughgarden 1982; Leal et al. 1998; Stuart et al.2014), strongly suggests a role for competition in contributing

Table 1: Examples of hypothesized mechanisms driving patterns of convergence and predictions from integrated analysis

Hypothesized underlyingmechanism

Biogeographic predictions

Predictions for patternsof coexistence ofconvergent forms

.236.027.111 on July 22, 2017 05s and Conditions (http://www.jou

Ecological predictions

Convergence due tochance

Convergence is independent ofbiogeography

Probability of convergence isindependent of communitycontext

Manipulating the density of a traitin a community does not changethe selective value of the trait

Convergence due toselection driven byphysical environment(e.g., climate, light)

Convergence is correlated withshared physical conditions

Probability of convergence isindependent of rangeoverlap with convergentspecies

Manipulating the density of a traitin a community does not changethe selective value of the trait

Convergence due tocompetition resulting inniche partitioning andcharacter displacement

Convergence may or may notbe correlated with physicalenvironment

Convergent forms evolve inallopatry but only wherethey are sympatric with acompetitor

Negative density-fitness relationship:the abundance of a phenotype inthe community decreases theselective value of that pheno-type. Convergent communitiesshould exhibitsimilar patterns of nichepartitioning

Convergence due tocompetition fornonsubstitutableresources

Convergence may be correlatedwith particular abioticconditions across phylogeny(e.g., an essential nutrient)

Convergent forms evolve insympatry

Resource limitation leads to selectionfor greater similarity in the sharedphenotype: supplementingresource decreases selection onthis trait

Convergence due tofacilitation/mutualism

Convergence may or may notbe correlated with physicalenvironment

Probability of convergence isindependent of rangeoverlap with convergentspecies

Average fitness is a positivefunction of the abundance ofthe mutualist and a negativefunction of the abundance ofsimilar competitors

Convergence due tocommensalism

Convergence may or may notbe correlated with physicalenvironment

Probability of convergence isindependent of rangeoverlap with convergentspecies

Average fitness is a positive functionof the abundance of thecommensal host

Convergence due topredation/parasitism

Convergence may or may notbe correlated with physicalenvironment

Convergent forms evolve inallopatry or sympatry butonly where they are sym-patric with an antagonist

Average fitness is a negativefunction of the abundance of theantagonist

Note: Linking pattern to process is a central challenge in comparative biology. However, in some cases, integrating multiple forms of inference can helpresearchers narrow possible pattern-to-process links. Here we provide several examples of how this framework could be applied to phylogenetic comparativestudies of convergence. It is important to note that while inferring past causation with absolute certainty is impossible, explicitly considering the predictions ofalternative hypotheses can help researchers identify mechanisms consistent with observed patterns. We provide several examples of mechanistic hypotheses,which are not necessarily mutually exclusive (a fact that should be accounted for in the design of a comparative study).

:14:28 AMrnals.uchicago.edu/t-and-c).

S24 The American Naturalist

to the repeated evolution of similar ecomorphs on differentislands (Mahler et al. 2013).

Other ecological mechanisms will leave distinct patternsin biogeographic patterns of convergent taxa. Cichlids inAfrican lakes are well known for replicated radiations indifferent lakes (Wagner et al. 2012), but Muschick et al.(2012) showed that within Lake Tanganyika, numerous con-vergent species co-occur within the lake. The pattern of sym-patric convergent taxa (see alsoKozak et al. 2009; Ingram andKai 2014) challenges the hypothesis of competition-drivencharacter displacement and might instead suggest compet-itive character convergence (MacArthur and Levins 1967;Abrams 1987; Scheffer and van Nes 2006). While a greatdeal of work is needed to validate the hypothesis that com-petition can drive convergence between coexisting taxaacross entire clades, the possibility highlights the need foran increased understanding of the expected biogeographicand macroevolutionary consequences of a broader range ofecological processes.

Revisiting the Fossil Record

The overwhelming majority of comparative phylogeneticstudies are conducted using only extant species. Inferencesfrom comparative analyses therefore suffer an unfortunatetemporal asymmetry, whereby estimates of both historicalpattern and process are associated with increasing levels ofuncertainty as one looks further back in time (Schluter et al.1997; Cunningham et al. 1998; Oakley and Cunningham2000; Losos 2011b). For this reason alone, fossil data canmake very large marginal improvements to the accuracy ofcomparative inference, and the incorporation of fossil infor-mation into comparative studies of convergence promises tohelp distinguish among alternative evolutionary models andrefine parameter estimates of key evolutionary processes. Re-cent years have witnessed encouraging progress in the merg-ing of paleontology and comparative phylogenetic methods,both with the development of integrative newmodels for phy-logenetic inference and divergence dating (e.g., Ronquist et al.2012; Heath et al. 2014; Drummond and Stadler 2016; Zhanget al. 2016) and for the fitting of comparative models of con-tinuous trait evolution (Slater et al. 2012; Slater and Harmon2013; Hunt and Slater 2016).

There has been enough progress to demonstrate thatfossil data can dramatically improve comparative infer-ence and in some cases shift the weight of evidence to al-ternative hypotheses (Slater et al. 2012; Mitchell 2015;Slater 2015; Hunt and Slater 2016). In studies of conver-gence, even one or a few fossil data points indicating traitvalues of ancestors may be critical in distinguishing be-tween alternative scenarios (e.g., the histories depicted infig. 1A, 1B). By anchoring ancestors in trait space, the in-corporation of fossil data can potentially separate conver-

This content downloaded from 132All use subject to University of Chicago Press Term

gent pattern from process and allow more powerful hy-pothesis tests than can be achieved using informationsolely about extant species. The integration of fossils intomolecular phylogenetic frameworks remains far from rou-tine, however, and a key hurdle is the often fragmentarynature of fossil data and the considerable uncertainty of-ten associated with phylogenetic placement of such fossils.Future efforts to address these challenges should yieldlarge dividends for the comparative study of convergence.

Conclusions

The phylogenetic comparative method provides a rich setof tools for answering questions about convergence, in-cluding many that are otherwise inaccessible to biologists.Models of the evolutionary process are at the core of allcomparative methods, however, and these models can pro-foundly influence the outcomes of comparative studies ofconvergence. This intrinsic link between pattern and processin phylogenetic comparative methods can be a liability if ig-nored but can be a powerful asset when models are explicitlyused to test hypotheses about the ecological and evolution-ary processes that give rise to phenotypic patterns. Futureprogress in the comparative study of convergence should re-sult from development of more realistic models of ecologicaland evolutionary processes, better integration of compara-tive study with complementary research on biogeographyand ecology, and the growing incorporation of fossil infor-mation into phylogenetic investigations currently dominatedby extant taxa.

Acknowledgments

We thank A. Agrawal for inviting us to contribute to thisspecial issue and for providing thoughtful feedback on earlydrafts of our manuscript. We also wish to thank T. Staytonas well as the presenters and many attendees of the 2016American Society of Naturalists Vice Presidential Sympo-sium for discussing these matters with us following our meet-ing presentation. Finally, we thank B. O’Meara, an anonymousreviewer, and members of the Mahler lab for insightful com-ments on our manuscript.

Literature Cited

Abrams, P. A. 1987. Alternative models of character displacement andniche shift. I. Adaptive shifts in resource usewhen there is competitionfor nutritionally nonsubstitutable resources. Evolution 41:651–661.

Alberch, P. 1982. Developmental constraints in evolutionary pro-cesses. Pages 313–332 in J. T. Bonner, ed. Evolution and develop-ment. Springer, Berlin.

Arbuckle, K., C. M. Bennett, and M. P. Speed. 2014. A simple mea-sure of the strength of convergent evolution. Methods in Ecologyand Evolution 5:685–693.

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Pattern, Process, and Convergence S25

Arbuckle, K., and M. P. Speed. 2016. Analyzing convergent evolu-tion: a practical guide to methods. Pages 23–36 in P. Pontarotti,ed. Evolutionary biology. Springer, Switzerland.

Arendt, J., and D. Reznick. 2008a. Convergence and parallelismreconsidered: what have we learned about the genetics of adapta-tion? Trends in Ecology and Evolution 23:26–32.

———. 2008b. Moving beyond phylogenetic assumptions aboutevolutionary convergence: response to Leander. Trends in Ecologyand Evolution 23:483–484.

Arnold, S. J. 1992. Constraints on phenotypic evolution. AmericanNaturalist 140(suppl.):S85–S107.

Bartoszek, K., J. Pienaar, P. Mostad, S. Andersson, and T. F. Hansen.2012. A phylogenetic comparative method for studying multivar-iate adaptation. Journal of Theoretical Biology 314:204–215.

Beaulieu, J. M., D.-C. Jhwueng, C. Boettiger, and B. C. O’Meara. 2012.Modeling stabilizing selection: expanding the Ornstein-Uhlenbeckmodel of adaptive evolution. Evolution 66:2369–2383.

Beaulieu, J. M., B. C. O’Meara, and M. J. Donoghue. 2013. Identifyinghidden rate changes in the evolution of a binary morphologicalcharacter: the evolution of plant habit in campanulid angiosperms.Systematic Biology 62:725–737.

Bell, M. A. 1987. Interacting evolutionary constraints in pelvic reduc-tion of threespine sticklebacks, Gasterosteus aculeatus (Pisces, Gaste-rosteidae). Biological Journal of the Linnean Society 31:347–382.

Boettiger, C., G. Coop, and P. Ralph. 2012. Is your phylogeny infor-mative? measuring the power of comparative methods. Evolution66:2240–2251.

Boucher, F. C., and V. Démery. 2016. Inferring bounded evolution inphenotypic characters from phylogenetic comparative data. Sys-tematic Biology 65:651–661.

Brakefield, P. M. 2011. Evo-devo and accounting for Darwin’s end-less forms. Philosophical Transactions of the Royal Society B366:2069–2075.

Bright, J. A., J. Marugán-Lobón, S. N. Cobb, and E. J. Rayfield. 2016.The shapes of bird beaks are highly controlled by nondietaryfactors. Proceedings of the National Academy of Sciences of theUSA 113:5352–5357.

Brooks, D. R., and D. A. McLennan. 1991. Phylogeny, ecology, andbehavior: a research program in comparative biology. Universityof Chicago Press, Chicago.

Butler, M. A., and A. A. King. 2004. Phylogenetic comparative anal-ysis: a modeling approach for adaptive evolution. American Nat-uralist 164:683–695.

Chapleau, F., P. H. Johansen, and M. Williamson. 1988. The distinc-tion between pattern and process in evolutionary biology: the useand abuse of the term “strategy.” Oikos 53:136–138.

Chesson, P. 2000. Mechanisms of maintenance of species diversity.Annual Review of Ecology and Systematics 31:343–366.

Clarke, M., G. H. Thomas, and R. P. Freckleton. 2017. Trait evolu-tion in adaptive radiations: modeling and measuring interspecificcompetition on phylogenies. American Naturalist 189:121–137.

Collar, D. C., J. S. Reece,M. E. Alfaro, P. C.Wainwright, and R. S. Mehta.2014. Imperfect morphological convergence: variable changes in cra-nial structures underlie transitions to durophagy inmoray eels. Amer-ican Naturalist 183:E168–E184.

Conway Morris, S. 2010. Evolution: like any other science it is predict-able. Philosophical Transactions of the Royal Society B 365:133–145.

Cooper, N., G. H. Thomas, and R. G. FitzJohn. 2016a. Shedding lighton the “dark side” of phylogenetic comparative methods. Methodsin Ecology and Evolution 7:693–699.

This content downloaded from 132All use subject to University of Chicago Press Term

Cooper, N., G. H. Thomas, C. Venditti, A. Meade, and R. P. Freckleton.2016b. A cautionary note on the use ofOrnstein-Uhlenbeckmodels inmacroevolutionary studies. Biological Journal of the Linnean Society118:64–77.

Cressler, C. E., M. A. Butler, and A. A. King. 2015. Detecting adap-tive evolution in phylogenetic comparative analysis using theOrnstein-Uhlenbeck model. Systematic Biology 64:953–968.

Cunningham, C. W., K. E. Omland, and T. H. Oakley. 1998.Reconstructing ancestral character states: a critical reappraisal.Trends in Ecology and Evolution 13:361–366.

Donoghue, M. J. 1989. Phylogenies and the analysis of evolutionarysequences, with examples from seed plants. Evolution 43:1137–1156.

Drummond, A. J., and T. Stadler. 2016. Bayesian phylogenetic esti-mation of fossil ages. Philosophical Transactions of the Royal So-ciety B 371:20150129.

Drury, J., J. Clavel, M. Manceau, and H. Morlon. 2016. Estimatingthe effect of competition on trait evolution using maximum like-lihood inference. Systematic Biology 65:700–710.

Eastman, J. M., D. Wegmann, C. Leuenberger, and L. J. Harmon.2013. Simpsonian “evolution by jumps” in an adaptive radiationof Anolis lizards. arXiv:1305.4216.

Eldredge, N., and J. Cracraft. 1980. Phylogenetic patterns and theevolutionary process: method and theory in comparative biology.Columbia University Press, New York.

Elliot, M. G., and A. Ø. Mooers. 2014. Inferring ancestral states withoutassuming neutrality or gradualism using a stable model of continuouscharacter evolution. BMC Evolutionary Biology 14:226.

Elmer, K. R., and A. Meyer. 2011. Adaptation in the age of ecologicalgenomics: insights from parallelism and convergence. Trends inEcology and Evolution 26:298–306.

Endler, J. A. 1981. An overview of the relationships between mimicryand crypsis. Biological Journal of the Linnean Society 16:25–31.

Esquerré, D., and J. S. Keogh. 2016. Parallel selective pressures driveconvergent diversification of phenotypes in pythons and boas.Ecology Letters 19:800–809.

Estes, S., and S. J. Arnold. 2007. Resolving the paradox of stasis:models with stabilizing selection explain evolutionary divergenceon all timescales. American Naturalist 169:227–244.

Felsenstein, J. 1985. Phylogenies and the comparative method. Amer-ican Naturalist 125:1–15.

———. 1988. Phylogenies and quantitative characters. Annual Re-view of Ecology and Systematics 19:445–471.

Fox, J. W., and D. A. Vasseur. 2008. Character convergence undercompetition for nutritionally essential resources. American Natu-ralist 172:667–680.

Freckleton, R. P., N. Cooper, and W. Jetz. 2011. Comparative meth-ods as a statistical fix: the dangers of ignoring an evolutionarymodel. American Naturalist 178:E10–E17.

Freckleton, R. P., and P. H. Harvey. 2006. Detecting non-Browniantrait evolution in adaptive radiations. PLoS Biology 4:e373.

Frédérich, B., L. Sorenson, F. Santini, G. J. Slater, and M. E. Alfaro.2013. Iterative ecological radiation and convergence during theevolutionary history of damselfishes (Pomacentridae). AmericanNaturalist 181:94–113.

Futuyma, D. J. 2005. Evolution. Sinauer, Sunderland, MA.Gould, S. J. 1980. The evolutionary biology of constraint. Daedalus1980:39–52.

———. 2002. The structure of evolutionary theory. Harvard Uni-versity Press, Cambridge, MA.

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S26 The American Naturalist

Gould, S. J., and R. C. Lewontin. 1979. The spandrels of San Marcoand the Panglossian paradigm: a critique of the adaptationistprogramme. Proceedings of the Royal Society B 205:581–598.

Gould, S. J., and E. S. Vrba. 1982. Exaptation—a missing term in thescience of form. Paleobiology 8:4–15.

Grundler, M. C., and D. L. Rabosky. 2014. Trophic divergence de-spite morphological convergence in a continental radiation ofsnakes. Proceedings of the Royal Society B 281:20140413.

Haldane, J. B. S. 1932. The causes of evolution. Longman, London.Hansen, T. F. 1997. Stabilizing selection and the comparative analy-sis of adaptation. Evolution 51:1341–1351.

———. 2012. Adaptive landscapes and macroevolutionary dynam-ics. Pages 205–226 in E. Svensson and R. Calsbeek, eds. The adaptivelandscape in evolutionary biology. Oxford University Press, Oxford.

———. 2014. Use and misuse of comparative methods in the studyof adaptation. Pages 351–379 in L. Z. Garamszegi, ed. Modernphylogenetic comparative methods and their application in evolu-tionary biology. Springer, Berlin.

Hansen, T. F., and E. P. Martins. 1996. Translating between micro-evolutionary process and macroevolutionary patterns: the correla-tion structure of interspecific data. Evolution 50:1404–1417.

Harmon, L. J., J. B. Losos, T. J. Davies, R. G. Gillespie, J. L. Gittleman,W. B. Jennings, K. H. Kozak, et al. 2010. Early bursts of body size andshape evolution are rare in comparative data. Evolution 64:2385–2396.

Harvey, P. H., and M. D. Pagel. 1991. The comparative method inevolutionary biology. Oxford University Press, Oxford.

Heath, T. A., J. P. Huelsenbeck, and T. Stadler. 2014. The fossilizedbirth-death process for coherent calibration of divergence-timeestimates. Proceedings of the National Academy of Sciences of theUSA 111:E2957–E2966.

Ho, L. S. T., and C. Ané. 2014. Intrinsic inference difficulties for traitevolution with Ornstein-Uhlenbeck models. Methods in Ecologyand Evolution 5:1133–1146.

Hoekstra, H. E., R. J. Hirschmann, R. A. Bundey, P. A. Insel, and J. P.Crossland. 2006. A single amino acid mutation contributes toadaptive beach mouse color pattern. Science 313:101–104.

Hoyal Cuthill, J. F., and M. Charleston. 2015. Wing patterning genesand coevolution of Müllerian mimicry in Heliconius butterflies:support from phylogeography, cophylogeny, and divergence times.Evolution 69:3082–3096.

Hubbell, S. P. 2001. The unified neutral theory of biodiversity andbiogeography. Monographs in Population Biology 32. PrincetonUniversity Press, Princeton, NJ.

Hunt, G. 2012. Measuring rates of phenotypic evolution and the in-separability of tempo and mode. Paleobiology 38:351–373.

Hunt, G., and G. Slater. 2016. Integrating paleontological and phy-logenetic approaches to macroevolution. Annual Review of Ecol-ogy, Evolution, and Systematics 47:189–213.

Ingram, T., and Y. Kai. 2014. The geography of morphological con-vergence in the radiations of Pacific Sebastes rockfishes. AmericanNaturalist 184:E115–E131.

Ingram, T., and D. L. Mahler. 2013. SURFACE: detecting convergentevolution from comparative data by fitting Ornstein-Uhlenbeckmodels with stepwise Akaike information criterion. Methods inEcology and Evolution 4:416–425.

Khabbazian, M., R. Kriebel, K. Rohe, and C. Ané. 2016. Fast and ac-curate detection of evolutionary shifts in Ornstein-Uhlenbeckmodels. Methods in Ecology and Evolution 7:811–824.

Kozak, K. H., R. W. Mendyk, and J. J. Wiens. 2009. Can parallel di-versification occur in sympatry? repeated patterns of body-size

This content downloaded from 132All use subject to University of Chicago Press Term

evolution in coexisting clades of North American salamanders.Evolution 63:1769–1784.

Lande, R. 1976. Natural selection and random genetic drift in phe-notypic evolution. Evolution 30:314–334.

Landis, M. J., J. G. Schraiber, and M. Liang. 2013. Phylogenetic anal-ysis using Lévy processes: finding jumps in the evolution of con-tinuous traits. Systematic Biology 62:193–204.

Leal, M., J. A. Rodríguez-Robles, and J. B. Losos. 1998. An experi-mental study of interspecific interactions between two Puerto RicanAnolis lizards. Oecologia 117:273–278.

Leander, B. S. 2008. Different modes of convergent evolution reflectphylogenetic distances: a reply to Arendt and Reznick. Trends inEcology and Evolution 23:481–482.

Losos, J. B. 1992. The evolution of convergent structure in CaribbeanAnolis communities. Systematic Biology 41:403–420.

———. 2009. Lizards in an evolutionary tree: ecology and adaptiveradiation of anoles. University of California Press, Berkeley.

———. 2011a. Convergence, adaptation and constraint. Evolution65:1827–1840.

———. 2011b. Seeing the forest for the trees: the limitations of phy-logenies in comparative biology. American Naturalist 177:709–727.

MacArthur, R., and R. Levins. 1967. The limiting similarity, convergence,and divergence of coexisting species. American Naturalist 101:377–385.

Maddison, W. P., and R. G. FitzJohn. 2015. The unsolved challengeto phylogenetic correlation tests for categorical characters. Sys-tematic Biology 64:127–136.

Mahler, D. L., and T. Ingram. 2014. Phylogenetic comparative meth-ods for studying clade-wide convergence. Pages 425–450 in L. Z.Garamszegi, ed. Modern phylogenetic comparative methods andtheir application in evolutionary biology. Springer, Berlin.

Mahler, D. L., T. Ingram, L. J. Revell, and J. B. Losos. 2013. Excep-tional convergence on the macroevolutionary landscape in islandlizard radiations. Science 341:292–295.

Mahler, D. L., L. J. Revell, R. E. Glor, and J. B. Losos. 2010. Ecological op-portunity and the rate of morphological evolution in the diversifica-tion of Greater Antillean anoles. Evolution 64:2731–2745.

Maraun, M., G. Erdmann, G. Schulz, R. A. Norton, S. Scheu, and K.Domes. 2009. Multiple convergent evolution of arboreal life inoribatid mites indicates the primacy of ecology. Proceedings ofthe Royal Society B 276:3219–3227.

Marazzi, B., C. Ané, M. F. Simon, A. Delgado-Salinas, M. Luckow,and M. J. Sanderson. 2012. Locating evolutionary precursors ona phylogenetic tree. Evolution 66:3918–3930.

Maynard Smith, J., R. Burian, S. Kauffman, P. Alberch, J. Campbell, B.Goodwin, R. Lande, et al. 1985. Developmental constraints and evolu-tion: a perspective from the Mountain Lake conference on develop-ment and evolution. Quarterly Review of Biology 60:265–287.

McCune, A. R. 1987. Lakes as laboratories of evolution: endemicfishes and environmental cyclicity. Palaios 2:446–454.

McCune, A. R., and R. L. Carlson. 2004. Twenty ways to lose yourbladder: common natural mutants in zebrafish and widespreadconvergence of swim bladder loss among teleost fishes. Evolutionand Development 6:246–259.

Meyer, J. R., D. T. Dobias, J. S. Weitz, J. E. Barrick, R. T. Quick, andR. E. Lenski. 2012. Repeatability and contingency in the evolutionof a key innovation in phage lambda. Science 335:428–432.

Mitchell, J. S. 2015. Extant-only comparative methods fail to recoverthe disparity preserved in the bird fossil record. Evolution 69:2414–2424.

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Pattern, Process, and Convergence S27

Moen, D. S., H. Morlon, and J. J. Wiens. 2016. Testing convergenceversus history: convergence dominates phenotypic evolution forover 150 million years in frogs. Systematic Biology 65:146–160.

Morlon, H. 2014. Phylogenetic approaches for studying diversifica-tion. Ecology Letters 17:508–525.

Muschick, M., A. Indermaur, and W. Salzburger. 2012. Convergentevolution within an adaptive radiation of cichlid fishes. CurrentBiology 22:2362–2368.

Nevo, E. 1979. Adaptive convergence and divergence of subterraneanmammals. Annual Review of Ecology and Systematics 10:269–308.

Nuismer, S. L., and L. J. Harmon. 2015. Predicting rates of interspecificinteraction from phylogenetic trees. Ecology Letters 18:17–27.

Oakley, T. H., and C. W. Cunningham. 2000. Independent contrastssucceed where ancestor reconstruction fails in a known bacterio-phage phylogeny. Evolution 54:397–405.

O’Meara, B. C., C. Ané, M. J. Sanderson, and P. C. Wainwright.2006. Testing for different rates of continuous trait evolution us-ing likelihood. Evolution 60:922–933.

O’Meara, B. C., and J. M. Beaulieu. 2014. Modelling stabilizing selec-tion: the attraction of Ornstein-Uhlenbeck models. Pages 381–393in L. Z. Garamszegi, ed. Modern phylogenetic comparative methodsand their application in evolutionary biology. Springer, Berlin.

Oster, G., and P. Alberch. 1982. Evolution and bifurcation of devel-opmental programs. Evolution 1982:444–459.

Pacala, S., and J. Roughgarden. 1982. Resource partitioning and in-terspecific competition in two two-species insular Anolis lizardcommunities. Science 217:444–446.

Pagel, M. D., and P. H. Harvey. 1989. Comparative methods for examin-ing adaptation depend on evolutionary models. Folia Primatologica53:203–220.

Pascoal, S., T. Cezard, A. Eik-Nes, K. Gharbi, J. Majewska, E. Payne,M. G. Ritchie, et al. 2014. Rapid convergent evolution in wildcrickets. Current Biology 24:1369–1374.

Pennell, M. W. 2014. Modern phylogenetic comparative methodsand their application in evolutionary biology: concepts and prac-tice. Systematic Biology 64:161–163.

Pennell, M. W., R. G. FitzJohn, W. K. Cornwell, and L. J. Harmon.2015. Model adequacy and the macroevolution of angiospermfunctional traits. American Naturalist 186:E33–E50.

Pennell, M. W., and L. J. Harmon. 2013. An integrative view of phy-logenetic comparative methods: connections to population genet-ics, community ecology, and paleobiology. Annals of the NewYork Academy of Sciences 1289:90–105.

Pennell, M. W., L. J. Harmon, and J. C. Uyeda. 2014. Is there roomfor punctuated equilibrium in macroevolution? Trends in Ecologyand Evolution 29:23–32.

Posada, D., and K. A. Crandall. 1998. Modeltest: testing the model ofDNA substitution. Bioinformatics 14:817–818.

Rabosky, D. L. 2012. Testing the time-for-speciation effect in the as-sembly of regional biotas. Methods in Ecology and Evolution3:224–233.

Rabosky, D. L., and E. E. Goldberg. 2015. Model inadequacy andmistaken inferences of trait-dependent speciation. Systematic Bi-ology 64:340–355.

Revell, L. J., L. J. Harmon, and D. C. Collar. 2008. Phylogenetic signal,evolutionary process, and rate. Systematic Biology 57:591–601.

Ronquist, F., S. Klopfstein, L. Vilhelmsen, S. Schulmeister, D. L.Murray, and A. P. Rasnitsyn. 2012. A total-evidence approachto dating with fossils, applied to the early radiation of the Hyme-noptera. Systematic Biology 61:973–999.

This content downloaded from 132All use subject to University of Chicago Press Term

Rosenblum, E. B., C. E. Parent, and E. E. Brandt. 2014. The molec-ular basis of phenotypic convergence. Annual Review of Ecology,Evolution, and Systematics 45:203–226.

Rosenblum, E. B., H. Römpler, T. Schöneberg, and H. E. Hoekstra.2010. Molecular and functional basis of phenotypic convergencein white lizards at White Sands. Proceedings of the NationalAcademy of Sciences of the USA 107:2113–2117.

Rosindell, J., L. J. Harmon, and R. S. Etienne. 2015. Unifying ecologyand macroevolution with individual-based theory. Ecology Letters18:472–482.

Scheffer, M., and E. H. van Nes. 2006. Self-organized similarity, theevolutionary emergence of groups of similar species. Proceedings ofthe National Academy of Sciences of the USA 103:6230–6235.

Schluter, D. 1996. Adaptive radiation along genetic lines of least re-sistance. Evolution 50:1766–1774.

———. 2000. The ecology of adaptive radiation. Oxford UniversityPress, Oxford.

Schluter, D., T. Price, A. Ø. Mooers, and D. Ludwig. 1997. Likelihoodof ancestor states in adaptive radiation. Evolution 51:1699–1711.

Scotland, R. W. 2011. What is parallelism? Evolution and Develop-ment 13:214–227.

Sidlauskas, B. 2008. Continuous and arrested morphological diversi-fication in sister clades of characiform fishes: a phylomorphospaceapproach. Evolution 62:3135–3156.

Siepielski, A. M., and C. W. Benkman. 2007. Convergent patterns inthe selection mosaic for two North American bird-dispersedpines. Ecological Monographs 77:203–220.

Slater, G. J. 2015. Iterative adaptive radiations of fossil canids showno evidence for diversity-dependent trait evolution. Proceedingsof the National Academy of Sciences of the USA 112:4897–4902.

Slater, G. J., and L. J. Harmon. 2013. Unifying fossils and phyloge-nies for comparative analyses of diversification and trait evolution.Methods in Ecology and Evolution 4:699–702.

Slater, G. J., L. J. Harmon, and M. E. Alfaro. 2012. Integrating fossilswith molecular phylogenies improves inferences of trait evolution.Evolution 66:3931–3944.

Slater, G. J., and M. W. Pennell. 2014. Robust regression and poste-rior predictive simulation increase power to detect early bursts oftrait evolution. Systematic Biology 63:293–308.

Speed, M. P., and K. Arbuckle. 2016. Quantification provides a con-ceptual basis for convergent evolution. Biological Reviews 92:815–829. doi:10.1111/brv.12257.

Spor, A., D. J. Kvitek, T. Nidelet, J. Martin, J. Legrand, C. Dillmann,A. Bourgais, et al. 2014. Phenotypic and genotypic convergencesare influenced by historical contingency and environment in yeast.Evolution 68:772–790.

Stayton, C. T. 2008. Is convergence surprising? an examination ofthe frequency of convergence in simulated datasets. Journal ofTheoretical Biology 252:1–14.

———. 2011. Biomechanics on the half shell: functional perfor-mance influences patterns of morphological variation in the emydidturtle carapace. Zoology 114:213–223.

———. 2015a. The definition, recognition, and interpretation of con-vergent evolution, and two new measures for quantifying and as-sessing the significance of convergence. Evolution 69:2140–2153.

———. 2015b. What does convergent evolution mean? the interpre-tation of convergence and its implications in the search for limitsto evolution. Interface Focus 5:20150039.

Stern, D. L. 2013. The genetic causes of convergent evolution. NatureReviews Genetics 14:751–764.

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S28 The American Naturalist

Streisfeld, M. A., and M. D. Rausher. 2011. Population genetics, plei-otropy, and the preferential fixation of mutations during adaptiveevolution. Evolution 65:629–642.

Stuart, Y. E., T. S. Campbell, P. A. Hohenlohe, R. G. Reynolds, L. J.Revell, and J. B. Losos. 2014. Rapid evolution of a native speciesfollowing invasion by a congener. Science 346:463–466.

Uyeda, J. C., T. F. Hansen, S. J. Arnold, and J. Pienaar. 2011. Themillion-year wait for macroevolutionary bursts. Proceedings ofthe National Academy of Sciences of the USA 108:15908–15913.

Uyeda, J. C., and L. J. Harmon. 2014. A novel Bayesian method forinferring and interpreting the dynamics of adaptive landscapesfrom phylogenetic comparative data. Systematic Biology 63:902–918.

Vellend, M. 2016. The theory of ecological communities. PrincetonUniversity Press, Princeton, NJ.

Venditti, C., A. Meade, and M. Pagel. 2011. Multiple routes to mam-malian diversity. Nature 479:393–396.

Wagner, C. E., L. J. Harmon, and O. Seehausen. 2012. Ecological op-portunity and sexual selection together predict adaptive radiation.Nature 487:366–369.

Wagner, P. J. 2000. Exhaustion of morphologic character statesamong fossil taxa. Evolution 54:365–386.

Wake, D. B. 1982. Functional and developmental constraints andopportunities in the evolution of feeding systems in urodeles.

“There are in all about twenty species of Draco, inhabiting the East IndPtychozoön.” Figured: Draco volans. From “Volant Adaptation in Vertebra

This content downloaded from 132All use subject to University of Chicago Press Term

Pages 51–66 in D. Mossakowski and G. Roth, eds. Environmentaladaptation and evolution. Fischer, New York.

Wake, D. B., and A. Larson. 1987. Multidimensional analysis of anevolving lineage. Science 238:42–48.

Wake, D. B., M. H. Wake, and C. D. Specht. 2011. Homoplasy: fromdetecting pattern to determining process and mechanism of evo-lution. Science 331:1032–1035.

Weber, M. G., and A. A. Agrawal. 2012. Phylogeny, ecology, and thecoupling of comparative and experimental approaches. Trends inEcology and Evolution 27:394–403.

Williams, E. E. 1983. Ecomorphs, faunas, island size, and diverse endpoints in island radiations of Anolis. Pages 326–370 in R. B. Huey,E. R. Pianka, and T. W. Schoener, eds. Lizard ecology: studies of amodel organism. Harvard University Press, Cambridge, MA.

Yeaman, S., K. A. Hodgins, K. E. Lotterhos, H. Suren, S. Nadeau, J. C.Degner, K. A. Nurkowski, et al. 2016. Convergent local adaptationto climate in distantly related conifers. Science 353:1431–1433.

Yoder, J. B., and S. L. Nuismer. 2010. When does coevolution pro-mote diversification? American Naturalist 176:802–817.

Zhang, C., T. Stadler, S. Klopfstein, T. A. Heath, and F. Ronquist.2016. Total-evidence dating under the fossilized birth-death pro-cess. Systematic Biology 65:228–249.

Symposium Editor: Anurag A. Agrawal

ies. The power of flight is not very great, but probably exceeds that oftes” by Richard S. Lull (The American Naturalist, 1906, 40:537–566).

.236.027.111 on July 22, 2017 05:14:28 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Convergently Evolved Toxic Secondary Metabolites in Plants

Drive the Parallel Molecular Evolution of Insect Resistance*

Georg Petschenka,1 Vera Wagschal,2 Michael von Tschirnhaus,3 Alexander Donath,4

and Susanne Dobler2,†

1. Institute for Insect Biotechnology, Justus-Liebig-Universität Giessen, Giessen, Germany; 2. Molecular Evolutionary Biology, Instituteof Zoology, Universität Hamburg, Hamburg, Germany; 3. Fakultät Biologie, Universität Bielefeld, Bielefeld, Germany; 4. ZoologischesForschungsmuseum Alexander Koenig, Zentrum für Molekulare Biodiversitätsforschung, Bonn, Germany

Online enhancements: appendix. Dryad data: http://dx.doi.org/10.5061/dryad.20191.

abstract: Natural selection imposed by natural toxins has led tostriking levels of convergent evolution at the molecular level. Cardiacglycosides represent a group of plant toxins that block the Na,K-ATPase, a vital membrane protein in animals. Several herbivorousinsects have convergently evolved resistant Na,K-ATPases, and insome species, convergent gene duplications have also arisen, likelyto cope with pleiotropic costs of resistance. To understand the geneticbasis and predictability of these adaptations, we studied five indepen-dent lineages of leaf-mining flies (Diptera: Agromyzidae). These flieshave colonized host plants in four botanical families that convergentlyevolved cardiac glycosides of two structural types: cardenolides andbufadienolides. We compared each of six fly species feeding on suchplants to a phylogenetically related but nonadapted species. Irre-spective of the type of cardiac glycoside in the host plant, five out ofsix exposed species displayed substitutions in the cardiac glycoside–binding site of the Na,K-ATPase that were previously described inother insect orders; in only one species was the gene duplicated. Invitro assays of nervous tissue extractions confirmed that the substitu-tions lead to increased resistance of the Na,K-ATPase. Our resultsdemonstrate that target site insensitivity of Na,K-ATPase is a commonresponse to dietary cardiac glycosides leading to highly predictable aminoacid changes; nonetheless, convergent evolution of gene duplication forthis multifunctional enzyme appears more constrained.

Keywords: cardiac glycosides, Na,K-ATPase resistance, target site in-sensitivity, Agromyzidae, phylogeny, gene duplication.

Introduction

Evolutionary responses to common selective pressures areideally suited to study the predictability of molecular adap-

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† Corresponding author; e-mail: susanne.dobler@uni-hamburg.de.ORCIDs: von Tschirnhaus, http://orcid.org/0000-0002-1903-4767; Donath,

http://orcid.org/0000-0001-5618-0547; Dobler, http://orcid.org/0000-0002-0635-7719.

Am. Nat. 2017. Vol. 190, pp. S29–S43. q 2017 by The University of Chicago.0003-0147/2017/190S1-57252$15.00. All rights reserved.DOI: 10.1086/691711

This content downloaded from 132All use subject to University of Chicago Press Term

tations, a major question in evolutionary genetics (Storz2016). Recent studies of species that acquired resistanceagainst toxins in their diet or the environment revealedstriking levels of convergent molecular adaptations. Herethe interaction of a noxious substance with a well-definedtarget site of a specific protein might facilitate the occur-rence of molecular convergence, as specific amino acid sub-stitutions can prevent docking of the toxin. Accordingly,across several classes of toxins, it has been shown that re-sistance was repeatedly achieved by substitutions of a fewamino acids at specific positions within the target site ofthe receptor. Examples include predatory snakes that con-sume toxic amphibians and have convergently evolved re-sistance to tetrodotoxin (Feldman et al. 2012), poison frogsthat acquire toxic alkaloids from arthropods to defend them-selves against predators (Tarvin et al. 2016), and diverseinsects that have evolved resistance to cyclodiene insecticides(Andreev et al. 1999) or pyrethroids (Rinkevich et al. 2013).An emergingmodel system for convergent evolution that

has been studied in unprecedented detail is resistance tocardiac glycosides in insects as well as in vertebrates. Car-diac glycosides are potent toxins that inhibit the ubiquitousanimal cation carrier Na,K-ATPase. They typically occurin plants but are also produced by some animals (e.g., toads,leaf beetles, and fireflies; Dobler et al. 2011; Agrawal et al.2012). Cardiac glycosides are composed of a steroidal skel-eton with a five-membered lactone ring at C17 in carde-nolides and a six-membered lactone ring in bufadienolides(fig. 1). Both types either occur as glycosides with one orseveral sugars attached to C3 or as genins (Malcolm 1991;Agrawal et al. 2012). By binding to the Na,K-ATPase, theyblock the enzyme in the phosphorylated state and thus im-pair various physiological functions such as maintenance ofan electrochemical gradient across the cell membrane andgeneration of neuronal action potentials (Horisberger 2004).Extensive mutagenesis experiments with the mammalian Na,K-ATPase and recent crystal structures revealed the specific

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S30 The American Naturalist

amino acid residues that are necessary for cardiac glycosidebinding (Price and Lingrel 1988; Croyle et al. 1997; Qiu et al.2005; Laursen et al. 2013, 2015). Thesedata point to the prom-inent role of several amino acids in the transmembrane seg-ments of the enzyme’s a-subunit that directly interact withthe cardiacglycosidemolecule and stabilize it in theextracellu-lar binding pocket (fig. 1).

Resistance to cardiac glycosides often relies on substitu-tions of amino acids in the binding pocket of the Na,K-ATPase. The first evidence was found in Danaus plexippus,the monarch butterfly (Vaughan and Jungreis 1977), which

This content downloaded from 132All use subject to University of Chicago Press Term

sequesters toxic cardenolides from its host plants (Asclepiasspecies) as a defense against predators. Later, the molecularbasis was unraveled (Holzinger et al. 1992), and more re-cently it was shown that target site insensitivity in the mon-arch butterfly is caused by two amino acid substitutions(Aardema et al. 2012; Dalla et al. 2013) that emerged con-secutively across the phylogeny of milkweed butterfliesleading to a stepwise increase of resistance (Petschenka et al.2013a). Target site insensitivity of the Na,K-ATPase has nowbeen reported in at least six orders of insects with multipleevolutionary origins (Al-Robai et al. 1993; Dobler et al. 2012,

Figure 1: A, Structure of the pig Na,K-ATPase with bound ouabain (Yatime et al. 2011; Protein Data Bank ID 3N23). Here, a, b, and g referto the subunits of the Na,K-ATPase, of which g has so far been described only for vertebrates. M p transmembrane domains, A p actuator do-main, N p nucleotide-binding domain, P p phosphorylation domain. B, Enlarged and rotated view of the ouabain-binding cavity withrelevant amino acids. A p alanine, E p glutamic acid, F p phenylalanine, G p glycine, I p isoleucine, N p asparagine, Q p glutamine,T p threonine, V p valine. C, Structure of cardiac glycosides. Top, hellebrin, a bufadienolide from Helleborus species; middle, ouabain, astandard cardenolide used for the physiological assays; bottom, calotropin, a common cardenolide from Asclepias species.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Parallel Evolution of Insect Resistance S31

2015; Zhen et al. 2012). While resistance can involve substi-tutions at least at five amino acid positions, the identical sub-stitution of histidine for asparagine at position 122 occurredat least six times independently (Dobler et al. 2015). A com-parable level of molecular convergence was found in cardiacglycoside–producing toads and their vertebrate predators (Uj-vari et al. 2015).

In insects, it was furthermore shown that resistance-conferring amino acid substitutions may be associated withduplications of the Na,K-ATPase gene (Zhen et al. 2012).Such duplications might alleviate negative pleiotropic ef-fects associated with increased resistance to cardiac gly-cosides. While resistance-conferring mutations may com-promise the pumping activity of the Na,K-ATPase (Dallaand Dobler 2016), other important functions of this en-zyme might not be impaired by the observed amino acidsubstitutions. Increasing evidence points to important rolesof the Na,K-ATPase as a signal transducer and morphoge-netic trigger, not only in vertebrates but also in insects(Liang et al. 2007; Paul et al. 2007; Li and Xie 2009). A bal-ance between more or less substituted copies of the Na,K-ATPase and tissue-specific expression patterns may allowadjusting between effective ion transport and these otherfunctions. Therefore, duplications of the Na,K-ATPase genein cardiac glycoside–adapted insects have been interpretedas an evolutionary strategy to reduce negative pleiotropic ef-fects (Zhen et al. 2012; Dalla and Dobler 2016).

Although cardiac glycosides are known to occur in 18 plantfamilies (Krenn and Kopp 1998; Agrawal et al. 2012), almostall work on resistance has focused on insect herbivores feed-ing on plants in the family Apocynaceae. Here we studiedthe parallel adaptation of five independent lineages of leaf-mining flies (Agromyzidae) adapted to cardiac glycoside–producing plants (six fly species adapted to cardiac glycosideplants, including one generalist species) compared with fiveclosely related fly species feeding on plants without cardiacglycosides. The host plants of these flies include species fromfour botanical families (Apocynaceae, Brassicaceae, Planta-ginaceae, and Ranunculaceae) in which cardiac glycosideshave themselves independently evolved. Thus, our data setcomprises multiple independent comparisons of animal hostuse and plant toxin production, a convergently evolved inter-action from the perspective of each partner.

Using our five independent contrasts of fly species, wetested how often amino acid substitutions evolved that havepreviously been assumed to-confer resistance of the Na,K-ATPase. We were especially interested to test for repeatedevolution of target site insensitivity by identical amino acidsubstitutions. To screen for such substitutions, we first an-alyzed Na,K-ATPase gene sequences amplified by reversetranscription polymerase chain reaction (RT-PCR). In spe-cies possessing putatively resistance-conferring amino acidsubstitutions, we further tested for the existence of dupli-

This content downloaded from 132All use subject to University of Chicago Press Term

cated Na,K-ATPase genes using a transcriptomic approach.Specifically, we wanted to investigate how strongly substitu-tions are associated with gene duplications. For three flyspecies that independently adapted to cardiac glycoside plantsand two outgroup species, we additionally assessed resistanceof their Na,K-ATPase in vitro and thereby confirmed thefunctionality of the observed molecular substitutions.

Material and Methods

Natural History of Flies

We compared six species of agromyzid taxa feeding on car-diac glycoside plants to related species whose host plantsare not known to contain these toxins. The species pairswere chosen according to a priori assumptions about phy-logenetic relationships based on morphology.Phytomyza digitalis Hering, 1925, versus Phytomyza cras-

siseta Zetterstedt, 1860.—We compared P. digitalis, whosehost plants are restricted to the genus Digitalis (Plantagina-ceae), includingD. grandiflora,D. lutea, andD. lanata (Hering1927, 1951; www.bladmineerders.nl), which all produce carde-nolides (Luckner andWichtl 2000), to P. crassiseta, a specialiston various species of Veronica (Plantaginaceae; Hering 1927;www.bladmineerders.nl) not known to produce cardenolides.Phytomyza hellebori Kaltenbach, 1872, versus Phytomyza

fallaciosa Brischke, 1880.—We also investigated P. hellebori,which exclusively uses Helleborus (Ranunculaceae) species ashosts (Hering 1951). Several species of Helleborus are knownto produce bufadienolides (e.g., hellebrin; Wissner and Ka-ting 1974). While H. foetidus seems to be the preferred hostplant at some locations (Ludwig 1907; www.bladmineerders.nl)and was reported to lack bufadienolides, this fly species alsouses other Helleborus species as hosts (e.g., H. niger, H. vi-ridis; Mortelmans et al. 2014; table 1) known to containbufadienolides (Wissner and Kating 1974; Glombitza et al.1989). The lack of cardiac glycosides in H. foetidus is ques-tionable, as we have a clear indication of the presence of suchcompounds based on inhibition of Na,K-ATPase (G. Pet-schenka, unpublished data). As a comparison for P. helle-bori, the Ranunculus feeder P. fallaciosa (Spencer 1990;www.bladmineerders.nl) was included.Napomyza scrophulariae Spencer, 1966, versus Napomyza

lateralis Fallén, 1823.—With respect to the other species weinvestigated, knowledge on the natural history of N. scro-phulariae is very limited. The species seems to be specializedon Digitalis purpurea but may also use other hosts (Spencer1966; www.ukflymines.co.uk). Nonetheless, the only speci-mens that were unambiguously determined as N. scrophu-lariae were all from D. purpurea (as were the specimens in-vestigated here; table 1); accordingly, we refer to this speciesas a specialist. As a phylogenetic comparison, we employedN. lateralis, a polyphagous species that has a preference forAsteraceae (Spencer 1990; www.bladmineerders.nl).

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Table

1:Dipteranspeciesused

inthisstud

yandplantsfrom

which

specim

enswerecollected

Flyspecies

Hostplant

Collectionsite,year

Use

GenBankaccessionno

.

Na,K-A

TPase

COI

28S

CAD

Chrom

atom

yiahorticola

Erysim

umcheiranthoides

Pevestorf,Germany,

2011

RT-PCR(1)

LT795081

LT795089

LT795087

LT795098

Liriom

yzaasclepiadis

Asclepias

syriaca

Ithaca,NY,2014,2015

Invitroassay,

transcriptom

eHE956747

HE862403

LT795104

LT795100

Liriom

yzaeupatorii

Eupatorium

cann

abinum

Giessen,Germany,

2015

Transcriptome

LT795109

LT795097

LT795105

LT795101

Napom

yzascroph

ulariae

Digitalispu

rpurea

Taunu

sstein,Waldk

atzenb

ach,

Germany,

2011

RT-PCR(2)

LT795080

LT795091

LT795086

LT795099

Napom

yzalateralis

Asteraceaesp.

Tangstedt,Germany,

2011

RT-PCR(2)

LT795079

EF1

04710

EF1

04884

EF1

04796

Phytom

yzacrassiseta

Veron

icasp.

Waldk

atzenb

ach,

Germany,

2011

RT-PCR(2)

LT795078

EU367549

LT795085

EU367638

Phytom

yzadigitalis

Digitalisgran

diflora

Mehrstetten,Germany,

2011

RT-PCR(2),in

vitroassay,

transcriptom

eLT

795083

LT795093

LT795106

LT795103

Phytom

yzafallaciosa

Ran

unculussp.

Taunu

sstein,Germany,

2011,2015

RT-PCR(2),transcriptom

eLT

795082

LT795094

LT795107

...

Phytom

yzahellebori

Helleborusfoetidus

(Mosbach:

native

popu

lation

;Ham

burg,

Jena:ornam

ental),Helleborus

niger(H

ochh

eim:ornam

ental)

Ham

burg,AHochh

eim/M

ain,

A

Jena,

A,CMosbach,A

,B

Germany,

2011,2012

RT-PCR(5),in

vitroassay,

transcriptom

eLT

795110,

LT795111

LT795095

LT795108

LT795102

Phytom

yzailicis

Ilex

aquifolia

Giessen,BHam

burg,A

,B

Germany,

2011,2016

RT-PCR(1),in

vitroassay

LT795077

EU367541

LT795084

EU367630

Drosophila

melan

ogaster

(oregonR

)...

Laboratory

cultu

reIn

vitroassay

AF0

44974

...

...

...

Note:Whenspecim

enswereobtained

atdifferentlocation

s,lettersin

superscriptindicate

where

specim

ensused

forreversetran

scriptionpo

lymerasechainreaction

(RT-PCR;a),Na,K-A

TPasein

vitroassays

(b),

ortran

scriptom

es(c)originated

from

.Num

bers

ofbiological

replicates

forreversetran

scriptionpo

lymerasechainreaction

–basedsequ

encingof

Na,K-A

TPasegenes

aregivenin

parentheses.

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:35 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Parallel Evolution of Insect Resistance S33

Liriomyza asclepiadis Spencer, 1969, versus Liriomyzaeupatorii Kaltenbach, 1873.—Another comparison involvedL. asclepiadis, a specialized feeder on Asclepias species (Spen-cer 1990), and an unidentified Liriomyza species (referred toas L. sp. Mexico) collected from Asclepias angustifolia that isdifferentiated in its cytochrome oxidase I (COI) sequencesfrom L. asclepiadis by 12.5% sequence divergence (Dobleret al. 2012). The Asclepias host plants of both species containcardenolides (Agrawal et al. 2012). While L. asclepiadis isspecialized on Asclepias, we have no information about thehost range of the unidentified Liriomyza species. As it seemsto be closely related to L. asclepiadis and no generalist leafminers are known to use Asclepias species (Spencer 1990),we refer to it as a specialist. We compared these two Lirio-myza species to L. eupatorii, which uses several genera ofAsteraceae as hosts (www.bladmineerders.nl).

Chromatomyia horticolaGoureau, 1851, versus Phytomyzailicis Curtis, 1846.—Last, we sequenced the Na,K-ATPasegene of C. horticola, a highly polyphagous species (www.bladmineerders.nl), which we have found on the cardenolide-producing Brassicaceae Erysimum cheiranthoides (Lei et al.1996). While all pairwise comparisons listed above were basedon an a priori assumed phylogenetic relationship between flyspecies adapted to cardiac glycosides and nonadapted spe-cies, here we arbitrarily selected the holly leaf miner, P. ilicis,a specialist on Ilex aquifolium, as a comparison for C. horticola.

Collection of Flies and Determination

Species of Agromyzidae were determined by means of themorphology of mines and host plant species (www.bladmineerders.nl, www.leafmines.co.uk; Hering 1927, 1935). Inaddition, species identities were confirmed based on adultcharacters and verified by sequence data compared to refer-ence sequences in GenBank. Host plant leaves containingmines were collected in the field (table 1) and kept underroom conditions until the emergence of flies. Upon hatchingand wing hardening, flies were frozen and stored at 2807C.In some cases, field-collected flies were also used for rearingto propagate specimens. Rearing was carried out in meshcages under ambient conditions on potted host plants orcut branches in water.Drosophilamelanogaster (Drosophilidae,strain oregonR) was kindly provided by theMax Planck Insti-tute for Developmental Biology, Tübingen, Germany.

RNA Extraction, RT-PCR, and Sequencing

For all fly species mentioned above, sequences were obtainedfor the mitochondrial COI gene and the nuclear 28S andcinnamyl alcohol dehydrogenase (CAD) genes to includethem in phylogenetic analysis and for their Na,K-ATPasegenes. All sequencing was based on RNA extractions, asthe Na,K-ATPase 1a gene consists of multiple exons sepa-

This content downloaded from 132All use subject to University of Chicago Press Term

rated by large intronic sequences (e.g., in D. melanogaster,the coding sequence of the commonest isoform of the Na,K-ATPase a stretches over 10.195 bp and 8 exons). ForRT-PCR, flies were decapitated and only heads were used,as their brains represent a rich source of Na,K-ATPase(Lebovitz et al. 1989). RNA was extracted with the RNeasykit (Qiagen, Hilden, Germany) and reverse transcribed (Su-perscript III, Invitrogen, Carlsbad, CA). Na,K-ATPase se-quences were amplified in standard PCR using the prim-ers given in Dobler et al. (2012). This yielded a fragment ofthe Na,K-ATPase covering the amino acids 89–805. Thenumbers refer to the mature protein of the pig (Sus scrofa)to enable comparison with the extensive literature on verte-brate Na,K-ATPase including crystal structures of the pro-tein (Yatime et al. 2011; Laursen et al. 2013, 2015). In addi-tion, gene sequences for COI, 28S, and CAD were obtainedforC. horticola, L. sp. (Mexico),N. lateralis,N. scrophulariae,and P. ilicis (see table 1) using primers given in Scheffer et al.(2007) and additional primers for COI given in Maus et al.(2001).

Transcriptome Analyses

Transcriptomes based on mRNA preparations were obtainedfor L. asclepiadis, L. eupatorii, P. digitalis, P. fallaciosa,and P. hellebori. Since Ranunculus sp. and Eupatorium can-nabinum are hosts for several species of leaf-mining flies,identification of L. eupatorii and P. fallaciosa based on hostplant and mine morphology was verified by barcoding in-dividual flies that were later pooled for transcriptome anal-yses. For this purpose, DNA of 2–3 legs of each individualwas extracted following the method detailed in Carvalhoet al. (2009). A PCR product for COI was generated withprimers S1634 and A2188 and sequenced to confirm speciesidentity before transcriptome sequencing. RNA extractionsof roughly 10 flies of L. eupatorii, P. digitalis, P. fallaciosa,and P. hellebori and of a single individual of L. asclepiadiswere obtained as above. mRNA-specific cDNA library con-struction and paired-end sequencing on an Illumina HiSeq2000 (Illumina, San Diego, CA) were carried out by StarSeq(Mainz, Germany) and yielded roughly 50 million readsfor each species. All transcriptomes, except for P. helle-bori, were assembled with Trinity 2.1.1 (Grabherr et al. 2011)using internal trimming with modified trimming options(SLIDINGWINDOW:4:30 LEADING:30 TRAILING:30MINLEN:50). Reads of P. hellebori were first trimmed withsickle, version 1.29 (Joshi and Fass 2011; min-length: 50 bp),and then assembled using trinityrnaseq_r20140413p1 (Grab-herr et al. 2011). Local Blast searches (Blastn 2.2.271 [Zhanget al. 2000] or Tblastn 2.2.271 [Altschul et al. 1997]) wereused to identify gene copies of the Na,K-ATPase a-subunitand the genes needed for the phylogenetic analysis. Na,K-ATPase sequences were also checked for alternative gene cop-

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S34 The American Naturalist

ies with CLC Genomics (Qiagen) by mapping all trimmedreads against reference sequences identified in the Trinityassemblies. Paralogs of the Na,K-ATPase are usually easyto recognize since variation between gene copies by far ex-ceeds allelic variation observed between individuals. Intra-individual variation accounts for only a few silent substi-tutions, and we extremely rarely observed nonsynonymoussubstitutions between individuals, the few cases being re-stricted to the large intracellular loop of the enzyme. Never-theless, we acknowledge that we can recognize only genecopies that differ by at least a few substitutions and are ex-pressed in the adult flies.

Phylogenetic Analyses

To test whether the chosen species pairs of Agromyzidaewere suitable phylogenetic comparisons, we used COI, 28S,and CAD sequences from Scheffer et al. (2007) andWinkleret al. (2009) and extended the data set with sequences ofthe flies investigated here (see table 1 for GenBank acces-sion numbers). From the original very extensive data set ofScheffer et al. (2007), we included all species of Liriomyza,Napomyza, and Phytomyza and three species of all majorgenera resolved between Agromyza and Phytomyza. Specif-ically, we chose three species of Agromyza, Aulagromyza,Calycomyza,Cerodontha,Ophiomya, andPhytoliriomyzaarc-tica. In addition, we used COI and CAD sequence data forPhytomyza ilicis and Phytomyza crassiseta from Winkleret al. (2009). In all cases where sequences from our ownRT-PCR or transcriptome analyses were available as well assequences from Scheffer et al. (2007) orWinkler et al. (2009),the sequences were identical or the differences negligible(few silent substitutions in the protein-coding genes).

Sequence alignment for COI (1,404 bp) and CAD (704 bp)was straight-forward and immediately congruent with thealignment by Scheffer et al. (2007) archived in TreeBASE(www.treebase.org, accession no. SN3150). In contrast, 28Sproved difficult to align so that we used the alignment ofScheffer and coauthors and manually added our sequencesto this alignment in MEGA7 (Kumar et al. 2016). This re-sulted in a total length of 823 bp and the exclusion of 168 basesdistributed in small groups along the gene that Scheffer andcolleagues had judged impossible to align unambiguously.The aligned concatenated sequences were subjected to max-imum likelihood analysis with RAxML (Stamatakis 2014) fit-ting a GTRGAMMAmodel of sequence evolution for each ofthe three gene partitions separately. Na,K-ATPase sequencesof the focal species were aligned with MEGA7 and subjectedto maximum likelihood analysis assuming a GTRGAMMAmodel of sequence evolution. For both data sets, bootstrap sup-port values were generated in 500 replicated analyses. Align-ments are archived in TreeBASE (www.treebase.org, acces-sion no. 20548).

This content downloaded from 132All use subject to University of Chicago Press Term

In Vitro Assays of Na,K-ATPase

Resistance of the Na,K-ATPase can be monitored in vitroby measuring the enzyme’s activity through spectrophoto-metric quantification of inorganic phosphate released fromadenosine triphosphate (ATP) hydrolysis under increasingconcentrations of cardiac glycosides. We assessed in vitroresistance of Na,K-ATPase to ouabain (a water-soluble stan-dard cardenolide) in all specialist agromyzid species feedingon cardiac glycoside–containing plants except for N. scro-phulariae and L. sp. (not enough material available). Forcomparison, we included the nonadapted agromyzid spe-cies P. ilicis and D. melanogaster as an outgroup. For Na,K-ATPase in vitro assays, flies were decapitated, and pooledheads (between 17–45) were homogenized in an all-glassgrinder (Wheaton) in 200–500 mL deionized water. Homog-enates were frozen at 2807C and freeze-dried. Lyophilisateswere stored at 2807C until use. Na,K-ATPase assays werecarried out as described previously (Petschenka et al. 2013a).Briefly, head extractions reconstituted with 300–400 mL de-ionized water were incubated at a concentration of 1028–1023

M ouabain, and Na,K-ATPase activity was measured as theamount of inorganic phosphate generated from ATP hy-drolysis compared to a noninhibited control. We used Origin-Pro 2016 (OriginLab, Northampton, MA) for nonlinear curvefitting. Top and bottom asymptotes were set to 100 and 0,respectively. For graphic display, replicate data (2–4 biolog-ical replicates per fly species) were fitted using the “concat-enate fit” option. Pairwise statistical comparisons of dose-response curves were carried out using the “compare datasets” option in OriginPro. For this function of the program,curves fitted through the means of the data have to be used.We evaluated P values of pairwise statistical tests after Holm-Bonferroni corrections. Due to the unusual shape of the curveobtained for P. hellebori, curve fitting using the means wasnot possible, and thus we excluded P. hellebori from statisti-cal analysis. Raw data are deposited in the Dryad Digital Re-pository: http://dx.doi.org/10.5061/dryad.20191 (Petschenkaet al. 2017).

Results

Phylogenetic Analyses

Analysis of a comprehensive sample of agromyzid flies in-cluding our focal species corroborated the suitability of thechosen phylogenetic comparisons. The three molecular mark-ers yielded a concatenated data set of 2,931 bp of 28S, COI,and CAD sequences and resulted in a well-supported treethat is in good agreement with the previous analysis of Schef-fer et al. (2007; fig. 2). All genera as represented here are re-solved as monophyletic groups, with the exception of theplacement of Chromatomyia horticola within the genus Phy-

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Parallel Evolution of Insect Resistance S35

tomyza. The lack of clear differentiation between Chromato-myia and Phytomyza species has been extensively demon-strated and discussed in Scheffer et al. (2007) and Winkleret al. (2009). More importantly, for the present study, ourphylogenetic analysis corroborated the five independent col-onizations of different plant lineages with cardiac glycosides,including the cardenolide type (in Asclepias, Digitalis, andErysimum) and the bufadienolide type (in Helleborus; fig. 2).In particular, the chosen species pairs, Phytomyza digitalisversus Phytomyza crassiseta, Phytomyza hellebori versus Phy-tomyza fallaciosa, Napomyza scrophulariae versus Napomyzalateralis, Liriomyza asclepiadis/Liriomyza sp. Mexico versusLiriomyza eupatorii, and C. horticola versus Phytomyza iliciswere confirmed as appropriate phylogenetic comparisonsthat contrast species adapted to cardiac glycosides to non-adapted relatives.

This content downloaded from 132All use subject to University of Chicago Press Term

Sequence Analysis of Na,K-ATPase Genes

Our molecular screening for target site insensitivity mediatedby mutations in the cardiac glycoside–binding pocket of theNa,K-ATPase gene revealed multiple convergently acquiredamino acid substitutions.No such substitutionswere detectedin four out of five species that are not exposed to dietary car-diac glycosides, rather these species maintained the ancestralamino acid residues present inDrosophila melanogaster at allpositions known to be part of the cardiac glycoside–bindingpocket (fig. A1, available online). In contrast, multiple substi-tutions of amino acid residues were observed at these posi-tions in five out of six species feeding on cardiac glycosidesplants. Specifically, two residues at the border of the first ex-tracellular loop, 111 and 122, which are well known to be crit-ical for cardiac glycoside resistance in other insects, were re-

Figure 2: Maximum likelihood analysis of COI, CAD, and 28S gene sequences of leaf-mining flies with GTRGAMMA models fitted to each genepartition. The data set was compiled from Scheffer et al. (2007) and enlarged by sequences from transcriptomes (asterisk), reverse transcription po-lymerase chain reaction (RT-PCR; double asterisk), or partly from Winkler et al. (2009), with additional sequences from RT-PCR (triple asterisk).Species representing our phylogenetic comparisons are boxed and connected by boldface lines, those feeding on cardiac glycoside–containing plantsare indicated “1CG,” and those from plants devoid of cardiac glycosides are indicated “2CG.” Plant names indicate the hosts from which the spec-imens were collected. Branches are drawn according to scale; numbers along branches indicate bootstrap values in 500 replicated analyses.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S36 The American Naturalist

peatedly exchanged (fig. 3). An exchange of glutamine 111 forleucine (Q111L), which we previously reported for the un-identified Liriomyza sp. collected from Asclepias angustifoliain Mexico (Dobler et al. 2012) was observed in two furtherspecies, P. digitalis and N. scrophulariae. The same exchangeis also present in the polyphagous N. lateralis and must haveoccurred in the common ancestor of both species, N. later-alis and N. scrophulariae. The substitution of histidine at po-sition 122 for the conserved asparagine detected earlier bydirect RT-PCR amplification in L. asclepiadis (Dobler et al.2012) was confirmed here by our transcriptome analysis. Thesame substitution was once more discovered in P. helleboriand coincides with an additional exchange of glutamine 111for histidine, a substitution that has not been previously re-ported in insects or vertebrates.

A duplication of the Na,K-ATPase gene could be detectedin only one of the agromyzid flies analyzed here. In P. hel-

This content downloaded from 132All use subject to University of Chicago Press Term

lebori, our transcriptome analyses and RT-PCR revealed ev-idence for a second gene copy. These two copies have a P dis-tance of 0.009, with seven nonsynonymous substitutions of21 total. In addition, all five individuals analyzed by PCRwere polymorphic at the same sites. One of the two genecopies features the conserved glutamine at position 111and the conserved asparagine at position 122, while the sec-ond gene copy has a histidine at both positions (figs. 3, A1).An analysis of the read frequencies in the transcriptomeshowed that the double-substituted gene copy is roughlyfourfold higher expressed than the one with the ancestralcondition. Other than these differences, the transcriptomesprovided evidence for only splice variants at the N-terminusoftheenzymeandformutuallyexclusiveexons(exon6)imme-diately downstream of the region analyzed here. These splicevariants correspond to the well-described situation in D. mela-nogaster (Palladino et al. 2003).

Figure 3: Maximum likelihood analysis of Na,K-ATPase sequences of Agromyzidae (GTRGAMMA model). The amino acid residues at po-sitions 111 and 122 are indicated along the branches and at the terminal nodes (numbering according to the mammalian Na,K-ATPase); thederived conditions are indicated in boldface and italic. Species on cardiac glycoside plants are boxed. The insert shows the relevant aminoacids in the enlarged cardiac glycoside–binding cavity of the Na,K-ATPase. H p histidine, L p leucine, N p asparagine, Q p glutamine.Sequences are from transcriptomes (asterisk) or reverse transcription polymerase chain reaction (double asterisk). Branch lengths are drawnaccording to scale; numbers along branches indicate bootstrap values in 500 replicated analyses.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Parallel Evolution of Insect Resistance S37

Na,K-ATPase Assays

Our in vitro assays of tissue-extracted fly Na,K-ATPasesrevealed different degrees of cardenolide resistance acrossthe fly species tested (fig. 4). Na,K-ATPase of the non-agromyzid outgroup species D. melanogaster was stronglyinhibited by ouabain (half maximal inhibitory concentra-tion [IC50] p 4:6# 1027 M). The Na,K-ATPase of theagromyzid species P. ilicis, which is not adapted to cardiacglycosides, showed a similar response to ouabain (IC50 p4:86# 1027 M), and statistical comparison revealed no dif-ference between inhibition of the P. ilicis andD.melanogasterNa,K-ATPase (F2, 8 p 3:802, P p :069). In contrast, Na,K-ATPases of P. digitalis and L. asclepiadis displayed 3 and30 times increased cardenolide resistance, respectively (P.digitalis: IC50 p 1:47# 1026 M; L. asclepiadis: 1:43#1025 M). Comparison of both dose-response curves to thecurve obtained from P. ilicis and comparisons among eachother revealed significant differences (P. digitalis vs. P. ilicis:F2, 8 p 100:664, P ! :001; L. asclepiadis vs. P. ilicis: F2, 8 p732:122, P ! :001; L. asclepiadis vs. P. digitalis: F2, 8 p504:26, P ! :001). Phytomyza hellebori was a special caseas ouabain inhibition of the P. hellebori Na,K-ATPase de-viated from the monophasic sigmoidal shape of the otherdose-response curves. The differing form of this curve re-sembled a two-component curve caused by two coexistingforms of Na,KATPase with different sensitivities to ouabain

This content downloaded from 132All use subject to University of Chicago Press Term

(Blanco and Mercer 1998). For fly-head preparations fromP. hellebori and P. digitalis, we verified that ouabain is notmetabolized under the conditions of our assay, demon-strating that the observed resistance is based on the bio-chemical properties of Na,K-ATPase and is not due todegradation of the inhibitor. Moreover, ouabain is knownnot to be metabolized by the Malpighian tubules of D.melanogaster (Torrie et al. 2004).

Discussion

The strategies of herbivorous insects to cope with toxic car-diac glycosides represent a striking case of convergent mo-lecular evolution (Dobler et al. 2012, 2015; Zhen et al. 2012)that allows insights into the origin and underlying rules ofadaptive amino acid substitutions (Storz 2016). The moststriking observation is the high frequency of occurrenceof one specific amino acid substitution in the Na,K-ATPase,N122H, that has evolved independently in six orders of phy-tophagous insects exposed to cardiac glycosides: in the Caeli-fera (S. Dobler, G. Petschenka, andV.Wagschal, unpublisheddata), Coleoptera (at least twice), Diptera, Hemiptera, Hyme-noptera, and Lepidoptera (Holzinger et al. 1992; Dobler et al.2012, 2015; Zhen et al. 2012). The present study increases thenumber of independent origins of theN122Hexchange, asweshow that this substitution also evolved twice independently

Figure 4: Sensitivity of dipteran Na,K-ATPases to ouabain in vitro. Data points represent the mean of 2–4 biological replicates (5SE); thenumber of actual replicates per species is given in parentheses after species names. Solid lines indicate species that are exposed to cardiacglycosides in their diets; dashed lines represent species that are not naturally exposed to the toxins. Dark blue p Phytomyza ilicis (4), green pDrosophila melanogaster (4), purple p Phytomyza digitalis (4), light blue p Phytomyza hellebori (4), orange p Liriomyza asclepiadis (2). Notethe deviating characteristics of P. hellebori.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S38 The American Naturalist

in the Agromyzidae (fig. 5). As we will discuss below, furtherpotentially resistance-conferring amino acid substitutions haverepeatedly evolved in the cardiac glycoside–binding pocket ofthe Na,K-ATPase in insects as well as in vertebrate predatorsexposed to cardiac glycosides in their diet.

As far as we can tell, a remarkable difference from ourinsects to other cases of convergent molecular evolutionconsists of the lack of intraspecific variability in resistanceto dietary cardiac glycosides. Neither sequence analyses

This content downloaded from 132All use subject to University of Chicago Press Term

of the Na,K-ATPase a gene (Aardema et al. 2012; Pierceet al. 2016; S. Dobler and V. Wagschal, unpublished data)nor in vitro enzyme assays testing for the resistance of thecorresponding protein to cardiac glycosides (fig. 4; Pets-chenka et al. 2013a; Bramer et al. 2015) so far revealed evi-dence for noticeable intraspecific variability. This contrastswith other well-investigated systems where the molecularfoundation of adaptive traits has been unraveled, such asthe blanched coloration displayed by lizards at White Sands

Figure 5: Compilation of available data on Na,K-ATPase sequences in insects adapted and nonadapted to cardiac glycosides. The overall treeand node ages correspond to the best-supported insect phylogeny to date (Misof et al. 2014) that has been complemented by phylogenetichypotheses presented in the Tree of Life Web Project (Maddison and Schulz 2007) and our own data. Species in black are not exposed tocardiac glycosides, species in orange are known to sequester cardiac glycosides from their host plants, and species in blue feed on cardiacglycosides plants but either are not known or not able to sequester cardiac glycosides. To the right are amino acid residues known to formthe cardiac glycoside–binding pocket of the Na,K-ATPase (see fig. 1), with potential resistance-conferring exchanges highlighted by grayboxes. The exchange of E312D (glutamic acid for aspartic acid) between vertebrates and all insects is considered negligible. Stars indicategene duplications. In Lygaeus kalmii and Oncopeltus fasciatus, both indicated copies are present plus a third less heavily substituted one fea-turing N122H and F786S exchanges. In Rhyssomatus lineaticollis, a second copy has the ancestral Q111 but a N122Y substitution. InChrysochus auratus, the second copy is ancestral except for a Q111L exchange (Zhen et al. 2012). In Phytomyza hellebori, the second genecopy has the ancestral state. For GenBank accession numbers, see table 1.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Parallel Evolution of Insect Resistance S39

(Rosenblum et al. 2010; this issue) and beachmice (Hoekstraet al. 2006; Mullen and Hoekstra 2008) or insect resistance topyrethroids (Rinkevich et al. 2013). In these cases, heritablevariability persists since the selective pressures vary betweenhabitats. In the case of adaptations to dietary cardiac gly-cosides, adoption of these feeding habits lies in the distantpast and the favorablemutations have apparently been drivento fixation (Aardema et al. 2012). A screen for intraspecificvariability might be most promising in generalist insect spe-cies if populations vary in host use, feeding locally on plantswith or without cardiac glycosides.

While we detected a striking amount of molecular con-vergence in resistance-conferring substitutions in the Na,K-ATPase, this strategy is not the only possible adaptationto avoid the toxic effect of dietary cardiac glycosides. Imper-meable guts that prevent the toxins from entering the body,or fast and efficient elimination from the hemolymph via theMalpighian tubules, represent possible alternatives and havebeen detected in nonadapted as well as cardiac glycoside–adapted insects (Scudder and Meredith 1982; Torrie et al.2004; Petschenka andDobler 2009; Petschenka and Agrawal2015). In addition, a protection of the sensitive neural tis-sue expressing high levels of Na,K-ATPase by active carriersin the insects’ blood-brain barrier plays an important role andis possibly sufficient to cope with small concentrations of he-molymph cardiac glycosides (Petschenka and Dobler 2009;Petschenka et al. 2013b). Thus, alternative evolutionary tra-jectories are possible to cope with dietary cardiac glycosides.

Remarkably, convergent molecular evolution of cardiacglycoside resistance by amino acid substitutions has alsoevolved in several vertebrates feeding on bufadienolide-producing toads (Jaisser et al. 1992; Ujvari et al. 2013,2015; Mohammadi et al. 2016). As in insects, amino acidsubstitutions in the first extracellular loop of the Na,K-ATPase mediate resistance to these toxins in their diet. Itis, however, noteworthy that in vertebrates adapted to car-diac glycosides, the substitution of N122H observed so of-ten in insects has never been described. Rather, other aminoacid replacements are present at position 122 (Price andLingrel 1988; Jaisser et al. 1992; Ujvari et al. 2013, 2015).Two reasons could account for these differences: First, thegenetic background of the remaining enzymemay differ be-tween insects and vertebrates, and thereby, the effect of spe-cific substitutions may also vary (e.g., Poelwijk et al. 2007;Storz 2016). Mutagenesis experiments that introduced theN122H exchange into the human and the Drosophila Na,K-ATPase in cell culture corroborate this idea (S. Dobler,S. Dalla, and F. List, unpublished data). Second, the toxinsthe two groups are exposed to may require different adap-tations. While the investigated vertebrates have evolved re-sistance to toad bufadienolides, most insects analyzed sofar are exposed to plant cardenolides. Recent crystallogra-phy, however, suggests that both types of toxins occupy

This content downloaded from 132All use subject to University of Chicago Press Term

the binding cavity in a similar fashion (Laursen et al. 2015).Our data further support that the optimal adaptation doesnot depend on the type of cardiac glycoside (or other aspectsof host plant variation), as a histidine at position 122 wasfound in Phytomyza hellebori exposed to bufadienolides (andthe same applies to the sawfly Monophadnus latus; Dobleret al. 2015), while Liriomyza asclepiadis, like the other speciesfeaturing the N122H exchange, is exposed to cardenolides inits Asclepias host plants (fig. 5).In many cases, in insects and vertebrates alike, substitu-

tions at position 122 are accompanied by exchanges at po-sition 111. The two residues Q111 and N122 are situated at theentrance of the cardiac glycoside–binding cavity and havelong been known to be of special importance for cardiacglycoside–binding affinity (Price and Lingrel 1988; Priceet al. 1990; Croyle et al. 1997). As the crystal structuresshow, binding of cardiac glycosides leads to a conforma-tional change bringing the aM1–M2 loop closer to the li-gand, and these two residues get close enough for hydrogenbonding (Yatime et al. 2011; Laursen et al. 2015). The re-placement of Q111 and N122 by residues not capable of hy-drogen bonding or having bulkier and charged side chainslike histidine apparently prohibit high-affinity cardiac gly-coside binding. Experimental evidence suggests that substi-tutions at positions 111 and 122 have synergistic functionaleffects on resistance (Price et al. 1990; Dobler et al. 2012;Dalla et al. 2013; Dalla and Dobler 2016).Compared to position 122, where we observed only ex-

changes of N122H, position 111 is more variable as to whichamino acid substitutions cause resistance to cardiac glyco-sides. At this position, leucine, valine, threonine, and glu-tamic acid have been repeatedly observed in cardiac glycoside–adapted insects (Labeyrie and Dobler 2004; Dobler et al.2012, 2015; Zhen et al. 2012), and here for the first timean exchange of Q111 for histidine is reported. Position 111is of importance in vertebrates adapted to cardiac glycosidesas well, and the observed exchanges partially match thosepresent in insects. In snakes and lizards known to feed ontoads, the Q111L exchange has evolved four times conver-gently, while toads, rodents, the European hedgehog, andthe frog Leptodactylus latrans display a substitution of argi-nine for Q111 (Price and Lingrel 1988; Jaisser et al. 1992;Ujvari et al. 2013; Ujvari et al. 2015). Though in vertebratesthe exchanges of Q111 are also coupled with substitutions atthe C-terminal end of the first extracellular loop of the Na,K-ATPase, these exchanges (at positions 119, 120, or 122) dif-fer from those observed in insects. Overall, this demonstratesan astonishingly high degree ofmolecular convergence acrossanimal phyla, but the use of different (presumably optimal)substitutions at the same time provides evidence for con-straints (or phylogenetic bias) that may be due to either thedifferences in the genetic background of the vertebrate versusinsect Na,K-ATPase or different physiological conditions.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S40 The American Naturalist

Although Q111L seems to be associated with resistancein vertebrates, the functional effect of this substitutionhas not been tested in isolation, that is, without the accom-panying substitutions at the positions mentioned above.While the observed resistance in Phytomyza digitalis coin-cides with Q111L, there is no evidence that this substitu-tion actually increases resistance of Na,K-ATPase in severallepidopteran insects, including nymphalid butterflies, arctiidmoths, and the oleander hawk moth (Petschenka et al. 2012,2013a; Dobler et al. 2015). Compared to butterflies andmoths, the situation in leaf-mining flies might be different,and the effect of Q111L needs to be investigated in more de-tail. Along the same lines, it is currently unclear why the Na,K-ATPase of the generalist Napomyza lateralis carries thissubstitution, although the species might not be exposed tocardiac glycosides in its diet.

In insects as well as in vertebrates, the available data sug-gest that amino acid substitutions in and at the border ofthe first extracellular loop (M1–M2) are much more likelyto evolve than exchanges in the remainder of the cardiacglycoside–binding cavity. According to the Na,K-ATPasecrystal structures, bulky and hydrophobic side chains ofthe transmembrane helices M4–M6 form a general dockingplatform for the steroid core of all cardiac glycosides, whiletheir lactone rings fit in a hydrophobic funnel of thesetransmembrane helices that is leading to the cation bindingsites of the enzyme (Laursen et al. 2015). The proximity tothe cation binding sites seems to create functional con-straints that may explain why amino acid substitutions inthese parts of the enzyme are so rare. So far they have beenobserved only in lygaeid bugs that possess three copies ofthe Na,K-ATPase a1 gene and in aphids where retrotrans-position seems to have created a second gene copy with al-tered properties (Zhen et al. 2012). Functional data supportthe notion that the amino acid exchanges observed in trans-membrane domains M5 and M6 of Oncopeltus fasciatus se-riously compromise the enzyme’s activity as an ion trans-porter (Dalla and Dobler 2016).

Testing the functional effects of the observed amino acidsubstitutions is critical for understanding the value of singlereplacements versus combinations as well as potential pleio-tropic constraints. Two approaches can be followed: en-zyme assays, as used here, to test for the sensitivity of thenative enzyme that is expressed at high levels in the nervoustissue of animals (Vaughan and Jungreis 1977; Moore andScudder 1986; Al-Robai 1993; Petschenka et al. 2012) orgenetic modification of Na,K-ATPase genes followed byevaluation of the physiological effects of the introducedmu-tations in enzymes expressed in cell culture (Price et al. 1990;Holzinger et al. 1992; Qiu et al. 2005; Dalla et al. 2013; Ujvariet al. 2013; Dalla and Dobler 2016). Both lines of evidencedemonstrate that the exchange of N122H causes a high levelof resistance to cardiac glycosides (Holzinger et al. 1992; Dalla

This content downloaded from 132All use subject to University of Chicago Press Term

et al. 2013; Petschenka et al. 2013a; Bramer et al. 2015). Inaccordance with these results obtained from genetically en-gineered Drosophila Na,K-ATPase, the inhibition curve ob-tained for L. asclepiadis shows that the single exchange ofN122H in an enzyme that is otherwise highly similar to thesensitiveDrosophilaNa,K-ATPase is sufficient to increase re-sistance by 31-fold.The heterogeneity of Na,K-ATPase sequences across the

fly species adapted to cardiac glycosides is mirrored by thedifferent sensitivities of their Na,K-ATPases to ouabain.Drosophila melanogaster and P. ilicis display similar sen-sitivities to ouabain, suggesting that the Na,K-ATPase ofP. ilicis likely represents the ancestral condition in the Agro-myzidae. The dose-response curve of P. digitalis Na,K-ATPase in comparison displays a threefold-higher resistanceto ouabain, thus supporting adaptation to its host plant’scardenolides. The highest in vitro resistance in our data setwas observed for L. asclepiadis. However, P. hellebori mayhave an even more resistant form of the Na,K-ATPase, butits interpretation is confounded by the coexistence of twodifferent forms of Na,K-ATPase. In P. hellebori, the dose-response curve can be interpreted as displaying the be-havior of two forms with differing sensitivity: while inhi-bition of the sensitive copy (with the ancestral Q111 and N122

amino acids) leads to a rapid activity decline at relativelylow concentrations of ouabain, the resistant copy (with H111

and H122) maintains a level of activity at millimolar ouabainconcentrations that is higher than for L. asclepiadis. Thisresembles the situation described for rat testis, where boththe a1 and a4 isoforms, with strongly differing ouabain sen-sitivities, are expressed together (Blanco and Mercer 1998).Gene duplications as observed here for P. hellebori could

buffer the detrimental effects of substitutions by counter-acting purifying selection on individual proteins and couldhave a direct advantage due to expression of higher genedoses (Soskine and Tawfik 2010; Kondrashov 2012). Othermodels, in contrast, suggest that gene duplication is a neu-tral event that may not be under selective pressure (Soskineand Tawfik 2010). For insects adapted to cardiac glycosides,previous analyses suggested that resistance-conferring mu-tations may go along with duplications of the Na,K-ATPasea1 gene (Zhen et al. 2012; indicated by stars in fig. 5). This isespecially prominent in lygaeid bugs where two duplica-tions gave rise to three gene copies that differ in the numberof substitutions, including unusual exchanges toward theC-terminus of the enzyme (Zhen et al. 2012). The Agro-myzidae P. hellebori provides a further example for a dupli-cation of the Na,K-ATPase gene that resulted in an unal-tered, ancestral form and a fourfold-higher expressed formwith two resistance-conferring mutations in the M1–M2domain (Q111H, N122H). As for the debate whether substi-tutions precede duplications or only follow later when theduplicate is freed from purifying selection (Soskine and Taw-

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Parallel Evolution of Insect Resistance S41

fik 2010), P. hellebori provides an example for an accu-mulation of mutations post duplication, as the ancestralcopy that is shared by the closely related P. fallaciosa is stillpresent. The duplication may have been driven by the ad-vantage of increased gene dosage in the face of toxins com-promising gene function and subfunctionalization of the du-plicates ensued later (Kondrashov 2012).

Duplications of the Na,K-ATPase gene can alleviate neg-ative pleiotropy imposed by the dual role of theNa,K-ATPasefor ion transport and morphogenesis (Xie and Askari 2002;Palladino et al. 2003; Horisberger 2004; Paul et al. 2007) andallow for resistance-conferring substitutions that compromiseion transport (Dalla and Dobler 2016). Nevertheless, in ouranalysis of Agromyzidae on cardiac glycoside plants, we de-tected four incidents of substitutions in the cardiac glycoside–binding pocket but only a single gene duplication event, de-spite the fact that all five independent lineages started fromthe same ancestral state. This suggests that in the case ofthe Na,K-ATPase, constraints on gene duplications seemto be stronger than constraints on amino acid substitutions,at least as long as substitutions occur only at the favored res-idues of the first extracellular loop and the bordering trans-membrane domains (M1–M2).We still know too little aboutadditional roles of insect Na,K-ATPase genes beyond iontransport (Paul et al. 2007), but the evidence accumulatingfor vertebrate Na,K-ATPases suggests that pleiotropic con-straints by Na,K-ATPase–initiated signaling cascades maybe immense (Xie and Askari 2002; Schoner and Scheiner-Bobis 2007; Li and Xie 2009). The functional constraints ofthis enzyme imposed by its dual role thus seem to enforcea strongly canalized evolutionary path.

Acknowledgments

We thank A. Agrawal for the opportunity to contributeto this symposium, for helpful discussions, and for his crit-ical reading of the manuscript. He, A. Hastings, and T. Züsthelped with the collection of Liriomyza asclepiadis. M. Fal-kenberg and R. Trusch supported the rediscovery of Phy-tomyza digitalis in Germany. L. M. Arcila Hernandez, A.Baudach, R. Halitschke, M. Schott, and S. Stiehler helpedwith collecting leaf-mining flies and data. A. Vilcinskas gen-erously provided materials for some parts of this study toG.P. The project was funded by a grant from the TempletonFoundation and Deutsche Forschungsgemeinschaft (DFG;Do 527/5-3) to S.D. G.P. was financially supported by theDFG (PE 2059/1, PE 2059/2).

Literature Cited

Aardema, M. L., Y. Zhen, and P. Andolfatto. 2012. The evolution ofcardenolide-resistant forms of Na1,K1-ATPase in Danainae but-terflies. Molecular Ecology 21:340–349.

This content downloaded from 132All use subject to University of Chicago Press Term

Agrawal, A. A., G. Petschenka, R. A. Bingham, M. G. Weber, and S.Rasman. 2012. Toxic cardenolides: chemical ecology and coevolu-tion of specialized plant-herbivore interactions. New Phytologist194:28–45.

Al-Robai, A. A. 1993. Different ouabain sensitivities of Na1/K1-ATPase from Poekilocerus bufonius tissues and a possible physio-logical cost. Comparative Biochemistry and Physiology B: Bio-chemistry and Molecular Biology 106:805–812.

Altschul, S. F., T. L. Madden, A. A. Schaffer, J. Zhang, Z. Zhang, W.Miller, and D. J. Lipman. 1997. Gapped BLAST and PSI-BLAST: anew generation of protein database search programs. Nucleic AcidsResearch 25:3389–3402.

Andreev, D., M. Kreitmann, T. W. Phillips, R. W. Beeman, and R. H.ffrench-Constant. 1999. Multiple origins of cyclodiene insecticideresistance in Tribolium castaneum (Coleoptera: Tenebrionidae).Journal of Molecular Evolution 48:615–624.

Blanco, G., and R. W. Mercer. 1998. Isozymes of the Na-K-ATPase:heterogeneity in structure, diversity in function. American Journalof Physiology-Renal Physiology 275:F633–F650.

Bramer, C., S. Dobler, J. Deckert, M. Stemmer, and G. Petschenka.2015. Na1/K1-ATPase resistance and cardenolide sequestration:basal adaptations to host plant toxins in the milkweed bugs (He-miptera: Lygaeidae: Lygaeinae). Proceedings of the Royal SocietyB 282, http://dx.doi.org/10.1098/rspb.2014.2346.

Carvalho, G. B., W. W. Ja, and S. Benzer. 2009. Non-lethal PCRgenotyping of single Drosophila. Biotechniques 46:312–314.

Croyle, M. L., A. L. Woo, and J. B. Lingrel. 1997. Extensive randommutagenesis analysis of the Na1/K1-ATPase a subunit identifiesknown and previously unidentified amino acid residues that alterouabain sensitivity—implications for ouabain binding. EuropeanJournal of Biochemistry 248:488–495.

Dalla, S., and S. Dobler. 2016. Gene duplications circumvent trade-offs in enzyme function: insect adaptation to toxic host plants.Evolution 70:2767–2777.

Dalla, S., H. G. P. Swarts, J. B. Koenderink, and S. Dobler. 2013.Amino acid substitutions of Na,K-ATPase–conferring decreasedsensitivity to cardenolides in insects compared to mammals. In-sect Biochemistry and Molecular Biology 43:1109–1115.

Dobler, S., S. Dalla, V.Wagschal, and A. A. Agrawal. 2012. Community-wide convergent evolution in insect adaptation to toxic cardenolidesby substitutions in the Na,K-ATPase. Proceedings of the NationalAcademy of Sciences of the USA 109:13040–13045.

Dobler, S., G. Petschenka, and H. C. Pankoke. 2011. Coping with toxicplant compounds: the insect’s perspective on iridoid glycosides andcardenolides. Phytochemistry 72:1591–1602.

Dobler, S., G. Petschenka, V. Wagschal, and L. Flacht. 2015. Conver-gent adaptive evolution: how insects master the challenge of cardiacglycoside–containing host plants. Entomologia Experimentalis etApplicata 157:30–39.

Feldman, C. R., E. D. Brodie Jr., E. D. Brodie 3rd, and M. E.Pfrender. 2012. Constraint shapes convergence in tetrodotoxin-resistant sodium channels of snakes. Proceedings of the NationalAcademy of Sciences of the USA 109:4556–4561.

Glombitza, K.-W., C. Kucera-Waldmann, and U. Fricke. 1989. Doroots of Helleborus niger contain cardioactive substances? PlantaMedica 55:107.

Grabherr, M. G., B. J. Haas, M. Yassour, J. Z. Levin, D. A. Thompson,I. Amit, X. Adiconis, et al. 2011. Trinity: reconstructing a full-lengthtranscriptome without a genome from RNA-Seq data. Nature Bio-technology 29:644–652.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S42 The American Naturalist

Hering, E. M. 1927. I. Agromyzidae (80. Familie). Die TierweltDeutschlands und der angrenzenden Meeresteile nach ihren Merk-malen und nach ihrer Lebensweise. Fischer, Jena.

———. 1935. Die BlattminenMittel- und Nord-Europas einschliesslichEnglands; Bestimmungstabellen aller von Insektenlarven der verschie-denen Ordnungen erzeugten Minen. G. Feller, Neubrandenburg.

———. 1951. Biology of the leaf miners. Dr. W. Junk, s-Gravenhage.Hoekstra, H. E., R. J. Hirschmann, R. A. Bundey, P. A. Insel, and J. P.Crossland. 2006. A single amino acid mutation contributes toadaptive beach mouse color pattern. Science 313:101–104.

Holzinger, F., C. Frick, and M. Wink. 1992. Molecular basis for theinsensitivity of the monarch (Danaus plexippus) to cardiac glyco-sides. FEBS Letters 314:477–480.

Horisberger, J. D. 2004. Recent insights into the structure and mech-anism of the sodium pump. Physiology 19:377–387.

Jaisser, F., C. M. Canessa, J. D. Horisberger, and B. C. Rossier. 1992.Primary sequence and functional expression of a novel ouabain-resistant Na,K-ATPase—the beta-subunit modulates potassiumactivation of the Na,K-pump. Journal of Biological Chemistry 267:16895–16903.

Joshi, N. A., and J. N. Fass. 2011. Sickle: a sliding-window, adaptive,quality-based trimming tool for FastQ files. Version 1.29. https://github.com/najoshi/sickle.

Kondrashov, F. A. 2012. Gene duplication as a mechanism of genomicadaptation to a changing environment. Proceedings of the RoyalSociety B 279:5048–5057.

Krenn, L., and B. Kopp. 1998. Bufadienolides from animal and plantsources. Phytochemistry 48:1–29.

Kumar, S., G. Stecher, and K. Tamura. 2016. MEGA7: Molecular Evo-lutionary Genetics Analysis. Version 7.0 for bigger datasets. Molec-ular Biology and Evolution 33:1870–1874.

Labeyrie, E., and S. Dobler. 2004. Molecular adaptation of Chry-sochus leaf beetles to toxic compounds in their food plants. Molec-ular Biology and Evolution 21:218–221.

Laursen, M., J. L. Gregersen, L. Yatime, P. Nissen, and N. U. Fedosova.2015. Structures and characterization of digoxin- and bufalin-bound Na1,K1-ATPase compared with the ouabain-bound com-plex. Proceedings of the National Academy of Sciences of theUSA 112:1755–1760.

Laursen, M., L. Yatime, P. Nissen, and N. U. Fedosova. 2013. Crystalstructure of the high-affinity Na1K1-ATPase-ouabain complexwith Mg21 bound in the cation binding site. Proceedings of theNational Academy of Sciences of the USA 110:10958–10963.

Lebovitz, R. M., K. Takeyasu, and D. M. Fambrough. 1989. Mole-cular characterization and expression of the (Na1 1 K1)-ATPasealpha-subunit in Drosophila melanogaster. Embo Journal 8:193–202.

Lei, Z.-H., S. Yahara, T. Nohara, T.-B. Shan, and J.-Z. Xiong. 1996.Cardenolides from Erysimum cheiranthoides. Phytochemistry 41:1187–1189.

Li, Z., and Z. Xie. 2009. The Na/K-ATPase/Src complex and cardio-tonic steroid-activated protein kinase cascades. Pflugers Archive457:635–644.

Liang, M., J. Tian, L. Liu, S. Pierre, J. Liu, J. Shapiro, and Z. J. Xie.2007. Identification of a pool of non-pumping Na/K-ATPase.Journal of Biological Chemistry 282:10585–10593.

Luckner, M., and M. Wichtl. 2000. Digitalis. Wissenschaftliche,Stuttgart.

Ludwig, F. 1907. Weiteres zur Biologie von Helleborus foetidus.Zeitschrift für wissenschaftliche Insektenbiologie 3:45–50.

This content downloaded from 132All use subject to University of Chicago Press Term

Maddison, D. R., and K. S. Schulz. 2007. The Tree of Life Web Proj-ect. http://tolweb.org/tree. Accessed August 30, 2016.

Malcolm, S. B. 1991. Cardenolide-mediated interactions betweenplants and herbivores. Pages 251–295 in G. A. Rosenthal andM. R. Berenbaum, eds. Herbivores: their interactions with second-ary plant metabolites. Academic Press, San Diego, CA.

Maus, C., K. Peschke, and S. Dobler. 2001. Phylogeny of Aleocharainferred from mitochondrial cytochrome oxidase sequences (Co-leoptera: Staphylinidae). Molecular Phylogenetics and Evolution18:202–216.

Misof, B., S. L. Liu, K. Meusemann, R. S. Peters, A. Donath, C.Mayer, P. B. Frandsen, et al. 2014. Phylogenomics resolves thetiming and pattern of insect evolution. Science 346:763–767.

Mohammadi, S., E. D. Brodie Jr., L. A. Neuman-Lee, and A. H.Savitzky. 2016. Mutations to the cardiotonic steroid binding siteof Na1/K1-ATPase are associated with high level of resistance togamabufotalin in a natricine snake. Toxicon 114:13–15.

Moore, L. V., and G. G. E. Scudder. 1986. Ouabain-resistant Na,K-ATPases and cardenolide tolerance in the large milkweed bug,Oncopeltus fasciatus. Journal of Insect Physiology 32:27–33.

Mortelmans, J., M. Boeraeve, W. Tamsyn, W. Proesmans, and D.Dekeukeleire. 2014. Thirteen new Agromyzidae for Belgium. Bul-letin de la Societé Royale Belge d’Entomologie 150:141–148.

Mullen, L. M., and H. E. Hoekstra. 2008. Natural selection along anenvironmental gradient: a classic cline in mouse pigmentation. Evo-lution 62:1555–1570.

Palladino, M. J., J. E. Bower, R. Kreber, and B. Ganetzky. 2003. Neuraldysfunction and neurodegeneration in DrosophilaNa1/K1 ATPasealpha subunit mutants. Journal of Neuroscience 23:1276–1286.

Paul, S. M., M. J. Palladino, and G. J. Beitel. 2007. A pump-independentfunction of the Na,K-ATPase is required for epithelial junction func-tion and tracheal tube-size control. Development 134:147–155.

Petschenka, G., and A. A. Agrawal. 2015. Milkweed butterfly resis-tance to plant toxins is linked to sequestration, not coping with atoxic diet. Proceedings of the Royal Society B 282:20151865.

Petschenka, G., and S. Dobler. 2009. Target-site sensitivity in a spe-cialized herbivore towards major toxic compounds of its host plant:the Na1K1-ATPase of the oleander hawk moth (Daphnis nerii) ishighly susceptible to cardenolides. Chemoecology 19:235–239.

Petschenka, G., S. Fandrich, N. Sander, V. Wagschal, M. Boppré, andS. Dobler. 2013a. Stepwise evolution of resistance to toxic carde-nolides via genetic substitutions in the Na1,K1-ATPase of milk-weed butterflies (Lepidoptera: Danaini). Evolution 67:2753–2761.

Petschenka, G., J. Offe, and S. Dobler. 2012. Physiological screeningfor target site insensitivity and localization of Na1/K1-ATPase incardenolide-adapted Lepidoptera. Journal of Insect Physiology58:607–612.

Petschenka, G., C. Pick, V. Wagschal, and S. Dobler. 2013b. Func-tional evidence for physiological mechanisms to circumvent neu-rotoxicity of cardenolides in an adapted and a non-adapted hawk-moth species. Proceedings of the Royal Society B 280:20123089.

Petschenka, G., V. Wagschal, M. von Tschirnhaus, A. Donath, and S.Dobler. 2017. Data from: Convergently evolved toxic secondarymetabolites in plants drive the parallel molecular evolution of insectresistance. American Naturalist, Dryad Digital Repository, http://dx.doi.org/10.5061/dryad.20191.

Pierce, A. A., J. C. De Roode, and L. Tao. 2016. Comparative geneticsof Na1/K1-ATPase in monarch butterfly populations with varyinghost plant toxicity. Biological Journal of the Linnean Society 119:194–200.

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Parallel Evolution of Insect Resistance S43

Poelwijk, F. J., D. J. Kiviet, D. M. Weinreich, and S. J. Tans. 2007.Empirical fitness landscapes reveal accessible evolutionary paths.Nature 445:383–386.

Price, E. M., and J. B. Lingrel. 1988. Structure-function relationshipsin the Na, K-ATPase alpha subunit: site-directed mutagenesis ofglutamine-111 to arginine and asparagine-122 to aspartic acidgenerates a ouabain-resistant enzyme. Biochemistry 27:8400–8408.

Price, E. M., D. A. Rice, and J. B. Lingrel. 1990. Structure-functionstudies of Na,K-ATPase: site-directed mutagenesis of the borderresidues from the H1-H2 extracellular domain of the a subunit.Journal of Biological Chemistry 265:6638–6641.

Qiu, L. Y., E. Krieger, G. Schaftenaar, H. G. P. Swarts, P. H. G. M.Willems, J. L. H. H. M. De Pont, and J. B. Koenderink. 2005. Re-construction of the complete ouabain-binding pocket of Na,K-ATPase in gastric H,K-ATPase by substitution of only seven aminoacids. Journal of Biological Chemistry 280:32349–32355.

Rinkevich, F. D., Y. Du, and K. Dong. 2013. Diversity and convergenceof sodium channel mutations involved in resistance to pyrethroids.Pesticide Biochemistry and Physiology 106:93–100.

Rosenblum, E. B., H. Rompler, T. Schoneberg, and H. E. Hoekstra.2010. Molecular and functional basis of phenotypic convergencein white lizards at White Sands. Proceedings of the National Acad-emy of Sciences of the USA 107:2113–2117.

Scheffer, S. J., I. S. Winkler, and B. M. Wiegmann. 2007. Phylogeneticrelationships within the leaf-mining flies (Diptera: Agromyzidae)inferred from sequence data from multiple genes. MolecularPhylogenetics and Evolution 42:756–775.

Schoner, W., and G. Scheiner-Bobis. 2007. Endogenous and exoge-nous cardiac glycosides: their roles in hypertension, salt metabo-lism, and cell growth. American Journal of Physiology–Cell Phys-iology 293:C509–C536.

Scudder, G. G. E., and J. Meredith. 1982. The permeability of themidgut of three insects to cardiac glycosides. Journal of InsectPhysiology 28:689–694.

Soskine, M., and D. S. Tawfik. 2010. Mutational effects and the evolu-tion of new protein functions. Nature Reviews Genetics 11:572–582.

Spencer, K. A. 1966. A clarification of the genus Napomyza West-wood (Diptera: Agromyzidae). Proceedings of the Royal Entomo-logical Society B 35:29–40.

———. 1990. Host specialization in the world Agromyzidae (Dip-tera). Kluwer Academic, Dordrecht.

Stamatakis, A. 2014. RAxML version 8: a tool for phylogenetic anal-ysis and post-analysis of large phylogenies. Bioinformatics 30:1312–1313.

This content downloaded from 132All use subject to University of Chicago Press Term

Storz, J. F. 2016. Causes of molecular convergence and parallelism inprotein evolution. Nature Reviews Genetics 17:239–250.

Tarvin, R. D., J. C. Santos, L. A. O’Connell, H. H. Zakon, and D. C.Cannatella. 2016. Convergent substitutions in a sodium channelsuggest multiple origins of toxin resistance in poison frogs. Molec-ular Biology and Evolution 33:1068–1081.

Torrie, L. S., J. C. Radford, T. D. Southall, L. Kean, A. J. Dinsmore,S. A. Davies, and J. A. Dow. 2004. Resolution of the insect ouabainparadox. Proceedings of the National Academy of Sciences of theUSA 101:13689–13693.

Ujvari, B., N. R. Casewell, K. Sunagar, K. Arbuckle, W. Wuster, N.Lo, D. O’Meally, et al. 2015. Widespread convergence in toxin re-sistance by predictable molecular evolution. Proceedings of theNational Academy of Sciences of the USA 112:11911–11916.

Ujvari, B., H. C. Mun, A. D. Conigrave, A. Bray, J. Osterkamp,P. Halling, and T. Madsen. 2013. Isolation breeds naivety: islandliving robs Australian varanid lizards of toad-toxin immunity viafour-base-pair mutation. Evolution 67:289–294.

Vaughan, G. L., and A.M. Jungreis. 1977. Insensitivity of lepidopterantissues to ouabain: physiological mechanisms for protection fromcardiac glycosides. Journal of Insect Physiology 23:585–589.

Winkler, I. S., S. J. Scheffer, and C. Mitter. 2009. Molecular phylog-eny and systematics of leaf-mining flies (Diptera: Agromyzidae):delimitation of Phytomyza Fallén sensu lato and included speciesgroups, with new insights on morphological and host-use evolu-tion. Systematic Entomology 34:260–292.

Wissner, W., and H. Kating. 1974. Botanische und PhytochemischeUntersuchungen an den europäischen und kleinasiatischen Artender Gattung Helleborus. Planta Medica 26:228–249.

Xie, Z., and A. Askari. 2002. Na1/K1-ATPase as a signal transducer.European Journal of Biochemistry 269:2434–2439.

Yatime, L., M. Laursen, J. P. Morth, M. Esman, P. Nissen, and N. U.Fedosova. 2011. Structural insights into the high affinity bindingof cardiotonic steroids to the Na1,K1-ATPase. Journal of StructuralBiology 174:296–306.

Zhang, Z., S. Schwartz, L. Wagner, and W. Miller. 2000. A greedyalgorithm for aligning DNA sequences. Journal of ComputationalBiology 7:203–214.

Zhen, Y., M. L. Aardema, E.M.Medina,M. Schumer, and P. Andolfatto.2012. Parallel molecular evolution in an herbivore community. Sci-ence 337:1634–1637.

Symposium Editor: Anurag A. Agrawal

.236.027.111 on July 22, 2017 05:14:35 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Convergent Phenotypic Evolution despite Contrasting

Demographic Histories in the Fauna of White Sands*

Erica Bree Rosenblum,1,2,† Christine E. Parent,1,‡,§ Eveline T. Diepeveen,1,‡,∥

Clay Noss,1 and Ke Bi2,3

1. Department of Environmental Science, Policy, and Management, University of California, Berkeley, California 94720;2. Museum of Vertebrate Zoology, University of California, Berkeley, California 94720; 3. Computational GenomicsResource Laboratory, University of California, Berkeley, California 94720

Online enhancements: appendixes.

abstract: When are evolutionary outcomes predictable? Cases ofconvergent evolution can shed light on when, why, and how differentspecies exhibit shared evolutionary trajectories. In particular, study-ing diverse species in a common environment can illuminate how dif-ferent factors facilitate or constrain adaptive evolution. Here we inte-grate studies of pattern and process in the fauna at White Sands (NewMexico) to understand the determinants of convergent evolution.Numerous animal species at White Sands exhibit phenotypic conver-gence in response to a novel—and shared—selective environment:geologically young gypsum dunes. We synthesize 15 years of researchon White Sands lizards to assess the contribution of natural selection,genetic architecture, and population demography to patterns of phe-notypic evolution. We also present new data for two species of WhiteSands arthropods,Ammobaenetes arenicolus andHabronattus ustula-tus. Overall, we find dramatic phenotypic convergence across diversespecies at White Sands. Although the direction of phenotypic responseis parallel, the magnitude of phenotypic response varies among species.We also find that species exhibit strikingly different demographicpatterns across the ecotone. The species with the most genetic structurebetween White Sands and dark-soil populations generally exhibit theleast phenotypic divergence, suggesting population demography as akey modulator of adaptation. Comparative studies are particularly im-portant for understanding the determinants of convergence in naturalsystems.

Keywords: White Sands, convergent evolution, jumping spiders,sand-treader crickets, restriction site–associated DNA sequencing.

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† Corresponding author; e-mail: rosenblum@berkeley.edu.‡ These authors contributed equally.§ Present address: Department of Biological Sciences, University of Idaho,Moscow, Idaho 83844.∥ Present address: Department of Bionanoscience, Delft University of Tech-nology, Delft, Netherlands.

ORCIDs: Parent, http://orcid.org/0000-0002-4378-6715; Diepeveen, http://orcid.org/0000-0002-0768-8528; Noss, http://orcid.org/0000-0001-5114-6155.

Am. Nat. 2017. Vol. 190, pp. S44–S56. q 2017 by The University of Chicago.0003-0147/2017/190S1-57238$15.00. All rights reserved.DOI: 10.1086/692138

This content downloaded from 132All use subject to University of Chicago Press Term

Understanding the factors that predict evolutionary responseto natural selection is a central goal of evolutionary biology.Instances of convergent evolution—the independent evolu-tion of phenotypic similarity in different lineages—provideunparalleled opportunities to study themechanisms of adap-tive trait evolution. Ultimately, whether different species ex-hibit a similar evolutionary response depends critically onthree interacting determinants (reviewed in Rosenblum et al.2014): (1) natural selection (e.g., whether themode, strength,and dimensionality of selection is similar across species),(2) genetic architecture of adaptive traits (e.g., whether her-itability, mutational effect sizes, and patterns of epistasis areparallel across species), and (3) population demography (e.g.,whether population size, patterns of population structure,and dynamics of gene flow are similar across species).A key challenge for understanding evolutionary predict-

ability—even for cases of convergent evolution—is disen-tangling the factors that can promote or hinder adaptive traitevolution. When different lineages in different environmentsare compared, patterns of phenotypic evolution can seem id-iosyncratic, and it can be difficult to distinguish the contri-butions of selection, demography, and genomic architectureto observed patterns. In contrast, cases of repeated evolutionwithin a single community provide exciting opportunities todifferentiate among the determinants of adaptive evolution.When distantly related species exhibit similar phenotypic re-sponses in a shared environment, we can address more nu-anced hypotheses about evolutionary predictability.Here, we integrate studies of pattern and process in the

White Sands system to understand the factors that pro-mote and hinder convergent evolution. We first synthesizewhat we have learned about the patterns of convergent evo-lution—and the mechanisms influencing that pattern—from 15 years of work onWhite Sands lizards. We then pre-sent new data on two invertebrate species as a next step towarda community-scale synthesis. Finally, we highlight outstand-ing questions about convergent evolution at White Sands,

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergent Evolution at White Sands S45

with particular attention to the interplay among natural se-lection, population demography, and genomic architecture.Throughout, we use a pattern-based definition of convergentevolution: the independent evolution of phenotypic similar-ity in different lineages. Defining convergence as a pattern—observed at the phenotypic level—allows us to be more ex-plicit as we endeavor to link process to pattern, mechanismto outcome, and cause to consequence (reviewed in Rosen-blum et al. 2014 and Stayton 2015).

White Sands as a Stage for Convergent Evolution

White Sands is a striking and geologically recent forma-tion in south-central New Mexico. The expansive gyp-sum dune system formed after the Last Glacial Maximum,with the bulk of the sedimentation deposition in the past2,000–7,000 years (Langford 2003; Kocurek et al. 2007).At ∼650 km2, White Sands is the largest gypsum dune fieldin the world. The gypsum sands contrast dramatically withthe brown soils of the surrounding ChihuahuanDesert, cre-ating a divergent selective environment. Many diverse ani-mal species have colonized White Sands and have con-vergently evolved blanched coloration in the dune habitat(fig. 1).

The best-studied example of phenotypic convergence atWhite Sands is in the lizard fauna. The three lizard speciesthat inhabit the heart of the dunes (Sceloporus cowlesi [East-ern fence lizard], Holbrookia maculata [lesser earless liz-ard], and Aspidoscelis inornata [little striped whiptail]) allexhibit dramatically blanched dorsal coloration. In contrast,populations of all three species exhibit darker dorsal color-ation in the rest of their ranges, typically similar to localsubstrate colors (Degenhardt et al. 1996). The White Sandslizards exhibit similar general patterns of color evolution.In all three species, the change in coloration is in the samedirection and is explained primarily by the brightness as-pect of color (i.e., changes in brightness explain 180% ofthe interpopulation variation in color; Rosenblum 2006).Evolution of dorsal brightness is not only in the same di-rection across species but also of roughly similar magnitudeacross species (Rosenblum 2006; Rosenblum and Harmon2011).While explaining less of total variation in dorsal color,changes in the hue and chroma aspects of coloration alsotrend in the same direction for all three species (Rosenblum2006; Robertson and Rosenblum 2009).

In addition to dorsal color variation, sexual-signaling patchesare also significantly different in color between dark and lightlizards in all three species (Robertson and Rosenblum 2009).These differences are not consistent across species, which isunsurprising, given that the location, color, and use of sexual-signaling color patches vary across species (e.g., blue ventralcolor in male S. cowlesi, seasonal orange throat color in fe-maleH. maculata, and blue head color of male A. inornata;

This content downloaded from 132All use subject to University of Chicago Press Term

Degenhardt et al. 1996). Differences in sexual-signaling colorpatches between dark and light color morphs could contrib-ute to observed preference for local mates in White Sands S.cowlesi and H. maculata (Rosenblum 2008; Hardwick et al.2013), but the potential for incipient speciation in WhiteSands lizards is outside the scope of our focus here on con-vergent evolution.In addition to convergent evolution of coloration in

White Sands lizards, the three lizard species also exhibitshifts in a variety of other phenotypes. Although detailedstudies of heritability for these traits have not beenconducted, we observe striking differences between dark-soil and White Sands populations in a myriad of traits, in-cluding morphology (e.g., body size, limb length; Rosen-blum and Harmon 2011), performance (e.g., sprint speed,bite force; Des Roches et al. 2013, 2016), resource use(e.g., diet, microhabitat; Des Roches et al. 2011, 2015,2016), and behavior (e.g., predator wariness, mate choice;Rosenblum 2008; Robertson et al. 2011; Hardwick et al.2013). For example, in all three species, White Sandslizards tend to have relatively larger heads size than dark-soil lizards (Rosenblum and Harmon 2011; Des Rocheset al. 2016). Larger head size, in turn, correlates with strongerbite force, and bite force is associated with dietary differencesbetween light and dark populations (Des Roches et al. 2016).Specifically, for all three species, White Sands lizards con-sume a more varied diet than dark-soil lizards (i.e., withstronger bite and wider gape, larger and harder-bodied preycan be included in the diet; e.g., Herrel et al. 2001; Des Rocheset al. 2015, 2016).

Factors Affecting Convergence in White Sands Lizards

Despite strong evidence for convergent evolution of color-ation and evidence for similar patterns of change for manyother phenotypes in the three White Sands lizard species,patterns are never completely identical across species. Weoften find that the direction of phenotypic change is con-cordant but the magnitude of change is different. For ex-ample, the lesser earless lizard (Holbrookia maculata) oftenexhibits the most pronounced differences between WhiteSands and dark-soil habitats, while Sceloporus cowlesi oftenexhibits the least (e.g., Rosenblum and Harmon 2011 fordorsal coloration; Des Roches et al. 2015 for diet; Des Rocheset al. 2013 for sprint speed).What explains similarities and differences among species

in the magnitude of divergence across the White Sands eco-tone? Here we evaluate the contribution of three key deter-minants that can influence the direction and magnitude ofphenotypic response across species: (1) natural selection,(2) genetic architecture of adaptive traits, and (3) demo-graphic context.

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Figu

re1:

Con

vergentevolutionof

blanched

coloration

inWhite

Sand

sfaun

a.The

topandbo

ttom

rowsshow

representative

phenotypes

from

White

Sand

sanddark-soilhabitats,re-

spectively.F

ocal

speciesforthisstud

yshow

n(from

leftto

right):Sceloporuscowlesi,A

spidoscleisinorna

ta,H

olbrookiamaculata,

Habrona

ttus

ustulatus,andAmmobaenetesarenicolus.

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:44 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergent Evolution at White Sands S47

Natural Selection

The White Sands environment clearly provides a dramaticbackdrop for natural selection. Classic observational workon reptile color variation has long posited the adaptive valueof substrate matching, especially for small diurnal lizards(e.g., Cott 1940; Norris and Lowe 1964). Experimental stud-ies have also demonstrated the functional importance ofsubstrate matching to avoid predation in a variety of taxa(e.g., Dice 1947; Reed and Janzen 1999). Several experimen-tal studies on lizards (e.g., Luke 1989) and small mammals(e.g., Kaufman 1973) specifically used avian predators thatare known to prey on lizards at White Sands, such as theloggerhead shrike and greater roadrunner (E.B.R., personalobservation). Our early common-garden rearing experimentsalso demonstrated that color variation in the White Sandslizards was not explained by ontogenetic or physiologicalplasticity (Rosenblum 2005). Moreover, there is evidence thatall three species experienced ecological release associatedwith colonizing the gypsum dunes (Des Roches et al. 2011).Specifically, fewer predators and competitors are found inWhite Sands, relative to dark-soil habitats, and all three focallizard species correspondingly exhibit higher abundances inthe gypsum habitat (i.e., density compensation; Des Rocheset al. 2011). Thus, the context for selection at White Sandsappears simple: heritable phenotypes, clear optimality crite-ria, an obvious agent of selection, and decreased predationand interspecific competition.

However, the dynamics of selection might be different fordifferent species, even if they inhabit a common macroenvi-ronment. For example, differences in microhabitat use orbehavior canmodulate exposure to predation and ultimatelythe strength of natural selection. To date, we know littleabout differences among species in the dynamics of selectionatWhite Sands. However, differences among the three lizardspecies in foraging mode, microhabitat use, and other traitscould affect dynamics of natural selection (e.g., Dixon 1967).For example, Aspidoscelis inornata is an active forager, whileH. maculata and S. cowlesi are sit-and-wait predators (e.g.,Degenhardt et al. 1996). Further, H. maculata typically usesopen, unvegetatedmicrohabitat,whileS. cowlesi ismore com-monly found in vegetated microhabitat (e.g., Hager 2001).Thus, it is possible, for example, that selection has beenstronger for optimal substrate matching inH.maculata, whichwould be consistent with the stronger phenotypic responsein this species.

Explicit tests are required to understandwhether dynamicsof selection differ among species at White Sands. Our initialefforts to quantify selection at White Sands involved con-ducting large-scale enclosure experiments. We constructedreplicated 100-m2 enclosures in the natural gypsum habitat(Hardwick et al. 2015). We then painted dozens of WhiteSands H. maculata to match the average dorsal color of the

This content downloaded from 132All use subject to University of Chicago Press Term

White Sands population and the dark-soil population.We re-leased painted lizards into the enclosures and scored survivalof substrate-matched versus that of substrate-unmatchedlizards. These experiments helped confirm the activity ofavian predation at White Sands but also revealed differencesin lizard survival rates across relatively fine spatial and tem-poral scales and also different survival patterns formales andfemales. An alternative promising approach is the use ofmark-recapture studies to assess differences in strength of se-lection across species. We conducted a multiyear effort tomark lizards with elastomer tags and assess traits correlatedwith survival in S. cowlesi (Des Roches et al. 2017), and com-parison with a similar H. maculata data set (S. Des Rochesand E. B. Rosenblum, unpublished data) promises to shedfurther light on differences in dynamics of selection amongspecies. Ultimately, the observational and experimental ap-proaches will complement each other in the effort to under-standwhetherWhite Sands species experience different selec-tion pressures despite sharing a common environment.

Genetic Architecture

Difference among species in the genetic architecture of func-tionally relevant traits can influence species’ response to se-lection. Many aspects of genetic architecture can influencethe probability of convergent evolution at the phenotypiclevel and the probability that similar geneticmechanisms un-derlie phenotypic convergence (reviewed in Rosenblum et al.2014).For White Sands lizards, convergent evolution of blanched

coloration is the most obvious and tractable trait for geneticdissection. Our early candidate gene studies revealed muta-tions associated with blanched color in all three lizard speciesin themelanocortin-1 receptor (Mc1r) gene (Rosenblum et al.2004). The protein produced by Mc1r is a key player in thevertebrate melanin synthesis pathway, and mutations inMc1r are known to be associated with color variation in avariety of other systems (e.g., Barsh 1996; Manceau et al.2010). Our early work showed a strong statistical associationbetween blanched coloration and a single Mc1r mutation ineach of the three White Sands lizards. All three mutationslead to amino acid substitutions, and all three substitutionsoccur in a transmembrane region of the protein, which isimportant for ligand binding, signal transduction, and struc-tural integrity of the receptor. Our subsequent functionalassays confirmed that theMc1rmutations in two of the threespecies (S. cowlesi and A. inornata) have important func-tional effects leading to decreased melanin production(i.e., by reducing receptor integration in S. cowlesi and re-ducing receptor signaling in A. inornata; Rosenblum et al.2010). Although data are less conclusive for the third species(H. maculata), we have not ruled out a role forMc1r in colorvariation in this species.

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S48 The American Naturalist

The molecular signature of selection at and around theMc1r gene is also striking—and similar—for S. cowlesiand A. inornata. We recently obtained population-level se-quence capture data for a ∼40-kb window of the chromo-somal region around the Mc1r gene (the gene itself is asingle exon !1 kb in length). We found no evidence for se-lection around Mc1r in dark-soil populations of S. cowlesiand A. inornata, but we found strong evidence for selectionaround Mc1r in White Sands populations of these species(Laurent et al. 2016). Moreover, estimates of the age ofthe blanched allele at White Sands were remarkably youngand similar across species (i.e., 1,200 and 900 years for S.cowlesi and A. inornata, respectively; Laurent et al. 2016).Thus, there are noteworthy mechanistic similarities under-lying convergent color evolution in at least two of the lizardspecies. Mutations at the same gene appear to have sweptthrough the populations during a similar time period, con-sistent with the age of the White Sands formation itself.

In addition to similarities in genetic architecture of color-ation in White Sands lizards, we find several important dif-ferences across species. One intriguing difference is the dom-inance of the blanched Mc1r allele in S. cowlesi but not inA. inornata. Allele frequencies in natural populations, pat-terns of association between color phenotype and differentgenotypic classes, and functional studies all suggest that theblanched allele is dominant in S. cowlesi but recessive in A.inornata (Rosenblum et al. 2010). Although both dominantand recessive alleles contribute to adaptation in natural sys-tems, allelic dominance can affect the visibility of adaptivealleles to selection, the likelihood ofmaladaptive gene swamp-ing via gene flow, and ultimately the probability of fixation(e.g., Orr and Betancourt 2001; Nuismer et al. 2012). Theconsequences of differences in allelic dominance on the adap-tive trajectories of S. cowlesi and A. inornata require furtherwork, as does understanding other genes that contribute toconvergent phenotypic evolution in the novel gypsum habi-tat. Variation atMc1r cannot explain all observed color var-iation in theWhite Sands system, suggesting that other genesare also involved in color variation. Moreover, we are stillworking to understand the molecular basis of convergenttraits other than color. Ultimately, identifying the genes andgene interactions that underlie convergence at White Sandswill allow us to understand how the genetic architectureof adaptive traits influences the direction and magnitude ofevolutionary change across species.

Population Demography

Differences in underlying population demography can alsocontribute to different evolutionary outcomes across species.A number of demographic factors—such as population size,time since colonization, and rates of gene flow—have likelyinfluenced species response to natural selection at White

This content downloaded from 132All use subject to University of Chicago Press Term

Sands. Moreover, differences among species in populationdemography may help explain differences among species inmagnitude of evolutionary change across the ecotone.Even our earliest molecular data suggested dramatic dif-

ferences in underlying population demography among thethree White Sands lizard species. Early mitochondrial datafor White Sands and dark-soil populations demonstratedthat patterns of population structure were different acrossspecies (Rosenblum 2006). For example, an analysis of mo-lecular variance showed the three lizard species on a spec-trum from nearly complete genetic separation between darkand light populations ofH. maculata (FST p 0:82) to nearlycomplete panmixis in A. inornata (FST p 0:09), with inter-mediate structure observed for S. cowlesi (FSTp 0:54). In thepast decade, we sampled additional populations and addi-tional regions of the genome. Our first analyses with multi-locus nuclear data (∼200 single-nucleotide polymorphisms[SNPs] for S. cowlesi and ∼50 amplified fragment lengthpolymorphism bands forH. maculata and A. inornata) largelysupported the mitochondrial patterns (Rosenblum et al. 2007;RosenblumandHarmon2011), again suggesting that blanchedcoloration could evolve under very different demographicscenarios. Our most recent data sets that rely on broadergenomic sampling (i.e., 120,000 SNPs for S. cowlesi and113,000 SNPs for A. inornata) show that adding data im-proves our ability to discriminate genetically among popula-tions but that differences in underlying population demogra-phy remain among species (Laurent et al. 2016). Specifically,the inferred split time between light and dark populationswas younger for A. inornata than for S. cowlesi, and therewas stronger evidence for ongoing migration after coloniza-tion for S. cowlesi (Laurent et al. 2016).Our data point to the possibility that gene flow may con-

strain local adaptation at White Sands. For example, wehave the most consistent and conclusive evidence for ongo-ing gene flow across the White Sands ecotone in S. cowlesi(Rosenblum et al. 2007; Rosenblum and Harmon 2011;Laurent et al. 2016). Sceloporus cowlesi is also the speciesthat tends to exhibit the least phenotypic divergence acrossthe ecotone (e.g., for color, body size, performance, and diet;Rosenblum 2006; Rosenblum and Harmon 2011; Des Rocheset al. 2013, 2015, 2016). Ultimately, understanding the rela-tionship between gene flow and local adaptation will be fa-cilitated by continued sampling of dark-soil populations torefine our understanding of the ancestry, time of coloniza-tion, and dynamics of ongoing gene flow for the White Sandspopulations.

New Insights from New Data: White Sands Arthropods

Thus far we have focused on convergent evolution in theWhite Sands lizard fauna, but several additional—and moredistantly related—species also exhibit blanched phenotypes

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergent Evolution at White Sands S49

in the gypsum habitat. Here we present new data on two otherdramatic examples of phenotypic convergence associatedwiththe colonization of White Sands in the terrestrial inverte-brate fauna: the sand-treader cricket Ammobaenetes areni-colus (Strohecker 1947) and the jumping spider Habronattusustulatus (Griswold 1979). Both species are characterizedby pale forms at White Sands and darker forms in the sur-rounding Chihuahuan Desert, convergent with the patternobserved for the lizards (fig. 1). By increasing the phyloge-netic breadth of our work, we can add generality to under-standing the factors that influence the direction and mag-nitude of evolutionary response across a shared ecotone.

We compared sand-treader crickets and jumping spidersfrom White Sands to those from nearby dark-soil sites. Tounderstand phenotypic response across the ecotone in thesespecies, we quantified coloration of the dorsal body surface,using spectrophotometry and photographic analysis. To un-derstand the demographic history for these populations, wecollected genetic data, including thousands of loci obtainedwith a double-digest restriction site–associated DNA sequenc-ing (RADseq) approach and the mitochondrial cytochrome coxidase subunit I (COI) gene. Sampling and methodologicaldetails are presented in appendix A, and detailed results arepresented in appendix B (apps. A, B available online). Herewe highlight key findings that relate to and extend our pre-vious work on White Sands lizards.

Arthropod Phenotypic Convergence

Consistent with patterns we have previously described forWhite Sands lizards (Rosenblum 2006; Rosenblum andHar-mon 2011), we found strong evidence for phenotypic con-vergence in the two lineages of terrestrial invertebrates. Pop-ulations of the sand-treader cricket A. arenicolus and thejumping spider H. ustulatus at White Sands exhibit signifi-cantly lighter dorsal body coloration than populations in thesurrounding Chihuahuan Desert (fig. 2). Although both in-vertebrate species exhibit the same direction of phenotypicchange, the magnitude of color divergence between WhiteSands and dark-soil individuals differs. Specifically, the spi-ders exhibit less divergence in dorsal coloration than thecrickets in comparisons of White Sands to dark-soil popula-tions (fig. 2). The spiders also exhibit sexual dichromatism,while the crickets do not. Specifically, bothmales and femalesare lighter in coloration at White Sands for both species, butmale spiders are significantly darker than female spiders inboth habitats (fig. 2).

Similar to the lizards in this system, natural selection forsubstrate matching is themost likely explanation for blanchedcoloration in White Sands invertebrates. Both A. arenicolusand H. ustulatus spend time exposed on the dunes. Jump-ing spiders are diurnal and active predators of other terres-trial invertebrates (Foelix 1982; Griswold 1987), and they are

This content downloaded from 132All use subject to University of Chicago Press Term

most commonly found in low vegetation or on sandy sub-strate near vegetative cover atWhite Sands. Sand-treader crick-ets burrow in the daytime and emerge to forage at night(Weissman 1997), and they are commonly found near theirburrows at the base of the sand dunes, typically in areas withsparse vegetation cover (E.B.R. and C.E.P., personal obser-vation). Despite differences in activity period, both speciesare likely targeted by a range of visually oriented predators.Therefore, dorsal color variation could be linked to selectionfor crypsis in A. arenicolus and H. ustulatus, as has beenfound in numerous other animal species (e.g., Vignieri et al.2010). Alternatively, dorsal color variation can serve roles inthermoregulation, mate choice, aggression, and immunity(e.g., Horth 2003; Fedorka et al. 2013; Roulin 2016). How-ever, most of these alternatives have little intuitive sup-port (e.g., because ambient and substrate temperatures aretypically lower at White Sands than in the surrounding Chi-huahuan Desert [Hager 2000], a thermoregulatory hypothe-sis would predict the opposite patterns, where darker bodycolor would be favored in the gypsum habitat). Althoughnatural selection for crypsis likely explains color differencesacross populations, sexual selection may play a role withinpopulations, particularly for the spiders, which—like theWhite Sands lizards—exhibit some sexual dichromatism.If dynamics of selection differ between the two inverte-

brate species, there could be an adaptive explanation for dif-ferences in phenotypic patterns. However, additional obser-vational and experimental studies will be required to assesswhether differences between the species (or between the sexes)in habitat use, life history, and exposure to predation mayhave contributed to differences in phenotypic patterns acrossthe ecotone. Experimental studies will also be particularlyuseful for assessing the potential for phenotypic plasticity tocontribute to color variation in A. arenicolus and H. ustulatusand for identifying genes contributing to color variation inthese species. Although both species are relatively difficultto breed in the lab, anecdotally we found that baby spidersborn in the lab from White Sands mothers were blanchedin color.

Arthropod Demographic History

Another possible explanation for differences in phenotypicresponse across the White Sands ecotone is differences inunderlying demography (e.g., population size, colonizationhistory, contemporary patterns of gene flow). Convergencein dorsal coloration is accompanied by conspicuously differ-ent demographic histories in A. arenicolus and H. ustulatus.The sand-treader cricketA. arenicolus exhibits strongpop-

ulation structure, while the jumping spider H. ustulatus ex-hibits little genetic differentiation across the White Sandsecotone. The cricket mitochondrial gene tree shows a well-supported monophyletic White Sands clade (fig. 3), and the

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Figu

re2:

DorsalcolorvariationforAmmobaenetesarenicolus

(A)andHabrona

ttus

ustulatus(B):meando

rsal

color(andstandard

errorof

themean)

forWhite

Sand

s(opensymbols)

anddark-soil(closedsymbo

ls)samples.C

ircles,squ

ares,and

triangles

representfemale,male,andsubstratesamples,respectively.

Notethat

Y-axesdiffer

becausecolorwas

quantified

differently

forthetwospecies(see

“Color

Quantification

”in

app.

A,availableon

line).PC1p

principalcompo

nent1.

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:44 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergent Evolution at White Sands S51

RADseq data show strong population structure, with WhiteSands and dark-soil individuals clearly in separate geneticclusters (fig. 4). In contrast, the spider mitochondrial genetree shows no clear clustering by habitat (fig. 3). Althoughthere is support for multiple clades in the spider tree, theseclades contain both White Sands and dark-soil individuals(with no substructure based on dark-soil collecting locality).Consistent with the mitochondrial data, the spider RADseqdata show largely overlapping clusters of White Sands anddark-soil individuals, with little structuring of genetic varia-tion by habitat (fig. 4).

Metrics of population differentiation also show strong dif-ferences in genetic patterns between the spider and cricketdata sets. For the mitochondrial data set, FST is more thanan order of magnitude higher for the crickets (FST p 0:45)than for the spiders (FST p 0:03; table 1). Moreover, themitochondrial data suggest dramatically reduced diversityin the White Sands cricket population but not in the spiderpopulation. For the crickets, both nucleotide diversity (p)

This content downloaded from 132All use subject to University of Chicago Press Term

and Watterson’s v are more than an order of magnitudesmaller for the White Sands population than for the dark-soil population (table 1), whereas for the spiders, p and v

are high and comparable for the White Sands and dark-soilpopulations (table 1). Mitochondrial nucleotide diversity ismore than two orders of magnitude higher in the WhiteSands spiders than in the White Sands crickets. Mitochon-drial and RADseq data give concordant results in the spidersbut show slightly different patterns in the crickets. Specifi-cally, for the crickets, p and v are substantially higher andFST is substantially lower for the RADseq data set than forthe mitochondrial data set, but both data sets indicate muchstronger population structure for the crickets than for thespiders.Our molecular results are consistent with different de-

mographic scenarios in the two focal invertebrate species.For the crickets, our results are consistent with a simplecolonization history (likely with a reduced population size),limited ongoing gene flow, and strong genetic divergence

C. gracilipes H. oregonensis

Figure 3: Contrasting demographic patterns based on mitochondrial cytochrome c oxidase subunit I gene sequences for Ammobaenetesarenicolus (A) and Habronattus ustulatus (B): Bayesian inference, with numbers at nodes representing posterior probabilities. The branchesfor the outgroup species are shortened by half, indicated by the “//” symbol. Open and filled symbols represent samples collected in WhiteSands and dark-soil habitats, respectively.

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Figu

re4:

Con

trasting

patterns

ofpo

pulation

structurebasedon

restrictionsite–associated

DNAsequ

encing

(RADseq)

data

forAmmobaenetesarenicolus(A

)andHabrona

ttus

ustulatus

(B).A,F

orA.a

renicolus,thepo

pulation

structureplot

(top)forKp

2show

sstrong

popu

lation

structure.Eachindividu

alisrepresentedby

averticalbarandlabeledas

dark

soil(black

horizontal

bar)

orWhite

Sand

s(w

hite

horizontal

bar).The

geneticprincipal-compo

nent

analysis(PCA)plot

(bottom

left)show

sno

overlapbetweendark-soil(filledcircles)

andWhite

Sand

s(opencircles)

individu

als.Ellipses

represent95%

confi

denceintervalsforthedark-soilandWhite

Sand

spo

pulation

s.The

jointsite

frequencyspectrum

(bottom

right)show

son

lylooselycorrelated

allele

frequencies.B,Fo

rH.ustulatus,thepo

pulation

structureplot

(top)show

sless

structureandmoreadmixture

betweenthetwopo

pulation

sforKp

2.There

issubstantialoverlap

betweenthedark-soiland

WhiteSand

sindividu

alsin

thegeneticPCAplot

(bottom

left),andthejointsitefrequencyspectrum

show

sstronglycorrelated

allelefrequencies

(bottom

right).

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:44 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergent Evolution at White Sands S53

across habitats. For the spiders, our results suggest admix-ture across the White Sands ecotone, and different scenarioscould underlie the lack of population structure, includingmultiple colonization events, a large recent founding popula-tion with substantial genetic variation, and high levels of on-going gene flow. Our results are consistent with patternsobserved in other groups of jumping spiders. Hybridization,introgression, interspecific gene flow, and incomplete line-age sorting are not uncommon in the Habronattus clade,and gene trees often fail to resolve geographic or phenotypicgroups (e.g., Maddison and McMahon 2000; Masta 2000;Hedin and Lowder 2009). To refine our understanding ofthe relationship between initial colonization history and on-going gene flow in these species, we will need more thoroughsampling throughout the Tularosa Basin (i.e., within speciesreplication) andmore explicit modeling of alternative demo-graphic scenarios.

Convergence across the Community at White Sands

Cases of convergent evolution have long been used as evi-dence that species can exhibit similar evolutionary responseswhen exposed to comparable selection pressures (e.g., Arendtand Reznick 2008; Losos 2011; McGhee 2011; Wake et al.2011; Conte et al. 2012). However, myriad factors influencewhether species will adapt along parallel trajectories. Study-ing evolutionary outcomes across diverse lineages providesreplicated variation in factors that can influence adaptive con-vergence (e.g., trait heritability, population size, gene flow,behavior). Further, comparisons across species in a single en-vironment (e.g., where colonization time is geologically con-strained and abiotic context is shared) can reduce complexityand analytical noise. Thus, we can better understandwhetheradaptation is constrained in similar ways across diverse spe-cies and which key factors modulate evolutionary response.

Our work thus far onWhite Sands lizards and arthopodsshows both shared and unique patterns across species ex-

This content downloaded from 132All use subject to University of Chicago Press Term

posed to a common ecotone. While the direction of colorevolution is typically convergent in the White Sands fauna,the magnitude of phenotypic change varies across species.Moreover, differences in phenotypic response appear to cor-relate with variation in underlying demographic patterns.Specifically, species with more genetic differentiation acrosstheWhite Sands ecotone (e.g.,Holbrookiamaculata andAm-mobaenetes arenicolus) typically exhibitmore phenotypic dif-ferentiation. Thus, population demography appears to be animportant modulator of phenotypic evolution in this sys-tem. Our research also suggests that subtle variation in thedynamics of natural selection and differences in genomicarchitecture across species can influence patterns of pheno-typic convergence.Moving forward, the White Sands system can be used to

address important outstanding questions about the interplayamong natural selection, genetic architecture, and popula-tion demography. For example, what is the role of phenotypicplasticity in promoting or hindering phenotypic divergence?How do different species experience the same environment?How similar are underlying molecular and functional mech-anisms of adaptation in closely versus distantly related spe-cies? How do levels of gene flow and strength of selectioninteract to determine phenotypic outcome?What null expec-tations are most appropriate to use when studying conver-gent evolution?Developing a general understanding of the factors that fa-

cilitate and constrain convergent evolution at White Sandswill be promoted by studying additional species. Other ani-mal species exhibit blanched forms on the dunes and canbe integrated into a broader community-scale study. For ex-ample, the moth Euxoa misturata and the pocket mousePerognathus flavescens also have blanched forms in the gyp-sum habitat that contrast with nearby dark forms. In addi-tion to these dramatic examples, White Sands is home tomultiple endemic species of Lepidoptera that are pale com-pared to close relatives (Metzler 2014).

Table 1: Estimates of the nucleotide diversity (p), Watterson’s theta (v), and FST for the mitochondrial andnuclear data sets

Ammobaenetes arenicolus

.236.027.111 on July s and Conditions (http

Habronattus ustulatus

Data set, measure

Dark soil White Sands Dark soil

22, 2017 05:14:44 AM://www.journals.uchica

White Sands

COI:

p .0030 .0002 .0294 .0261 v .0048 .0003 .0177 .0154 FST .4541 .0295

RAD sites:

p .0213 .0198 .0210 .0220 v .0316 .0215 .0336 .0347 FST .0312 .0140

Note: COI p cytochrome c oxidase subunit I gene; RAD p restriction site–associated DNA sequencing.

go.edu/t-and-c).

S54 The American Naturalist

Although White Sands is a relatively depauperate eco-system, the species that have survived on the dunes typi-cally exhibit thriving populations. For example in the her-petofauna, 28 species of reptiles and amphibians are foundin the immediate vicinity (Prival and Goode 2005), but onlythree of these species are found commonly in the heart of thedunes. These three species exhibit dramatic density compen-sation, with much higher local population densities at WhiteSands than in nearby dark-soil environments (Des Rocheset al. 2011). Relatively high densities allow for a diversity ofresearch approaches to be applied at White Sands. Integrat-ing natural history studies, manipulative lab and field exper-iments, detailed phenotypic analysis, genomic sequencing,and demographic modeling across the entire White Sandscommunity will ultimately shed light on the factors govern-ing the probability—and the degree—of convergent evolu-tion in natural systems.

Acknowledgments

We thank White Sands National Monument, White SandsMissile Range, and Jornada Long-Term Ecological ResearchStation for access to field sites and research facilitation overthe past 15 years. In particular we are grateful to J. Anderson,D. Burkett, D. Bustos, B. Conrod, and P. Cutler. We thankM. Girard for help in the field and K. Mathews for help withcare of spiders in the lab. We also thank D. Elias, M. Girard,D. Lightfoot, E. Metzler, and the Rosenblum lab for pro-ductive conversations about White Sands invertebrates. Wethank H. Hoekstra for sharing RADseq adapters, and wethank A. Agrawal, L. Mahler, K. Zamudio, and members ofthe Rosenblum lab for helpful comments on the manuscript.This work was supported by a National Science FoundationCAREER grant to E.B.R. (DEB-1054062) and a research grantfrom the Basler Stiftung für Biologische Forschung to E.T.D.Sequencing was performed by the Vincent J. Coates Geno-mics Sequencing Laboratory at the University of California,Berkeley, which is supported by National Institutes of HealthS10 Instrumentation Grants S10RR029668 and S10RR027303.Animal work was approved by the Animal Care and UseCommittee at the University of California, Berkeley (protocol2014-11-6857), or the University of Idaho (protocol 2009-37).

Literature Cited

Arendt, J., and D. Reznick. 2008. Convergence and parallelism recon-sidered: what have we learned about the genetics of adaptation? Trendsin Ecology and Evolution 23:26–32.

Barsh, G. S. 1996. The genetics of pigmentation: from fancy genes tocomplex traits. Trends in Genetics 12:299–305.

Conte, G. L., M. E. Arnegard, C. L. Peichel, and D. Schluter. 2012.The probability of genetic parallelism and convergence in naturalpopulations. Proceedings of the Royal Society B 279:5039–5047.

Cott, H. B. 1940. Adaptive coloration in animals. Methuen, London.

This content downloaded from 132All use subject to University of Chicago Press Term

Degenhardt, W. G., C. W. Painter, and A. H. Price. 1996. Amphibiansand reptiles of New Mexico. University of New Mexico Press, Al-buquerque.

Des Roches, S., M. S. Brinkmeyer, L. J. Harmon, and E. B. Rosenblum.2015. Ecological release and directional change inWhite Sands lizardtrophic ecomorphology. Evolutionary Ecology 29:1–16.

Des Roches, S., L. J. Harmon, and E. B. Rosenblum. 2016. Coloniza-tion of a novel depauperate habitat leads to trophic niche shifts inthree desert lizard species. Oikos 125:343–353.

Des Roches, S., J. M. Robertson, L. J. Harmon, and E. B. Rosenblum.2011. Ecological release in White Sands lizards. Ecology and Evo-lution 1:571–578.

Des Roches, S., R. Sollman, K. Calhoun, A. Rothstein, and E. B. Ro-senblum. 2017. Survival by genotype: patterns atMc1r are not blackand white at the White Sands ecotone. Molecular Ecology 26:320–329.

Des Roches, S., J. Torresdal, T. W. Morgan, L. J. Harmon, and E. B.Rosenblum. 2013. Beyond black and white: divergent behavior andperformance in three rapidly evolving lizard species at White Sands.Biological Journal of the Linnean Society 111:169–182.

Dice, L. R. 1947. Effectiveness of selection by owls of deer mice(Peromyscus maniculatus) which contrast in color with their back-ground. Contributions from the Laboratory of Vertebrate Biologyof the University of Michigan 34:1–20.

Dixon, J. R. 1967. Aspects of the biology of the lizards of the WhiteSands, New Mexico. Los Angeles County Museum Contributionsin Science 129:1–22.

Fedorka, K. M., V. Lee, and W. E. Winterhalter. 2013. Thermal en-vironment shapes cuticle melanism and melanin-based immunityin the ground cricket Allonemobius socius. Evolutionary Ecology27:521–531.

Foelix, R. 1982. Biology of spiders. Harvard University Press, Cam-bridge, MA.

Griswold, C. E. 1979. New species of Pellenes from California(Araneae: Salticidae). Journal of Arachnology 7:129–138.

———. 1987. A revision of the jumping spider genus HabronattusF. O. P.-Cambridge (Araneae: Salticidae), with phenetic and cla-distic analyses. University of California Publications in Entomol-ogy 107. University of California Press, Berkeley.

Hager, S. B. 2000. Variation in body temperature and thermoregula-tory behavior between two populations of the lesser earless lizard,Holbrookia maculata. Contemporary Herpetology 2000:1. http://jnah.cnah.org/ch/ch/2000/1/index.htm.

———. 2001. Microhabitat use and activity patterns of Holbrookiamaculata and Sceloporus undulatus at White Sands National Mon-ument, New Mexico. Journal of Herpetology 35:326–330.

Hardwick, K. M., L. Harmon, S. D. Hardwick, and E. B. Rosenblum.2015. When field experiments yield unexpected results: lessonslearned from measuring selection in White Sands lizards. PLoSONE 10(2):e0118560. doi:10.1371/journal.pone.0118560.

Hardwick, K. M., J. M. Robertson, and E. B. Rosenblum. 2013. Asym-metrical mate preference in recently adapted White Sands and blacklava populations of Sceloporus undulatus. Current Zoology 59:20–30.

Hedin, M., and M. C. Lowder. 2009. Phylogeography of the Habronat-tus amicus species complex (Araneae: Salticidae) of western NorthAmerica, with evidence for localized asymmetrical mitochondrialintrogression. Zootaxa 2307:39–60.

Herrel, A., R. Van Damme, B. Vanhooydonck, and F. De Vree. 2001.The implications of bite performance for diet in two species oflacertid lizards. Canadian Journal of Zoology 79:662–670.

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Convergent Evolution at White Sands S55

Horth, L. 2003. Melanic body colour and aggressive mating behaviourare correlated traits in male mosquitofish (Gambusia hotbrooki).Proceedings of the Royal Society B 270(1519):1033–1040.

Kaufman, D. W. 1973. Shrike prey selection: color or conspicuous-ness? Auk 90:204–206.

Kocurek, G., M. Carr, R. Ewing, K. G. Havholm, Y. C. Nagar, andA. K. Singhvi. 2007. White Sands Dune Field, New Mexico: age,dune dynamics and recent accumulations. Sedimentary Geology197:313–331.

Langford, R. P. 2003. The Holocene history of the White Sands dunefield and influences on eolian deflation and playa lakes. QuaternaryInternational 104:31–39.

Laurent, S., S. P. Pfeifer, M. L. Settles, S. S. Hunter, K. M. Hardwick,L. Ormond, V. Sousa, J. J. Jensen, and E. B. Rosenblum. 2016. Thepopulation genomics of rapid adaptation: disentangling signaturesof selection and demography in White Sands lizards. MolecularEcology 25:306–323.

Losos, J. B. 2011. Convergence, adaptation, and constraint. Evolution65:1827–1840.

Luke, C. A. 1989. Color as a phenotypically plastic character in theside-blotched lizard, Uta stansburiana. PhD diss. University ofCalifornia, Berkeley.

Maddison, W., and M. McMahon. 2000. Divergence and reticulationamong montane populations of a jumping spider (Habronattuspugillis Griswold). Systematic Biology 49:400–421.

Manceau, M. C., V. Domingues, C. R. Linnen, E. B. Rosenblum, andH. E. Hoekstra. 2010. The genetic basis of phenotypic conver-gence: lessons from pigmentation. Philosophical Transactions ofthe Royal Society B 365:2439–2450.

Masta, S. E. 2000. Phylogeography of the jumping spiderHabronattuspugillis (Araneae: Salticidae): recent vicariance of sky island popu-lations? Evolution 54:1699–1711.

McGhee, G. R. 2011. Convergent evolution: limited forms most beauti-ful. Vienna Series in Theoretical Biology. MIT Press, Cambridge, MA.

Metzler, E. H. 2014. The remarkable endemism of moths at WhiteSands National Monument in New Mexico, USA, with special em-phasis on Gelechioidea (Lepidoptera). Journal of Asia-Pacific Bio-diversity 7:e1–e5.

Norris, K., and C. H. Lowe. 1964. An analysis of background color-matching in amphibians and reptiles. Ecology 45:565–580.

Nuismer, S. L., A. MacPherson, and E. B. Rosenblum. 2012. Crossingthe threshold: gene flow, dominance, and the critical level ofstanding genetic variation required for adaptation to novel envi-ronments. Journal of Evolutionary Biology 25:2665–2671.

Orr, H. A., and A. J. Betancourt. 2001. Haldane’s sieve and adapta-tion from the standing genetic variation. Genetics 157:875–884.

Prival, D., andM. Goode. 2005. Chihuahuan Desert national parks rep-tile and amphibian inventory. Final report. University of Arizona,Tucson.

Reed, W. L., and F. J. Janzen. 1999. Natural selection by avian pred-ators on size and colour of a freshwater snail (Pomacea flagellata).Biological Journal of the Linnean Society 67:331–342.

Robertson, J., K. Hoversten, M. Grundler, T. Poorten, D. Hews, andE. B. Rosenblum. 2011. Colonization of novel White Sands habitatassociated with changes in lizard anti-predator behavior. Biolog-ical Journal of the Linnean Society 103:657–667.

Robertson, J., and E. B. Rosenblum. 2009. Rapid divergence of socialsignal coloration across the White Sands ecotone for three lizardspecies under strong natural selection. Biological Journal of theLinnean Society 98:243–255.

This content downloaded from 132All use subject to University of Chicago Press Term

Rosenblum, E. B. 2005. The role of phenotypic plasticity in colorvariation in Tularosa Basin lizards. Copeia 2005:586–596.

———. 2006. Convergent evolution and divergent selection: lizardsat the White Sands ecotone. American Naturalist 167:1–15.

———. 2008. Preference for local mates in a recently diverged pop-ulation of the lesser earless lizard (Holbrookia maculata) at WhiteSands. Journal of Herpetology 42:572–583.

Rosenblum, E. B., and L. J. Harmon. 2011. “Same same but differ-ent”: replicated ecological speciation at White Sands. Evolution65:946–960.

Rosenblum, E. B., M. Hickerson, and C. Moritz. 2007. A multilocusperspective on colonization accompanied by selection and geneflow. Evolution 61:2971–2985.

Rosenblum, E. B., H. E. Hoekstra, and M. W. Nachman. 2004. Adap-tive reptile color variation and the evolution of the Mc1r gene.Evolution 58:1794–1808.

Rosenblum, E. B., C. E. Parent, and E. Brandt. 2014. The molecularbasis of phenotypic convergence. Annual Reviews of Ecology,Evolution, and Systematics 45:203–226.

Rosenblum, E. B., H. Römpler, T. Schöneberg, and H. E. Hoekstra.2010. Molecular and functional basis of phenotypic convergencein white lizards at White Sands. Proceedings of the National Acad-emy of Sciences of the USA 107:2113–2117.

Roulin, A. 2016. Condition-dependence, pleiotropy and the handi-cap principle of sexual selection in melanin-based colouration. Bi-ological Reviews 91:328–348.

Stayton, C. T. 2015. The definition, recognition, and interpretation ofconvergent evolution, and two new measures for quantifying andassessing the significance of convergence. Evolution 69:2140–2153.

Strohecker, H. F. 1947. Some southwestern Gryllacrididae (Orthop-tera). Annals of the Entomological Society of America 40:241–246.

Vignieri, S. N., J. G. Larson, and H. E. Hoekstra. 2010. The selectiveadvantage of crypsis in mice. Evolution 64:2153–2158.

Wake, D. B., M. H. Wake, and C. D. Specht. 2011. Homoplasy: fromdetecting pattern to determining process and mechanism of evo-lution. Science 331:1032–1035.

Weissman, M. J. 1997. Natural history of the giant sand treader camelcricket, Daihinibaenetes giganteus Tinkham (Orthoptera: Rhaphi-dophoridae). Journal of Orthoptera Research 6:33–48.

References Cited Only in the Online Appendixes

Bi, K., T. Linderoth, D. Vanderpool, J. M. Good, R. Nielsen, andC. Moritz. 2013. Unlocking the vault: next generation museum pop-ulation genomics. Molecular Ecology 22:6018–6032.

Darriba, D., G. L. Taboada, R. Doallo, and D. Posada. 2012. jModelTest2:moremodels, newheuristics andparallel computing.NatureMeth-ods 9:772.

Evanno, G., S. Regnaut, and J. Goudet. 2005. Detecting the numberof clusters of individuals using the software structure: a simulationstudy. Molecular Ecology 14:2611–2620.

Folmer, O., M. Black, W. Hoeh, R. Lutz, and R. Vrijenhoek. 1994.DNA primers for amplification of mitochondrial cytochrome coxidase subunit I from diverse metazoan invertebrates. MolecularMarine Biology and Biotechnology 3:294–299.

Fumagalli, M., F. G. Vieira, T. S. Korneliussen, T. Linderoth, E. Huerta-Sánchez, A. Albrechtsen, and R. Nielsen. 2013. Quantifying popula-tion genetic differentiation from next-generation sequencing data.Genetics 195:979–992.

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S56 The American Naturalist

Gregory, T. R., and D. P. Shorthouse. 2003. Genome sizes of spiders.Journal of Heredity 94:285–290.

Guindon, S., and O. Gascuel. 2003. A simple, fast, and accurate algo-rithm to estimate large phylogenies by maximum likelihood. Sys-tematic Biology 52:696–704.

Hedin, M. C. 1997. Speciational history in a diverse clade of habitat-specialized spiders (Araneae: Nesticidae: Nesticus): inferences fromgeographic-based sampling. Evolution 51:1929–1945.

Huelsenbeck, J. P., and F. Ronquist. 2001. MrBayes: Bayesian infer-ence of phylogenetic trees. Bioinformatics 17:754–755.

Korneliussen, T. S., A. Albrechtsen, and R. Nielsen. 2014. ANGSD:analysis of next generation sequencing data. BMC Bioinformatics15:356. doi:10.1186/s12859-014-0356-4.

Li, H., B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer,G. Marth, G. Abecasis, R. Durbin, and 1000 Genome Project DataProcessing Subgroup. 2009. The sequence alignment/map (SAM)format and SAMtools. Bioinformatics 25:2078–2079.

Librado, P., and J. Rozas. 2009. DnaSP v5: a software for comprehen-sive analysis of DNA polymorphism data. Bioinformatics 25:1451–1452.

McKenna, A., M. Hanna, E. Banks, A. Sivachenko, K. Cibulskis,A. Kernytsky, K. Garimella, et al. 2010. The Genome Analysis Tool-kit: a MapReduce framework for analyzing next-generation DNAsequencing data. Genome Research 20:1297–1303.

Nielsen, R., T. Korneliussen, A. Albrechtsen, Y. Li, and J. Wang. 2012.SNP calling, genotype calling, and sample allele frequency estima-tion from new-generation sequencing data. PLoSONE 7(7):e37558.doi:10.1371/journal.pone.0037558.

Nielsen, R., J. S. Paul, A. Albrechtsen, and Y. S. Song. 2011. Geno-type and SNP calling from next-generation sequencing data. Na-ture Reviews Genetics 12:443–451.

O’Grady, P. M., J. B. Clark, and M. G. Kidwell. 1998. Phylogeny ofthe Drosophila saltans species group based on combined analysisof nuclear and mitochondrial DNA sequences. Molecular Biologyand Evolution 15:656–664.

Peterson, B. K., J. N.Weber, E. H. Kay, H. S. Fisher, and H. E. Hoekstra.2012. Double digest RADseq: an inexpensive method for de novo

“The farmer should know his true insect friends as well as his insec(Carabidæ, from Carabus, the name of the typical genus) which prey oendar” (The American Naturalist, 1868, 2:110–111).

This content downloaded from 132All use subject to University of Chicago Press Term

SNP discovery and genotyping in model and non-model species.PLoS ONE 7(5):e37135. doi:10.1371/journal.pone.0037135.

Posada, D. 2008. jModelTest: phylogenetic model averaging. Molec-ular Biology and Evolution 25:1253–1256.

Posnien, N., V. Zeng, E. E. Schwager, M. Pechmann, M. Hilbrant,J. D. Keefe, W. G. M. Damen, N.-M. Prpic, A. P. McGregor, andC. G. Extavour. 2014. A comprehensive reference transcriptomeresource for the common house spider Parasteatoda tepidariorum.PLoS ONE 9(8):e104885. doi:10.1371/journal.pone.0104885.

R Development Core Team. 2014. R: a language and environmentfor statistical computing. R Foundation for Statistical Computing,Vienna, Austria. http://www.R-project.org/.

Ronquist, F., and J. P. Huelsenbeck. 2003. MrBayes 3: Bayesian phy-logenetic inference under mixed models. Bioinformatics 19:1572–1574.

Sanggaard, K. W., J. S. Bechsgaard, X. Fang, J. Duan, T. F. Dyrlund,V. Gupta, X. Jiang, et al. 2014. Spider genomes provide insight intocomposition and evolution of venom and silk. Nature Communica-tions 5:3765. doi:10.1038/ncomms4765.

Schneider, C. A., W. S. Rasband, and K. W. Eliceiri. 2012. NIH Imageto ImageJ: 25 years of image analysis. Nature Methods 9:671–675.

Simon, C., F. Frati, A. Beckenbach, B. Crespi, H. Liu, and P. Flook.1994. Evolution, weighting, and phylogenetic utility of mitochon-drial gene sequences and a compilation of conserved polymerasechain reaction primers. Annals of the Entomological Society ofAmerica 87:651–701.

Skotte, L., T. S. Korneliussen, and A. Albrechtsen. 2013. Estimatingindividual admixture proportions from next generation sequenc-ing data. Genetics 195:693–702.

Zhang, J.-X., and W. P. Maddison. 2013. Molecular phylogeny, diver-gence times and biogeography of spiders of the subfamily Euoph-ryinae (Araneae: Salticidae). Molecular Phylogenetics and Evolution68:81–92.

Symposium Editor: Anurag A. Agrawal

t foes. We introduce to our readers a large family of ground-beetlesn those insects largely injurious to crops.” From “Entomological Cal-

.236.027.111 on July 22, 2017 05:14:44 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Convergence and Divergence in a Long-Term

Experiment with Bacteria*

Richard E. Lenski†

Department of Microbiology and Molecular Genetics, Michigan State University, East Lansing, Michigan 48824

abstract: Suitably designed experiments offer the possibility ofquantifying evolutionary convergence because the fraction of repli-cate populations that converge is known. Here I review an experimentwith Escherichia coli, in which 12 populations were founded from thesame ancestral strain and have evolved for almost 30 years and morethan 65,000 generations under the same conditions. The tension be-tween divergence and convergence has been a major focus of this ex-periment. I summarize analyses of competitive fitness, correlated re-sponses to different environments, cell morphology, the capacity touse a previously untapped resource, mutation rates, genomic changes,and within-population polymorphisms. These analyses reveal conver-gence, divergence, and often a complicated mix thereof. Complica-tions include concordance in the direction of evolutionary changewith sustained quantitative variation among populations, and the po-tential for a given trait to exhibit divergence on one timescale and con-vergence on another. Despite these complications, which also occurin nature, experiments provide a powerful way to study evolutionaryconvergence based on analyzing replicate lineages that experience thesame environment.

Keywords: adaptation, Escherichia coli, experimental evolution, fit-ness, mutation rate, parallel evolution.

Introduction

Against the backdrop of the diversity of life, nature is re-plete with examples of parallel and convergent phenotypicevolution. Beyond compiling examples of convergence, how-ever, it is difficult to devise a framework for quantifying theextent of convergence or predicting when it is likely, or un-likely, to arise. One complication is the possibility of report-ing bias, or whatmight be called the “denominator problem.”For example, camera-like eyes have evolved independentlyin many lineages, including not only vertebrates and cepha-lopod mollusks but also some gastropod mollusks, marineannelids, spiders, and cubozoan jellyfish—the last of which

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† E-mail: lenski@msu.edu.

ORCIDs: Lenski, http://orcid.org/0000-0002-1064-8375.

Am. Nat. 2017. Vol. 190, pp. S57–S68. q 2017 by The University of Chicago.0003-0147/2017/190S1-57225$15.00. All rights reserved.DOI: 10.1086/691209

This content downloaded from 132All use subject to University of Chicago Press Term

strikingly lack a recognizable brain (Conway Morris 2003,pp. 151–158). But in howmany lineages did camera-like eyesfail to evolve? Even if we had that number, how should we ac-count for differences in, say, the sizes and ages of lineageswhen trying to formulate and test hypotheses about ecologicalor genetic factors that might have predisposed a lineage toevolve camera-like eyes? And in those cases where such fea-tures did not evolve, despite seemingly comparable opportu-nities, did some other difference in circumstances promotedivergence instead of convergence?Although lacking the grandeur of nature, laboratory ex-

periments offer the possibility of quantifying convergencebecause the denominator—the number of replicate popula-tions that do or do not evolve the same way—is known.Moreover, the sizes and ages of lineages are known in mostexperiments. Therefore, describing the repeatability of evo-lution seems simple in principle, although in practice com-plications may arise. For example, divergent outcomes overthe short run could become convergent over the long run,or vice versa, depending on the rates of phenotypic transi-tion and their sensitivity to prior events that promote or im-pede the transitions. Also, one might well see a mixture ofconvergent and divergent responses if one examines manydifferent traits.In recent years, evolution experiments have been used to

address a wide range of questions (Elena and Lenski 2003;Garland and Rose 2009; Kawecki et al. 2012; Barrick andLenski 2013), and the issue of evolutionary repeatabilityhas been a focus of many of these studies. I will not attemptto review the entire field of experimental evolution here, oreven those studies focused on repeatability, but I will beginby brieflymentioning three studies that call attention to im-portant issues. First, Burke et al. (2010) found evidence ofrepeatable genomic evolution across replicate Drosophilamelanogaster populations that had been selected for accel-erated development. The replicates all derived from a singleancestral population that harbored substantial variation.Thus, the observed repeatability of the genomic changesprobably stemmed from standing genetic variation—thatis, collateral evolution in the sense of Stern (2013)—andnot parallel evolution based on independent mutational

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S58 The American Naturalist

events. Second, in a study of the bacteriophage fX174,Wichman et al. (1999) found that almost half of the inde-pendent mutations that reached high frequency in two rep-licate populations were identical, indicating extensive par-allelism at the level of the DNA sequence itself. In otherstudies, however, repeated phenotypic evolution usuallyresults from different mutations in the same gene or dif-ferent genes that affect the same trait. Third, using a high-resolution random–bar code approach, Levy et al. (2015)monitored thousands of lineages defined by new beneficialmutations within a single population of Saccharomyces cere-visiae. Some of these mutations undoubtedly affected thesame gene, and perhaps even the same nucleotide, eventhough they were independent events, as indicated by differ-ent bar codes. Thus, parallel evolution can occur not onlyacross replicate populations but also across multiple lineageswithin a single population.

Instead of reviewing the entire field, I will focus on oneexperiment that I have directed for nearly three decadesand where the repeatability of evolution—the tension be-tween divergence and convergence—has been at the fore-front since it began. In fact, one reason I called this an “evo-lution experiment,” as opposed to the then-more-familiar“selection experiment,” was to emphasize the fact that anychanges that occurred during this experiment encompassedthe origin, as well as the fate, of genetic variation—in essence,highlighting the distinction between parallel and collateralevolution that Stern (2013) also emphasized. In addition topresenting what I think are interesting examples of evolutionin action, I hope tomake clear three facets of convergence anddivergence that emerge from this experiment. First, we haveobserved striking examples of phenotypic and genomic con-vergence across the replicate populations. Second, we havealso seen striking cases of divergence, despite the fact thatthe populations started from the same genotype and havebeen evolving in identical physical environments. Third, thismixture of convergence and divergence, when played outover time and across many traits, defies easy summationand instead requires a more nuanced description. As a con-sequence, assessing the repeatability of evolution is bothmore interesting and more difficult than I imagined whenI began the experiment. And if things become this compli-cated in a set of small, closed, identical flasks, how muchmore challenging it must be to study the repeatability ofevolution in the natural world.

In the next section, I briefly describe the basic structureand methods of the long-term evolution experiment (LTEE)with Escherichia coli. I then summarize some of the pheno-typic and genetic analyses that reveal convergence, diver-gence, and the complicated mix thereof. I conclude bydiscussing some of the challenges that arise in interpretingthe data that result from even such a simple experiment asthe LTEE, and by suggesting a few questions that might be

This content downloaded from 132All use subject to University of Chicago Press Term

addressed by evolution experiments to better understandthe fascinating tension between chance and necessity inevolving systems.Before moving to the experiment, I should briefly discuss

a semantic issue that I hope will not cause confusion. In pre-vious presentations on the LTEE, I have usually referred torepeated evolutionary changes as examples of parallelism,rather than convergence. I have done so because the repli-cate populations started with the same ancestral state, andso the changes are, geometrically speaking, parallel andnot convergent. This distinction is clear when the relevantstates are discrete (e.g., the identity of a particular nucleo-tide in a DNA sequence) and the history of changes isknown precisely. But for traits that might change repeatedly,and especially for quantitative traits (e.g., size or fitness), wetypically measure the net effect of many changes, some ofwhichmight be parallel, others divergent, and yet others con-vergent following prior divergence. In general, our knowl-edge of the steps along the way becomes less certain as thetime elapsed since the identical ancestral states gets longer.Nonetheless, I think it is fair to say that many of us are mostinterested in the push and pull of evolutionary forces thatpromote divergence versus convergence. Therefore, in thisreview, I often refer broadly to divergence and convergence,with the understanding that convergence may, dependingon context, include some changes that are, in a strict sense,parallel.

Experimental Overview

The LTEE has 12 populations, each founded from the sameancestral strain of Escherichia coli B except for a geneticmarker that distinguishes two sets of six populations each(Lenski et al. 1991). The marker serves two purposes. First,it provides an indication of possible cross contamination,which would compromise the independence of the replicatepopulations. Second, it allows one to compete a samplefrom an evolved population against the ancestral strain thatpossesses the alternative marker, or to compete samplesfrom two reciprocally marked populations against one an-other. Each population started from a single haploid cell,which was achieved by plating colonies (with each onethe outgrowth of a single cell) and then using a different col-ony to found each population. As a result, all of the geneticvariation available for evolution was generated by new mu-tations, and no variation was identical by descent acrossreplicate populations, thus excluding collateral evolution(Stern 2013) as a potential source of convergence. Eachpopulation exists in a small Erlenmeyer flask that is heldin a shaking incubator at 377C, where it grows in 10 mLof a medium with glucose as the limiting resource. Everyday, someone removes 1% of the volume from each flaskand transfers it to 10 mL of fresh medium. The 100-fold

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

A Long-Term Experiment with Bacteria S59

dilution and regrowth allow log2100 ≈ 6:64 doublings, orgenerations, per day, which we round to 100 generationsfor every 15 transfers. Each day, the bacteria experience alag phase followed by exponential growth, depletion of thelimiting glucose, and a stationary phase where they sit untilthe next transfer. Even as the bacteria have evolved to growfaster than their progenitors, the number of generationseach day is set by the dilution and regrowth. The populationsize initially fluctuated between∼5# 106 and∼5# 108 cells;those numbers have declined somewhat as the bacteria evolvedlarger individual cells.

The LTEE started in 1988, and it has now been runningfor over 65,000 generations. There have been occasionalinterruptions along the way, such asmoving the experimentacross the country and various accidents. Even more im-portant than their rapid growth, which allows an experi-ment to encompass so many generations, bacteria can befrozen and then revived. Every 100 generations in the earlyhistory of the LTEE, and then every 500 generations since,samples have been stored at2807C, providing a frozen fos-sil record. These samples serve two critical functions. First,in the event of an interruption or accident, the LTEE can berestarted without going back to the beginning. Second, onecan directly compete or compare samples that lived at dif-ferent points in time—in essence, time travel. The long du-ration of the LTEE not only provides an impressive numberof generations but also means that new technologies, includ-ing the ability to sequence entire genomes, have emergedsince it began.

Convergence and Divergence

Fitness in the LTEE Environment

All of the populations have become more fit, according toassays in which the evolved bacteria compete against the re-ciprocally marked ancestral strain in the same environmentas the LTEE has run (Lenski and Travisano 1994; Wiseret al. 2013). Before every fitness assay, both competitorshave been frozen, revived, and acclimated to that environ-ment. After 50,000 generations, a typical population has afitness of ∼1.7 relative to the ancestor, meaning that theevolved bacteria grow about 70% faster during head-to-head competition (Wiser et al. 2013). Moreover, the fitnesstrajectories are similar in their overall form, with muchgreater improvements in the early generations than in laterones (fig. 1).

On the other hand, there is significant among-populationvariation in fitness—that is, heterogeneity greater than onewould expect from measurement error (Lenski et al. 2015).Moreover, from 40,000 to 60,000 generations the relative-fitness ranks of the populations were strongly conserved,even as they all continued to experience significant fitness

This content downloaded from 132All use subject to University of Chicago Press Term

gains. The among-population variance in fitness shows nosign of declining with time. The square root of that vari-ance—in essence, the genetic standard deviation—has hov-ered around a few percent, which is substantially less thanone-tenth of the average fitness increase relative to the an-cestor. Some, but not all, of the persistent variation is asso-ciated with evolved differences in mutation rates (Lenskiet al. 2015), as described below. Also, one unlucky popu-lation, called Ara11, appears to have gotten stuck in a lessproductive region of the adaptive landscape than all of theothers, having both the lowest fitness and the slowest rateof improvement from 40,000 to 60,000 generations (Lenskiet al. 2015).

Correlated Responses to Other Environments

One can also run competition assays between the evolvedpopulations and their ancestor in environments that differfrom the LTEE environment. In contrast to the high andrelatively homogeneous fitness values measured in the LTEEenvironment, fitness levels tend to be lower and more vari-able in other environments (Travisano et al. 1995b). For ex-ample, when clones sampled at generation 2,000 competedagainst the ancestor in an environment where maltose re-placed glucose, on average they were no more fit than theancestor. However, the evolved bacteria were extremely het-erogeneous, with some almost as fit in maltose as in glucoseand others much less fit than the ancestor when competingfor maltose.With lactose replacing glucose, the evolved lineswere on average as fit as they were in glucose, but the geneticvariation in fitness was even greater than it was in maltose.In short, the correlated responses to alternative environ-ments have tended to be more variable than the direct re-sponse to selection.

0 10 20 30 40 501.0

1.2

1.4

1.6

1.8

2.0

Time (thousands of generations)

Fitn

ess

rela

tive

to a

nces

tor

Figure 1: Trajectories of mean fitness for the 12 independentlyevolving populations of the long-term evolution experiment. Eachcurve shows the best fit of a power-law model to the data fromone population; three trajectories are truncated because of difficul-ties in performing the competitions used to estimate fitness. Dataare from Wiser et al. (2013).

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S60 The American Naturalist

But this tendency toward greater divergence in novel en-vironments has some wrinkles. For example, all 12 popula-tions completely lost their ability to grow on another sugar,ribose, over the first 2,000 generations of the LTEE (Coo-per et al. 2001). In this case, the parallel losses resultedfrom deletion mutations that were demonstrably beneficialin the glucose-based medium and, moreover, occurred spon-taneously at an exceptionally high rate. The high mutationrate was caused by a transposable element adjacent to theribose operon that transposed into the operon and thenunderwent homologous recombination, leading to a dele-tion. These losses were so consistent across the replicatepopulations as to generate convergent, not divergent, out-comes.

There is also a sort of gray area that exists between diver-gence and convergence. Over time, many of the LTEE pop-ulations have lost their ability to grow on maltose; after20,000 generations, clones from over half of them couldnot use that sugar (Pelosi et al. 2006). So one could say thatthe populations underwent divergent responses to maltoseearly in the LTEE but are now undergoing convergent evo-lution toward the maltose-negative state. Thus, both themetric we use—quantitative variation in performance ver-sus discrete losses of function—and the timescale over whichevolution takes place can affect whether we describe certainpatterns as divergent or convergent.

One of the unexpected responses to an alternative envi-ronment is that many of the populations have gained resis-tance to infection by the phage Lambda (Meyer et al. 2010).That resistance is surprising because the bacteria in theLTEE environment are never exposed to Lambda or anyother virus. Conventional wisdom about trade-offs is thatresistance has a fitness cost, which would lead one to expectthat evolution in the absence of parasites should lead toincreased susceptibility, not resistance. In fact, the evolvedresistance to Lambda has the same genetic basis as the al-tered performance onmaltose. It so happens that mutationsin a gene calledmalT result in reduced expression of a genethat encodes a porin protein through which maltose crossesthe outer membrane of the Escherichia coli cell. That sameprotein, called LamB, is also the receptor to which Lambdabinds to initiate an infection. Mutations in malT provide ademonstrable fitness advantage in the glucose-based LTEEenvironment (Pelosi et al. 2006), probably because they re-duce the costs of unnecessary expression of LamB and otherproteins regulated by malT, and the resistance to phageLambda is simply a pleiotropic effect of these cost-savingmutations.

Cell Size and Yield

Had I thought in advance about how the size of the bacte-rial cells would change during the LTEE, I would have

This content downloaded from 132All use subject to University of Chicago Press Term

predicted they should become smaller. All else being equal,smaller cells have more surface area relative to volume,which seems advantageous in a resource-limited environ-ment. However, that is not what has happened. Instead,all 12 populations produce much larger individual cellsthan the ancestral strain (Lenski and Travisano 1994; Vasiet al. 1994; Lenski andMongold 2000). The adaptive signif-icance, if any, of the larger cells remains unclear. One po-tential explanation is that faster-growing cells tend to belarger, with each cell having more chromosomal copies, ri-bosomes, and so on. The evolved cells grow faster than theancestor, so perhaps they are larger simply because they aregrowing faster. We tested this hypothesis by forcing evolvedand ancestral clones to grow at the same rate in separatechemostats, wherewe could control the growth rate by settingthe dilution rate (Mongold and Lenski 1996). We confirmedthat faster-growing cells are larger for both the ancestral andthe evolved bacteria. However, the evolved bacteria exhibit asteeper increase in cell size with growth rate, such that theyproduce larger cells than the ancestral strain at all growthrates. Thus, this hypothesis correctly predicts the directionof the change in cell size, but it does not account for the mag-nitude of the change. Another possibility—one that remainshypothetical—is that larger cells have an advantage in the“feast and famine” transfer regime of the LTEE because theyare, in effect, larger sponges and thus able to acquire glucosefaster than smaller ones. The larger cells might thereby se-quester that limiting resource and pass it (or its by-products)along as they divide. In any case, all of the populations pro-duce much larger cells than their common ancestor, so thatthey have evolved in the same direction, although their re-sulting cell volumes are quite variable (Lenski and Travisano1994).There has also been substantial divergence in cell shape,

with most lineages clearly rod shaped but others somewhatmore spherical (Lenski and Mongold 2000). Also, as thecells have evolved to become larger, their numerical yieldat the end of the daily growth cycle has decreased (Vasiet al. 1994; Lenski and Mongold 2000). That is, there arefewer, but larger, cells than there were at the outset of theLTEE. Nonetheless, the product of cell number and averagecell volume has increased in all 12 populations, whichmeans that the biovolume yield increased in parallel as well.Moreover, there was a near-perfect inverse relation betweenthe numerical yield and the average cell size of the popula-tions at 10,000 generations (Lenski and Mongold 2000).The constancy of that product indicates that the changesin yield were not only qualitatively but also quantitativelyconvergent. Once again, however, different analyses of cellmorphology—the particular traits measured as well as thedirection versus the magnitude of evolved changes—mayaffect whether one is more impressed by the convergenceor by the divergence of the replicate populations.

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

A Long-Term Experiment with Bacteria S61

A New Metabolic Capability

The LTEE culture medium contains glucose as the limitingresource, but there is another large pool of carbon and en-ergy in the medium that the cells cannot consume. In par-ticular, the medium contains citrate as a chelating agent,and its concentration is much higher than that of glucose.However, a near-universal feature of E. coli as a species isthat it cannot grow on citrate in oxygen-containing envi-ronments, because the cells are unable to take up citratefrom the environment. So every day, after the bacteria con-sumed all of the available glucose, another source of carbonand energy sat there untapped. That situation continued forover 30,000 generations, but then one population, calledAra23, began to use the citrate (Blount et al. 2008), re-sulting in a culture that was much more turbid than theothers (fig. 2A). At first, I thought that a citrate-using con-taminant of another species was responsible, and so thepopulation was restarted from its most recent frozen sam-ple. When the citrate users reemerged, we analyzed geneticmarkers that confirmed that they were indeed E. coli cellsdescended from the strain used to start the LTEE.

This content downloaded from 132All use subject to University of Chicago Press Term

At first, we struggled to understand what had changedgenetically that enabled the cells to grow on citrate. How-ever, we ran what we call “replay” experiments to distin-guish between two hypotheses. According to one, the newCit1 function resulted from some rare mutation—for ex-ample, an inversion that had to occur at two exact pointsin the genome—that could have happened at any time inthat population’s history and would have yielded the sameability to grow on citrate. Alternatively, the origin of thisfunction was contingent on one or more earlier “potentiat-ing” mutations that, by themselves, conferred no capacityto use citrate but, with the right subsequent mutation, en-abled its consumption. Several large replay experiments sup-ported the hypothesis of historical contingency: neither theancestor nor any clone tested from population Ara23 before20,000 generations produced any Cit1 mutants, but laterclones generated many such mutants (Blount et al. 2008).Soon it became affordable to sequence whole genomes

and identify the mutations responsible for the new Cit1

function (Blount et al. 2012). This analysis revealed thatthe mutation that produced the first Cit1 cell in populationAra23 was a tandem duplication of a segment that includes

A

B

Figure 2: Evolution of a new function in population Ara23. A, Increase in optical density (i.e., turbidity) when some cells evolved the ca-pacity to use citrate. Redrawn from Blount et al. (2008). B, Tandem duplication leading to two copies of the citT gene, with the second copycoming under the control of one of the promoter elements, shown as arrows, downstream of the ancestral copy. Redrawn from Blount et al.(2012).

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S62 The American Naturalist

the citT gene, which encodes a protein used to transport cit-rate and some other carboxylic acids under anoxic condi-tions. However, the effect of the duplication is not a dou-bling of the basal expression level; in fact, there was nodetectable expression of the ancestral copy of citT (Blountet al. 2012). Instead, the duplication brought together thenew copy of that gene with a promoter region that wasdownstream of the original copy (fig. 2B), which led to theexpression of the citT-encoded transporter during aerobicgrowth just as the glucose was depleted. All of the Cit1 mu-tants found in the replay experiments also had rearrange-ments of their DNA that appeared to bring together the citTgene and either the same or other promoters, although innone of the replays was the rearrangement identical to theone in the original Ara23 population. The genomic dataalso showed that the new module with the combined regu-latory domain and structural gene underwent further du-plications that improved growth on citrate. By cloning themodule into a plasmid that could be moved into other ge-netic backgrounds, we found strong epistasis between thosebackgrounds and the module. The new module conferredsome capacity to grow on citrate even in the ancestral strain;however, growth was much weaker in that background andseveral others than it was in the background of a late-generation clone closely related to the lineage in which thenew function evolved (Blount et al. 2012). These results thusconfirmed the importance of potentiating mutations in set-ting the stage for the subsequent evolution of the Cit1 phe-notype. Later genetic analyses identified a specific mutationin another gene that played an important role in that poten-tiation (Quandt et al. 2015).

After the bacteria began to use the citrate, it caused im-portant changes to the ecology of the system. Because ofthe high concentration of citrate in the medium, the popu-lation density increased several-fold. Also, the Cit1 bacteriadid not drive the Cit2 bacteria to extinction; instead, thetwo lineages coexisted for over 10,000 generations (Blountet al. 2012; Turner et al. 2015a). The Cit2 cells could persistbecause they had an advantage in competition for the glu-cose; in particular, they transitioned from stationary phaseto renewed growth after transfer into fresh medium fasterthan did the Cit1 cells (Blount et al. 2008). This differenceprobably occurred because the Cit1 cells enter stationaryphase after depleting the citrate and then switch over tothe more valuable glucose in the fresh medium, whereasthe Cit2 cells used glucose both before they entered station-ary phase and when they recommenced growth. Eventually,however, the Cit2 lineage became extinct. The cause of thatextinction is unclear, but it appears to have been causedby some transient perturbation to the experiment (Turneret al. 2015a).

In fact, the Ara23 ecosystem became even more compli-cated than one might expect in going from one resource

This content downloaded from 132All use subject to University of Chicago Press Term

(glucose) to two (glucose and citrate) as a consequence ofhow the new function works. In particular, the citT-encodedprotein is an antiporter: in the course of importing onemol-ecule into the cell, it exports another molecule into the en-vironment. Thus, as the citrate is consumed, other, less valu-able but still usable sources of carbon and energy—succinate,fumarate, andmalate—accumulate in themedium, leading toa system with not one, not two, but several distinct resources.Both the Cit1 and Cit2 lineages further evolved in ways thatimproved their growth on these by-products of citrate trans-port (Turner et al. 2015b).As of this writing, at 65,000 generations, no other LTEE

population has evolved the ability to grow on the citrate.The evolution of this function is therefore a conspicuous ex-ample of divergence under identical conditions from thesame initial genetic state. It will be interesting to see whetherany of the other 11 populations make this transition in theyears, decades, and centuries ahead. If some or all of themdo, then what was once seen as divergence might eventuallybe called convergence. Having replicate populations evolveunder identical conditions solves the denominator problemwith respect to the number of lineages at risk, but it doesnot address the open-ended nature of the time at risk forsome particular change to occur.

Mutation Rates

Six of the 12 LTEE populations evolved point-mutation ratesroughly 100-fold higher than that of their ancestor (Snie-gowski et al. 1997; Tenaillon et al. 2016). In four cases thehypermutability resulted from mutations that affect DNAmismatch repair, and in two cases it was caused by muta-tions affecting enzymes that remove oxidized nucleotides.Hence, the set of populations as a whole have diverged intheir mutation rates. However, one could also say that sixpopulations converged on higher mutation rates while di-verging in terms of the biochemical basis for their hyper-mutability. In any case, these changes have had importantconsequences. For example, the hypermutable lineages haveaccumulated many more mutations, and a higher propor-tion of synonymous substitutions, than those that retainedthe low ancestral mutation rate (Tenaillon et al. 2016). Also,the populations that evolved hypermutability early in theLTEE have reached higher fitness levels than the other pop-ulations (Wiser et al. 2013; Lenski et al. 2015), although thefitness differences are far smaller than the differences inmutation rate owing to clonal interference (i.e., competitionamong lineages with different beneficial mutations) and tothe increased load of deleterious mutations. Moreover, asa consequence of the increased mutational load and the de-clining opportunity for further adaptation, five of the hyper-mutable populations subsequently evolved lower mutationrates (Wielgoss et al. 2013; Tenaillon et al. 2016). Thus, we

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

A Long-Term Experiment with Bacteria S63

see evidence of convergence not only in the mutation-ratephenotype but also in its temporal trajectory.

At this point, one might be wondering whether the line-age in population Ara23 that gained the ability to grow oncitrate had previously evolved hypermutability. In fact, ithad not. However, it evolved a mutator phenotype not longafter it acquired that function. Moreover, Ara23 is the onlypopulation that became hypermutable but, so far, has not atleast partially compensated for that change. As a conse-quence of its new ability to use citrate, this population hasadditional opportunities for adaptation, apparently overrid-ing the benefit that would accrue from a reduced load of del-eterious mutations.

In addition to the six populations that evolved higherrates of point mutation, another population—Ara11, thelaggard with respect to its fitness trajectory—evolved a dif-ferent type of hypermutability (Lenski et al. 2015; Tenaillonet al. 2016). In this case, one of the several transposable ele-ments present in the ancestral genome evolved heightenedactivity that caused a substantial increase in the number ofinsertion mutations.

Genomic Changes

When the LTEE began in 1988, not a single bacterial ge-nome had been fully sequenced. In one recent paper, 264 ge-nomes sampled from the LTEE were fully sequenced andanalyzed, including two representatives from all 12 popu-lations at 11 time points through generation 50,000 (Te-naillon et al. 2016). As discussed above, some of the LTEEpopulations evolved hypermutability, and those changesobviously affect genomic evolution. If we exclude the pop-ulations that became hypermutable (including the one withincreased insertions), the genomes sampled from the otherpopulations averaged ∼76 total mutations after 50,000 gen-erations (fig. 3). Of that total, ∼56% were point mutations;various types of insertions and deletions account for most ofthe rest.

Most mutations that occurred during the LTEE were pre-sumably neutral or deleterious, but those that gave somecompetitive advantage were more likely to reach high fre-quency and thus be present in the genomes that were se-quenced. Even so, neutral and deleterious mutations canhitchhike with beneficial driver mutations; they may alsobe present in a sampled genome as the result of a recentmutation or random drift. However, several lines of evi-dence indicate that most of the mutations observed in thenonmutator populations, and in the others before they be-came hypermutable, were beneficial (Tenaillon et al. 2016).First, the trajectory for the number of mutations over timewas better fit by a model with both neutral and beneficialmutations, which are expected to increase linearly and at adeclining rate, respectively, than by models without both

This content downloaded from 132All use subject to University of Chicago Press Term

terms. Second, nonsynonymous and intergenic point muta-tions were found in substantial excess relative to the expec-tation based on synonymous substitutions, after adjust-ment for the number of sites at risk for each class. Third,insertions and deletions, as well as nonsynonymous andintergenic point mutations, were seen more often than ex-pected in comparison to a mutation-accumulation experi-ment, in which passage through single-cell bottleneckslargely eliminated the effects of natural selection.Moreover, there was strong convergent evolution at the

level of genes. Fifty-seven protein-coding genes (out of14,000) had two or more independent nonsynonymousmutations across all of the nonmutator and premutatorlineages combined (Tenaillon et al. 2016). These 57 genescomprise only 2% of the total length of protein-codinggenes, but they harbored fully 50% of the nonsynonymousmutations in those genomes. That degree of convergenceis extremely unlikely (P ! 102143), according to simulationsin which mutations were randomly distributed among theprotein-coding genes according to length (Tenaillon et al.2016). Some of the genes showing the strongest convergenceencode proteins with core regulatory and metabolic func-tions. Two such genes, spoT andmalT, were first discoveredfrom changes in gene-expression profiles, through tran-scriptomics and proteomics, respectively, and then tracingback to the causal mutations (Cooper et al. 2003; Pelosi et al.2006). Another gene, topA, was found after changes in DNAsuperhelicity were observed (Crozat et al. 2010). Severalother genes showing convergent evolution encode proteinsinvolved in peptidoglycan synthesis, and mutations therecontribute to the evolved changes in cell size and shape(Philippe et al. 2009). There was also a significant, thoughweaker, signal of convergent evolution in the lineages thatevolved hypermutability (Tenaillon et al. 2016). That sig-nal was weaker because the beneficial mutations in those

Time (thousands of generations)

Tota

l mut

atio

ns

0 10 20 30 40 50

0

20

40

60

80

100

Figure 3: Trajectories for the total number of mutations in genomessampled from five long-term evolution experiment populations thatdid not evolve any type of hypermutability. Each curve shows the fitof a two-parameter model, with terms representing beneficial andneutral mutations, to numbers obtained from two clones at eachof 11 time points. Data are from Tenaillon et al. (2016).

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S64 The American Naturalist

lineages were diluted in a sea of neutral or weakly deleteri-ous hitchhikers and drifters.

Deletion and insertion mutations also exhibited substan-tial convergence in the LTEE (Tenaillon et al. 2016). How-ever, it is more difficult to interpret these events, for tworeasons. First, many of the parallel indels potentially affectmultiple genes. Second, some of them seem to reflect hyper-mutability caused by transposable elements and homolo-gous recombination involving repeated sequences, ratherthan convergence caused by natural selection. In any case,many of the gene regions most often affected by indels inthe LTEE appear to have been acquired by horizontal trans-fer in the distant past. The disruption or deletion of suchgenes may be effectively neutral if their products are unusedin the LTEE environment, or those mutations might bebeneficial if the genes and their products impose some met-abolic burden on the cells.

Trajectory of Genetic Parallelism

While the statistical signal for genetic convergence is ex-tremely strong, what can be said about its magnitude andhow it changes over time? In a review of published studies,Conte et al. (2012) sought to quantify the probability thatthe same genes were reused in cases of parallel and conver-gent evolution of various phenotypes in nature. Dependingon the methods used in the underlying studies, they foundthat the same genes were involved in about one-third toone-half of the cases they examined. However, this proba-bility varied, depending on how long ago the lineages haddiverged—there was a greater chance that the same geneswere responsible for parallel changes in the most recentlydiverged lineages.

As noted in the introduction, one of the advantages of anexperiment like the LTEE is that we know the denominator(i.e., the number of populations evolving under a given setof conditions) when it comes to analyzing convergence. Wealso know precisely how long populations have been sepa-rated since their common ancestor. Therefore, it is feasibleto address the same issues examined by Conte et al. (2012),albeit in a very different system. However, there are em-pirical challenges: the evolution of hypermutable lineagesdilutes and thus obscures the driver mutations, and somemutational events affect multiple genes. Moreover, fitnessis an extremely broad and integrative phenotype relative to,say, floral pigmentation or resistance to some toxin. None-theless, I sought to address the issue as follows. I focused onnonsynonymous mutations in protein-coding genes, usingthe data in Tenaillon et al. (2016). I used all 12 populationsthrough 2,000 generations, because none had evolved hyper-mutability to that point; I used only the six populations thatretained the ancestral point-mutation rate through 50,000generations for the later samples. Using a Web-based tool

This content downloaded from 132All use subject to University of Chicago Press Term

to screen and sort mutations (http://barricklab.org/shiny/LTEE-Ecoli/), I asked, for each population at each of the11 timepoints, howmany genes had one ormore nonsynony-mous substitutions in at least one of the two sequencedclones. That number rose from ∼2, on average, at 500 gen-erations to ∼32 at 50,000 generations. Then, for every pair ofpopulations, I asked how many genes had nonsynonymousmutations in both of the samples from the same generation.That number increased steadily, from ∼0.3 to ∼5.8, on aver-age, after 500 and 50,000 generations, respectively. I then cal-culated the proportional overlap in the affected genes for eachpopulation pair at each time point as 2nab=(na 1 nb), wherena and nb are the total number of affected genes for each pop-ulation and nab is the number they share. This index of par-allelism was ∼15% after 500 generations, it rose to ∼33% at2,000 and 5,000 generations, and it then declined graduallyto ∼18% by 50,000 generations.Conte et al. (2012) suggest two possible explanations for

the declining probability of reusing the same genes at longerdivergence times. First, more recently diverged lineages aremore likely to share genetic variants that could fuel parallelevolution (i.e., collateral evolution, in the sense of Stern2013). This issue is not relevant to the LTEE, however, be-cause there was no shared variation at the start. Second, thegenetic backgrounds in which new alleles must function di-verge over time, introducing the potential for epistatic in-teractions to alter the spectrum of mutations that producea particular beneficial effect and thereby generate historicalcontingencies. This explanation is certainly plausible forthe LTEE as well, and indeed it is implicated in the evolu-tion of citrate utilization (Blount et al. 2008).Two other explanations are also relevant to the LTEE, es-

pecially for such an integrative trait as fitness. First, thereare probably more genes that can yield mutations that con-fer small fitness gains than there are genes that can producemutations that confer large gains. Because of the dynamicsof selection, including clonal interference between asexuallineages that possess different beneficial mutations, thereis a tendency for beneficial mutations of large effect to fixbefore those that confer smaller benefits (Gerrish and Lenski1998; Wiser et al. 2013). As a consequence, adaptation mayfollow more similar paths—involving those few genes withthe largest benefits—early in the LTEE, before fanning outalong more diverse paths in later generations. Second, thefraction of accumulatedmutations that are beneficial declinesover time, with a corresponding increase in the proportionof neutral or weakly deleterious mutations (Tenaillon et al.2016). While most mutations that affect the same gene in apair of populations are beneficial, those parallel changes be-come a smaller fraction of the total mutations over time. Thiseffectmightwell account formuch of the apparent increase inthe diversity of genetic pathways over time. The accumula-tion of neutral or weakly deleterious mutations could also

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

A Long-Term Experiment with Bacteria S65

generate historical contingencies if some of them provide“stepping stones” that enable adaptations that would other-wise not be accessible by a strictly hill-climbing process (Co-vert et al. 2013).

Stable Polymorphisms

Bacteria reproduce asexually, and those in the LTEE lackany means for horizontal gene transfer. Unlike some bacte-rial species, E. coli is not naturally transformable, and thestrain used to found the LTEE does not carry any plasmidsor functional phages that could move genes between cellsby conjugation or transduction, respectively. Bacteria arealso haploid, and so polymorphisms maintained by hetero-zygote advantage are not possible. Moreover, the LTEEenvironment is simple, thereby limiting—but not eliminat-ing—the opportunity for distinct ecotypes to evolve and sta-bly coexist. In short, one might expect evolution in the LTEEto occur by sequential selective sweeps of beneficial muta-tions, either singly or as cohorts (Lang et al. 2013; Maddam-setti et al. 2015), such that any polymorphisms are transient,not persistent.

I already mentioned one exception to this expectation,namely, the coexistence between the Cit1 and Cit2 lineagesin population Ara23 after the former gained the new abil-ity to use the exogenously supplied citrate. Although theevolution of that ability was surprising, the fact that twolineages could stably coexist on two resources—glucoseand citrate—is not. However, another population, Ara22,produced a polymorphism that arose much earlier andhas persisted far longer (Rozen and Lenski 2000; Le Gacet al. 2012; Tenaillon et al. 2016). The polymorphism wasfirst detected through heritable differences in colony mor-phology, with two morphotypes dubbed “L” and “S” for theirlarge and small colonies, respectively. Experiments in whichthe two types from the same generation were mixed at dif-ferent initial frequencies demonstrate a negative frequency-dependent interaction that supports coexistence (Rozen andLenski 2000; Le Gac et al. 2012). Other experiments haveshown that the L type grows faster on glucose, although bothtypes grow substantially faster than the ancestral strain. How-ever, the S type can also grow in spent medium where L hadpreviously grown but can grow no further (Rozen and Lenski2000). Recent work indicates that acetate, which is excretedby both types during growth on glucose, is used more effec-tively by and supports the S population (Großkopf et al.2016).

Although the two ecotypes converge on a stable equilib-rium over the course of several weeks, over longer periodstheir relative abundance has shifted dramatically (Rozenand Lenski 2000), with their ratio varying by almost twoorders of magnitude over the first 20,000 generations (fig. 4).These shifts reflect the ongoing evolution of the two line-

This content downloaded from 132All use subject to University of Chicago Press Term

ages—one lineage gets a beneficial mutation that drives theequilibrium ratio in its favor, and then the other does, andso on (Le Gac et al. 2012). However, neither lineage has ac-quired a beneficial mutation that is sufficient to overcomethe advantage when rare that keeps the other lineage frombecoming extinct. That is, a sustained polymorphism de-pends not only on the effect sizes of the beneficial mutationsthat arise in the contending lineages but also on the strengthof the frequency-dependent interaction that gives each eco-type its advantage when rare (Maddamsetti et al. 2015).In population Ara21, two genetically distinct lineages

coexisted from about 7,000 to 15,000 generations, and com-petitions between representatives of the two lineages indi-cate a negative frequency-dependent interaction (Maddam-setti et al. 2015). There were also dramatic back-and-forthfluctuations in the relative abundance of the two lineages,similar to those seen in Ara22. However, in Ara21, onelineage eventually accumulated beneficial mutations thatdelivered a knockout blow, driving the other to extinction(Maddamsetti et al. 2015). Genome sequencing has also un-covered distinct lineages that coexisted in other populationsfor over 10,000 generations (Blount et al. 2012; Tenaillonet al. 2016). These deeply diverged lineages suggest that neg-ative frequency-dependent interactions, whether transientor sustained, have evolved in at least several of the LTEEpopulations.

Overview and Outlook

I began this review by suggesting that evolution experimentsoffer the potential to quantify rigorously the extent of con-vergence and divergence, because the number of lineages

10.90.80.70.60.50.40.30.20.1

00 5000 10,000 15,000 20,000

Generation

Freq

uenc

y of

S m

orph

Figure 4: Fluctuations in the frequency of the S type in populationAra22. Two ecotypes, L and S, have coexisted for many thousandsof generations, but their relative abundances have changed substan-tially as beneficial mutations arose in their respective lineages. Re-printed from Rozen and Lenski (2000).

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S66 The American Naturalist

that might or might not change in a particular way is knownand because the conditions can be made essentially identicalwith respect to both the environment and the initial geneticstate. Using the Escherichia coli LTEE as an example, I haveshown how this approach has allowed my team to quantifythe among-population variation in fitness, test for parallelmutations in genes, and characterize the patterns of changein many other traits. When I look across the sweep of theseanalyses—from genetics and genomics through physiologyand morphology to performance and ecology—I am fasci-nated by both the convergences and the divergences of thereplicate populations. While we have seen many strikingexamples of convergence, the more traits we study and themore deeply we examine them, we see that each populationis unique and follows its own evolutionary path. For exam-ple, all populations evolved to produce much larger cellsthan the common ancestor, and yet the extent of the sizechanges and the resulting shapes of the cells show consider-able variation.

Another complication arises if we examine the extent ofconvergence and divergence over time. For example, six ofthe LTEE populations evolved hypermutator phenotypes,and six did not. But perhaps all of them will eventually be-come mutators. Or perhaps the mutator lineages will re-evolve lower mutation rates, as some of them have alreadydone. Similarly, only one population has evolved the abilityto grow on citrate even after 60,000 generations, but per-haps all of them will have gained this ability after 600,000or 6,000,000 generations. Or maybe some of them willevolve other new abilities, such as horizontal gene transferor predation on other cells. In short, we know that the pop-ulations started out identical, and we know that they nowdiffer from one another despite having undergone manysimilar changes, but it is unclear whether the divergenceswill persist indefinitely or might yet prove to be transient.

My hope is that the LTEE will continue for many scien-tific generations (Fox and Lenski 2015), providing furtherinsights into the core tension between chance and necessityin evolution. Of course, these E. coli populations are differ-ent in important respects from the systems that most evo-lutionary biologists study. Their asexual reproduction leadsto clonal interference, which might promote convergencebecause the relatively few mutations with the largest bene-ficial effects are even more likely to fix first than they are inrecombining populations. On the other hand, asexual re-production may promote divergence if it allows syner-gistically beneficial combinations of mutations to fix morereadily. Also, analyses of the LTEE populations have often fo-cused on extremely broad and integrative phenotypic traits,including cell size and competitive fitness, where the numberof genes with relevant mutations is likely to be much largerthan that for specific traits, such as pigmentation and toxinresistance, that are typically studied by evolutionary biolo-

This content downloaded from 132All use subject to University of Chicago Press Term

gists working with plants and animals. The mere handful ofgenes that underlie similarly specific traits that have changedduring the LTEE, such as DNA superhelicity (Crozat et al.2010) and hypermutability (Sniegowski et al. 1997; Wiel-goss et al. 2013; Tenaillon et al. 2016), might provide morerelevant comparisons with typical studies of evolutionaryconvergence.In any case, other evolution experiments will be neces-

sary to illuminate important factors that could influencethe likelihood of convergent versus divergent responses.Does recombination accelerate adaptive evolution, anddoes it promote convergence above and beyond the effectof any shared variants present at the start of an experiment(Azevedo et al. 2006; Cooper 2007; McDonald et al. 2016)?Does environmental complexity, such as that produced byspatial structure, multiple resources, or coevolutionary in-teractions, promote not only diversity within populationsbut also divergence across replicate lineages (Korona et al.1994; Rainey and Travisano 1998; Cooper and Lenski 2010;Paterson et al. 2010; Nahum et al. 2015)? How do popula-tion size and initial genotypes, whether sampled from na-ture or from other experiments, affect the propensity forsubsequent divergence (Travisano et al. 1995a; Burch andChao 1999; Moore and Woods 2006; Weinreich et al. 2006;Perfeito et al. 2007; Woods et al. 2011; Lindsey et al. 2013;Kryazhimskiy et al. 2014)? In short, what genetic and envi-ronmental factors determine whether fitness landscapes aresmooth or rugged, and how do they interact to promote di-vergent or convergent solutions to a given set of challenges(De Visser and Krug 2014;Wang et al. 2016)?With so manyquestions to answer, the tension between evolutionary con-vergence and divergence—between the repeatable and theunique—should be a source of fascination and discoveryfar into the future.

Acknowledgments

I thank everyone who has worked on the LTEE over thecourse of nearly 30 years, including especially N. Hajela,who has been the lab manager and technician for almost20 years; D. Schneider, who has worked to understand thegenetic changes in the LTEE lines for nearly two decades;and Z. Blount, who has studied the causes and consequencesof citrate utilization for over a decade. I thank A. Agrawal forinviting me to write this review; D. Schluter for a suggestionthat led to the section on the trajectory of genetic parallel-ism; J. Barrick for developing the tool that facilitated theanalyses in that section; and Z. Blount, O. Tenaillon, andM. Wiser for preparing figures. The LTEE is currently sup-ported by the National Science Foundation (DEB-1451740),the BEACON Center for the Study of Evolution in Action(DBI-0939454), and Michigan State University.

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

A Long-Term Experiment with Bacteria S67

Literature Cited

Azevedo, R. B. R., R. Lohau, S. Srinivasan, K. K. Dang, and C. L.Burch. 2006. Sexual reproduction selects for robustness and neg-ative epistasis in artificial gene networks. Nature 440:87–90.

Barrick, J. E., and R. E. Lenski. 2013. Genome dynamics during ex-perimental evolution. Nature Reviews Genetics 14:827–839.

Blount, Z. D., J. E. Barrick, C. J. Davidson, and R. E. Lenski. 2012.Genomic analysis of a key innovation in an experimental Esche-richia coli population. Nature 489:513–518.

Blount, Z. D., C. Z. Borland, and R. E. Lenski. 2008. Historical con-tingency and the evolution of a key innovation in an experimentalpopulation of Escherichia coli. Proceedings of the National Acad-emy of Sciences of the USA 105:7899–7906.

Burch, C. L., and L. Chao. 1999. Evolution by small steps and ruggedlandscapes in the RNA virus f6. Genetics 151:921–927.

Burke, M. K., J. P. Dunham, P. Shahrestani, K. R. Thornton, M. R.Rose, and A. D. Long. 2010. Genome-wide analysis of a long-termevolution experiment with Drosophila. Nature 467:587–590.

Conte, G. L., M. E. Arnegard, C. L. Peichel, and D. Schluter. 2012.The probability of genetic parallelism and convergence in naturalpopulations. Proceedings of the Royal Society B 279:5039–5047.

Conway Morris, S. 2003. Life’s solution: inevitable humans in a lonelyuniverse. Cambridge University Press, Cambridge.

Cooper, T. F. 2007. Recombination speeds adaptation by reducingcompetition between beneficial mutations in populations of Esch-erichia coli. PLoS Biology 5(9):e225. doi:10.1371/journal.pbio.0050225.

Cooper, T. F., and R. E. Lenski. 2010. Experimental evolution with E.coli in diverse resource environments. I. Fluctuating environmentspromote divergence of replicate populations. BMC EvolutionaryBiology 10:11. doi:10.1186/1471-2148-10-11.

Cooper, T. F., D. E. Rozen, and R. E. Lenski. 2003. Parallel changesin gene expression after 20,000 generations of evolution in Esche-richia coli. Proceedings of the National Academy of Sciences of theUSA 100:1072–1077.

Cooper, V. S., D. Schneider, M. Blot, and R. E. Lenski. 2001. Mech-anisms causing rapid and parallel losses of ribose catabolism inevolving populations of E. coli B. Journal of Bacteriology 183:2834–2841.

Covert, A. W., III, R. E. Lenski, C. O. Wilke, and C. Ofria. 2013. Ex-periments on the role of deleterious mutations as stepping stonesin adaptive evolution. Proceedings of the National Academy ofSciences of the USA 110:E3171–E3178.

Crozat, E., C. Winkworth, J. Gaffé, P. F. Hallin, M. A. Riley, R. E.Lenski, and D. Schneider. 2010. Parallel genetic and phenotypicevolution of DNA superhelicity in experimental populations ofEscherichia coli. Molecular Biology and Evolution 27:2113–2128.

De Visser, J. A. G. M., and J. Krug. 2014. Empirical fitness land-scapes and the predictability of evolution. Nature Reviews Genet-ics 15:480–490.

Elena, S. F., and R. E. Lenski. 2003. Evolution experiments withmicro-organisms: the dynamics and genetic bases of adaptation. NatureReviews Genetics 4:457–469.

Fox, J. W., and R. E. Lenski. 2015. From here to eternity—the theoryand practice of a really long experiment. PLoS Biology 13(6):e1002185. doi:10.1371/journal.pbio.1002185.

Garland, T., Jr., and M. R. Rose, eds. 2009. Experimental evolution:concepts, methods, and applications of selection experiments.University of California Press, Berkeley.

This content downloaded from 132All use subject to University of Chicago Press Term

Gerrish, P. J., and R. E. Lenski. 1998. The fate of competing benefi-cial mutations in an asexual population. Genetica 102/103:127–144.

Großkopf, T., J. Consuegra, J. Gaffé, J. Willison, R. E. Lenski, O. S.Soyer, and D. Schneider. 2016. Metabolic modelling in a dynamicevolutionary framework predicts adaptive diversification of bacte-ria in a long-term evolution experiment. BMC Evolutionary Biol-ogy 16:163. doi:10.1186/s12862-016-0733-x.

Kawecki, T. J., R. E. Lenski, D. Ebert, B. Hollis, I. Olivieri, and M. C.Whitlock. 2012. Experimental evolution. Trends in Ecology andEvolution 27:547–560.

Korona, R., C. H. Nakatsu, L. J. Forney, and R. E. Lenski. 1994. Evi-dence for multiple adaptive peaks from populations of bacteriaevolving in a structured habitat. Proceedings of the National Acad-emy of Sciences of the USA 91:9037–9041.

Kryazhimskiy, S., D. P. Rice, E. R. Jerison, and M. M. Desai. 2014.Global epistasis makes adaptation predictable despite sequence-level stochasticity. Science 344:1519–1522.

Lang, G. I., D. P. Rice, M. J. Hickman, E. Sodergren, G. M. Wein-stock, D. Botstein, and M. M. Desai. 2013. Pervasive genetic hitch-hiking and clonal interference in forty evolving yeast populations.Nature 500:571–574.

Le Gac, M., J. Plucain, T. Hindré, R. E. Lenski, and D. Schneider.2012. Ecological and evolutionary dynamics of coexisting lineagesduring a long-term experiment with Escherichia coli. Proceedingsof the National Academy of Sciences of the USA 109:9487–9492.

Lenski, R. E., and J. A. Mongold. 2000. Cell size, shape, and fitness inevolving populations of bacteria. Pages 221–235 in J. H. Brownand G. B. West, eds. Scaling in biology. Oxford University Press,Oxford.

Lenski, R. E., M. R. Rose, S. C. Simpson, and S. C. Tadler. 1991.Long-term experimental evolution in Escherichia coli. I. Adapta-tion and divergence during 2,000 generations. American Natural-ist 138:1315–1341.

Lenski, R. E., andM. Travisano. 1994. Dynamics of adaptation and di-versification: a 10,000-generation experiment with bacterial popu-lations. Proceedings of the National Academy of Sciences of theUSA 91:6808–6814.

Lenski, R. E., M. J. Wiser, N. Ribeck, Z. D. Blount, J. R. Nahum, J. J.Morris, L. Zaman, et al. 2015. Sustained fitness gains and variabilityin fitness trajectories in the long-term evolution experiment withEscherichia coli. Proceedings of the Royal Society B 282:20152292.doi:10.1098/rspb.2015.2292.

Levy, S. F., J. R. Blundell, S. Venkataram, D. A. Petrov, D. S. Fisher,and G. Sherlock. 2015. Quantitative evolutionary dynamics usinghigh-resolution lineage tracking. Nature 519:181–186.

Lindsey, H. A., J. Gallie, S. Taylor, and B. Kerr. 2013. Evolutionaryrescue from extinction is contingent on a lower rate of environ-mental change. Nature 494:463–467.

Maddamsetti, R., R. E. Lenski, and J. E. Barrick. 2015. Adaptation,clonal interference, and frequency-dependent interactions in a long-term evolution experiment with Escherichia coli. Genetics 200:619–631.

McDonald, M. J., D. P. Rice, and M. M. Desai. 2016. Sex speeds ad-aptation by altering the dynamics of molecular evolution. Nature531:233–236.

Meyer, J. R., A. A. Agrawal, R. T. Quick, D. T. Dobias, D. Schneider,and R. E. Lenski. 2010. Parallel changes in host resistance to viralinfection during 45,000 generations of relaxed selection. Evolution64:3024–3034.

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S68 The American Naturalist

Mongold, J. A., and R. E. Lenski. 1996. Experimental rejection of anonadaptive explanation for increased cell size in Escherichia coli.Journal of Bacteriology 178:5333–5334.

Moore, F. B.-G., and R. Woods. 2006. Tempo and constraint of adaptiveevolution in Escherichia coli (Enterobacteriaceae, Enterobacteriales).Biological Journal of the Linnean Society 88:403–411.

Nahum, J. R., P. Godfrey-Smith, B. N. Harding, J. H. Marcus, J.Carlson-Stevermer, and B. Kerr. 2015. A tortoise-hare patternseen in adapting structured and unstructured populations suggestsa rugged fitness landscape in bacteria. Proceedings of the NationalAcademy of Sciences of the USA 112:7530–7535.

Paterson, S., T. Vogwill, A. Buckling, R. Benmayor, A. J. Spiers, N. R.Thomson, M. Quail, et al. 2010. Antagonistic coevolution acceleratesmolecular evolution. Nature 464:275–278.

Pelosi, L., L. Kühn, D. Guetta, J. Garin, J. Geiselmann, R. E. Lenski,and D. Schneider. 2006. Parallel changes in global protein profilesduring long-term experimental evolution in Escherichia coli. Ge-netics 173:1851–1869.

Perfeito, L., L. Fernandes, C. Mota, and I. Gordo. 2007. Adaptive mu-tations in bacteria: high rate and small effects. Science 317:813–815.

Philippe, N., L. Pelosi, R. E. Lenski, and D. Schneider. 2009. Evolu-tion of penicillin-binding protein 2 concentration and cell shapeduring a long-term experiment with Escherichia coli. Journal ofBacteriology 191:909–921.

Quandt, E. M., J. Gollihar, Z. D. Blount, A. D. Ellington, G. Geor-giou, and J. E. Barrick. 2015. Fine-tuning citrate synthase flux po-tentiates and refines metabolic innovation in the Lenski evolutionexperiment. eLife 4:e09696. doi:10.7554/eLife.09696.

Rainey, P. B., and M. Travisano. 1998. Adaptive radiation in a het-erogeneous environment. Nature 394:69:72.

Rozen, D. E., and R. E. Lenski. 2000. Long-term experimental evolu-tion in Escherichia coli. VIII. Dynamics of a balanced polymor-phism. American Naturalist 155:24–35.

Sniegowski, P. D., P. J. Gerrish, and R. E. Lenski. 1997. Evolution ofhigh mutation rates in experimental populations of Escherichiacoli. Nature 387:703–705.

Stern, D. L. 2013. The genetic causes of convergent evolution. NatureReviews Genetics 14:751–764.

Tenaillon, O., J. E. Barrick, N. Ribeck, D. E. Deatherage, J. L.Blanchard, A. Dasgupta, G. C. Wu, et al. 2016. Tempo and modeof genome evolution in a 50,000-generation experiment. Nature536:165–170.

“The Bill-fish (Scomberesox Storerii), . . . which but fifteen years sindevouring pursuers, has gradually decreased, till at the present time itthe past year there was not one specimen seen at Provincetown.” Fromsachusetts” by James H. Blake (The American Naturalist, 1870, 4:513–

This content downloaded from 132All use subject to University of Chicago Press Term

Travisano, M., J. A. Mongold, A. F. Bennett, and R. E. Lenski. 1995a.Experimental tests of the roles of adaptation, chance, and historyin evolution. Science 267:87–90.

Travisano, M., F. Vasi, and R. E. Lenski. 1995b. Long-term experi-mental evolution in Escherichia coli. III. Variation among replicatepopulations in correlated responses to novel environments. Evolu-tion 49:189–200.

Turner, C. B., Z. D. Blount, and R. E. Lenski. 2015a. Replaying evo-lution to test the cause of extinction of one ecotype in an experi-mentally evolved population. PLoS ONE 10(11):e0142050. doi:10.1371/journal.pone.0142050.

Turner, C. B., Z. D. Blount, D. H. Mitchell, and R. E. Lenski. 2015b.Evolution and coexistence in response to a key innovation in along-term evolution experiment with Escherichia coli. bioRxiv020958. doi:10.1101/020958.

Vasi, F., M. Travisano, and R. E. Lenski. 1994. Long-term experimen-tal evolution in Escherichia coli. II. Changes in life-history traitsduring adaptation to a seasonal environment. American Naturalist144:432–456.

Wang, Y., C. D. Arenas, D. M. Stoebel, K. Flynn, E. Knapp, M. M.Dillon, A. Wünsche, et al. 2016. Benefit of transferred mutationsis better predicted by the fitness of recipients than by their ecolog-ical or genetic relatedness. Proceedings of the National Academyof Sciences of the USA 113:5047–5052.

Weinreich, D. M., N. F. Delaney, M. A. DePristo, and D. L. Hartl.2006. Darwinian evolution can follow only very few mutationalpaths to fitter proteins. Science 312:111–114.

Wichman, H. A., M. R. Badgett, L. A. Scott, C. M. Boulianne, andJ. J. Bull. 1999. Different trajectories of parallel evolution duringviral adaptation. Science 285:422–424.

Wielgoss, S., J. E. Barrick, O. Tenaillon, M. J. Wiser, W. J. Dittmar, S.Cruveiller, B. Chane-Woon-Ming, C. Médigue, R. E. Lenski, andD. Schneider. 2013. Mutation rate dynamics in a bacterial popu-lation balance evolvability and genetic load. Proceedings of theNational Academy of Sciences of the USA 110:222–227.

Wiser, M. J., N. Ribeck, and R. E. Lenski. 2013. Long-term dynamicsof adaptation in asexual populations. Science 342:1364–1367.

Woods, R. J., J. E. Barrick, T. F. Cooper, U. Shrestha, M. R. Kauth,and R. E. Lenski. 2011. Second-order selection for evolvabilityin a large Escherichia coli population. Science 331:1433–1436.

Symposium Editor: Anurag Agrawal

ce I saw stranded on the shore by the thousands, driven in by itshas nearly, if not quite, been driven away, and I think that during“The Habits and Migrations of Some of the Marine Fishes of Mas-

521).

.236.027.111 on July 22, 2017 05:14:46 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Evolutionary Scenarios and Primate Natural History*

Harry W. Greene†

Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, New York 14853

abstract: Scenarios summarize evolutionary patterns and processesby interpreting organismal traits and their natural history correlates ina phylogenetic context. They are constructed by (1) describing pheno-types (including physiology and behavior), ideally with attention to for-mative roles of development, experience, and culture; (2) inferring homol-ogies, homoplasies, ancestral character states, and their transformationswith phylogenetic analyses; and (3) integrating those components withecological and other ancillary data. At their best, evolutionary scenariosare factually dense narratives that entail no known falsehoods; their em-pirical and methodological shortcomings are transparent, they mightbe rejected based on new discoveries, and their potential ideological pit-falls are flagged for scrutiny. They are exemplified here by homoplasticforaging with percussive tools by humans, chimpanzees, capuchins, andmacaques; homoplastic hunting with spears by humans and chimpan-zees; and private experiences (e.g., sense of fairness, grief) among diverseanimals, the homologous or homoplastic status of which often remainsunexplored. Although scenarios are problematic when used to bolsterpolitical agendas, if constructed carefully and regarded skeptically, theycan synthesize knowledge, inspire research, engender public understand-ing of evolution, enrich ethical debates, and provide a deeper historicalcontext for conservation, including nature appreciation.

Keywords: homology, homoplasy, scenarios, tools, weapons, privateexperiences.

Introduction

This paper stems from the 2016 American Society of Natu-ralists Vice Presidential Symposium, for which speakers wereasked to consider homology and homoplasy in the context ofbig questions in biology. Taking that charge literally, and in-spired by recent discoveries in anthropology and primatol-ogy, I first chose: Who are we, and over the long haul, howand why have we come to be this way? These are timelesslyintriguing, often controversial puzzles for anthropologists andevolutionary biologists, as well as for philosophers and otherscholars in the humanities. But how far can we go in pursu-ing answers, and by way of framing a more manageable

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† E-mail: hwg5@cornell.edu.

ORCIDs: Greene, http://orcid.org/0000-0002-2160-9355.

Am. Nat. 2017. Vol. 190, pp. S69–S86. q 2017 by The University of Chicago.0003-0147/2017/190S1-57348$15.00. All rights reserved.DOI: 10.1086/692830

This content downloaded from 132All use subject to University of Chicago Press Term

topic, what roles might be played by phylogenetic system-atics and natural history?As cognitive and cultural organisms, who we are is sub-

stantially contingent on what we do and with whom we as-sociate, rather than simply what we look like, inside and out.How humans came to be as we are is a matter of member-ships in families, societies, andmore ancient cultural and ge-nealogical lineages; why we have certain characteristics thusreflects, at least partly, causal relationships among those com-plex intrinsic and extrinsic influences. All that said, for evo-lutionary and organismal biologists, understanding the whos,hows, and whys of other species, including of our closest kinamong other primates, involves comparing similar attributesamong related taxa—assessing homologies and homoplasiesfor morphological, physiological, and behavioral traits in thecontext of ecological correlates (for a pioneering synthesis,see Brooks andMcLennan 1991; amongmany taxon-focusedexamples: Kappeler and Heymann 1996, for primates; Losos2009, for lizards, Cavender-Bares 2016, for oaks; for broaderperspectives, see Autumn et al. 2002; Losos 2011). The result-ing summaries are narrative in structure and thereby vulner-able to abridgement, selectivity, simplification, and other po-tential distortions (e.g., Landau 1984; O’Hara 1992; Cartmill2002); consequently, because primate evolution involves ourown stories, we should not be surprised that evidentiary andideological controversies plague them (e.g., Caplan 1978; Se-gerstråle 2000; Pinker 2002; Borofsky 2004; Prindle 2009;Marks 2015).Accordingly, this paper explores the content, legitimacy,

and relevance of what Eldredge (1979) termed phylogeneticscenarios, with special reference to primates (for their phylog-eny and classification, see fig. 1). My treatment emphasizescomparative analyses for inferring historical changes in mor-phology, physiology, and behavior (e.g., Greene 1986, p. 3; Au-tumn et al. 2002; Ereshefsky 2007, pp. 668–670; Losos 2011),rather than denying a place for game-theoretical and othercomplementary approaches (e.g., Dugatkin and Reeve 2000;BorgerhoffMulder and Beheim 2011). I first recount the con-ceptualization and critical early reception of what soon cameto be called evolutionary scenarios; second, examine the sta-tus of behavioral, ecological, archaeological, and other data intheir construction; third, exemplify some of their potentialphylogenetic and evidentiary breadth with three case stu-

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S70 The American Naturalist

dies; and finally, delve into their implications for broaderconcerns.

Evolutionary Scenarios Defined,Condemned, and Defended

I use excerpts to reconstruct the origin and early reception ofscenario as a term in evolutionary biology because decades-old references are likely obscure to many readers, some sub-sequent coverage ismisleading, and the original authors’word-ings are instructive. As it happened, this innovation occurredamid the following contexts: (1) claims (Atz 1970; Klopfer1973) and counterclaims (e.g., Greene and Burghardt 1978;Greene 1994, 1999) that behavior cannot be homologized anddoes not evolve (addressed in the next section); (2) Wilson’s(1975) book Sociobiology: The New Synthesis, in which thechapter on humans was only quasi-phylogenetic (see pp. 550–551) and spawned a highly public controversy (e.g., Caplan1978; Segerstråle 2000; Prindle 2009); (3) claims (Lewontin

This content downloaded from 132All use subject to University of Chicago Press Term

1978; Gould and Lewontin 1979; Cracraft 1981) and coun-terclaims (e.g.,Mayr 1983;Wanntorp 1983; BaumandLarson1991; Larson and Losos 1996; Alcock 1998) that adaptation ispoorly substantiated, overemphasized as an evolutionary force,and at best difficult to study from macroevolutionary per-spectives (also discussed in the next section); and (4) the riseof phylogenetic systematics (e.g., Eldredge and Cracraft 1980;Hull 1988), which beyond morphology soon encompassedphysiology, behavior, and ecology (e.g.,Wanntorp 1983; Brooks1985; Donoghue 1989; Brooks and McLennan 1991)—andwas itself a manifestation of the still-ongoing nineteenth-century historicization of biology (de Queiroz 1988; O’Hara1988; Quinn 2016a).These four threads were scarcely entwined, at least overtly,

in the 1970s and early 1980s. However, Gould’s and Lewon-tin’s pejorative framing of adaptationism and the adaptation-ist program occurred in the context of their structuralistperspectives (see below) and disregard for ecology, stronglymotivated as well by Sociobiology’s chapter on humans and

Homo(humans)

other australopithecines

Australopithecus afarensis

Ardipithecus ramidus

Pan(chimpanzees, bonobos)

Gorillinae(gorillas)

Pongidae(orangutans)

Hylobatidae(gibbons)

Cercopithecoidea(Old World monkeys)

Platyrrhini(New World monkeys)

Tarsiiformes(tarsiers)

Strepsirrhini(lemurs, lorises, bushbabies)

Prim

ates

Hap

lorh

ini

Ant

hrop

oide

a

Cat

arrh

ini

Hom

inid

ae

Hom

inin

ae

Hom

inin

i

Figure 1: Phylogenetic relationships (left) and classification (right) of primates, based on Wilson and Reeder (2005), Cartmill and Smith (2011),and Fleagle (2013).

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S71

perhaps earlier concerns about ethological themes in Nazi-erapublications (Alcock 1998; Klopfer 1999; Segerstråle 2000;Gould 2002, pp. 42–43; Burkhardt 2005; Allmon 2009; All-mon et al. 2009; Prindle 2009). Eldredge’s (1979) original con-ceptualization of scenarios cited neitherWilson’s (1975) booknor his critics, but it echoed the latter’s reservations:

[A] phylogenetic scenario [is]. . . a [phylogenetic] treewith an overlay of adaptationist narrative. . . . [A] clad-ogram must precede the construction of a tree . . . a treebefore a scenario. . . . [T]o be debated are (1) the proce-dures for cladogram, tree, and scenario construction and(2) whether or not the additional assumptions . . . [forthemost] complex level of analysis are worth it. (Eldredge1979, p. 168)

Scenarios are inductive narratives (the best are alsoseductive) concocted to explain how some particular con-figuration of events . . . took place. . . . [Their] hallmark[is] . . . the analysis of the adaptive significance of evolu-tionary changes in size, form, and structure. . . . [They]are engrossing and concern themselves with the applica-tion of what we think we know to the real world, as pre-served in the fossil record . . . [but they are] mostly fairytales . . . a maze of untestable propositions concerning se-lection, function, niche utilization, and community inte-gration, and alas, do not generally represent good science.(Eldredge 1979, pp. 192–193)

[Scenarios] are easily capable of testing and refuta-tion. . . . [T]he problem here is that we expect scenar-ios, even more than trees, to be wrong in detail. . . . Thisad hoc changing of hypothesis content rightly infuriatesstrict adherents of hypothetico-deductive methodology inscience and is the feature of scenarios that most horrifiesthem. (Eldredge 1979, p. 194)

A year later, Eldredge’s coauthor of punctuated equilib-rium theory (Eldredge and Gould 1972), after six pages ofpraise for a paleontologist’s fictionalized account of IceAge humans, offered this stinging critique:

Kurtén’s [1980] novel is a more appropriate place thanthe professional literature itself for discussing many ofthe truly scientific issues that swirl about theNeandertal-Cro-Magnon debate. Evolutionary biology has been se-verely hampered by a speculative style of argument thatrecords anatomy and ecology and then tries to constructhistorical or adaptive explanations for why this bonelooks like that or this creature lived here. These specu-lations have been charitably called “scenarios”; they areoften more contemptuously, and rightly, labeled “stories”(or “just-so stories” if they rely on the fallacious assump-

This content downloaded from 132All use subject to University of Chicago Press Term

tion that everything exists for a purpose). Scientists knowthat these tales are stories; unfortunately, they are pre-sented in the professional literature where they are takentoo seriously and literally. Then they become “facts” andenter the popular literature, often in such socially dubiousforms as the ancestral killer ape who absolves us from re-sponsibility for our current nastiness, or as the “innate”male dominance that justifies cultural sexism as the markof nature. (Gould 1980, pp. xv–xvi)

Lewontin, Gould’s antiadaptationist coauthor, was evenmore cynical about understanding, let alone portraying,the deep history of behavior, later opining, “It might be inter-esting to know how cognition (whatever that is) arose andspread and changed, but we cannot know. Tough luck”(Lewontin 1998, p. 130; also see Burghardt 2004). Criticismsof adaptationism, however, were soon countered with concep-tual andmethodological advances (e.g.,Wanntorp et al. 1990;Baum and Larson 1991; Brooks andMcLennan 1991; Brooks1996; Larson and Losos 1996; Alcock 1998; Autumn et al.2002; see below), while taxon-focused biologists elaboratedthe notion of scenarios with nonhuman examples, emphasiz-ing explicit phylogenetic analyses, empirical support, andfalsification of alternative versions (e.g., Greene 1983, 1992;Wake 1992; Desutter-Grandcolas 1995). The term scenarionow is used routinely in diverse contexts (e.g., Sessions et al.2016), including anthropology (e.g., Dominy 2015).Nonetheless, almost four decades after Eldredge defined

scenarios within the context of phylogenetic systematics,these quotations from a prominent anthropologist’s book,Tales of the Ex-Apes: HowWeThink aboutHuman Evolution,illustrate contempt that I suspect is common in some circles:

We can study what a feature does, and we can study howit got there, but to ask what is it for is to decorate thescientific question with a lot of metaphysical accessoriesthat it just doesn’t need . . . to assume that there is a rea-son for it—a deterministic, selective regime for the fea-ture; a particular optimal solution to a problem. But ac-tually, there may be no reasons for some things, justnaturalistic causes and uses, and a lot of random noise;life may be more like clothes than like saws. (Marks2015, p. 62)

[T]hat we can do something does not mean that weevolved to do it, a well-known fallacy known as adapta-tionism. (Marks 2015, p. 99)

Yet even for such a fundamental feature as bipedality,we know far more about how it evolved than why itevolved . . . what parts of the body changed, and howthey changed. . . that’s boring, because it’s anatomy.What we can’t tell you is why it happened. That is his-

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S72 The American Naturalist

tory, that is origin myth, that is interesting. . . . The factthat apes sometimes stand erect when threatening eachother . . . that humans can run longer distances thanapes . . . that having your eyes higher off the groundallows you to see farther—and many others—have beensuggested as scenarios for the evolution of bipedalism.That is to say, it must have been good for something.What these propositions all share is the property of futil-ity. If there was an advantage to bipedalism, we can’t tellwhat it was, from our vantage point of 5 or 6 millionyears later. . . . We have to bracket the question “Whydid we become bipedal?” and set it off from more empiri-cally based scientific discourse. (Marks 2015, pp. 117–118)

Marks’s concerns are evidently imbedded in a more gen-eral disdain for approaches that to him, as with Gould andLewontin (e.g., Segerstråle 2000; Allmon 2009; Prindle 2009),smack of dangerous biological determinism:

The “pop science” origin myth of human evolution ob-serves the genetic intimacy of humans and apes, appliesthe cultural assumption that genetic relationships are themost important relationships, and concludes our identitycan easily be established from our ancestors. But . . . thefact that your ancestors may have been peasants or slavesdoes not make you a peasant or slave. We find the culturalidea of reducing identity to ancestry to be morally repug-nant. The reason is simple, that we are different fromour ancestors, and our identity is established dialectically,recognizing that we are simultaneously both composedof their DNA and yet different from them. (Marks 2015,p. 110)

But, of course, clothes often do have design properties,their advantages contingent on particular environmentaland cultural conditions—try exploring the Arctic in a sunga—and that people are not now slaves does not guarantee theirancestors were not, nor that knowing they once were wouldnot prove useful for, say, socially progressive political advocacy.Moreover, although we might never know precisely why bi-pedalism evolved in our lineage, in terms of particular selec-tive forces, data do constrain the plausibility of alternativeevolutionary scenarios for hominins. Ardi (Ardipithecusramidus) and Lucy (Australopithecus afarensis) are excitingfossils because (assuming the former is a hominin) they con-firm that our lineage, after diverging from one leading tobonobos (Pan paniscus) and chimpanzees (Pan troglodytes),evolved bipedal skeletal attributes before we got big brains,not vice versa (White et al. 2009; Wood and Harrison 2011;Fleagle 2013). And as for the alleged futility of scenarios, wedo have rich paleoecological and archaeological contexts forinterpreting primate evolution, including the inference thatour lineage diversified in more open habitats than those an-

This content downloaded from 132All use subject to University of Chicago Press Term

cestrally occupied by hominins (e.g., Harrison 2011; O’Brienet al. 2013; Werdelin and Lewis 2013; Domínguez-Rodrigo2014).EvenEldredge, in his original formulation, offered a guarded

reprieve and suggestions for progress:

I no longer oppose . . . scenarios (which are, after all,the most fun). . . . As long as we understand preciselywhat we are doing . . . [and have] an adequate grasp ofthe probability we are wrong and what assumptions wehave added along the way, there no longer seems to meto be any reason for anyone to tell anyone else what notto do. (Eldredge 1979, p. 169)

[We can] improve scenarios: (1) by basing themmore explicitly on [phylogenetic] trees and (2) by elimi-nating some of the more purely speculative elements. . . .[A]ny explicit statement about mode or rate of selectionwould be idle and fatuous. . . . Adaptation and especiallynatural selection have no real value in the elaboration ofmost scenarios and are two examples of excess conceptualbaggage. (Eldredge 1979, p. 193)

[Scenarios give] us ideas—lower level hypotheses wemight very well be able to test (and which we might nothave formulated. . .)—which force us to stretch our imag-inations. . . . They are certainly more fun to constructthan a mere tree or a dry cladogram . . . as long as weexplicitly realize how we build scenarios and what theirstatus is as scientific propositions, we should continueto build them, hoping to find them scientifically, as wellas spiritually, uplifting and rewarding. (Eldredge 1979,pp. 194–195)

Gould’s critique likewise ended with a justification forscenarios—but apparently only when they are constructedby evolutionary biologists with opportunities to publishnovels:

Yet these stories have a role in science. They probe therange of alternatives; they channel thought into the con-struction of testable hypotheses; they serve as tentativeframeworks for the ordering of observations. But theyare stories. So why not treat them as such [by presentingthem in fiction], get all the benefits and pleasures, andavoid the wrangles that arise from their usual, inappro-priate placement? (Gould 1980, p. xvi)

I conclude that phylogenetic scenario arose under a cloudof skepticism, owing to concerns about evidence, scientificstatus, and political misuse—the former two including allega-tions that we never can knowmuch about ancestral behaviorsand ecological contexts, the latter with respect to modernhuman affairs. Nonetheless, building on earlier discussions(e.g., Greene 1983, 1992; Wake 1992), as well as Eldredge’s

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S73

(1979) and Gould’s (1980) own musings, the concept canbe expanded as follows: Evolutionary scenarios summarizeinferred patterns and processes by overlaying phenotypesand their ecological correlates on phylogenetic trees and theninterpreting the results in narrative format. At their best,scenarios are factually dense, and unlike Kurtén’s (1980) pre-historical fiction, they entail no unidentified speculation, letalone known falsehoods; their methodological and empiricalshortcomings are obvious, they might be rejected based onnew discoveries, and their ideological pitfalls are flagged forscrutiny. They are constructed by (1) describing phenotypes(including physiology and behavior), with attention, whenpossible and applicable, to formative roles of developmentalgenetics, experience, and culture; (2) distinguishing homol-ogy from homoplasy in the context of a phylogenetic tree,such that ancestral phenotypic states and subsequent trans-formations can be inferred; and (3) integrating the forgoingcomponents with ancillary data (e.g., from paleoecology, ar-chaeology) in a manner that is transparent in terms of facts,uncertainties, and ideological implications.

As such, evolutionary scenarios can range from explicitlyspeculative (e.g., canid-human relationships; Treves and Bo-nancic 2016) to extensively corroborated, from a few para-graphs in length (e.g., alcohol consumption by primates; Do-miny 2015) to monographic, at which point they may belabeled theories (e.g., play: Burghardt 2005; snakes and pri-mates: Isbell 2009;fire and cooking:Wrangham2009;mother-hood: Hrdy 2011; aggression and hunting: Pickering 2013;weaponry and battle: Emlen 2014). I also reckon that care-fully constructed and skeptically regarded, scenarios can playpositive roles in anthropology, biology, and society moregenerally—whether or not they are regarded as scientific,testable, or falsifiable (e.g., Eldredge 1979; Cartmill 2002; Fitz-hugh 2016) and despite risks of political pitfalls (e.g., Gould1980; Klopfer 1999; Burkhardt 2005; Marks 2015). As forcareful construction and skeptical reception, important con-cerns include the nature of resemblances, windows on an-cient behavior and ecology, and macroevolutionary perspec-tives on adaptation.

Elaborating on Three Core Issues

Resemblance, Homology, and Homoplasy

Homology, granting some important conceptual nuances, isresemblance due to common ancestry (e.g., Hall 1994; Mc-Cune and Schimenti 2012). In older literature, it was con-trasted with convergence, the latter earlier called analogy,and often defined as independently evolved similarities dueto similar functional or ecological roles. The rise of phyloge-netic systematics included preference for homoplasy as anoverarching term for independently evolved resemblances,whether due to convergence, parallelism, or reversals; anal-

This content downloaded from 132All use subject to University of Chicago Press Term

ogy has since been widely abandoned as a term in evolution-ary biology (e.g., Eldredge and Cracraft 1980; Sanderson andHufford 1996; Wake et al. 2011; Quinn 2016b; but see Ere-shefsky 2012). Moreover, some researchers prefer that ho-moplasy not be defined with respect to selection, such thatfrom this structuralist (or formalist) standpoint, processesfavoring or constraining independent evolution of similarphenotypes are a matter for separate investigations and in-ference, rather than unsupported assumptions (e.g., Cracraft1981; Wake 1996; Gould 2002, pp. 1076–1081).Evaluating resemblances as homology or homoplasy—

elucidatingphenotypic transformationsduring lineage diver-sification—initially entailed parsimonious congruence of char-acter state data in the context of a cladogram (e.g., Eldredgeand Cracraft 1980; Hall 1994; Sanderson and Hufford 1996);ancestral-state reconstruction now also relies on likelihoodanalyses and hypothesis-testing frameworks, statistical aspectsof which will continue to evolve (e.g., Griffiths et al. 2015).Regardless of analytic techniques, a core task is delimitationof characters and, ideally, their generative mechanisms, suchthat when possible we analyze the phylogeny of ontogenies(deQueiroz 1985)—thus encompassing developmental geno-mics, individual experience, phenotypic plasticity, and, whenapplicable, culture (e.g., Greene 1994; Ereshefsky 2007; Mc-Cune and Schimenti 2012; Foster 2013; Hall 2013; for theontogeny of prehensile-tail use in two platyrrhines, see Bezan-son 2012; of locomotion and postures in wild chimpanzees,see Sarringhaus et al. 2014). Another core challenge is ade-quate phylogenetic sampling, such that component lineagesare neither under- nor overrepresented; in particular, we shouldbe cautious of incorporating only one member (e.g., chim-panzees) of a multispecies sister group as representative ofcharacter states for the ancestor of both lineages (e.g., Bate-man 1996; for hominids, Cartmill and Smith 2011, p. 104;Wood and Harrison 2011; Pickering 2014).An important topic of ongoing exploration, conceptually

and analytically, is phylogenetic analyses of culture and itstangible indicators (material culture; e.g.,Gray et al 2007;Tëm-kin and Eldredge 2007; Collard et al. 2007, 2008, 2011;O’Brienet al. 2012), as well as placing that research within broaderframeworks of gene-culture evolution (e.g., Richerson andBoyd 2005). Of course, individual and cultural experienceshape much of primate behavior, complicating comparativeassessments—toothbrushes are recent inventions, so, parsi-mony notwithstanding, observations of geographically dispa-rate and ethnically diverse people cleaning their teeth with sim-ilar motor patterns need not imply homologous resemblanceto the behavior of ancestral Homo sapiens 1200,000 yearsago. On the other hand, tooth picking with twigs and floss-ing with hair appear sporadically among Old and NewWorldnonhuman anthropoids (McGrew and Tutin 1973; Watanabeet al. 2007; Haslam and Falótico 2015), thus ancient procliv-ities for dental tool use might underlie modern oral hygiene

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S74 The American Naturalist

practices. And in any case, the emergence of nonhuman ar-chaeology raises prospects that we might someday study tran-sitional material cultures among primate lineages (Haslam et al.2009; see also below).

Behavior and Ecology as PhenotypicAttributes with Histories?

Behavior is what animals do and do not do, visibly and oth-erwise; besides external manifestations (e.g., postures, move-ments, vocalizations, immobility), it can include surveillance,cognition, emotions, and other private experiences (Burg-hardt 1997; see below). As such, behavioral systems reflectneurosensory, neuromuscular, neurohormonal, integrative,and mental processes, themselves influenced—as are mor-phology and physiology—in complex ways by genetics, de-velopment, experience, and, in some species, culture (Burg-hardt and Bowers 2017). Although early ethologists treatedmotor patterns comparatively in the same way as anatomy(e.g., Lorenz 1941; Daanje 1950), Atz (1970) asserted thateven those overt, often stereotyped behaviors are too difficultto describe, too variable, too subject to selection and conver-gence, and unsubstantiated by fossils; thus, he argued, behav-ioral homologies are at best difficult to detect and should notbe expected above the level of genera. In fact, none of thesecriticisms can be generalized, unsupported claims to the con-trary notwithstanding (e.g., Klopfer 1999, pp. 124–125), andwell-corroborated behavioral homologies, judged by the samesorts of evidence used for morphology and physiology, char-acterize even some large, ancient clades (e.g., Greene andBurghardt 1978; Wenzel 1992; de Queiroz and Wimberger1993; Greene 1994, 1999; Proctor 1996; Ereshefsky 2007).

The most comprehensive source for inferences about an-cient behavior likely always will be phylogenetic analyses andancestral-state reconstructions, based on comparison amongliving taxa (e.g., oviposition in frogs: Zamudio et al. 2016; so-cial systems in primates: Di Fiore and Rendall 1994; Shultzet al. 2011) and subject to the same limitations and inno-vations as for other phenotypic attributes (e.g., Cooper et al.2016; for hominins, see, e.g., Wood and Harrison 2011). Aswithmorphology andphysiology, character delineation, anal-yses of variation within and among taxa, and phylogeneticsampling breadth are core challenges—namely, detecting reg-ularities and modularity in behavior, avoiding functionallyand ideologically laden terminology, and redressing the scar-city for many taxa of detailed, high-quality descriptions (e.g.,Drummond 1981; Gowaty 1981; Proctor 1996; Greene 1999,2005; Scholes 2008). Important related issues include sourcesof information for ancient and internal behavior, and the sta-tus of ecological data in phylogenetic analyses and scenarioconstruction.

Trace fossils and morphological correlates based on livingtaxa also provide relatively common evidence for the behav-

This content downloaded from 132All use subject to University of Chicago Press Term

ior of extinct animals (Seilacher 2007). Classic examples ofthe former include Jurassic social dinosaur trackways (Ostrom1972) and the 3.7-million-year-old Laetoli footprints, evi-dently made by two adult Australopithecus afarensis with achild in tow (Leakey 1981; White and Suwa 1987). Anatom-ical indicators of ancient behavior are exemplified by associ-ation of an enlarged gyrus in 10-million-year-old otter brainendocasts with that structure and tactile hand sensitivity inextant lutrines (Radinsky 1968); of bipedal skeletal featuresamong hominins (Fleagle 2013, pp. 364–365); and of innova-tion, social learning, and tool use with brain size amongprimates (Reader and Laland 2002). So-called fossilized be-havior deserves skeptical appraisal, however, because animalsusually do not die during routine activities—granted, the pos-ture of a Cretaceous Oviraptor dinosaur over eggs so closelyresembles that of incubating modern birds as likely to reflecthomology (Norell et al. 1995); on the other hand, a fossil earlyamniote with a congener in its throat, rather than indicativeof cannibalism, probably choked attempting atypical tail-firstingestion (Greene 1994). Other records of ancient behaviorcome from prehistoric artist-ethologists (e.g., Guthrie 2005;Greene 2013, p. 236), evidenced by 130,000-year-old repre-sentations of courtship in cave lions (Panthera atrox; Yama-guchi et al. 2004), and for hominines, archaeological sources(for an extended example, see Pickering 2013).With respect to activities that are internal and not ob-

servable directly, recall that Tinbergen (1963) codified fourquestions for ethologists: How does behavior develop, howis it controlled, what is its function or ecological role, andwhatis its evolutionary history? In posing a fifth question—Whatare the private experiences of animals?—Burghardt (1997)called for investigations of cognition, emotion, intention, con-sciousness, awareness, and other internal behaviorwithinTin-bergen’s framework. As such, private experiences might bestudied across taxa by combining critical anthropomorphism(Burghardt 1991, 2016) with laboratory (e.g., Brosnan and deWaal 2003) and field (e.g., Holekamp et al. 2007) experi-ments, as well as with brain imaging, genomics, and otheremerging technologies (Preuss 2012). Private experiencesthereby can be subjected to phylogenetic analysis, and indeeddeWaal (2016a) recently has summarized a career’s worth ofprimate research in that context. Core problems, once again,include adequate phylogenetic sampling and defining char-acters (e.g., for insight, see Shettleworth 2012; for psycho-logical homologies, see Ereshefsky 2007)—the latter perhapsentailing trade-offs between, on the one hand, stringent cri-teria and “killjoy explanations” (Shettleworth 2010), andon the other, more open-minded skepticism and optimisticprospects for detecting trans-species mental homologies (e.g.,Cartmill 2000; Rivas and Burghardt 2002).Ecological correlates of behavior, whether biotic (e.g., pred-

ators, competitors) or abiotic (e.g., climate, soil type), arenot themselves attributes of organisms, passed from genera-

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S75

tion to generation. They thus arguably should not be ana-lyzed phylogenetically (Grandcolas et al. 2011), although pro-pensities to choose particular prey and microhabitats, forexample, might be transmitted genetically and/or culturally,such that they can be treated as behavioral traits (e.g.,Burghardt 1967). Likewise, a phylogenetic analysis of mam-mals in which incidence of lethal violence, a population-leveltrait, served as proxy for aggression (Gómez et al. 2016) mightprove controversial, the more so because that word is subjectto diverse, sometimes ambiguous usage. In any case, avail-ability of ecological data for interpreting human evolutioncontinues to grow, encompassing not only more traditionalarchaeological and paleontological methods (e.g., Harrison2011; Pickering 2013) but also stable isotope analyses andother technological innovations (e.g., Cerling et al. 2013;Yeaves et al. 2013; Katoh et al. 2016).

Adaptationist “Stories” as Macroevolutionary Hypotheses

Even today some might argue, paraphrasing Brown (1982,p. 886), that most evolutionary biologists are interested innatural populations instead of fossils, in variation among in-dividuals, rather than among species and higher taxa. I be-lieve instead that many anthropologists and biologists, as wellas members of the public, are fascinated by big questions likethose that began this essay and, consequently, will be inter-ested in evolutionary scenarios. A key issue, then, is just whatcan we infer about adaptations as historical phenomena, asopposed to reflecting natural selectionwithin populations to-day? Three points bear emphasis.

First, behavioral character states can be characteristic ofspecies and higher taxa, rather than variable within popula-tions (e.g., Greene 1994, 1999; Proctor 1996); they thus canbe corroborated as homologous or homoplastic at deeperlevels in evolutionary history and used to construct scenarios(e.g., Greene 1986, 1992; Coddington 1988; for social behav-ior in primates, see Shultz et al. 2011). Moreover, as withother phenotypic attributes, hypotheses of homology andhomoplasy are falsifiable if their constituent claims proveinconsistent with new evidence frompaleontology, archaeol-ogy, and other ancillary sources (e.g., Greene 1986; Larsonand Losos 1996; Wood and Harrison 2011).

Second, inferring natural selection from phylogenetic pat-terns depends on evidence for its general prevalence as aforce in evolution and resulting assumptions about pastpopulation-level processes. Some evolutionary biologists be-lieve selection must have caused most if not all phyloge-netic correlations between phenotypes and ecological circum-stances (e.g., Mayr 1983; Coddington 1988; Baum and Larson1991; Losos 2011), while others are more minimalistic (e.g.,Cracraft 1981; Greene 1986; Autumn et al. 2002)—in anycase, selection evidently underlies numerous well-studied ex-amples of homoplasy (e.g., Losos 2011; Langin et al. 2015).

This content downloaded from 132All use subject to University of Chicago Press Term

Third, irrespective of whetherwe infer that selection shapedancient adaptations, widespread correlations of homoplasieswith shifts in functional and ecological contexts imply law-like evolutionary patterns (e.g., Brooks 1996; Losos 2011; Ag-rawal 2017)—the expensive tissue hypothesis, for example,first proposed for hominins, is supported by a correlation be-tween large brains and small guts in amphibians (Liao et al.2016). Moreover, phylogenetic analyses can reveal whethernovel forms, functions, and ecological roles arose simulta-neously or sequentially, thus distinguishing between adap-tive and exaptive explanations (Gould andVrba 1982; Greene1986). Conversely, absence of homoplasymight reflect lineage-wide constraints (e.g., life history and socioecological traitsof primates; Kappeler and Heymann 1996). And althoughCartmill (2002) argued that prospects are slim for homo-plasies revealing much about humans, some astonishingwindows on behavioral evolution in primates have openedduring the past decade, as illustrated below.

Evolutionary Scenarios and (Mainly) Primate Behavior

Percussive Tools and Extractive Foraging

Bonobos and chimpanzees compose our living sister groupwithin hominine catarrhines, and their use of tools has beenincreasingly appreciated over the past half century (e.g.,Goodall 1964; McGrew 1992). Recent discoveries, however,reveal that although bonobos rarely use stone hammers andanvils for extractive foraging (Roffman et al. 2015), chimpsshow complexity and cultural variation in that behavior (e.g.,Haslam et al. 2009; Lycett et al. 2009). Moreover, percussivetool use also occurs in a cercopithecoid catarrhine, the Bur-mese island subspecies of long-tailed macaque (Macacafascicularis aurea), which feeds on nuts and shellfish (e.g.,Carpenter 1887; Gumert et al. 2009), and in a platyrrhine,the bearded capuchin (Sapajus libidinosus), which harvestsfruits and nuts (fig. 2A; e.g., Fragaszy et al. 2004; Falóticoand Ottoni 2016; Luncz et al. 2016). Similar behavior thushas originated four times among primates, including withinhominins; most intriguingly, gracile (Cebus) and robust (Sa-pajus) capuchins diverged several million years ago, and thelatter sometimes inhabit savanna woodlands, have enlargedbrains, and excel at tool use, reminiscent of the Pan-Homosplit (Haslam et al. 2009; Lynch Alfaro et al. 2012). Each ofthe three nonhuman species is subject to ongoing research,yielding rich comparative details about ontogeny, controlmechanisms, ecological roles, and cultural transmission ofpercussive tool use (e.g., Visalberghi et al. 2015; Cardoso andOttoni 2016).Material culture now confirms percussive tools for an un-

identified (likely pre-Homo) hominin 3.3 million years ago(Harmand et al. 2015), chimps ∼4,300 years ago (Mercadoret al. 2007), Burmesemacaques1100 years ago (Haslam et al.

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

AB

Figu

re2:

A,B

earded

capu

chin

withston

ehammer,p

oisedto

crackanu

t.B,N

onhum

anprim

atearchaeologicalsitein

Brazil,withston

ehammer

insitu

(arrow

),used

1600yearsagoby

abeardedcapu

chin.Pho

tocredits:P.Oxford(A

);M.Haslam

(B).

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:50 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S77

2016a), and bearded capuchins 1600 years ago (fig. 2B; Has-lam et al. 2016b). A deeper, denser archaeological recordmightreduce independent origins to three if the common ancestorof Pan and Homo used percussive tools (Hoevers 2015),to two if ancestral catarrhines were implicated, or even—asseems unlikely given the phylogenetic distance among thosetaxa—to one if percussive tools characterized the ancestorof catarrhines and platyrrhines. Moreover, developmentalstudies of that behavior might reveal homologous precur-sors across anthropoids or an even larger clade (e.g., stonehandling: Huffman et al. 2010; object manipulation and/orcognitive abilities: Hoevers 2015), such that percussive tooluse could have arisen as multiple parallelisms, perhaps eachtime encouraged by particular ecological conditions (e.g.,terrestriality; Meulman et al. 2012).

In any case, when bearded capuchins pound stones to-gether, perhaps for nutritional powder, they produce sharp-edged flakes that are indistinguishable from early hominintools (Proffitt et al. 2016). This discovery suggests that stonesinadvertently flaked by nonhuman primates might be erro-neously interpreted as tools produced by hominins, as wellas that early hominins could have adopted similar incidentalflakes as prototools, including projectiles, in a foraging con-text. Also, because large brains and problem solving are phylo-genetically correlated among primates and carnivores (Readerand Laland 2002; Benson-Amram et al. 2016), similar under-lying cognitive innovationsmight bemore broadly homoplas-tic among mammals, including for percussive tool use (e.g.,for sea otters; Fujii et al. 2015).

Lances and Javelins

Probing with sticks (e.g., for termites) and aiming projectiles(e.g., stones) are widespread practices among anthropoids,but fashioned weapons (modified to inflict bodily injury) werelong believed restricted to our lineage, as documented by a richarchaeological and historical record (e.g., Pickering 2013;Emlen 2014). Spears are long-shafted, penetrating weapons,of which themost ancient known for hominins are∼500,000-year-old hafted, stone-tipped lances (thrusting spears; Wilkinset al. 2012); a 400,000-year-old wooden spear tip (deliverymode unknown; Schoch et al. 2015); 300,000-year-oldwoodenlances and javelins (light throwing spears; fig. 3A; Thieme 1997;Schoch et al. 2015); and 280,000-year-old hafted, stone-tipped throwing spears (Sahle et al. 2013). The ancient jav-elins are remarkable for having been made with stone toolsfrom trunks of small spruce (Picea sp.) and pine (Pinussylvestris) trees, with their shaft and tips carved offset fromthe softer central axis of the wood and then polished (Schochet al. 2015). Nonetheless, savanna-woodland chimpanzeesin Senegal break off selected saplings, roughly sharpen anend with their teeth, and use those lances to puncture their

This content downloaded from 132All use subject to University of Chicago Press Term

prey, lesser bushbabies (Galago senegalensis), in tree cavities(fig. 3B; Roach 2008; Pruetz et al. 2015). Also intriguingly,bearded capuchins break off and thin the tips of sticks, withwhich they probe for honey and small prey animals, as wellas jab at dangerous snakes and their mimics (e.g., a false jara-raca pit viper [Xenodon merremi]; fig. 4; Falótico and Ottoni2014).Once again, character descriptions and nomenclature are

central to phylogenetic analyses and interpretation, and ac-cording to Roach (2008) and Pickering (2013), some otherresearchers were critical of Pruetz and Bertolani’s (2007)calling the savanna-woodland chimps’modified tools spearsand their behavior hunting (perhaps because bushbabies of-ten are killed by females, and prey are small compared to thecolobus monkeys [Piliocolobus sp.] taken by forest-dwellingchimps). Like Roach (2008) and Pickering (2013), I regardboth words, employed conventionally, as appropriate for theweapons made by early humans and by chimps, whereas thethinned sticks with which bearded capuchins poke (but evi-dently do not pierce) snakes can be termed protolances. Otherplausible antecedents of defensive weapons in hominins, de-serving of further study in that context (see Crabb and Elizaga2008), include unfashioned stick clubs and projectiles usedby free-living white-faced capuchins (Cebus capucinus) againsta terciopelo pit viper (Bothrops asper; Boinski 1988) and bysemifree tufted capuchins (Sapajus sp.) against a leopard tor-toise (Stigmochelys pardalis; Hamilton and Fragaszy 2014).Spear making has arisen repeatedly in primate history,

yielding lances, javelins, and hafted projectiles withinHomo;lances in savanna-woodland chimps; and, arguably, proto-lances in bearded capuchins. In all three instances, thoseweapons increase the distance between primates and prey orotherwise dangerous adversaries, diminishing the risk of fatalretaliation and fostering emotional distance (Pickering 2013).Like human hunters, savanna-woodland chimps appear in-tently composed while trying to skewer their victims, whereasforest chimps are highly agitatedwhile hand-capturingmon-keys—a difference used to argue that hunting and aggressionwere decoupled early in hominin evolution (Pickering 2013).Furthermore, also consistent with the savanna-woodland hy-pothesis of human ancestry (e.g., Pickering and Domínguez-Rodrigo 2010; Fleagle 2013; Pickering 2014), chimps in thathabitat have converged with early Homo by taking shelterin caves, behaving as if familiar with fire, and at least to myeye, being more sparsely haired than forest-dwelling individ-uals of their species (Pruetz and LaDuke 2010; see images inRoach 2008).

Private Experiences

Internal activities are increasingly subject to phylogeneticperspectives, involving considerations of, among other be-haviors, consciousness (e.g., Cartmill 2000), cognition (e.g.,

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

A

B

Figure 3: A, Wooden spear and horse bones from Germany, dated at ∼300,000 years ago; note pointed tip (arrow, upper right) and thickerarea (arrow, center) typical of javelins. B, Field sample of lances fashioned by chimpanzees for hunting bushbabies in Senegal. Photo credits:P. Pfarr, State Agency for the Heritage of Lower Saxony (A); J. Pruetz (B).

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:50 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S79

MacLean et al. 2012), emotion (e.g., Clark 2013), and musi-cality (Honig and Ploeger 2012) in diverse animals; of socialintelligence among primates and carnivores (Reader and Laland2002; Holekamp et al. 2007; Benson-Amram et al. 2016); offairness, empathy, and other putative precursors of moral-ity among anthropoids (Brosnan and deWaal 2003; deWaal2016a, 2016b); and of psychological attributes more gener-ally (Ereshefsky 2007). Although such studies typically ad-dress private experiences with reference to homology and ho-moplasy, central to understanding behavioral evolution, thatperspective is not yet true of popular writing on the subject.

Among four recent widely and favorably reviewed books,only Are We Smart Enough to Know How Smart Animals

This content downloaded from 132All use subject to University of Chicago Press Term

Are? (de Waal 2016a) mentions homology; that word andhomoplasy are lacking from the indices and, as far as I cantell, otherwise unmentioned inAnimal Wise: HowWe KnowAnimals Think and Feel (Morrell 2013),HowAnimals Grieve(King 2013), and Beyond Words: What Animals Think andFeel (Safina 2015). Whether thinking, grieving, and feelingmight not be the same behaviors across species as phyloge-netically disparate as whales, elephants, and primates—andthus, perhaps, not have shared a common evolutionary der-ivation—is variously explicit (e.g., King 2013, pp. 7–8) or notin the other volumes. And although those authors are moti-vated by conservation and animal welfare concerns (as am I),they do not grapple in detail with how whether private expe-

Figure 4: Bearded capuchin jabbing with protolance (upper arrow) at plausibly dangerous pit viper mimic (lower arrow) in a rock crevice(from video). Photo credit: T. Falótico.

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S80 The American Naturalist

riences in other species are homologous or homoplastic withour ownmight affect ethical debates on those topics (see, e.g.,Cartmill 2000).

Beyond matters of definition and historical explanation,robust evolutionary scenarios for private experiences will re-quire denser and broader sampling than currently available.In terms of denser representation among relatively basal pri-mates, given logistical and ethical restrictions on captive re-search, field experiments might more likely yield new in-sights (e.g., for strepsirrhines: Huebner and Fichtel 2015; fortarsiers: Gursky 2005; for a cercopithecoid: Isbell and Etting2017). As for phylogenetic breadth, many of the species dis-cussed by King (2013), Morrell (2013), Safina (2015), andde Waal (2016a) are mammals, and with more distant taxait becomes more difficult to imagine their inner worlds, letalone experimentally address them. However, if we could ac-cess the private experiences of eyelid-less, limbless creatureslike snakes—“in which we detect no joy and no emotion”(Skutch 1980, p. 257)—that effort might inspire new waysto illuminate more familiar and accessible species (Burghardt1977, 1991).

Discussion: Now What?

Almost four decades after Eldredge (1979) conceptualizedevolutionary scenarios, I judge that their shortcomings havebeen substantially overstated. With respect to the deep his-tory of primates, including of our own species, we have anever-growing empirical basis for studying the evolution ofmorphology, physiology, and behavior in the context of ecol-ogy and archaeology—but then again, back in the 1970s and1980s, who would have imagined material culture for non-human primates dating back more than four millennia orthat savanna-woodland chimps fashion crude spears? Now,as a generalization and granting further refinements of ana-lytic methods, scenario construction seems unproblematic—limited primarily by lack of phylogenies and natural historydetails for many taxa, and, for better or worse, by restrictionson invasive neurobiological studies of primates (Preuss 2012).

Nevertheless, concerns that evolutionary scenarios can beused to support ideological and political agendas sometimesare justified (e.g., Klopfer 1999; Marks 2015). As narratives,scenarios indeed are vulnerable to distortions and self-servingrhetoric, including by critics; they also are susceptible to in-terpretive excesses and lapses (for verbatim examples, seePrindle 2009, pp. 135–138), as well as faulty interpretationbecause of obscure terminology and concepts. Christina Cau-terucci’s (2016) blog post asks, for example, “Why dowe idol-ize chimps when we could be imitating female bonobos?”—but even setting aside why wemight aspire to a close relative’ssocial system (e.g., Cartmill 2002), how many lay readers willunderstand that we are equally related to bonobos and chimps,such that as a generalization and all else equal, neither species

This content downloaded from 132All use subject to University of Chicago Press Term

is better or worse as a model for thinking about ourselves?(For a critical overview favoring chimps, see Pickering andDomínguez-Rodrigo 2010.) In fact, robustly inferring ances-tral states for hominins could require observations of addi-tional taxa, including extinct species, and for some attributesmight never be possible (Wood and Harrison 2011). Mini-mizing political problems surely will follow, at least in part,from transparency in scenario presentation with regard toshaky underpinnings, as well as from admissions of bias bycritics and proponents of diverse viewpoints (for extendeddiscussions, see Borofsky 2004; Prindle 2009; Dreger 2011).As even Eldredge (1979) andGould (1980) acknowledged,

and as exemplified by the above examples, scenarios can playpositive roles in anthropology, biology, and society moregenerally. By highlighting homologies and homoplasies inthe diversification of clades, we focus on the controversialextent to which adaptation, constraint, stasis, and repeatedevolution of sameness have characterized life’s history (e.g.,Brooks 1996; Kappeler and Heymann 1996; Wake 1996;Losos 2011; Wake et al. 2011; Edwards and Donoghue 2013;Bridgham 2016; Agrawal 2017). Moreover, by laying out de-tailed case studies of Darwin’s descent with modification,scenarios can inspire public understanding of evolution andour long-term roles in the current extinction crisis, as wellas inform management strategies and appreciation for bio-diversity (e.g., Brooks and McLennan 2010; Greene 2013,pp. 183–187; Werdelin 2013; Cavender-Bares 2016; Sullivanet al. 2017)—wemight well regard island macaques, beardedcapuchins, and savanna-woodland chimps as especially wor-thy of conservation, given their homoplastic resemblances tohominins, the possibility of deeper homologies shared amongus, and thus the contributions of those other primates to illu-minating our own past. Likewise, evolutionary scenarios fora sense of fairness, grief, and other private experiences, byclarifying the nature and origins of those phenomena, couldenrich our deliberations about ethical treatment of otherspecies.With that last point inmind and a nod to deWaal’s (1999)

notion of anthropodenial—“blindness to the human-like char-acteristics of animals, or the animal-like characteristics ofourselves” (p. 258)—I will close with two koan-like tensions.First, some anthropologists and biologists, myself included,strive to confirm private experiences similar to our own inother species, all the while yearning to justify treating thembetter (e.g., Greene et al. 2002, pp. 198–199; King 2013; Mor-rell 2013; Safina 2015; de Waal 2016a). Other scholars (andperhaps some of the same ones) insist on human exception-alism and minimize evolution’s relevance for knowing our-selves (e.g., Lewontin 1998; Klopfer 1999; Marks 2015). Itis as if we could be inspired by two metrics, yet ignore theirfundamental irreconcilability—onewherebynonhuman ani-mals resemble us ever more closely with each new discovery,the other with which, no matter those revelations, we always

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S81

must be profoundly unlike even bonobos and chimps. Sec-ond, having spent a career advocating for snakes, I long haveregarded their persecution as ignorant, gratuitous, and shame-ful—especially by the likes of Albert Schweitzer (who fa-mously “revered all life”) and Alexander Skutch (an acclaimedornithologist), both of whom routinely killed those reptiles(Greene 2013, pp. 125, 181).Now, though, I ponder suchneg-ative biases, my own motivations, and prospects for con-serving serpents in the light of some 75million years of mor-tal conflict between our two lineages (Isbell 2009; Headlandand Greene 2011). Well-corroborated scenarios of primateevolution, however uncomfortable to contemplate, will likelyplay important roles in clarifying these and other contempo-rary dilemmas.

Note Added in Proof

While this article was in press, I encountered Tattersall andEldredge’s (1977) earlier formulation of scenario in a phylo-genetic context; although not cited by Eldredge (1979), myquotes herein from the latter reference faithfully reflect viewsadvanced in the former.

Acknowledgments

The American Naturalistwas the journal I most eagerly readas a graduate student andmost aspired to publish in as a soleauthor, and I am grateful to A. Agrawal for organizing thesymposium that finally led to realizing that goal. I also thankG. M. Burghardt, my PhD advisor, for more than 40 years ofmentorship; D. B. Wake and M. H. Wake for their friend-ship, scholarly examples, and support at Berkeley; W. D.Allmon,W. Bemis,M. Bezanson, C. J. Campbell, D. Cundall,K. de Queiroz, F. B. de Waal, H. Drummond, T. Falótico,T. N. Headland, K. R. Hill, D. M. Hillis, K. E. Holekamp,L. A. Isbell, R. E. Keen, B. J. King, J. B. Losos, K. C.MacKinnon, J. Marks, G. Mayer, A. R. McCune, G. B. Pauly,T. R. Pickering, J. Pruetz, A. Quinn, J. P. Sparks, P. Weldon,A. D. Yoder, and K. R. Zamudio for feedback and/or inspira-tion; E. Behrens, R. Bish, T. Falótico, M. Haslam, G. Legant,N. A. Mason, A. Meyer, R. C. Nali, P. Oxford, J. Pruetz, andT. Terberger for assistance with figures; an anonymous re-viewer for cogent criticisms; and the National Science Foun-dation for support from Opportunities for Promoting Un-derstanding through Synthesis grant 1354156.

Literature Cited

Agrawal, A. A. 2017. Toward a predictive framework for convergentevolution: integrating natural history, genetic mechanisms, andconsequences for the diversity of life. American Naturalist 190(suppl.):S1–S12.

Alcock, J. 1998. Unpunctuated equilibrium in the natural history essaysof Stephen Jay Gould. Evolution and Human Behavior 19:321–336.

This content downloaded from 132All use subject to University of Chicago Press Term

Allmon, W. D. 2009. The structure of Gould: happenstance, human-ism, history, and the unity of his view of life. Pages 3–68 in W. D.Allmon, P. H. Kelley, and R. M. Ross, eds. Stephen Jay Gould: re-flections on his view of life. Oxford University Press, Oxford.

Allmon, W. D., P. J. Morris, and L. C. Ivany. 2009. A tree growsin Queens: Stephen Jay Gould and ecology. Pages 147–170 inW. D. Allmon, P. H. Kelley, and R. M. Ross, eds. Stephen Jay Gould:reflections on his view of life. Oxford University Press, Oxford.

Atz, J. 1970. The application of the idea of homology to behavior.Pages 53–74 in L. R. Aronson, E. Tobach, D. S. Lehrman, and J. S.Rosenblatt, eds. Development and evolution of behavior. Freeman,San Francisco.

Autumn, K., M. J. Ryan, and D. B. Wake. 2002. Integrating historicalandmechanistic biology enhances the study of adaptation. QuarterlyReview of Biology 77:383–408.

Bateman, R. M. 1996. Nonfloral homoplasy and evolutionary scenariosin living and fossil land plants. Pages 91–130 inM. J. Sanderson andL. Hufford, eds. Homoplasy: the recurrence of similarity in evolution.Academic Press, San Diego, CA.

Baum, D. A., and A. Larson. 1991. Adaptation reviewed: a phylogeneticmethodology for studying character macroevolution. Systematic Zo-ology 40:1–18.

Benson-Amram, S., B. Dantzer, G. Stricker, E. M. Swanson, and K. E.Holekamp. 2016. Brain size predicts problem-solving abilities inmammalian carnivores. Proceedings of the National Academy ofSciences of the USA 113:2532–2537.

Bezanson, M. 2012. The ontogeny of prehensile-tail use in Cebuscapucinus and Alouatta palliata. American Journal of Primatology74:770–782.

Boinski, S. 1988. Use of a club by a white-faced capuchin (Cebus ca-pucinus) to attack a venomous snake (Bothrops asper). AmericanJournal of Primatology 14:177–179.

Borgerhoff Mulder, M., and B. A. Beheim. 2011. Understanding thenature of wealth and its effects on human fitness. PhilosophicalTransactions of the Royal Society B 366:344–356.

Borofsky, R., ed. 2004. Yanomami: the fierce controversy. Universityof California Press, Berkeley.

Bridgham, J. T. 2016. Predicting the basis of convergent evolution.Science 354:289.

Brooks, D. R. 1985. Historical ecology: a new approach to studying theevolution of ecological associations. Annals of the Missouri Botan-ical Garden 72:660–680.

———. 1996. Explanations of homoplasy at different levels of bio-logical organization. Pages 3–36 inM. J. Sanderson and L. Hufford,eds. Homoplasy: the recurrence of similarity in evolution. AcademicPress, San Diego, CA.

Brooks, D. R., and D. A. McLennan. 1991. Phylogeny, ecology, andbehavior: a research program in comparative biology. University ofChicago Press, Chicago.

———. 2010. The biodiversity crisis: lessons from phylogenetic sagas.Evolution Education and Outreach 3:558–562.

Brosnan, S. F., and F. B. M. deWaal. 2003. Monkeys reject unequalpay. Nature 425:297–299.

Brown, J. L. 1982. The adaptationist program. Science 217:884, 886.Burghardt, G. M. 1967. Chemical-cue prey preferences of inexperi-enced snakes: comparative aspects. Science 157:718–721.

———. 1977. Of iguanas and dinosaurs: social behavior and com-munication in neonate reptiles. American Zoologist 17:177–190.

———. 1991. Cognitive ethology and critical anthropomorphism: asnake with two heads and hognose snakes that play dead. Pages 53–

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S82 The American Naturalist

90 in C. A. Ristau, ed. Cognitive ethology: the minds of other ani-mals. Lawrence Erlbaum, Hillsdale, NJ.

———. 1997. Amending Tinbergen: a fifth aim for ethology. Pages254–276 in R. W. Mitchell, N. S. Thompson, and H. L. Miles, eds.Anthropomorphism, anecdotes, and animals. State University ofNew York Press, Albany.

———. 2004. Play: how evolution can explain the most mysteriousbehavior of all. Pages 231–246 in A. Moya and E. Font, eds. Evo-lution: from molecules to ecosystems. Oxford University Press, Ox-ford.

———. 2005. The genesis of animal play: testing the limits. MIT Press,Cambridge, MA.

———. 2016. Mediating claims through critical anthropomor-phism: commentary on Key on fish pain. Animal Sentience 2016.024:1–4.

Burghardt, G. M., and R. I. Bowers. 2017. From instinct to behavior sys-tems: an integrated approach to ethological psychology. Pages 333–364 in J. Call, ed. APA handbook of comparative psychology. Vol 1.Basic concepts, methods, neural substrate, and behavior. AmericanPsychological Association, Washington, DC.

Burkhardt, R. W., Jr. 2005. Patterns of behavior: Konrad Lorenz, NikoTinbergen, and the founding of ethology. University of ChicagoPress, Chicago.

Caplan, A. L., ed. 1978. The sociobiology debate: readings on the eth-ical and scientific issues concerning sociobiology. Harper and Row,New York.

Cardoso, R. M., and E. B. Ottoni. 2016. The effects of tradition on prob-lem solving by twowild populations of bearded capuchinmonkeys ina probing task. Biology Letters 12:20160604.

Carpenter, A. 1887. Monkey opening oyster. Nature (May 19):53.Cartmill, M. 2000. Animal consciousness: some philosophical, method-ological, and evolutionary problems. American Zoologist 40:835–846.

———. 2002. Paleoanthropology: science or mythological charter.Journal of Anthropological Research 58:183–201.

Cartmill, M., and F. H. Smith. 2011. The human lineage. Wiley Black-well, New York.

Cauterucci, C. 2016. Why do we idolize chimps when we could be im-itating female bonobos? Slate: XX Factor: What women really think(Sept. 14). http://www.slate.com/blogs/xx_factor/2016/09/14/why_do_we_idolize_chimps_when_we_could_be_imitating_feminist_bonobos.html. Accessed October 16, 2016.

Cavender-Bares, J. 2016. Diversity, distribution and ecosystem servicesof the North American oaks. International Oaks 27:37–48.

Cerling, T. E., F. K. Manthi, E. N Mbjua, L. N. Leakey, M. G. Leakey,R. E. Leakey, F. H. Brown, et al. 2013. Stable isotope-based dietreconstructions of Turkana Basin hominins. Proceedings of the Na-tional Academy of Sciences of the USA 110:10501–10506.

Clark, J. A. 2013. Intersections between development and evolutionin the classification of emotions. Development and Psychobiology55:67–75.

Coddington, J. A 1988. Cladistic tests of adaptational hypotheses.Cladistics 4:3–22.

Collard,M., B. Buchanan, A. Ruttle, andM. J. O’Brien. 2011. Niche con-struction and the toolkits of hunter-gatherers and food gatherers.Biological Theory 6:251–259.

Collard, M., and S. J. Shennan. 2008. Patterns, process, and parsimony:studying cultural evolution with analytic techniques from evolution-ary biology. Pages 17–33 in M. T. Stark, B. J. Bower, and L. Horne,

This content downloaded from 132All use subject to University of Chicago Press Term

eds. Cultural transmission andmaterial culture: breaking downbound-aries. University of Arizona Press, Tucson.

Collard, M., S. J. Shennan, B. Buchanan, and R. A. Bentley. 2007. Evo-lutionary biological methods and cultural data. Pages 203–223 inR. A. Bentley, H. D. G. Maschner, and C. Chippindale, eds. Hand-book of archaeological theories. Altamira, Lanham, MD.

Cooper, N., G. H. Thomas, and R. G. FitzJohn. 2016. Shedding light onthe “dark side” of phylogenetic comparative methods. Methods inEcology and Evolution 7:693–699.

Crabb, P. B., and A. Elizaga. 2008. The adaptive value of tool-aided de-fense against wild animal attacks. Aggressive Behavior 34:633–638.

Cracraft, J. 1981. The use of functional and adaptive criteria in phy-logenetic systematics. American Zoologist 21:21–36.

Daanje, 1950. On locomotor movements in birds and the intentionmovements and behavior derived from them. Behaviour 3:49–98.

de Queiroz, A., and P. H. Wimberger. 1993. The usefulness of behav-ior for phylogeny estimation: levels of homoplasy in behavioral andmorphological characters. Evolution 47:46–60.

de Queiroz, K. 1985. The ontogenetic method for determining char-acter polarity and its relevance to phylogenetic systematics. System-atic Zoology 34:280–299.

———. 1988. Systematics and the Darwinian revolution. Philosophyof Science 55:238–259.

Desutter-Grandcolas, L. 1995. Toward the knowledge of the evolution-ary biology of phalangopsid crickets (Orthoptera: Grylloidea: Phal-angopsidae): data, questions, and evolutionary scenarios. Journal ofOrthopteran Research 1995(4):163–175.

de Waal, F. 2016a. Are we smart enough to know how smart animalsare? W. W. Norton, New York.

———. 2016b. Apes know what others believe. Science 254:39–40.de Waal, F. B. M. 1999. Anthropomorphism and anthropodenial: con-sistency in our thinking about humans and other animals. Philosoph-ical Topics 27:255–280.

Di Fiore, A., and D. Rendall. 1994. Evolution of social organization: areappraisal for primates by using phylogenetic methods. Proceedingsof the National Academy of Sciences of the USA 91:9941–9945.

Domínguez-Rodrigo, M. 2014. Is the “savannah hypothesis” a deadconcept for explaining the emergence of the earliest hominins? Cur-rent Anthropology 55:59–81.

Dominy, N. J. 2015. Ferment in the family tree. Proceedings of theNational Academy of Sciences of the USA 112:308–309.

Donoghue, M. J. 1989. Phylogenies and the analysis of evolutionary se-quences, with examples from seed plants. Evolution 43:1137–1156.

Dreger, A. 2011. Darkness’s descent on the American AnthropologicalAssociation: a cautionary tale. Human Nature 22:225–246.

Drummond, H. 1981. The nature and descriptions of behavior patterns.Pages 1–33 in P. P. G. Bateson and P. Klopfer, eds. Perspectives inethology. Vol. 4. Plenum, New York.

Dugatkin, L. A., and H. K. Reeve, eds. 2000. Game theory and animalbehavior. Oxford University Press, Oxford.

Edwards, E. J., and M. J. Donoghue. 2013. Is it easy to move and easyto evolve? evolutionary accessibility and adaptation. Journal of Ex-perimental Botany 64:4047–4052.

Eldredge, N. 1979. Cladism and common sense. Pages 165–198 inJ. Cracraft and N. Eldredge, eds. Phylogenetic analysis and paleon-tology. Columbia University Press, New York.

Eldredge, N., and J. Cracraft. 1980. Phylogenetic patterns and the evo-lutionary process: method and theory in comparative biology. Co-lumbia University Press, New York.

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S83

Eldredge, N., and S. J. Gould. 1972. Punctuated equilibria: an alter-native to phyletic gradualism. Pages 193–223 in T. J. M. Schopf,ed. Models in paleobiology. Freeman Cooper, San Francisco.

Emlen, D. J. 2014. Animal weapons: the evolution of battle. H. Holt,New York.

Ereshefsky, M. 2007. Psychological categories as homologies: lessonsfrom ethology. Biology and Philosophy 22:659–674.

———. 2012. Homology thinking. Biology and Philosophy 27:381–400.

Falótico, T., and E. B. Ottoni. 2014. Sexual bias in probe tool man-ufacture and use by wild bearded capuchin monkeys. BehaviouralProcesses 108:117–122.

———. 2016. The manifold use of pounding stone tools by wild ca-puchin monkeys of Serra da Capivara National Park, Brazil. Behav-iour 153:421–442.

Fitzhugh, K. 2016. Phylogenetic hypotheses: neither testable nor fal-sifiable. Evolutionary Biology 43:257–266.

Fleagle, J. G. 2013. Primate adaptation and evolution. 3rd ed. AcademicPress, San Diego, CA.

Foster, S. A. 2013. Evolution of behavioural phenotypes: influencesof ancestry and expression. Animal Behaviour 85:1061–1075.

Fragaszy, D., P. Izar, E. Visaberghi, E. B. Ottoni, and M. Gomes deOliveira. 2004. Wild capuchin monkeys (Cebus libidinosus) use an-vils and stone pounding tools. American Journal of Primatology64:359–366.

Fujii, J. A., K. Ralls, and M. T. Tinker. 2015. Ecological drivers of var-iation in tool-use frequency across sea otter populations. BehavioralEcology 26:519–526.

Gómez, J. M., M. Verdú, A. Gonzáles-Megías, and M. Méndez. 2016.The phylogenetic roots of human lethal violence. Nature 538:233–237.

Goodall, J. 1964. Tool-using and aimed throwing in a community offree-living chimpanzees. Nature 201:1264–1266.

Gould, S. J. 1980. Introduction. Pages ix-xviii in B. Kurtén. Dance ofthe tiger: a novel of the Ice Age. Random House, New York.

———. 2002. The structure of evolutionary theory. Harvard Univer-sity Press, Cambridge, MA.

Gould, S. J., and R. C. Lewontin. 1979. The spandrels of San Marcoand the Panglossian paradigm: a critique of the adaptationist pro-gramme. Proceedings of the Royal Society B 205:581–598.

Gould, S. J., and E. Vrba. 1982. Exaptation—a missing term in thescience of form. Paleobiology 8:4–15.

Gowaty, P. A. 1981. Sexual terms in sociobiology: emotionally evoca-tive and, paradoxically, jargon. Animal Behaviour 30:630–631.

Grandcolas, P., R. Nattier, F. Legendre, and R. Pellens. 2011. Map-ping extrinsic traits such as extinction risk or modelled [sic] biocli-matic niches on a phylogeny: does it make sense at all? Cladistics 27:181–185.

Gray, R. D., S. J. Greenhill, and R. M. Ross. 2007. The pleasures andperils of Darwinizing culture (with phylogenies). Biological Theory2:360–375.

Greene, H. W. 1983. Dietary correlates of the origin and radiation ofsnakes. American Zoologist 23:431–441.

———. 1986. Diet and arboreality in the emerald monitor, Varanusprasinus, with comments on the study of adaptation. Fieldiana,Zoology (New Series) 31:1–12.

———. 1992. The behavioral and ecological context for pitviper evo-lution. Pages 107–117 in J. A. Campbell and E. D. Brodie Jr., eds.Biology of the pitvipers. Selva, Tyler, TX.

This content downloaded from 132All use subject to University of Chicago Press Term

———. 1994. Homology and behavioral repertoires. Pages 369–391in B. K. Hall, ed. Homology: the hierarchical basis of comparativebiology. Academic Press, San Diego, CA.

———. 1999. Natural history and behavioural homology. Pages 173–188 in G. R. Bock and G. Cardew, eds. Homology (Novartis Founda-tion Symposium 222). J. Wiley and Sons, Chichester.

———. 2005. Organisms in nature as a central focus for biology.Trends in Ecology and Evolution 20:23–27.

———. 2013. Tracks and shadows: field biology as art. University ofCalifornia Press, Berkeley.

Greene, H. W., and G. M. Burghardt. 1978. Behavior and phylog-eny: constriction in ancient and modern snakes. Science 200:74–77.

Greene, H. W., P. May, D. L. Hardy Sr., J. Sciturro, and T. Farrell.2002. Parental behavior by vipers. Pages 179–205 in G. W. Schuett,M. Höggren, M. E. Douglas, and H. W. Greene, eds. Biology of thevipers. Eagle Mountain, Eagle Mountain, UT.

Griffiths, O. W., D. G. Blackburn, M. C. Brandley, J. U. Van Dyke,C. M. Whittington, and M. B. Thompson. 2015. Ancestral state re-constructions require biological evidence to test evolutionary hypoth-eses: a case study examining the evolution of reproductive mode insquamate reptiles. Journal of Experimental Zoology (Molecular andDevelopmental Evolution) 324B:493–503.

Gumert, M. D., M. Kluck, and S. Malaivigitnond. 2009. The physicalcharacteristics and usage patterns of stone ax and pounding ham-mers used by long-tailed macaques in the Andaman Sea region ofThailand. American Journal of Primatology 71:594–608.

Gursky, S. 2005. Predator mobbing in Tarsius spectrum. InternationalJournal of Primatology 26:207–221.

Guthrie, R. D. 2005. The nature of Paleolithic art. University of ChicagoPress, Chicago.

Hall, B. K., ed. 1994. Homology: the hierarchical basis of compara-tive biology. Academic Press, San Diego, CA.

———. 2013. Homology, homoplasy, novelty, and behavior. Devel-opmental Psychobiology 55:4–12.

Hamilton, C., and D. M. Fragaszy. 2014. Observation of weapon usein a group of semi-free tufted capuchins (Sapajus spp.). Neotrop-ical Primates 21:199–201.

Harmand, S., J. E. Lewis, C. S. Feibel, C. J. Lepre, S. Prat, A. Lenoble,X. Boes, et al. 2015. 3.3-million-year-old stone tools from Lomekwi 3,West Turkana, Kenya. Nature 521:310–315.

Harrison, T., ed. 2011. Paleontology and geology of Laetoli: humanevolution in context. Springer, New York.

Haslam, M., and T. Falótico. 2015. Nasal probe and toothpick use bya wild female bearded capuchin (Sapajus libidinosus). Primates 56:211–214.

Haslam, M., A. Hernandez-Aguilar, V. Ling, S. Carvalho, I. de la Torre,A. DeStefano, A. Du, et al. 2009. Primate archaeology. Nature 460:339–344.

Haslam, M., L. Luncz, A. Pascual-Garrido, T. Falótico, S. Malaivijit-nond, and M. Gumert. 2016a. Archaeological excavation of wildmacaque tone tools. Journal of Human Evolution 96:134–138.

Haslam, M., L. V. Luncz, R. A. Staff, F. Bradshaw, E. B. Ottoni, andT. Falótico. 2016b. Pre-Colombian monkey tools. Current Biology26:R515–R522.

Headland, T. N., and H. W. Greene. 2011. Hunter-gatherers andother primates as prey, predators, and competitors of snakes. Pro-ceedings of the National Academy of Sciences of the USA 108:20865–20866, E1470–E1474.

Hoevers, E. 2015. Tools go back in time. Nature 521:294–295.

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S84 The American Naturalist

Holekamp, K., S. T. Sakai, and B. L. Lundrigan. 2007. Social intelligencein the spotted hyena (Crocuta crocuta). Philosophical Transactionsof the Royal Society B 362:523–538.

Honing,H., andA. Ploeger. 2012. Cognition and the evolution ofmusic:pitfalls and prospects. Topics in Cognitive Science 4:513–524.

Hrdy, S. B. 2011. Mothers and others: the evolutionary origins ofmutual understanding. Belknap, Cambridge, MA.

Huebner, F., and C. Fichtel. 2015. Innovation and behavioral flexibilityin wild red-fronted lemurs (Eulemur rufifrons). Animal Cognition 18:777–787.

Huffman, M. A., J.-B. Leca, and C. A. D. Nahallage. 2010. CulturedJapanese macaques: a multidisciplinary approach to stone handlingbehavior and its implications for the evolution of behavioral tra-dition in non-human primates. Pages 185–213 in N. Nakagawa,M. Nakamichi, andH. Sugiura, eds. The Japanesemacaques. Springer,Berlin.

Hull, D. L. 1988. Science as a process: an evolutionary account of thesocial and conceptual development of science. University of ChicagoPress, Chicago.

Isbell, L. A. 2009. The fruit, the tree, and the serpent: why we see sowell. Harvard University Press, Cambridge, MA.

Isbell, L. A., and S. F. Etting. 2017. Scales drive visual detection, atten-tion, and memory of snakes in wild vervet monkeys (Chlorocebuspygerythrus). Primates 58:121–129.

Kappeler, P. M., and E. W. Heymann. 1996. Nonconvergence in theevolution of primate life history and socioecology. Biological Jour-nal of the Linnean Society 59:297–326.

Katoh, S., Y. Beyene, T. Itaya, H. Hyodo,M. Hyodo, K. Yagi, C. Gouzu,et al. 2016. New geological and palaeontological age constraint forthe gorilla-human lineage split. Nature 530:215–218.

King, B. J. 2013. How animals grieve. University of Chicago Press,Chicago.

Klopfer, P. H. 1973. Does behavior evolve? Annals of the New YorkAcademy of Sciences 223:113–119.

———. 1999. Politics and people in ethology: personal reflections onthe study of animal behavior. Bucknell University Press, Lewisburg,PA.

Kurtén, B. 1980. Dance of the tiger: a novel of the Ice Age. RandomHouse, New York.

Landau, M. 1984. Human evolution as narrative: have hero mythsand folktales influenced our interpretations of the evolutionarypast? American Scientist 72:262–268.

Langin, K. M., T. S. Sillett, W. C. Funk, S. A. Morrison, M. A.Desrosier, and C. K. Ghalambor. 2015. Islands within an island:repeated adaptive divergence in a single population. Evolution 69:653–665.

Larson, A., and J. B. Losos. 1996. Phylogenetic systematics of adapta-tion. Pages 187–220 inM. R. Rose and G. V. Lauder, eds. Adaptation.Academic Press, San Diego, CA.

Leakey, M. D. 1981. Discoveries at Laetoli in northern Tanzania. Pro-ceedings of the Geological Association 92:81–86.

Lewontin, R. C. 1978. Adaptation. Scientific American 293:212–228.———. 1998. The evolution of cognition: questions we will never an-swer. Pages 107–132 inD. Scarborough and S. Sternberg, eds. An in-vitation to cognitive science. Vol. 4. Methods, models, and conceptualissues. MIT Press, Cambridge, MA.

Liao,W. B., S. L. Lou, Y. Zeng, andA. Kotrscal. 2016. Large brains, smallguts: the expensive tissue hypothesis supportedwithin anurans.Amer-ican Naturalist 188:693–700.

This content downloaded from 132All use subject to University of Chicago Press Term

Lorenz, K. 1941. Vergleichende Bewegungsstudien bei Anatiden.Journal für Ornithologie 89:194–294.

Losos, J. B. 2009. Lizards in an evolutionary tree: ecology and adap-tive radiation of anoles. University of California Press, Berkeley.

———. 2011. Convergence, adaptation, and constraint. Evolution 65:1827–1840.

Luncz, L. V., T. Falatico A. Pascual-Garrido, A. Corat, H. Mosley, andM. Haslam. 2016. Wild capuchin monkeys adjust stone tools ac-cording to changing nut properties. Scientific Reports 6:33089.

Lycett, S. J., M. Collard, and W. C. McGrew. 2009. Cladistic analyses ofbehavioural variation in wild Pan troglodytes: exploring the chim-panzee cultural hypothesis. Journal of Human Evolution 57: 337–349.

Lynch Alfaro, J. W., J. de Sousa e Silva Jr., and A. B. Rylands. 2012. Howdifferent are robust and gracile capuchin monkeys? an argument forthe use of Sapajus and Cebus. American Journal of Primatology 74:273–286.

MacLean, E. L., L. J. Matthews, B. A. Hare, C. L. Nunn, R. C. Anderson,F. Aureli, E. M. Brannon, et al. 2012. How does cognition evolve? phy-logenetic comparative psychology. Animal Cognition 15:223–238.

Marks, J. 2015. Tales of the ex-apes: how we think about human evo-lution. University of California Press, Berkeley.

Mayr, E. 1983. How to carry out the adaptationist program. Amer-ican Naturalist 121:324–334.

McCune, A. R., and J. C. Schimenti. 2012. Using genetic networksand homology to understand the evolution of phenotypic traits.Current Genomics 13:74–84.

McGrew, W. C. 1992. Chimpanzee material culture: implications forhuman evolution. Cambridge University Press, Cambridge.

McGrew,W. C., and C. E. G. Tutin. 1973. Chimpanzee tool use in den-tal grooming. Nature 241:477–478.

Mercador, J., H. Barton, J. Gillespie, J. Harris, S. Kuhn, R. Tyler, andC. Boesch. 2007. 4,300 year-old chimpanzee sites and the originsof percussive stone technology. Proceedings of the National Acad-emy of Sciences of the USA 104:3043–3048.

Meulman, E. J. M., C. M. Sanz, E. Visalberghi, and C. P. van Schaik.2012. The role of terrestriality in promoting primate technology.Evolutionary Anthropology 21:58–68.

Morrell, V. 2013. Animal wise: how we know animals think and feel.Crown, New York.

Norell, M. A., J. M. Clark, L. M. Chaippe, and D. Dashzeveg. 1995. Anesting dinosaur. Nature 378:774–776.

O’Brien, M. J., B. Buchanan, M. Collard, and M. T. Boulanger. 2012.Cultural cladistics and the early prehistory of North America.Pages 23–42 in P. Pontarotti, ed. Evolutionary biology: mecha-nisms and trends. Springer, Berlin.

O’Brien, M. J., M. Collard, B. Buchanan, and M. T. Boulanger. 2013.Trees, thickets, or something in between? recent theoretical andempirical work in cultural phylogeny. Israeli Journal of Ecologyand Evolution 59:45–61.

O’Hara, R. 1988. Homage to Clio: or, toward an historical philoso-phy for evolutionary biology. Systematic Zoology 37:142–155.

———. 1992. Telling the tree: narrative representation and thestudy of evolutionary history. Biology and Philosophy 7:135–160.

Ostrom, J. 1972. Were some dinosaurs gregarious? Palaeogeography,Palaeoclimatology, Palaeoecology 11:287–301.

Pickering, T. R. 2013. Rough and tumble: aggression, hunting, andhuman evolution. University of California Press, Berkeley.

———. 2014. Comments on Domínguez-Rodrigo (2014). CurrentAnthropology 55:73–74.

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Scenarios and Natural History S85

Pickering, T. R., and M. Domínguez-Rodrigo. 2010. Chimpanzeereferents and the emergence of human hunting. Open Anthropol-ogy Journal 3:94–100.

Pinker, S. 2002. The blank slate: the modern denial of human nature.Penguin, New York.

Preuss, T. M. 2012. Human brain evolution: from gene discovery tophenotype discovery. Proceedings of the National Academy ofSciences of the USA 109:10709–10716.

Prindle, D. F. 2009. Stephen Jay Gould and the politics of evolution.Prometheus, Amherst, NY.

Proctor, H. C. 1996. Behavioral characters and homoplasy: percep-tion versus practice. Pages 131–149 in M. J. Sanderson and L.Hufford, eds. Homoplasy: the recurrence of similarity in evolu-tion. Academic Press, San Diego, CA.

Proffitt, T., L. V. Luncz, T. Falótico, E. B. Ottoni, I. de la Torre, andM. Haslam. 2016.Wildmonkeys flake stone tools. Nature 539:85–88.

Pruetz, J., and P. Bertolani. 2007. Savanna chimpanzees, Pan troglodytesverus, hunt with tools. Current Biology 17:412–417.

Pruetz, J. D., P. Bertolani, K. Boyer Ontl, S. Lindshield, M. Shelley, andE. G. Wessling. 2015. New evidence on the tool-assisted huntingexhibited by chimpanzees (Pan troglodytes verus) in a savannah hab-itat at Fongoli, Senegal. Royal Society Open Science 2:140507.

Pruetz, J. D., and T. C. Laduke. 2010. Reaction to fire by savannachimpanzees (Pan troglodytes verus) at Fongoli, Senegal: concep-tualization of “fire behavior” and the case for a chimpanzee model.American Journal of Physical Anthropology 141:646–650.

Quinn, A. 2016a. William Whewell’s philosophy of architecture andthe historicization of biology. Studies in History and Philosophyof Biological and Biomedical Sciences 59:11–19.

———. 2016b. Charles Girard: relationships and representation innineteenth century systematics. Journal of the History of Biology,http://dx.doi.org/10.1007/s10739-016-9447-6.

Radinsky, L. B. 1968. Evolution of somatosensory specialization inotter brains. Journal of Comparative Neurology 134:495–505.

Reader, S. M., and K. N. Laland. 2002. Social intelligence, innova-tion, and enhanced brain size in primates. Proceedings of the Na-tional Academy of Sciences of the USA 99:4436–4441.

Richerson, P. J., and R. Boyd. 2005. Not by genes alone: how culturetransformed human evolution. University of Chicago Press, Chicago.

Rivas, J. A., and G. M. Burghardt. 2002. Crotalomorphism: a meta-phor for understanding anthropomorphism by omission. Pages 9–17 in M. Bekoff, C. Allen, and G. M. Burghardt, eds. The cognitiveanimal: empirical and theoretical perspectives on animal cogni-tion. MIT Press, Cambridge, MA.

Roach, M. 2008. Almost human. National Geographic 213:127–144.Roffman, I., S. Savage-Rumbaugh, E. Rupert-Pugh, A. Stadler, A.Ronen, and E. Nevo. 2015. Preparation and use of varied naturaltools for extractive foraging by bonobos (Pan paniscus). AmericanJournal of Physical Anthropology 158:78–91.

Safina, C. 2015. Beyond words: what animals think and feel. H. Holt,New York.

Sahle, Y., W. K. Hutchings, D. R. Braun, J. C. Sealy, L. E. Morgan,A. Negash, and B. Atnafu. 2013. Earliest stone-tipped projectiles fromthe Ethiopian Rift date to 1279,000 years ago. PLoS ONE 8:e78092.

Sanderson, M. J., and L. Hufford, eds. 1996. Homoplasy: the recur-rence of similarity in evolution. Academic Press, San Diego, CA.

Sarringhaus, L. A., L. M. MacLatchy, and J. C. Mitani. 2014. Locomo-tor and postural development of wild chimpanzees. Journal of Hu-man Evolution 66:29–38.

This content downloaded from 132All use subject to University of Chicago Press Term

Schoch, W. H., G. Bigga, U. Böhner, P. Richter, and T. Terberger.2015. New insights on the wooden weapons from the Paleolithicsite of Schöenigen. Journal of Human Evolution 89:214–225.

Scholes, E. 2008. Courtship ethology of Wahnes’ parotia Parotiawahnesi (Aves: Paradisaeidae). Journal of Ethology 26:79–91.

Segerstråle, U. 2000. Defenders of the truth: the battle for science in thesociobiology debate and beyond. Oxford University Press, Oxford.

Seilacher, A. 2007. Trace fossil analysis. Springer, Berlin.Sessions, S. K., L. BizjakMali, D.M.Green, V. Trifona, andM. Ferguson-Smith. 2016. Evidence for sex chromosome turnover in proteidsalamanders. Cytogenetics and Genome Research 148:305–313.

Shettleworth, S. J. 2010. Clever animals and killjoy explanations incomparative psychology. Trends in Cognitive Sciences 14:477–481.

———. 2012. Do animals have insight, and what is insight anyway?Canadian Journal of Experimental Psychology 66:217–226.

Shultz, S., C. Opie, and Q. D. Atkinson. 2011. Stepwise evolution ofstable sociality in primates. Nature 479:219–222.

Skutch, A. F. 1980. A naturalist on a tropical farm. University ofCalifornia Press, Berkeley.

Sullivan, A. P., D.W. Bird, and G. H. Perry. 2017. Human behavior as along-term ecological driver of non-human evolution. Nature Ecologyand Evolution, http://dx.doi.org/10.1038/s41559-016-0065.

Tattersall, I., and N. Eldredge. 1977. Fact, theory, and fantasy in hu-man paleontology: controversy in the study of human evolutionreflects inadequacies in the formulation of hypotheses more thanit does the supposed inadequacies of the fossil record. AmericanScientist 65:204–211.

Tëmkin, I., and N. Eldredge. 2007. Phylogenetics andmaterial culture.Current Anthropology 48:146–154.

Thieme, H. 1997. Lower Paleolithic hunting spears from northernGermany. Nature 385:807–810.

Tinbergen, N. 1963. On aims and methods of ethology. Zeitschriftfür Tierpsychologie 20:410–429.

Treves, A., and C. Bonancic. 2016. Humanity’s dual response to dogsand wolves. Trends in Ecology and Evolution 31:489–491.

Visalberghi, E., G. Sirianni, D. Fragaszy, and C. Boesch. 2015. Per-cussive tool use by Taï Western chimpanzees and Fazenda BoaVista bearded capuchin monkeys. Philosophical Transactions ofthe Royal Society B 370(1682):20140351.

Wake, D. B. 1996. Introduction. Pages xvii–xxv in M. J. Sandersonand L. Hufford, eds. Homoplasy: the recurrence of similarity inevolution. Academic Press, San Diego, CA.

Wake, D. B., M. H. Wake, and C. D. Specht. 2011. Homoplasy: fromdetecting pattern to determining process and mechanism of evo-lution. Science 331:1032–1035.

Wake, M. H. 1992. Evolutionary scenarios, homology and conver-gence of structural specializations for vertebrate viviparity. Amer-ican Zoologist 32:256–263.

Wanntorp, H.-E. 1983. Historical constraints in adaptation theory:traits and non-traits. Oikos 41:157–160.

Wanntorp, H.-E., D. R. Brooks, T. Nilsson, S. Nylin, F. Qonquist, S. C.Stearns, and N. Wedell. 1990. Phylogenetic approaches in ecology.Oikos 57:119–132.

Watanabe, K., N. Urasopon, and S. Malaivijitnond. 2007. Long-tailedmacaques use human hair as dental floss. American Journal of Pri-matology 69:940–944.

Wenzel, J. W. 1992. Behavioral homology and phylogeny. AnnualReviews of Ecology and Systematics 23:361–381.

Werdelin, L. 2013. King of beasts. Scientific American (Nov.):36–39.

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S86 The American Naturalist

Werdelin, L., and M. E. Lewis. 2013. Temporal change in functionalrichness and evenness in the East African Plio-Pleistocene carnivo-ran guild. PLoS ONE 8:e57944.

White, T. D., B. Asfaw, Y. Beyene, Y. Haile-Selasse, C. O. Lovejoy,G. Suwa, and G. WoldeGabriel. 2009. Ardipithecus ramidus andthe paleobiology of early hominids. Science 326:75–86.

White, T. D., and G. Suwa. 1987. Hominid footprints at Laetoli: factsand interpretations. American Journal of Physical Anthropology 72:485–514.

Wilkins, J., B. J. Schoville, K. S. Brown, and M. Chazan. 2012. Evi-dence for early hafted hunting technology. Science 338:942–946.

Wilson, D. E., and D.M. Reeder. 2005.Mammal species of the world: ataxonomic and geographic reference. 3rd ed. Johns Hopkins Uni-versity Press, Baltimore.

Wilson, E. O. 1975. Sociobiology: the new synthesis. Harvard Uni-versity Press, Cambridge, MA.

“Then follows a general ‘popular’ account of the forms of apes, theirlanguage, reproduction, education, rearing of young, diseases, life in confinthe head of a five-year-old tschego in the Dresden Zoological Garden, whAnimal Life (The American Naturalist, 1877, 11:557–559).

This content downloaded from 132All use subject to University of Chicago Press Term

Wood, B., and T. Harrison. 2011. The evolutionary context of thefirst hominins. Nature 470:347–352.

Wrangham, R. 2009. Catching fire: how cooking made us human.Basic, New York.

Yamaguchi, N., A. Cooper, L. Werdelin, and D. W. Macdonald. 2004.Evolution of the mane and group-living in the lion (Panthera leo):a review. Journal of Zoology 263:329–342.

Yeaves, J. D., N. J. Dominy, P. L. Koch, and M. Mangel. 2013. Func-tional morphology, stable isotopes, and human evolution: a modelof consilience. Evolution 68:190–203.

Zamudio, K. R., R. C. Bell, R. C. Nali, C. F. B. Haddad, and C. P. A.Prado. 2016. Polyandry, predation, and the evolution of frog re-productive modes. American Naturalist 188(suppl.):S41–S61.

Symposium Editor: Anurag A. Agrawal

geographical distribution, dwelling-places, food, motions, social life,ement, and of the apes figured on the Egyptian temples. . . . [Figured:]ich was brought from the Loango coast.” From the review of Brehm’s

.236.027.111 on July 22, 2017 05:14:50 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Convergence, Consilience, and the Evolution

of Temperate Deciduous Forests*

Erika J. Edwards,1,† David S. Chatelet,1,‡ Bo-Chang Chen,2 Jin Yao Ong,3 Shuichiro Tagane,4

Hironobu Kanemitsu,5 Kazuki Tagawa,5 Kentaro Teramoto,5 Brian Park,6 Kuo-Fang Chung,2,7

Jer-Ming Hu,3 Tetsukazu Yahara,4 and Michael J. Donoghue6

1. Department of Ecology and Evolutionary Biology, Brown University, Providence, Rhode Island 02912; 2. School of Forestry and ResourceConservation, National Taiwan University, Taipei 10617, Taiwan; 3. Institute of Ecology and Evolutionary Biology, National TaiwanUniversity, Taipei 10617, Taiwan; 4. Department of Biology, Kyushu University, 744 Motooka, Fukuoka 819-0395, Japan; 5. GraduateSchool of Systems Life Sciences, Kyushu University, 744 Motooka, Fukuoka 819-0395, Japan; 6. Department of Ecology and EvolutionaryBiology, Yale University, New Haven, Connecticut 06520; 7. Biodiversity Research Center, Academia Sinica, Taipei 11529, Taiwan

Dryad data: http://dx.doi.org/10.5061/dryad.k505s.

abstract: The deciduous habit of northern temperate trees andshrubs provides one of the most obvious examples of convergent evo-lution, but how did it evolve? Hypotheses based on the fossil recordposit that deciduousness evolved first in response to drought or dark-ness and preadapted certain lineages as cold climates spread. An alter-native is that evergreens first established in freezing environments andlater evolved the deciduous habit. We monitored phenological pat-terns of 20 species of Viburnum spanning tropical, lucidophyllous(subtropical montane and warm temperate), and cool temperate Asianforests. In lucidophyllous forests, all viburnums were evergreen plantsthat exhibited coordinated leaf flushes with the onset of the rainy sea-son but varied greatly in the timing of leaf senescence. In contrast, de-ciduous species exhibited tight coordination of both flushing and se-nescence, and we found a perfect correlation between the deciduoushabit and prolonged annual freezing. In contrast to previous stepwisehypotheses, a consilience of independent lines of evidence supports alockstepmodel in which deciduousness evolved in situ, in parallel, andconcurrent with a gradual cooling climate. A pervasive selective forcecombined with the elevated evolutionary accessibility of a particularresponse may explain the massive convergence of adaptive strategiesthat characterizes the world’s biomes.

Keywords: Viburnum, biome assembly, freezing tolerance, leaf habit,climate change, phylogeny.

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† Corresponding author; e-mail: erika_edwards@brown.edu.‡ Present address: Biomedical Imaging Unit, University of Southampton, South-ampton SO16 6YD, United Kingdom.ORCIDs: Kanemitsu, http://orcid.org/0000-0002-4332-2152; Hu, http://orcid

.org/0000-0003-2739-9077.

Am. Nat. 2017. Vol. 190, pp. S87–S104. q 2017 by The University of Chicago.0003-0147/2017/190S1-57320$15.00. All rights reserved.DOI: 10.1086/692627

This content downloaded from 132All use subject to University of Chicago Press Term

Introduction

Biologists have long been attracted to convergent evolution,and for good reason, as it provides powerful evidence ofnatural selection on organismal performance. Many con-vergent traits reflect organismal interactions, both mutual-istic and antagonistic (Fenster et al. 2004; Agrawal andFishbein 2006; Wilson et al. 2007), but past climate changehas undoubtedly also driven convergence on a global scale.For example, the latter half of the Cenozoic witnessed theformation and spread of deserts and grasslands, resultingin the repeated evolution of succulent life-forms and C4

photosynthesis (Edwards et al. 2010; Arakaki et al. 2011).Convergence also provides insights into the relative evo-

lutionary accessibility of certain phenotypes and how struc-tural features of organismsmay influence their evolutionaryresponse (Sanderson and Hufford 1996; Donoghue and Ree2000; Christin et al. 2013). The degree to which a particularcharacter repeatedly emerges must be a function of boththe pervasiveness of the selection pressure(s) and the rela-tive ease of its evolution. As climate change is experiencedby virtually all organisms in a region concurrently, it makessense that it would be one of the most powerful agents ofmassive convergent evolution. However, whether suchmas-sive convergence is the outcome depends on whether a givenadaptive response is also the most evolutionarily accessible(e.g., Weinreich 2006; Meyer et al. 2012; Agrawal 2017) inmultiple lineages. The combination of a ubiquitous selec-tive force with the evolutionary accessibility of a particularadaptation virtually guarantees rampant convergence.On the surface of it, massive convergence would appear

to lend itself perfectly to comparative phylogenetic analy-ses. It is only through a phylogenetic lens that convergencecan be identified, after all, and a multitude of comparativemethods have been developed to measure convergence and

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S88 The American Naturalist

the factors that are correlated with it, perhaps causally. Incharacters with moderate levels of convergence, phyloge-netic approaches can work well, but for extremely labiletraits, uncertainties in phylogenetic inference can becomelimiting (Schultz et al. 1996; Schluter et al. 1997). This sit-uation is complicated further by the observation that con-vergence itself is often distinctly concentrated in particularregions of the tree of life (Edwards and Donoghue 2013).For example, C4 photosynthesis has evolved over 60 timesin plants, but fully two-thirds of these origins occur in justtwo flowering plant lineages, the grasses and the Caryophyl-lales (Sage et al. 2011), and even within grasses, C4 originsare clustered yet again in a clade that includes roughly halfof all grass species (Christin et al. 2013). Furthermore, ifevolutionary responses are rapid enough, it becomes im-possible to identify the evolutionary sequences and transi-tional conditions that might provide us with clues aboutcause and effect, especially about evolutionary events thathappened long ago. Faced with these difficulties, convinc-ing answers may not be achieved using phylogenies alone(Christin et al. 2010; Hancock and Edwards 2014). Prog-ress, then, depends on the integration of phylogenetic stud-ies with other lines of evidence (Weber and Agrawal 2012;Olson and Arroyo-Santos 2015) and, as we highlight, per-haps by further deconstruction of the traits and potentialselective factors of interest.

Here we focus on the evolution of the deciduous leafhabit, one of the most obvious and highly convergent adap-tations of woody plants to temperate environments that ex-perience an annual period of prolonged freezing temper-atures. Deciduous plants shed their leaves with the onset ofthe cold period and remain leafless until they flush a newset of leaves in the spring. This habit dominates northernhardwood forests and has evolved independently many timesin distantly related plant groups (e.g., maples and oaks). It isimportant to note that deciduousness has also evolved manytimes in settings that we will not consider here, particularlyin dry tropical forests where there is also a season unfavor-able for growth—in this case, determined by drought, notcold (Murphy and Lugo 1986). We do not know of any at-tempts to infer the number of independent origins of the de-ciduous leaf habit in angiosperms, but considering the ex-tremely broad phylogenetic distribution of deciduous plantsand its demonstrated lability within smaller lineages (e.g.,Schmerler et al. 2012), we guess that it has evolvedmore fre-quently than other well-known convergent plant traits (e.g.,162 origins of C4 photosynthesis [Sage et al. 2011];1130 or-igins of bilaterally symmetrical flowers [Reyes et al. 2016]).

As a case study, we analyze the evolution of deciduous-ness in the woody plant clade Viburnum, combining stan-dard phylogenetic analyses with a phenological field studyalong a latitudinal transect in eastern Asia. Independentof the traditional evergreen versus deciduous leaf habit cat-

This content downloaded from 132All use subject to University of Chicago Press Term

egories, our fieldwork revealed a natural decomposition ofphenological behavior into two distinct and potentially in-dependently evolving elements: the flushing of new leavesand the senescing of old ones. It also exposed the inadequacyof the standard binary breakdown of biome or habitat typeinto tropical versus temperate forests. Specifically, we found itimportant to also recognize the existence of subtropical mon-tane and warm temperate lucidophyllous forests (Tang 2010;Tang et al. 2013). These Asian forests are characterized bythe dominance of broad-leaved angiosperm trees (typicallyLauraceae and Fagaceae) in a climate that is distinctly mon-soonal in comparison to tropical rainforests but where freez-ing is infrequent andwinter temperatures aremild comparedto cold temperate forests. The Viburnum species that westudied in these forests show patterns of leaf flushing and se-nescence that provide important clues to the transition be-tween the archetypal evergreen and deciduous conditions.Considered in isolation, our phylogenetic and field stud-

ies each provide insight into the evolution of deciduousnessbut not overwhelming support for a particular evolutionarypathway. Considered together, however, the consilience ofevidence leads us to propose a model in which the decidu-ous habit evolves quite rapidly, in lockstep with a gradual insitu transition to a routinely freezing climate. We hypothe-size that the novel behavior that emerged with the decidu-ous habit was the tight coordination of leaf senescence, per-haps as a means of ensuring the resorption of nutrientsprior to leaf death by freezing (Feild et al. 2001; Niinemetsand Tamm 2005). To our surprise, this lockstep model ap-pears to be novel, as prior theories have envisioned a dis-tinctly stepwise process. As we review below, these entail ei-ther the evolution of the deciduous habit, first in responseto a period of drought or darkness (with establishment infreezing climates emerging only later), or the establishmentof evergreen plants in freezing climates, followed later bythe evolution of deciduousness. We suggest that the step-wise perspective that has explicitly or implicitly orientedmany phylogenetic studies of adaptation may be inappro-priate or even misleading when studying past evolutionaryresponses to the gradual climate changes that have shapedthe assembly of the world’s biomes.

Background Information

Evolution of the Deciduous Habit in Temperate Forests

In general, the deciduous leaf habit in woody plants is un-derstood to be a response to long periods of time that areunfavorable for growth, which can result from episodes offreezing or drought (Chabot andHicks 1982). By definition,annually deciduous plants must have leaf life spans of lessthan 12 months, with many cold temperate woody plantsaveraging no more than 6 or 7 months (Reich et al. 1992;

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Consilience, Convergence, and Phylogeny S89

van Ommen Kloeke et al. 2012). Deciduous plants also tendto fall on the fast or resource-acquisitive end of the leaf eco-nomic spectrum (Wright et al. 2004), and their leaves aretypically thinner, with higher nitrogen contents and higherphotosynthetic rates than evergreen species. Soil fertilityhas thus been identified as an important secondary variable:nutrient-poor, cold sites tend to have a higher representa-tion of broad-leaved evergreens, presumably because leafnitrogen contents are too low to maintain the higher pho-tosynthetic rates required to balance investment costs in ashort-lived leaf. These arguments, revolving around trade-offs in leaf carbon and nutrient economy, are well developedelsewhere (Chabot and Hicks 1982; Kikuzawa 1991; Givnish2002), and our aim here is not to elaborate on the adaptivesignificance of the deciduous leaf habit. Instead, we simplyassume that deciduousness is adaptive for all of the reasonsabove and consider a related question that has received farless attention, namely, how the deciduous habit evolved froman evergreen state. In particular, we analyze leaf flushing andleaf senescence as two potentially independent behaviorsthat became coordinated with the emergence of the decid-uous habit.

Climate Change and the Origin ofTemperate Deciduous Forests

In spite of its importance as a modern biome, there are rel-atively few hypotheses about the origin of broad-leaved de-ciduous forests. Inferences based on fossil leaves indicatethat deciduous angiosperm species have existed for a longtime, appearing as elements of fossil floras from the LateCretaceous through the Cenozoic (Wolfe 1987). There isevidence for the dominance of the deciduous habit in thehighest northern latitudes (polar deciduous forests) duringthe Late Cretaceous, when global temperatures were signif-icantly warmer than today. These polar forests persistedthrough the Paleocene and Eocene periods, with broad-leaved evergreen forests occupying significant areas in mid-latitude and tropical regions (Axelrod 1966; Wolfe andUpchurch 1986; Wolfe 1987). Post-Eocene global coolingshifted many of these midlatitude zones toward the tropics,but it was not until the late Miocene that the modern distri-bution of cool temperate deciduous forests was established(Tiffney and Manchester 2001; Utescher and Mosbrugger2007; Pound et al. 2011).

Axelrod (1966) proposed that deciduousness first evolvedin a warm, subtropical climate zone that experienced mildwinter drought. He hypothesized that such forests werewidespread throughout the midlatitudes of the NorthernHemisphere in the Late Cretaceous and early Cenozoic butbecame restricted during theMiocene. As modern analogues,he specifically cited the lucidophyllous forests of Asia (par-ticularly Taiwan) and the cloud forests of southern Mexico,

This content downloaded from 132All use subject to University of Chicago Press Term

which experience strong seasonality in both rainfall andtemperature but rarely freeze. Axelrod hypothesized thatthe deciduous leaf habit first appeared in such forests andthat this later enabled these plants to expand into the freez-ing zone.Wolfe and Upchurch (1986) andWolfe (1987) explained

the origin of deciduous forests in another way. They pro-posed that deciduousness evolved in situ at high latitudes,when even areas at 807N did not experience significantfreezing. However, winters at such high latitudes were asdark as they are today, and they hypothesized that the de-ciduous habit evolved there as a response to many monthswith little or no sunlight. At the Cretaceous-Paleogene (K-Pg)boundary, the brief impact winter was creditedwith reorganiz-ing plant communities at lower latitudes, andWolfe and col-leagues suggested that it was during this time that polar de-ciduous lineages migrated south and became established aselements of midlatitude floras. Only during the late Mio-cene cooling did they outcompete warm-adapted evergreenlineages to establish a new deciduous biome.It is noteworthy that both the Axelrod and Wolfe hy-

potheses rest on the idea that deciduousness was a preadap-tation to freezing—that is, it evolved first as a response tosome other stress but then allowed lineages to tolerate freez-ing climates when they were eventually exposed to them.Neither author explained why they preferred such an orderof events as opposed to the perhaps simpler hypothesis thatdeciduousness evolved directly in response to freezing. Morerecently, Zanne et al. (2014), without reference to these olderhypotheses, addressed the samequestionusing amegaphylog-eny for angiosperms and concluded that most lineages firstbecame established in the freezing zone as evergreen plantsand only later evolved the deciduous habit. Elsewhere wereanalyzed the data of Zanne et al. and raised doubts abouttheir conclusions (Edwards et al. 2015), but we highlightthis study here because it focused directly on freezing as adriver. Like Axelrod and Wolfe, Zanne et al. envisioned astepwise evolutionary process but with a reversed order ofevents. In the terminology of Zanne et al. (2014), Axelrodand Wolfe favored a trait-first hypothesis, whereas they fa-vored a climate-first hypothesis.

The Viburnum Study System

Viburnum (Adoxaceae, Dipsacales, Campanulidae) is anangiosperm clade of ∼165 species of shrubs and small treesthat is broadly distributed (and widely cultivated) aroundtheNorthernHemisphere.Viburnum species all occupyme-sic forests, but they have adapted to a range of climatic con-ditions and forest environments (fig. 1). At one end of thespectrum, multiple (and distantly related) species in tropicalforests are never subjected to freezing temperatures. At theother end, upwards of 10 Viburnum lineages appear to have

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S90 The American Naturalist

evolved independently into cool temperate forests that expe-rience consecutive months of below freezing temperatures.In Asia, members of several clades occupy lucidophyllousforests, and members of the Neotropical Oreinotinus cladeoccupy montane cloud forests, with reduced temperatureseasonality and only rare and intermittent freezing temper-atures.

Evolutionary shifts between these different environmentshave been accompanied by transitions in leafing habit (fig. 1).In tropical, lucidophyllous, and cloud forests, Viburnumplants are most often evergreen, while, with very few ex-

This content downloaded from 132All use subject to University of Chicago Press Term

ceptions, in cool temperate forests they are seasonally de-ciduous. Of the 165 species of Viburnum that we currentlyrecognize, we consider 84 to be evergreen and 81 to be de-ciduous. Eight of the well-supported major clades withinViburnum include both evergreen and deciduous species,implying a minimum of eight evolutionary shifts. However,we infer no fewer than 20 transitions in leaf habit using anyof our recently published phylogenetic trees (Clement et al.2014; Spriggs et al. 2015; Eaton et al. 2017), including someclear-cut evolutionary shifts in both directions. For exam-ple, the deciduous species Viburnum plicatum appears to

acerifolium

acutifolium

amplificatum

anabaptista

atrocyaneum

australe

ayavacense

beccarii

betulifolium

blandum

brachyandrum

brachybotryum

bracteatum

buddleifolium

burejaeticum

carlesii

cassinoides

chingii

cinnamomifolium

clemensiae

colebrookeanum

coriaceum

costaricanum

cotinifolium

cylindricum

davidii

dentatum

dilatatum

discolor

disjunctum

edule

elatum

ellipticum

erosum

erubescens

farreri

flavescens

foetens

foetidum

furcatumgrandiflorum

hallii

hartwegii

henryi

hispidulum

hupehense

ichangense

inopinatum

integrifolium

jamesonii

japonicum

jucundum

kansuense

koreanum

lancifolium

lantana

lantanoides

lasiophyllum

lautum

lentago

lepidotulum

lobophyllum

lutescens

luzonicum

macrocephalum

melanocarpum

microcarpummicrophyllum

molle

mongolicum

mullaha

nervosum

nudum

obtectum

obtusatum

odoratissimum

oliganthum

opulus

orientale

parvifolium

phlebotrichum

pichinchense

plicatum

propinquum

prunifoliumpunctatum

rafinesquianum

recognitum

rhytidophyllum

rigidum

rufidulum

sambucinum

sargentii

scabrellum

schensianum

seemenii

sempervirens

sieboldii

stellato-tomentosum

stipitatum

subalpinum

sulcatum

suspensum

sympodiale

taitoense

taiwanianum

tashiroi

tinoides

tinus

toronis

trilobum

triphyllum

undulatum

urceolatum

utile

veitchii

venustum

vernicosum

villosum

wrightii

evergreendeciduous

tropicallucidophyllouscold temperatecloud forestuncertain

minimum temperature of the coldest week

20.6 °C-22 °C 0 °C

restricted to New World

Figure 1: Phylogeny, climatic envelope, forest habitat, and leaf habit of Viburnum. A 120-taxon phylogeny, pruned from Spriggs et al.(2015). Branch colors represent inferred ancestral minimum weekly temperatures (BIO6). Pie charts at internal nodes are ancestral stateestimates of leaf habit, inferred using the threshold model (Revell 2012). Species at tips are coded by leaf habit in the left column (evergreenvs. deciduous) and habitat in the right column (tropical, lucidophyllous, cold temperate, or cloud forests). Species with names in boldfacewere included in the phenological monitoring study.

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Consilience, Convergence, and Phylogeny S91

have been derived from evergreen ancestors within the Lu-tescentia clade. Likewise, the evergreen species Viburnumrhytidophyllum and Viburnum utile appear to have origi-nated from deciduous ancestors within the Euviburnumclade, as did Viburnum sempervirens and its several ever-green relatives within Succotinus.

As we have shown previously (Schmerler et al. 2012),these repeated evolutionary shifts in forest environmentand leafing behavior have been accompanied by specificchanges in leaf form. Adaptation to cool temperate forestsat more northerly latitudes has entailed coordinated shiftsto rounder leaf shapes with more marginal teeth or lobes.Marked seasonal heteroblasty in the temperate species (thedevelopment of preformed leaves with more derived leafshapes and neoformed leaveswithmore tropical shapes) sug-gested to us that the evolution of temperate leaf forms mayrelate to the development of leaf primordia within the restingbuds of deciduous species (Edwards et al. 2016). In any case,Viburnum provides clear evidence, directly in line with globaltrends (Bailey and Sinnott 1916; Wolfe 1995; Royer andWilf2006; Peppe et al. 2011), of repeated evolutionary changes inboth leaf habit and leaf form in connection with shifts be-tween tropical and temperate forests.

Material and Methods

Phenology. To monitor Viburnum leafing behavior across awide range of environments, we established a latitudinaltransect spanning 277 latitude, from montane tropical for-ests in Kinabalu National Park, Sabah, Malaysia, at 67N totemperate forests in the mountains surrounding Fukuoka,Japan, on Kyushu Island at 337N (fig. 2). Sites were chosento represent different forest types (tropical rain forest inBorneo, both lucidophyllous and cool temperate forests inTaiwan and Japan) and also to maximize the number ofco-occurring Viburnum species in each location. From Mayto June 2013, we selected a total of 19 populations represent-ing 18 species in Taiwan and Japan, and in February 2014 weadded two species in Borneo. For logistical reasons, not allpopulations were monitored for the same length of time.The Borneo populations were monitored for 13 consecutivemonths, the Japanese populations for 10–19 consecutivemonths, and most of the Taiwanese populations for 25 con-secutive months. We achieved our best sampling in Taiwan,where wewere also able tomonitor two populations ofVibur-num luzonicum that spanned its elevational range (120 m,2,100 m).

In each population of each of our focal species, we taggedthree to six individuals (mostly four individuals per species)for monitoring. We marked four branches per individualand made biweekly to monthly census trips to record thepersistence of each leaf on these branches, the emergenceof new leaves, and all events of leaf senescence using meth-

This content downloaded from 132All use subject to University of Chicago Press Term

ods described in Edwards et al. (2014). At the start of thestudy, each branch usually had only a few apical meristems,but due to the growth architecture of most Viburnum spe-cies (stems typically terminate in an inflorescence, withgrowth recommencing from two subtending buds; Don-oghue 1981; Edwards et al. 2014), by the end of the studyeach branch consisted of multiple independent leafy shoots,sometimes bearing over 100 leaves.Viburnum-wide climate data, leaf habit, and forest type.

We downloaded georeferenced locality data for Viburnumfrom the following databases: Global Biodiversity Informa-tion Facility (http://www.gbif.org/), Chinese Virtual Her-barium (http://www.cvh.org.cn/), Interactive AgriculturalEcological Atlas of Russia and Neighboring Countries (http://www.agroatlas.ru/), Plant DNA Bank in Korea (http://pdbk.korea.ac.kr/), and Virtual Viburnum (http://viburnum.peabody.yale.edu/). We also georeferenced occurrence rec-ords fromHara’s (1983) treatment of theViburnum speciesof Japan. The raw data were filtered to remove spuriousrecords. Specifically, we matched taxon names with thosein Spriggs et al. (2015) and removed records with coordi-nates at 07, 07 (and those from botanical gardens and her-baria) and records of the same species with identical coor-dinates. We also plotted the records for each species usingthe R package ggmap (Kahle and Wickham 2013) and re-moved points outside of the known geographic ranges ofthese species based on floristic treatments and herbariumspecimens examined in preparation for our worldwidemono-graph of Viburnum.We used the CliMond database (Kriticos et al. 2012) at a

10-min resolution (∼18 km2 at the equator) to extract es-timates of climate for each locality in the filtered data set us-ing the R package raster (Hijmans and van Etten 2012).This spatial grain best matches the georeferencing precisionof our data set. The species locality data were then spatiallyrarefied so that there was no more than one occurrence perclimate grid cell using SDMtoolbox (Brown 2014), whichincreased the evenness of coverage across the species’ entirerange. Species with fewer than three grid cell occurrenceswere removed from the data set. We calculated the meanminimum temperature of the coldest week (BIO6) and sea-sonality of precipitation (BIO15) for each species. We notethat although CliMond estimates BIO6 as a weekly mea-sure, their variable is interpolated from monthly tempera-ture data and is strongly correlated with the more commonlyused BIO6 variable, the mean minimum temperature of thecoldest month (R2 p 0:995). In all, our data set consists of7,718 records for 120 Viburnum species. We also used theCliMond database to extract mean monthly temperatureand precipitation data for the sites of our monitored popu-lations.We scored each species in our data set as (1) belong-

ing in the traditional evergreen or deciduous category and

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S92 The American Naturalist

(2) occurring in tropical, lucidophyllous, cool temperate, orcloud forests. These scores were based on monographic andfloristic treatments of Viburnum (e.g., Rehder 1908; Killipand Smith 1931; Morton 1933; Kern 1951; Hara 1983; Yangand Malecot 2011). For the majority of these species wehave first-hand knowledge based on our own field and her-barium studies, but for Japanese species we also relied onHara (1983) and for Chinese species we followed Yang andMalecot (2011). We were uncertain about assignment to for-est type for seven species and scored these as unknown.

Analysis of phenological data. We recorded leaf gain andloss using hand-drawn sketches of branches, and dates ofemergence and disappearance of each leaf were transcribedinto a spreadsheet. The total number of leaves gained or lostat each census was summed across branches within an in-

This content downloaded from 132All use subject to University of Chicago Press Term

dividual and averaged across individuals within a popula-tion. Only a small subset of leaves were observed for theirentire life span in the evergreen species in our sample; themajority of leaves were either present when we initiatedthe study and senesced partway through the monitoring pe-riod or emerged during the study and were still present atthe end. Population-level patterns in flushing and senes-cence were visualized using the ribbon function in ggplots2(Wickham 2016) in R. Most species appeared to exhibitstrong seasonality in leaf flushing and sometimes also in se-nescence. To evaluate the statistical significance of thesepatterns, for each species we compared our observed datawith a null distribution of phenological events. Because thetiming of our population visits was not perfectly spacedthroughout the year, we could not assume a flat null distri-

Figure 2: Locations of the phenological monitoring transect in eastern Asia. Insets provide greater resolution of our different populationlocations. Top inset, Sites situated around Fukuoka, Japan, on the island of Kyushu. Middle inset, Location of our low-elevation populationof Viburnum luzonicum in the suburbs of Taipei, Taiwan. Bottom inset, The majority of our Taiwan monitoring sites scattered around theHehuanshan Mountain district of the Central Mountain Range. In both Japan and Taiwan, sites occupied both freezing and nonfreezingclimates.

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Consilience, Convergence, and Phylogeny S93

bution; just by chance we would expect to see more eventsin spring and fall seasons, when our visitation rates wereslightly higher. For each species, we pooled all visitationdates (as Julian days) and generated 1,000 data sets eachof new leaf and senesced leaf observations, sampled with re-placement from the pool of Julian day visits, with the totalnumber of events equal to that in the observed data. Wethen tested whether observed distributions were statisticallydifferent from the null distribution using a Mann-Whitneytest. All data and scripts for analyses are publicly availablein a GitHub repository (https://github.com/ejedwards/amnat_2017/phenology/null_analyses).

Analysis of Viburnum-wide climate data, leaf habit, andforest type. We conducted comparative analyses using themaximum-clade credibility phylogeny from Spriggs et al.(2015). We pruned the phylogeny to the 120 species (ofthe 165 species ofViburnum) represented in our data set us-ing the drop.tip function in ape (Paradis et al. 2004). Thesespecies represent all of the named clades in the comprehen-sive phylogenetic classification of Viburnum developed byClement et al. (2014). We conducted several analyses to testthe relationship between leaf habit and climate. First, we es-timated the significance of the relationship between thesetraits using phylogenetic independent contrasts (Felsenstein1985) as implemented for use with a discrete and continu-ous character in the function crunch in the R package caper(Orme 2013), treating leaf habit as the dependent variableand either BIO6 or BIO15 as the independent, continuousvariable. Second, we inferred evolutionary shifts in leaf habitusing the threshold model first proposed by Wright (1934)and developed as an explicitly phylogenetic method by Fel-senstein (2012). The threshold model assumes that discretecharacter changes are governed by shifts in an underlying,unobserved continuous variable called the liability. As thevalue of the liability reaches a particular threshold, the bi-nary character changes. This is an appealing approach to ex-amine possible evolutionary relationships between observ-able continuous and discrete characters by evaluating therelationship between the continuous character and the esti-mated liability parameter. We estimated the liability valuesunderlying shifts in leaf habit using the function thresh-BAYES in phytools (Revell 2012), running Markov chainMonte Carlo for 106 generations, sampling every 1,000, anddiscarding the first 20% of steps as burn-in. We regressedestimated liability values to inferred ancestral estimates ofBIO6 and BIO15. Third, we utilized standard methods forassessing the correlated evolution of two binary characterswith likelihood ratio tests (Pagel 1994). In this case, we con-verted BIO6 into a binary character, binning species intowhether their mean value of BIO6 was greater or less than07C. Because six species with mean values just above or justbelow 07Cwere assigned to categories incongruent with ourfield knowledge, we created an additional binary variable

This content downloaded from 132All use subject to University of Chicago Press Term

where we adjusted these assignments. We ran Pagel testsusing both the curated and the uncurated binary variable.All data and scripts for analyses are available in a GitHubrepository (https://github.com/ejedwards/amnat_2017/final_analyses/phylogeny). All data used in all analyses are avail-able in theDryadDigital Repository: http://dx.doi.org/10.5061/dryad.k505s (Edwards et al. 2017).

Results

Phenology. The two tropical Viburnum species that we fol-lowed in Borneo (Viburnum clemensiae, Viburnum verni-cosum) are evergreen and exhibited little leaf turnover dur-ing the course of our study (fig. 3). This was especially truefor leaf senescence, with some individuals losing no morethan a single leaf over the course of 13 months. Viburnumclemensiae produced leaves from October to April and noleaves at all during the months of May–September. In con-trast, V. vernicosum produced leaves throughout the yeardespite the fact that these two species occupy the samemon-tane tropical forest, with individuals less than 1 km apart.The very low rates of leaf turnover in both species neces-sitates a much longer period of data collection to make ro-bust observations of seasonality.In both Taiwan and Japan, we included evergreen and

deciduous species in our sample, though the ratios wereinverted: in Taiwan we monitored three deciduous speciesand six evergreen, while in Japan we monitored six decid-uous and three evergreen. In each location, one of the ever-green species is perhaps better described as a leaf exchanger;Viburnum luzonicum (Taiwan) and Viburnum erosum (Ja-pan) plants shed almost all of their leaves as they produceda new flush.Across all species and all sites in Taiwan and Japan,

there was a striking coordination of the timing of new leafflushes; all species, whether deciduous or evergreen, pro-duced a significant flush of new leaves in the months ofMarch, April, and May (fig. 3; table A1). In sharp contrast,senescence patterns were more variable across species andacross the seasons. In deciduous species, as expected, leafsenescence was highly seasonal and varied little betweenspecies or locations; all deciduous species demonstratedcoordinated leaf shedding as winter approached, typicallyin November and December. In evergreen species, senes-cence periods were typically much longer, occurring inthe months preceding, during, and after spring leaf flushes(fig. 3; e.g., Viburnum propinquum, Viburnum taitoense).The tightest coordination of flushing and senescence ap-peared in V. luzonicum and V. erosum, but as noted, thesedo not fit neatly into the strictly evergreen or deciduouscategory. Indeed, they may have a short leafless periodin one year but not the next (e.g., V. luzonicum was evergreen

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

0246 0246

Apr

201

4Ju

l 201

4O

ct 2

014

Jan

2015

204060 204060 204060

0102030 0102030 0102030 0102030 0102030 0102030

0102030 0102030 01020304080 4080 4080 4080 4080 4080

V. c

lem

ensi

ae

V. v

erni

cosu

m

V. sy

mpod

iale

V. pa

rvifo

lium

V. be

tulifo

lium

V. ta

iwan

ianu

m

V. ta

itoen

se

V. p

ropi

nquu

m

V. lu

zoni

cum

*

V. in

tegr

ifoliu

m

V. o

dora

tissim

um

V. ja

poni

cum

V. e

rosu

m*

V. a

wab

uki

V. fu

rcat

um

V. d

ilata

tum

V. b

itchi

uens

e

V. p

hleb

otric

hum

V. si

ebol

dii

V. u

rceo

latu

mB

orne

o –

ever

gree

n

Taiw

an –

dec

iduo

usTa

iwan

– e

verg

reen

Japa

n –

deci

duou

s

Japa

n –

ever

gree

n

Jul 2

013

Jan

2014

Jul 2

014

Jan

2015

Jul 2

015

Jul 2

013

Jan 2

014

Jul 2

014

Jan 2

015

Jul 2

015

Jul 2

013

Jan

2014

Jul 2

014

Jul 2

013

Jan

2014

Jul 2

014

Jan

2015

Figu

re3:

Flushing

andsenescence

patterns

of20

Viburnu

mspecieslivingalon

gatrop

ical-tem

perategradient.Lines

representthe

meannu

mberof

leaves

gained

orlostperobservational

period

across

threeto

sixindividu

alsperspecies;shadingrepresents

51SE

.Green

lines

indicate

new

leaf

flushes,andyello

wlin

esrepresentleaf

senescence.

Speciesaregrou

pedby

geograph

icalarea

(Borneo,T

aiwan,Japan)andleaf

habit(evergreen,d

eciduo

us).Asterisks

indicate

speciesthat

may

bebetter

describedas

leaf

exchangers

rather

than

strictlyevergreen.

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:54 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

Consilience, Convergence, and Phylogeny S95

in 2014 but deciduous in 2015; V. erosum was deciduousin 2014 but evergreen in 2015).

The significance of these visual patterns was confirmedby comparison to our null distributions of flushing and se-nescence (table A1). All deciduous species exhibited highlysignificant clustering of both flushing and senescence. Flush-ing patterns were all significant in our evergreen species;in contrast, in most cases senescence was indistinguishablefrom our null distributions (table A1; see additional filesand analyses at https://github.com/ejedwards/amnat_2017/final_analyses/phenology/null_analyses). The three exceptionswere Viburnum japonicum in Japan and V. propinquumand Viburnum integrifolium in Taiwan. These species allexhibit a similar senescence pattern, with a period of se-nescence directly preceding and then directly following thespring leaf flush (fig. 3). We note that both V. japonicumand V. integrifolium belong to the Succotinus clade alongwith V. erosum and V. luzonicum and likely evolved an ev-ergreen habit from a deciduous ancestral state.

Correlation of phenology with climate. We chose sitesin both Taiwan and Japan that spanned freezing and non-freezing climates, reflecting differences in elevation. In bothcountries, there is a pronounced monsoonal climate, withthe heaviest rains falling during the warm summer months.As precipitation is indistinguishable across the freezing/nonfreezing boundary, we were able to compare phenologyin contrasting temperature regimes while holding rainfallpatterns constant (fig. 4).

As noted above, leaf flushes were coordinated across allsites and species in both freezing and nonfreezing zones.These flushes occurred in the spring as temperatureswarmedand monsoonal rains returned. Due to the correlation be-tween temperature and rainfall, it is not possible at this timeto discern which climatic factor the flush is primarily re-sponding to. Ancillary evidence that rainfall may be the pri-mary driver is provided by our two populations of V. lu-zonicum (fig. 5). Our high-elevation site showed a strongmonsoonal rainfall pattern and, correspondingly, a well-defined spring flush; our low-elevation site experiencedpronounced temperature seasonality but with high rainfallyear-round, and its leaf flushing patterns were considerablymore erratic (fig. 5).

In contrast to flushing, we found evidence for a single en-vironmental driver of coordinated senescence and decidu-ousness: in both Taiwan and Japan, our deciduous speciesall lived in areas experiencing routinely freezing temper-atures during the coldest months of January and February(fig. 4). In contrast, all of our evergreen species lived in areaswhere minimum monthly mean temperatures remainedabove freezing year-round. It is important to note that thisdoes not mean that they never experienced occasional frosts,only that the mean monthly minimum temperatures neverfell below 07C. Fortunately, our sampling allows us to sep-

This content downloaded from 132All use subject to University of Chicago Press Term

arate the influence of temperature from the influence ofdrought, as our evergreen and deciduous species in theseareas all experienced similar periods of relative droughtfrom November to February (fig. 4).The clear association between the deciduous leaf habit

and regular exposure to freezing temperatures but notchanges in precipitation seasonality was corroborated inourViburnum-wide analyses. Figure 1 maps themeanmin-imum temperature of the coldest week (BIO6) on a Vibur-num phylogeny derived from Spriggs et al. (2015), togetherwith ancestral reconstruction of leaf habit inferred from thethreshold analyses. This implies that the first viburnumslived in warm climates without significant freezing and thatthere were multiple independent transitions into colder (andthen back again into warmer) environments. Scanning thedistribution of the leaf habit and forest type characters givesthe impression of a strong correlation: species living inwarmer climates tend to be evergreen and to occupy trop-ical or lucidophyllous forests, while species living in colderclimates are generally deciduous and occupy cool temperateforests. This is confirmed by our independent contrast anal-yses, which showed a significant relationship between leafhabit and BIO6 (R2 p 0:607, Pp :00001) but not BIO15(R2 p 0:037, P p :3785). Likewise, our threshBAYES anal-ysis yielded a strong positive relationship between leaf habitand BIO6 (R2 p 0:868, P p 2:2e216) but not BIO15 (R2 p1:6e25, P p :9647). Similarly, our binary tests strongly fa-vored a model of correlated evolution of leaf habit and hab-itat (uncurated BIO6 binary: DAIC p 58:56, P p 1:22e213;curated BIO6 binary: DAIC p 73:2, P p 9:63e217).

Discussion

Flushing, senescence, and emergence of the deciduous habit.The deciduous habit so characteristic of Northern Hemi-sphere temperate forests involves two distinct behaviors:in the spring, there is a rapid and dramatic flush of newleaves, and in the fall, there is an equally dramatic, coor-dinated, and often colorful display of leaf senescence. Bymonitoring the flushing and senescing of leaves as indepen-dent behaviors in both evergreen and deciduous Viburnumspecies, we can begin to understand how these two aspectsof phenology relate to different climatic factors and howtheir coordination may have evolved along an evergreen-to-deciduous evolutionary trajectory.The preliminary data onViburnum phenology presented

here clearly document several patterns. Unfortunately, thevery low leaf turnover in the essentially aseasonal tropicalsite (Borneo) limits interpretation of that data set for now.We note only that our single year of data collection providessome evidence of seasonality to new leaf production (fig. 3)and that periods of leaf flushing are not uncommon in tropical

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

200

400

600

Taiw

an –

dec

iduo

us

5101020

Taiw

an –

eve

rgre

enJa

pan

– ev

ergr

een

Japa

n –

deci

duou

s

200

400

600

0.01

5

0.01

0

0.00

5

0.02

0.01

0.04

0.03

0.02

0.01

0.10

0.05

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

200

400

600

01020

00

200

400

600 1020 030

Figu

re4:

Season

aldrou

ght,freezing

,andthecoordination

offlushingandsenescence.B

oxplotsrepresentmeanmon

thly

precipitation(inmm),andlin

eplotsrepresentmeanmon

thly

maxim

umtemperature

(red)andminim

umtemperature

(blue)

in7C.D

ensity

curves

below

theclim

ateplotswerederivedfrom

poolingallnew

(green)andsenesced(orange)obser-

vation

events

from

allspecieswithinthat

category

(e.g.,evergreenJapaninclud

esViburnu

mjaponicum,V

iburnu

maw

abuk

i,andViburnu

merosum

).Clim

atedata

arefrom

onerep-

resentativefieldsite

percategory.Incaseswhere

wehadmultiplelocation

sof

evergreenor

decidu

ouspo

pulation

swithinacoun

try,theirclim

atediagramswerevirtually

indistinguish-

able.

S96

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:54 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

010203040 010203040

Jul 2

013

Jan

2014

Jul 2

014

Jan

2015

Jul 2

015

leaves flushed or senescedA B

0102030 0

200

400

600 0102030 0

200

400

600

J

F

M

A

M

J

J

A

S

O

N

D

J

F

M

A

M

J

J

A

S

O

N

D

Figu

re5:

Variation

inph

enologicalbehavior

inViburnum

luzonicum.L

ines

onleft-handgraphs

representthemeannu

mberof

leaves

gained

orlostperob

servationalperiod

across

four

tosixindividu

alsperspecies;shadingrepresents51SE

.Green

lines

representnewleaf

flushes,andyello

wlin

esrepresentleaf

senescence.C

limatediagramson

therigh

taresimilarto

figure

4.A,V

.luzonicum

popu

lation

intheDanfengregion

justou

tsideof

Taipei,Taiwan,atan

elevationof

120m.B

,V.luzon

icum

popu

lation

intheHehuanshanarea

oftheCentral

Mou

ntainsof

Taiwan

atan

elevationof

2,124m.

S97

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:54 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).

S98 The American Naturalist

plants more generally, perhaps in response to herbivore pres-sure (Coley and Barone 1996).We focus instead on our obser-vations in Taiwan and Japan, where all of our sites are mesicbut with strong rainfall seasonality. Here it is clear that a pro-nounced and coordinated leaf flush is present in all species.Senescence patterns, on the other hand, are quite varied, andleaf drop appears to be less concentrated in time. Althoughour evergreen species varied considerably, most exhibited pat-terns that were indistinguishable from a random null distribu-tion (table A1). The senescence period could be quite extendedin these species, and the intensity (i.e., the absolute numberof leaves gained or lost per event) of senescence was generallyreduced compared with leaf flushing. Our results are remark-ably similar to an earlier study of phenology in lucidophyllousJapanese forests, wheremost species produced a single springflush of leaves but presented at least four different senescencepatterns (Nitta and Ohsawa 1997). In marked contrast, ourdeciduous species all had short and coordinated senescenceperiods that commonly matched their spring flush in inten-sity. Senescence also occurred concurrently, in the late fall,in all of these species. Owing to the high level of homoplasy,the directionality of character change in Viburnum is still dif-ficult to infer in some areas, but our sample of nine deciduousspecies likely captures a minimum of five independent originsof the deciduous habit (i.e., within the Euviburnum, Pseudo-tinus, Urceolata, Solenotinus, and Succotinus clades; Clementet al. 2014; Spriggs et al. 2015). Therefore, this provides a clearcase of the convergent evolution of coordinated leaf senes-cence.

Several species stand out as notable exceptions. Two, Vi-burnum luzonicum and Viburnum erosum, are what we de-scribe as leaf exchangers, as their senescence events directlypreceded or were concurrent with their flushes. The rangeofV. luzonicum extends towarm temperate forests in south-ern China and south to the Philippines. In contrast, we sam-pled V. erosum in the warmest part of its range in Japan; itextends northward into colder forests, where it is certainlydeciduous (Hara 1983). Two other exceptions are Vibur-num integrifolium and Viburnum japonicum, two evergreenspecies that exhibit senescence patterns that appear to beslightly concentrated in two bouts, one preceding and onefollowing their annual leaf flush. Interestingly, all of thesespecies are nested within the large eastern Asian Succotinusclade and have close relatives in cool temperate forests(Clement et al. 2014; Spriggs et al. 2015).We infer that thesespecies have extended their ranges back into warmer climatesand have shifted (at least in the southern parts of their geo-graphic ranges) in the direction of an evergreen habit. With-out a phylogenetic context, one might consider a leaf ex-changing phenology to be an intermediate phenotype in thetransition from an evergreen to a deciduous habit. However,in the case of Viburnum it seems more likely to be an inter-mediate phenotype in the transition from a deciduous to an

This content downloaded from 132All use subject to University of Chicago Press Term

evergreen habit. This makes sense given that a coordinatedsenescence period had already evolved in deciduous species;in the absence of freezing, senescence may have become de-layed until it effectively co-occurred with new leaf produc-tion. A single peak of senescence might have been disruptedin shifting to a more fully evergreen condition, as evidencedby V. integrifolium and V. japonicum.A close comparison of the climates occupied by our ever-

green versus deciduous Viburnum species highlights theclimatic factor that most likely drove the initial evolutionof coordinated senescence. All of our sites in Taiwan and Ja-pan experience an annual dry season beginning inOctober–November, but in the evergreen species this is not associatedwith coordinated leaf loss; instead, coordinated senescenceappears only when plants are subjected to prolonged freez-ing. All of our deciduous species and only our deciduousspecies occur in locations where minimum monthly tem-peratures routinely dip below freezing in winter (January–February; fig. 4). This perfect correlation in our 20-speciesdata set was corroborated by our Viburnum-wide phyloge-netic analyses, which strongly supported minimum tempera-ture of the coldest week (BIO6) as a significant correlate ofshifts in leaf habit across the clade and strongly rejected anyinfluence of precipitation seasonality (BIO15). Taken together,our data indicate that periods of leaf flushing are universal inViburnum, perhaps even in the tropical forest species (wheremore data are clearly needed). A more concentrated burst ofleaf flushing may have evolved first in lucidophyllous foreststhat experience monsoonal rainfall patterns with a period ofdrought in the cool season. Specifically, we suggest that theinitiation of flushingmight be triggered by the onset of springrains (generally inApril–May), though given the obvious cor-relation, it is difficult to disentangle the relative effects of rain-fall versus temperature or even rainfall versus lengtheningphotoperiod at that time of year. In the case of leaf senescence,on the other hand, the environmental trigger seems quiteclear. Our data suggest that a regular and prolonged periodof freezing temperatures was the primary driver of highly co-ordinated leaf senescence and consequently the emergence ofa fully deciduous leaf habit.How do our results square with expectations from the

Wolfe, Axelrod, and Zanne et al. hypotheses outlined above?Although we cannot rule out Wolfe’s hypothesis that dark-ness at high latitudes favored the evolution of deciduousnessin the Late Cretaceous in some plant lineages (Wolfe andUpchurch 1986;Wolfe 1987), we doubt that this explanationapplies to Viburnum. Our phylogenetic analyses, the Vibur-num fossil record, and the absolute dates that we have in-ferred (Spriggs et al. 2015) are not consistent with Cretaceousorigins of deciduousness in Viburnum. Instead, most originsprobably occurred in the late Miocene, when temperate veg-etation was becoming more common at lower latitudes(Utescher and Mosbrugger 2007; Pound et al. 2011).

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Consilience, Convergence, and Phylogeny S99

Axelrod’s hypothesis that deciduousness evolved in warmbutmonsoonal forests with seasonal drought (Axelrod 1966)does not apply neatly to Viburnum either, considering boththe phylogenetic and phenological analyses. Breaking decid-uousness into two components—leaf flushing and leaf senes-cence—allows us to consider the possibility that these mayhave had separate evolutionary causes. Temporally restrictedand coordinated leaf flushingmight have evolved in responseto the strong rainfall seasonality of monsoonal climates, withthe onset of leafing possibly controlled by the onset of rainsor increasing temperatures in the spring. Restricted and co-ordinated leaf senescence, on the other hand, might havebeen driven by exposure to freezing temperatures. Droughtis potentially also a factor in senescence, as Axelrod sup-posed, but as we have shown, our evergreen species also ex-perience similar periods of significant drought in the fallwithout dropping their leaves. Freezing, not drought, seemsto be most directly related to coordinated senescence in ourdeciduous species.

We concur with a simple physiological explanation for adirect link between freezing and deciduousness: a coordi-nated period of senescence allows plants to remobilize keynutrients prior to the inevitable death of leaves by predict-able freezing (Feild et al. 2001; Keskitalo et al. 2005; Niine-mets and Tamm 2005). In all of the varied senescence pat-terns observed in our evergreen species, it is worth notingthat no peaks in senescence coincide with the onset of thedry season; instead, they are scattered throughout the rainyseason. It appears that viburnums can weather dry periodsby either tolerating lower water potentials or maintaininglow stomatal conductance. Amodest drought tolerance com-bined with interannual variation in the strength of droughtperiods may render drought a weaker selection pressure forcoordinated senescence, as many leavesmay be able to persistand continue to function during the following rainy season. Itis important to note, of course, that there are evergreenwoodyplants in the northern temperate zone (e.g., Rhododenron,Ilex) and that these species instead produce freezing-tolerantleaves with longer life spans.With very few exceptions (fig. 1),Viburnum has not evolved this alternative strategy, sug-gesting that deciduousnesswas themore evolutionarily acces-sible adaptation in this lineage. Investigating the very fewinstances of evergreen viburnums that experience freezing(e.g.,Viburnum rhytidophyllum, which is nested within a coldtemperate deciduous clade) might provide insights into therelative advantages of these two distinct ecological strategiesto the cold temperate zone.

In proposing that deciduousness evolved directly as anadaptation to predictable and prolonged freezing, we mightappear to be more aligned with the hypothesis of Zanneet al. (2014). These authors concluded that many lineagesfirst became established in the freezing zone as evergreenplants and that deciduousness evolved later. Our data, how-

This content downloaded from 132All use subject to University of Chicago Press Term

ever, do not support this order of events. One expectation ofsuch a stepwise model is the existence of one or more ances-trally evergreen clades living in freezing climates, with oneor more deciduous species nested within it. However, thereis no unambiguous pattern of this type in Viburnum phy-logeny. Furthermore, in the species that we monitored, ever-greens never occur in forests that experience predictablefreezing. On these grounds we judge the Zanne et al. (2014)explanation to be unlikely in this case, although the hypoth-esis could be salvaged by imagining the systematic extinc-tion of evergreen species in cold climates.An alternative model of evolution. As we have empha-

sized, the previously proposed hypotheses envision a se-quence of steps ending in the association of cold climateswith a deciduous habit. A slightly different view seems muchsimpler in the face of the evidence that we have presented forViburnum. We might consider the possibility that the decid-uous habit and in particular the coordinated senescence ofleaves in the fall evolved in some plant lineages directly in re-sponse to freezing. By “directly”wemean that the senescencecomponent of the deciduous habit evolved in situ as popula-tions were subjected to a gradually cooling climate and theestablishment of a prolonged and predictable period of freez-ing during the winter. We envision that this change in cli-mate took place in fits and starts over the course of thousandsto millions of years (Zachos 2001; Herbert et al. 2016), andthis is a timescale during which the final steps in the evolu-tion of a fully deciduous habit could certainly have evolved.Under this hypothesis the deciduous habit is not a preadap-tation that first originated as the solution to some other prob-lem, nor did it evolve in lineages that got there first andadapted later. Instead, we are suggesting that climate changeand evolutionary adaptation occurred together, effectively inlockstep with one another. This scenario is especially realisticfor an adaptation related to leafing phenology, which wehave shown to be quite sensitive to climate even within a sin-gle species across its range (fig. 5).The distinction we are making between stepwise and

lockstep may seem like a subtle one, but we think it is cru-cially important in considering biome assembly and thecauses of concerted convergence. At the very least it war-rants greater care in our use of phrases like “moved into”or “shifted into” in relation to biome transitions. Likewise,the Zanne et al. (2014) terms “climate first” and “trait first”quite explicitly describe a stepwise process—in other words,that one thing happened first and the other thing happenedlater. In the case of the emergence of the temperate decidu-ous biome, it appears that a widespread and floristically richwarm evergreen forest (the boreotropical flora sensu Wolfe1975; Tiffney 1985a, 1985b) was subjected to late Miocenecooling, eventually resulting in a climate with annual periodsof freezing temperatures. Themany existing lineages of woodyplants in these forests would have experienced this in situ di-

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S100 The American Naturalist

rectional trend. Some of them probably fared poorly, eithergoing extinct or shifting their ranges accordingly, but manyothers simply kept up with the changing times and stayedput.

What factors determine the winners and losers as largegeographical regions gradually become reconfigured andnew biomes emerge? In our case, the deciduous habit mighthave been more accessible to species that already producedresting buds and that flushed their leaves annually in re-sponse to variation in precipitation or temperature. Thekey step in such lineages would be the consolidation of leafsenescence, and it is this, we argue, that likely was a directresponse to freezing, not drought. Indeed, such predispo-sitions likely explain the repeated evolution of the decidu-ous habit within particular lineages and in general the clus-tering of the origin of certain traits in particular regions ofthe tree of life (Edwards and Donoghue 2013; Donoghueand Edwards 2014). Viburnum provides an excellent exam-ple, with possibly as many as 10 separate origins of decidu-ousness, but we suspect that other familiar temperatelineages that were present in boreotropical forests (e.g.,Acer, Carya, Nyssa, Hamamelis, Sassafras, Cornus; Tiffney1985a), when analyzed carefully, will show similar patternsof in situ convergence.

It is certainly the case that biome assembly also reflectsother processes, including preadaptation, habitat filtering(cf. Ackerly 2004), and even occasional long-distance dis-persal events (Pennington and Dick 2004; Crisp et al. 2009).We also do not doubt that all of these pathways have con-tributed to the assembly of temperate deciduous forests.However, we believe that these other phenomena may havehad a relatively minor influence as compared to the in situadaptation of multiple lineages to gradual climate change.We suggest that such concurrent responses may largely un-derpin the assembly of new biomes and that this processmay be responsible for the extremely high number of ori-gins of certain traits.

Although we have concentrated here on direct shifts be-tween mesic tropical and temperate forests, we fully appre-ciate that there have likely been other biome pathways andin particular that some members of cool temperate forestsmay have been derived within lineages that occupied sea-sonally dry tropical forests (e.g., possibly Celtis). Such apathwaywould be farmore consistent with a stepwisemodel,where deciduousness evolved first in response to one vari-able (drought) and served later as a preadaptation for an-other (cold). Nevertheless, in both pathways (i.e., with an-cestors in either mesic forests or dry forests) the final stepsin the process may have been similar. In both cases, senes-cence may have been relatively uncoordinated at first, as inthe case of dry forests, where leaves tend to senesce at differ-ent times depending largely on individual plant water status(e.g., Reich and Borchert 1984). However, freezing events are

This content downloaded from 132All use subject to University of Chicago Press Term

experienced by all individuals simultaneously, and routineexposure to prolonged cold could have tightly coordinatedthe senescence period to amore condensed time period acrossthe community.Consilience and integration in phylogenetic biology. How

do we justify our lockstep model? If we approached thiswith only our phylogenetic tools, we would not be entirelysatisfied. The high lability of both the deciduous habit andclimate makes it difficult to infer with any confidence manyof the deeper and potentially most relevant transitions(fig. 1). We do know that deciduousness and freezing cli-mate are very tightly correlated across our tree, so muchso that changes in both of them very often co-occur alongthe same branches, making it difficult to establish causeand effect (Baum and Donoghue 2001). If there had in factbeen a repeated order of events, history has largely erased itssignal. We might also be frustrated by our inability to ana-lyze both characters as continuous variables, which wouldhave provided a more direct test of a simultaneous, lockstepevolutionary model. Transforming habitat into a continu-ous climate variable was possible by using the CliMond da-tabase, but assigning continuous values to species along anevergreen-deciduous continuum is logistically and concep-tually more difficult and would require detailed phenolog-ical studies of every species, not just the 20 that we havepresented here.But what if we consider these conclusions together with

the results from the phenological study? Here we also seeevidence of a very tight association between freezing in par-ticular and the deciduous habit (fig. 4) and of the extremelability of phenological patterns, even among populationswithin a single species (fig. 5). These observations are com-pletely consistent with the strong correlation across the Vi-burnum phylogeny, and together these lines of evidencesuggest to us that the “simultaneous” changes that we seein phylogenetic analyses may often be real simultaneouschanges rather than artifacts of extinction. If individual spe-cies can be evergreen in one part of their range and decid-uous in another (as in V. erosum and V. luzonicum), grad-ual and concurrent emergence of the deciduous habit as afreezing climate becomes established begins to seem highlylikely.The study of phylogenetic comparative methods appears

to be shifting into a more introspective phase, with renewedscrutiny of the limitations of our inferences (Edwards et al.2015; Maddison and FitzJohn 2015; Pennell et al. 2015;Wright et al. 2015). Critiques typically end with a call forbetter models, and we agree completely that there is muchroom for improvement. However, at the end of the day, weneed to remind ourselves of the extreme difficulty of con-vincingly inferring evolutionary events in the distant pastwith data usually only from extant organisms, especially whenconvergence is rampant and densely clustered in particular

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Consilience, Convergence, and Phylogeny S101

clades. Under these circumstances, phylogenetic insightswill be most powerful when interpreted alongside other,quite independent lines of evidence until (hopefully) con-silience emerges (Weber and Agrawal 2012; Olson andArroyo-Santos 2015).

From this perspective, we believe that research con-ducted at the scale of Viburnum will continue to be highlyproductive, precisely because the results of multiple inde-pendent analyses canmost effectively be brought into align-ment in such model lineages. For example, the data pre-sented here are relevant to questions about the deciduousleaf habit, but they are also directly connected to our studiesof leaf form and function (Chatelet et al. 2013; Scoffoni et al.2016), growth architecture and leaf life span (Edwards et al.2014), and even flowering phenology (L. Garrison, M. Don-oghue, and E. Edwards, unpublished manuscript) and woodanatomy (D. Chatelet, M. Donoghue, and E. Edwards, un-published manuscript). Consequently, we have the poten-tial to establish connections between what might at first ap-pear to be unrelated phenomena. It is this form of synthesisthat we find most inspiring and most likely to provide newconceptual breakthroughs as we face the formidable chal-

This content downloaded from 132All use subject to University of Chicago Press Term

lenge of inferring evolutionary history. Immersion in themessy details of particular lineages, far from being a distrac-tion,may be themost fertile ground for achieving consilience—the moment when multiple, sometimes weakly supportedfindings all point to the same conclusion.

Acknowledgments

We thank A. Agrawal for his invitation to participate in the2016 American Society of Naturalists Vice PresidentialSymposium and for his patience in handling our manu-script. We appreciated insightful comments by D. Ackerly,T. Pennington, and two other reviewers, which greatly im-proved our manuscript. We are also grateful to J. Pereiraand the staff of Kinabalu National Park for their help withphenological monitoring, E. Johnson for help with datacleansing and management, and J. de Vos, J. Gauthier, M.Olson, and members of the Edwards and Donoghue labsfor many thoughtful discussions. This work was supportedby National Science Foundation grants IOS-1257262 toE.J.E. and IOS-1256706 to M.J.D.

APPENDIX

Supplementary Table

Table A1: Significance of leaf flushing and senescence patterns in 20 species of Viburnum along a latitudinal gradient in Asia

Species

Location Leaf habit Event

.23s a

Totalleaves

6.027.111 onnd Conditions

Mean Julian day(observed)

July 22, 2017 05:14:5 (http://www.journals.

Mean Julian day(null)

4 AMuchicago.edu/t-and-c).

P

V. awabuki

Japan Evergreen Leaf flush 41 89 194 2.60E-12 V. awabuki Japan Evergreen Leaf drop 48 176 195 .239 V. betulifolium Taiwan Deciduous Leaf flush 211 118 187 3.20E-25 V. betulifolium Taiwan Deciduous Leaf drop 195 336 187 1.90E-88 V. bitchiuense Japan Deciduous Leaf flush 353 102 207 4.00E-102 V. bitchiuense Japan Deciduous Leaf drop 523 305 207 2.60E-115 V. clemensiae Borneo Evergreen Leaf flush 68 110 169 4.20E-06 V. clemensiae Borneo Evergreen Leaf drop 15 183 170 4.80E-01 V. dilatatum Japan Deciduous Leaf flush 353 106 189 6.57E-39 V. dilatatum Japan Deciduous Leaf drop 352 311 189 1.39E-96 V. erosum Japan Leaf exchanger Leaf flush 240 91 195 3.54E-65 V. erosum Japan Leaf exchanger Leaf drop 205 276 195 2.67E-34 V. furcatum Japan Deciduous Leaf flush 229 134 205 1.41E-32 V. furcatum Japan Deciduous Leaf drop 341 311 205 6.47E-106 V. integrifolium Taiwan Evergreen Leaf flush 154 126 187 4.65E-12 V. integrifolium* Taiwan Evergreen Leaf drop 109 170 188 2.00E-02 V. japonicum Japan Evergreen Leaf flush 134 81 195 3.01E-47 V. japonicum* Japan Evergreen Leaf drop 95 156 195 3.28E-04 V. luzonicum Taiwan Leaf exchanger Leaf flush 205 89 187 4.31E-56 V. luzonicum Taiwan Leaf exchanger Leaf drop 133 154 187 3.37E-07 V. odoratissimum Taiwan Evergreen Leaf flush 220 151 187 1.35E-08 V. odoratissimum Taiwan Evergreen Leaf drop 91 180 186 4.26E-01

S102 The American Naturalist

Table A1 (Continued )

Species

A

Location

ll use subject to

Leaf habit

This content download University of Chicago

Event

ed from 132.23 Press Terms a

Totalleaves

6.027.111 onnd Conditions

Mean Julian day(observed)

July 22, 2017 05:14:5 (http://www.journals.

Mean Julian day(null)

4 AMuchicago.edu/t-and-c).

P

V. parvifolium

Taiwan Deciduous Leaf flush 405 99 187 2.37E-86 V. parvifolium Taiwan Deciduous Leaf drop 396 332 187 1.95E-174 V. phlebotrichum Japan Deciduous Leaf flush 198 97 202 4.50E-62 V. phlebotrichum Japan Deciduous Leaf drop 196 306 202 1.14E-57 V. propinquum Taiwan Evergreen Leaf flush 295 122 187 8.12E-33 V. propinquum* Taiwan Evergreen Leaf drop 125 165 187 4.10E-02 V. sieboldii Japan Deciduous Leaf flush 333 100 215 3.70E-105 V. sieboldii Japan Deciduous Leaf drop 332 277 215 5.33E-27 V. sympodiale Taiwan Deciduous Leaf flush 508 121 187 3.73E-53 V sympodiale Taiwan Deciduous Leaf drop 399 328 187 1.31E-156 V. taitoense Taiwan Evergreen Leaf flush 273 105 187 9.60E-47 V. taitoense Taiwan Evergreen Leaf drop 109 180 187 7.10E-01 V. taiwanianum Taiwan Evergreen Leaf flush 326 145 187 1.77E-11 V. taiwanianum Taiwan Evergreen Leaf drop 245 208 187 4.00E-01 V. urceolatum Japan Deciduous Leaf flush 96 111 203 1.62E-19 V. urceolatum Japan Deciduous Leaf drop 96 271 204 1.95E-09 V. vernicosum Borneo Evergreen Leaf flush 126 117 169 3.25E-07 V. vernicosum Borneo Evergreen Leaf drop 17 199 170 3.60E-01

Note: Species in boldface exhibited leaf senescence patterns statistically indistinguishable from our generated null distributions. Evergreen species markedwith an asterisk exhibited senescence patterns significantly different from the null at P ! .05. All species exhibited highly significant leaf flushes, occurringearlier in the year than the null expectation.

Literature Cited

Ackerly, D. D. 2004. Adaptation, niche conservatism, and conver-gence: comparative studies of leaf evolution in the California chap-arral. American Naturalist 163:654–671.

Agrawal, A. A. 2017. Toward a predictive framework for convergentevolution: integrating natural history, genetic mechanisms, andconsequences for the diversity of life. American Naturalist 190(suppl.):S1–S12.

Agrawal, A. A., and M. Fishbein. 2006. Plant defense syndromes.Ecology 87:S132–S149.

Arakaki, M., P. A. Christin, R. Nyffeler, A. Lendel, U. Eggli, R. M.Ogburn, E. Spriggs, M. J. Moore, and E. J. Edwards. 2011. Contem-poraneous and recent radiations of the world’s major succulentplant lineages. Proceedings of the National Academy of Sciencesof the USA 108:8379–8384.

Axelrod, D. I. 1966. Origin of deciduous and evergreen habits in tem-perate forests. Evolution 20:1–15.

Bailey, I. W., and E. W. Sinnott. 1916. The climatic distribution of cer-tain types of angiosperm leaves. American Journal of Botany 3:24–39.

Baum, D. A., and M. J. Donoghue. 2001. A likelihood framework forthe phylogenetic analysis of adaptation. Pages 22–44 in S. H. Or-zack and E. Sober, eds. Adaptation and optimality. Cambridge Uni-versity Press, Cambridge.

Brown, J. L. 2014. SDMtoolbox: a python-based GIS toolkit for land-scape genetic, biogeographic and species distribution model analy-ses. Methods in Ecology and Evolution 5:694–700.

Chabot, B., and D. Hicks. 1982. The ecology of leaf life spans. An-nual Review of Ecology and Systematics 13:229–259.

Chatelet, D. S., W. L. Clement, L. Sack, M. J. Donoghue, and E. J.Edwards. 2013. The evolution of photosynthetic anatomy in Vi-burnum (Adoxaceae). International Journal of Plant Sciences 174:1277–1291.

Christin, P. A., R. P. Freckleton, and C. P. Osborne. 2010. Canphylogenetics identify C4 origins and reversals? Trends in Ecologyand Evolution 25:403–409.

Christin, P. A., C. P. Osborne, D. S. Chatelet, J. T. Columbus, G.Besnard, T. R. Hodkinson, L. M. Garrison, M. S. Vorontsova, andE. J. Edwards. 2013. Anatomical enablers and the evolution of C4

photosynthesis in grasses. Proceedings of the National Academyof Sciences of the USA 110:1381–1386.

Clement, W. L., M. Arakaki, P. W. Sweeney, E. J. Edwards, and M. J.Donoghue. 2014. A chloroplast tree for Viburnum (Adoxaceae)and its implications for phylogenetic classification and characterevolution. American Journal of Botany 101:1029–1049.

Coley, P. D., and J. A. Barone. 1996. Herbivory and plant defenses intropical forests. Annual Review of Ecology and Systematics 27:305–335.

Crisp, M. D., M. T. K. Arroyo, L. G. Cook, M. A. Gandolfo, G. J.Jordan, M. S. McGlone, P. H. Weston, M. Westoby, P. Wilf, andH. P. Linder. 2009. Phylogenetic biome conservatism on a globalscale. Nature 458:754–756.

Donoghue, M. J. 1981. Growth patterns in woody plants with exam-ples from the genus Viburnum. Arnoldia 41:2–23.

Donoghue, M. J., and E. J. Edwards. 2014. Biome shifts and niche evo-lution in plants. Annual Review of Ecology, Evolution, and System-atics 45:547–572.

Donoghue, M. J., and R. H. Ree. 2000. Homoplasy and developmen-tal constraint: a model and an example from plants. American Zo-ologist 40:759–769.

Eaton, D. A. R., E. L. Spriggs, B. Park, and M. J. Donoghue. 2017.Misconceptions on missing data in RAD-seq phylogenetics with adeep-scale example from flowering plants. Systematic Biology 66:399–412.

Edwards, E. J., D. S. Chatelet, B. Chen, J. Y. Ong, S. Tagane, J. Kane-mitsu, K. Tagawa, et al. 2017. Data from: Convergence, consil-

Consilience, Convergence, and Phylogeny S103

ience, and the evolution of temperate deciduous forests. AmericanNaturalist, Dryad Digital Respository, http://dx.doi.org/10.5061/dryad.k505s.

Edwards, E. J., D. S. Chatelet, L. Sack, and M. J. Donoghue. 2014.Leaf life span and the leaf economic spectrum in the context ofwhole plant architecture. Journal of Ecology 102:328–336.

Edwards, E. J., J. M. de Vos, and M. J. Donoghue. 2015. Doubtfulpathways to cold tolerance in plants. Nature 521:E5–E6.

Edwards, E. J., and M. J. Donoghue. 2013. Is it easy to move and easyto evolve? evolutionary accessibility and adaptation. Journal of Ex-perimental Botany 64:4047–4052.

Edwards, E. J, C. Osborne, C. Stromberg, S. Smith, C4 Grasses Con-sortium. 2010. The origins of C4 grasslands: integrating evolution-ary and ecosystem science. Science 328:587–591.

Edwards, E. J., E. L. Spriggs, D. S. Chatelet, and M. J. Donoghue.2016. Unpacking a century-old mystery: winter buds and the lat-itudinal gradient in leaf form. American Journal of Botany 103:975–978.

Feild, T. S., D. W. Lee, and N. M. Holbrook. 2001. Why leaves turnred in autumn: the role of anthocyanins in senescing leaves of red-osier dogwood. Plant Physiology 127:566–574.

Felsenstein, J. 1985. Phylogenies and the comparative method.American Naturalist 125:1–15.

———. 2012. A comparative method for both discrete and contin-uous characters using the threshold model. American Naturalist179:145–156.

Fenster, C. B., W. S. Armbruster, P. Wilson, M. R. Dudash, and J. D.Thomson. 2004. Pollination syndromes and floral specialization.Annual Review of Ecology, Evolution, and Systematics 35:375–403.

Givnish, T. 2002. Adaptive significance of evergreen vs. deciduousleaves: solving the triple paradox. Silva Fennica 36:703–743.

Hancock, L., and E. J. Edwards. 2014. Phylogeny and the inference ofevolutionary trajectories. Journal of Experimental Botany 65:3491–3498.

Hara, H. 1983. A revision of the Caprifoliaceae of Japan with refer-ence to allied plants in other districts and the Adoxaceae. Acade-mia Scientific, Tokyo.

Herbert, T. D., K. T. Lawrence, A. Tzanova, L. C. Peterson, R. Caballero-Gill, and C. S. Kelly. 2016. Late Miocene global cooling and the rise ofmodern ecosystems. Nature Geosciences 9:843–847.

Hijmans, R. J., and J. van Etten. 2012. raster: geographic analysis andmodeling with raster data. R package version 2.0-12. http://cran.r-project.org/packagepraster.

Kahle, D., and H. Wickham. 2013. ggmap: spatial visualization withggplot2. R Journal 5:144–161.

Kern, J. H. 1951. The genus Viburnum (Caprifoliaceae) in Malaysia.Reinwardtia 1:107–170.

Keskitalo, J., G. Bergquist, P. Gardeström, and S. Jansson. 2005. Acellular timetable of autumn senescence. Plant Physiology 139:1635–1648.

Kikuzawa, K. 1991. A cost-benefit analysis of leaf habit and leaf lon-gevity of trees and their geographical pattern. American Naturalist138:1250–1263.

Killip, E. P., and A. C. Smith. 1931. The South American species ofViburnum. Bulletin of the Torrey Botanical Club 57:245–258.

Kriticos, D. J., B. L. Webber, A. Leriche, N. Ota, J. Bathols, I. Mac-adam, and J. K. Scott. 2012. CliMond: global high resolution his-torical and future scenario climate surfaces for bioclimatic model-ling. Methods in Ecology and Evolution 3:53–64.

This content downloaded from 132All use subject to University of Chicago Press Term

Maddison, W. P., and R. G. FitzJohn. 2015. The unsolved challengeto phylogenetic correlation tests for categorical characters. Sys-tematic Biology 64:127–136.

Meyer, J. R., D. T. Dobias, J. S. Weitz, J. E. Barrick, R. T. Quick, andR. E. Lenski. 2012. Repeatability and contingency in the evolutionof a key innovation in phage lambda. Science 335:428–432.

Morton, C. V. 1933. The Mexican and Central American species ofViburnum. Contributions of the US National Herbarium 26:339–366.

Murphy, P. G., and A. E. Lugo. 1986. Ecology of tropical dry forest.Annual Review of Ecology and Systematics 17:67–88.

Niinemets, Ü., and Ü. Tamm. 2005. Species differences in timing ofleaf fall and foliage chemistry modify nutrient resorption efficiencyin deciduous temperate forest stands. Tree Physiology 25:1001–1014.

Nitta, I., and M. Ohsawa. 1997. Leaf dynamics and shoot phenologyof eleven warm-temperate evergreen broad-leaved trees near theirnorthern limit in central Japan. Plant Ecology 130:71–88.

Olson, M. E., and A. Arroyo-Santos. 2015. How to study adaptation(and why to do it that way). Quarterly Review of Biology 90:167–191.

Orme, D. 2013. caper: comparative analysis of phylogenetics andevolution in R. R package version 0.5.2. http://cran.r-project.org/packagepcaper.

Pagel, M. 1994. Detecting correlated evolution on phylogenies: a gen-eral method for the comparative analysis of discrete characters.Proceedings of the Royal Society B 255:37–45.

Paradis, E., J. Claude, and K. Strimmer. 2004. APE: analyses of phy-logenetics and evolution in R language. Bioinformatics 20:289–290.

Pennell, M. W., R. G. FitzJohn, W. K. Cornwell, and L. J. Harmon.2015. Model adequacy and the macroevolution of angiospermfunctional traits. American Naturalist 186:E33–E50.

Pennington, R. T., and C. W. Dick. 2004. The role of immigrants inthe assembly of the South American rainforest tree flora. Philo-sophical Transactions of the Royal Society B 359:1611–1622.

Peppe, D., D. Royer, B. Cariglino, S. Oliver, S. Newman, E. Leight, R.Enikolopov, et al. 2011. Sensitivity of leaf size and shape to cli-mate: global patterns and paleoclimatic applications. New Phytol-ogist 190:724–739.

Pound, M. J., A. M. Haywood, U. Salzmann, J. B. Riding, D. J. Lunt,and S. J. Hunter. 2011. A Tortonian (Late Miocene, 11.61–7.25Ma) global vegetation reconstruction. Palaeogeography, Palaeocli-matology, Palaeoecology 300:29–45.

Rehder, A. 1908. The viburnums of eastern Asia. Pages 105–116 inC. S. Sargent, ed. Trees and shrubs. II. Vol. 2. Houghton Mifflin,Boston.

Reich, P. B., and R. Borchert. 1984. Water stress and tree phenologyin a tropical dry forest in the lowlands of Costa Rica. Journal ofEcology 72:61–74.

Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1992. Leaf life-spanin relation to leaf, plant, and stand characteristics among diverseecosystems. Ecological Monographs 62:365–392.

Revell, L. J. 2012. phytools: an R package for phylogenetic compar-ative biology (and other things). Methods in Ecology and Evolu-tion 3:217–223.

Reyes, E., H. Sauquet, and S. Nadot. 2016. Perianth symmetry changedat least 199 times in angiosperm evolution. Taxon 65:945–964.

Royer, D. L., and P. Wilf. 2006. Why do toothed leaves correlate withcold climates? gas exchange at leaf margins provides new insightsinto a classic paleotemperature proxy. International Journal of PlantSciences 167:11–18.

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S104 The American Naturalist

Sage, R. F., P. A. Christin, and E. J. Edwards. 2011. The C4 plantlineages of planet Earth. Journal of Experimental Botany 62:3155–3169.

Sanderson, M. J., and L. Hufford, eds. 1996. Homoplasy: the recur-rence of similarity in evolution. Academic Press, San Diego, CA.

Schluter, D., T. Price, A. O. Mooers, and D. Ludwig. 1997. Likeli-hood of ancestor states in adaptive radiation. Evolution 51:1699.

Schmerler, S., W. Clement, D. Chatelet, J. Beaulieu, L. Sack, M. J.Donoghue, and E. J. Edwards. 2012. Evolution of leaf form cor-relates with tropical-temperate transitions in Viburnum (Adoxaceae).Proceedings of the Royal Society B 279:3905–3913.

Schultz, T. R., R. B. Cocroft, and G. A. Churchill. 1996. The recon-struction of ancestral character states. Evolution 50:504–511.

Scoffoni, C., D. S. Chatelet, J. Pasquet-Kok, M. Rawls, M. J. Donoghue,E. J. Edwards, and L. Sack. 2016. Hydraulic basis for the evolution ofphotosynthetic productivity. Nature Plants 2:16072.

Spriggs, E. L., W. L. Clement, P. W. Sweeney, S. Madriñán, E. J.Edwards, and M. J. Donoghue. 2015. Temperate radiations anddying embers of a tropical past: the diversification of Viburnum.New Phytologist 207:340–354.

Tang, C. Q. 2010. Subtropical montane evergreen broad-leaved forestsof Yunnan, China: diversity, succession dynamics, human influ-ence. Frontiers in Earth Science China 4:22–32.

Tang, C. Q., C.-R. Chiou, C.-T. Lin, J.-R. Lin, C.-F. Hsieh, J.-W. Tang,W.-H. Su, and X. Hou. 2013. Plant diversity patterns in subtropicalevergreen broad-leaved forests of Yunnan and Taiwan. EcologicalResearch 28:81–92.

Tiffney, B. H. 1985a. Perspectives on the origin of the floristic sim-ilarity between eastern Asia and eastern North America. Journalof the Arnold Arboretum 66:73–94.

———. 1985b. The Eocene North Atlantic land bridge: its impor-tance in Tertiary and modern phytogeography of the NorthernHemisphere. Journal of the Arnold Arboretum 66:243–273.

Tiffney, B. H., and S. R. Manchester. 2001. The use of geological andpaleontological evidence in evaluating plant phylogeographic hy-potheses in the Northern Hemisphere Tertiary. InternationalJournal of Plant Sciences 162(suppl.):S3–S17.

Utescher, T., and V. Mosbrugger. 2007. Eocene vegetation patternsreconstructed from plant diversity—a global perspective. Palaeo-geography, Palaeoclimatology, Palaeoecology 247:243–271.

van Ommen Kloeke, A. E. E., J. C. Douma, J. C. Ordoñez, P. B. Reich,and P. M. van Bodegom. 2012. Global quantification of contrastingleaf life span strategies for deciduous and evergreen species in re-sponse to environmental conditions. Global Ecology and Biogeog-raphy 21:224–235.

This content downloaded from 132All use subject to University of Chicago Press Term

Weber, M. G., and A. A. Agrawal. 2012. Phylogeny, ecology, and thecoupling of comparative and experimental approaches. Trends inEcology and Evolution 27:394–403.

Weinreich, D. M., N. F. Delaney, M. A. DePristo, and D. L. Hartl.2006. Darwinian evolution can follow only very few mutationalpaths to fitter proteins. Science 312:111–114.

Wickham, H. 2016. Getting started with ggplot2. Pages 11–31 inggplot2. Springer, New York.

Wilson, P., A. D. Wolfe, W. S. Armbruster, and J. D. Thomson. 2007.Constrained lability in floral evolution: counting convergent originsof hummingbird pollination in Penstemon and Keckiella. New Phy-tologist 176:883–890.

Wolfe, J. 1995. Paleoclimatic estimates from Tertiary leaf assem-blages. Annual Review of Earth and Planetary Sciences 23:119–142.

Wolfe, J. A. 1975. Some aspects of plant geography of the NorthernHemisphere during the Late Cretaceous and Tertiary. Annals ofthe Missouri Botanical Garden 62:264–279.

———. 1987. Late Cretaceous-Cenozoic history of deciduousnessand the terminal Cretaceous event. Paleobiology 13:215–226.

Wolfe, J. A., and G. R. Upchurch. 1986. Vegetation, climatic and flo-ral changes at the Cretaceous-Tertiary boundary. Nature 324:148–152.

Wright, A. M., K. M. Lyons, and M. C. Brandley. 2015. Which camefirst: the lizard or the egg? robustness in phylogenetic reconstruc-tion of ancestral states. Journal of Experimental Zoology 324B:504–516.

Wright, I. J., P. B. Reich, M. Westoby, D. D. Ackerly, Z. Baruch, F.Bongers, J. Cavender-Bares, et al. 2004. The worldwide leaf eco-nomics spectrum. Nature 428:821–827.

Wright, S. 1934. An analysis of variability in number of digits in aninbred strain of guinea pigs. Genetics 19:506–536.

Yang, Q.-E., and V. Malécot. 2011. Viburnum. Pages 570–611 in Z.-Y.Wu, P. H. Raven, and D. Y. Hong, eds. Flora of China. Vol. 19.Lentibulariaceae through Dipsacaceae. Missouri Botanical Gar-den, St. Louis.

Zachos, J., M. Pagani, L. Sloan, E. Thomas, and K. Billups. 2001.Trends, rhythms, and aberrations in global climate 65 Ma to pres-ent. Science 292:686–693.

Zanne, A. E., D. C. Tank, W. K. Cornwell, J. M. Eastman, S. A. Smith,R. G. FitzJohn, D. J. McGlinn, et al. 2014. Three keys to the radia-tion of angiosperms into freezing environments. Nature 506:89–92.

Symposium Editor: Anurag A. Agrawal

.236.027.111 on July 22, 2017 05:14:54 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

vol . 1 90 , supplement the amer ican natural i st august 20 1 7

Sympos ium

Geographical Variation in Community Divergence:

Insights from Tropical Forest Monodominance

by Ectomycorrhizal Trees*

Tadashi Fukami,1,† Mifuyu Nakajima,1 Claire Fortunel,2 Paul V. A. Fine,3

Christopher Baraloto,4 Sabrina E. Russo,5 and Kabir G. Peay1

1. Department of Biology, Stanford University, Stanford, California 94305; 2. Department of Ecology and Evolutionary Biology, Universityof California, Los Angeles, California 90095; 3. Department of Integrative Biology and University and Jepson Herbaria, University ofCalifornia, Berkeley, California 94720; 4. L’Institut National de la Recherche Agronomique, Unité Mixte de Recherche Ecologie des Forêtsde Guyane, Kourou, French Guiana; and Department of Biological Sciences and International Center for Tropical Botany, FloridaInternational University, Miami, Florida 33199; 5. School of Biological Sciences, University of Nebraska, Lincoln, Nebraska 68588

Online enhancements: supplementary material. Dryad data: http://dx.doi.org/10.5061/dryad.c0kr7.

abstract: Convergence occurs in both species traits and commu-nity structure, but how convergence at the two scales influences eachother remains unclear. To address this question, we focus on tropicalforest monodominance, in which a single, often ectomycorrhizal (EM)tree species occasionally dominates forest stands within a landscapeotherwise characterized by diverse communities of arbuscular mycor-rhizal (AM) trees. Such monodominance is a striking potential ex-ample of community divergence resulting in alternative stable states.However, it is observed only in some tropical regions. A diverse suiteof AM and EM trees locally codominate forest stands elsewhere. Wedevelop a hypothesis to explain this geographical difference using asimulation model of plant community assembly. Simulation resultssuggest that in a region with a few EM species (e.g., South America),EM trees experience strong selection for convergent traits that matchthe abiotic conditions of the environment. Consequently, EM speciessuccessfully compete against other species to form monodominantstands via positive plant-soil feedbacks. By contrast, in a region withmany EM species (e.g., Southeast Asia), species maintain divergenttraits because of complex plant-soil feedbacks, with no species havingtraits that enablemonodominance. An analysis of plant trait data fromBorneo and Peruvian Amazon was inconclusive. Overall, this workhighlights the utility of geographical comparison in understanding therelationship between trait convergence and community convergence.

Keywords: community assembly, mycorrhizae, plant-soil feedback,plant traits, priority effects, species pools.

* This issue originated as the 2016 Vice Presidential Symposium presented atthe annual meetings of the American Society of Naturalists.† Corresponding author; e-mail: fukamit@stanford.edu.ORCIDs: Fukami, http://orcid.org/0000-0001-5654-4785; Nakajima, http://

orcid.org/0000-0001-7778-2029.

Am. Nat. 2017. Vol. 190, pp. S105–S122. q 2017 by The University ofChicago. 0003-0147/2017/190S1-57233$15.00. All rights reserved.DOI: 10.1086/692439

This content downloaded from 132All use subject to University of Chicago Press Term

Introduction

At the species level, convergence is defined by distantly re-lated species sharing similar traits, but convergence is alsopossible at the community level, in which distantly locatedcommunities develop to have similar species abundance dis-tribution, functional group composition, and other aspectsof community structure (Fukami 2009). These forms ofconvergence are thought to result from the predictable re-sponse of independently evolved species (e.g., Reich et al.1997; Conte et al. 2012) and separately assembled commu-nities (e.g., Samuels and Drake 1997; Li et al. 2016) to simi-lar environmental conditions. Identifying factors that pro-mote or prevent convergence is therefore fundamental tothe understanding of predictability at both the species andcommunity levels of biological organization (Stern 2013).Despite this duality of convergence, the link between traitconvergence and community convergence remains poorlyinvestigated (Cavender-Bares et al. 2009). It may seem ob-vious that trait convergence automatically translates intocommunity convergence (Melville et al. 2006). As we arguein this article, however, the link may not be so straightfor-ward, particularly when multiple lineages of species con-verge in traits among themselves but diverge frommembersof other lineages, which can promote divergence—ratherthan convergence—of community structure.How does community convergence and divergence take

place? Insights on this question can be found as early asClements, who developed the climax concept of plant suc-cession (e.g., Clements 1936). According to this well-knownconcept, communities converge to a predictable species com-position determined by the abiotic environment (but seeGleason 1927). More recently, increasing evidence suggeststhat community divergence—in the form of alternative sta-

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S106 The American Naturalist

ble states—may be more common than previously thought(Schröder et al. 2005), but understanding when alternativestable states emerge remains a major challenge for commu-nity ecologists (Petraitis 2013). Local communities are con-sidered to be in alternative stable states when they divergein species composition, even though the communities sharethe same environmental conditions and the same speciespool (Fukami 2015). This divergence is driven by priority ef-fects, in which the order or initial abundance in which spe-cies arrive influences the effects that species have on one an-other in local communities (Palmgren 1926; Sutherland1974; Drake 1991). As factors determining when alternativestable states occur, more attention has been paid to the abi-otic conditions of local habitats than to the trait values of po-tential colonists in the regional species pool (but see, e.g.,Fox 1987; Fargione et al. 2003). Consequently, the connec-tion between convergence in species traits and in communitystructure remains poorly understood. In this article, we ex-plore this connection, using tropical trees and their associa-tion with mycorrhizal fungi as an illustrative example.

Mycorrhizal Association and Forest Monodominance

A classic case of convergent evolution in plants is their as-sociation with mycorrhizal fungi (Brundrett 2002). Thereare at least 19 evolutionarily independent origins of ecto-mycorrhizal (EM) symbiosis in plants (Koele et al. 2012)and more than 60 in fungi (Tedersoo et al. 2010). In addi-tion, plants associated with EM fungi may potentially expe-rience common selective pressure and, because of it, undergofurther convergent evolution in other traits, particularly thoserelated toresourceeconomy(e.g.,Read1991;Cornelissen et al.2001; Phillips et al. 2013; but see Koele et al. 2012).

As a result, EM host plant species may form a species al-liance (sensu Van Nes and Scheffer 2004). That is, EM hostplants may modify the local environment to make it morefavorable to the members of their own alliance (EM plantspecies) than to those of other plants, which are often asso-ciated with arbuscular mycorrhizal (AM) fungi. Thus, if bychance a locality were initially dominated by EM plants, itmight persist as such owing to positive feedbacks resultingfrom this alliance. Consequently, it is possible that localplant communities diverge as alternative stable states, eachdeveloping as either EM or AMdominated as a result of pri-ority effects, as we explain in more detail below.

A particularly striking example of such alternative stablestates is seen in the phenomenon of tree species monodom-inance in tropical forests, in which forests otherwise char-acterized by highly diverse plant communities are dottedby occasional stands in which one tree species is far moredominant than any other (Richards 1952; Janzen 1974; Con-nell and Lowman 1989; Hart et al. 1989; Torti et al. 2001). Intropical forests, such stands can range in area from one

This content downloaded from 132All use subject to University of Chicago Press Term

to several thousand hectares (Peh et al. 2011). The term“monodominance” has been variously defined in the trop-ical forest literature, but here we consider those cases inwhich one species contributes more than 60% of total basalarea (Hart et al. 1989) or more than 25% of total stem den-sity of a forest stand that is surrounded by mixed forest.There are a number of hypothesized mechanisms by

which monodominance can emerge (Torti et al. 2001; Fred-erickson et al. 2005; McGuire et al. 2008; Peh et al. 2011),but one factor that is frequently associated with mono-dominance is the mycorrhizal status of the plants (Hartet al. 1989). Tree species that host any type of mycorrhizacan form monodominant stands. For example, multiple spe-cies from primarily AM plant families (Brundrett 2009)—such as the Apocynaceae, Burseraceae, Euphorbiaceae, Faba-ceae, Lauraceae, Moraceae (reviewed in Peh et al. 2011), andRubiaceae (Frederickson et al. 2005)—formmonodominance.However, monodominance-forming species are dispropor-tionately from EM lineages within the Dipterocarpaceae, Fa-baceae, Fagaceae, and Juglandaceae (reviewed in Peh et al.2011; Smith et al. 2013; Corrales et al. 2016). One definingfeature of monodominant stands of EM host tree species inNeotropical rainforests is that they are embedded in forestwith a more even abundance distribution of primarily AMhost trees. EM host monodominance in these forests is par-ticularly dramatic, given that far fewer tree species form EMcompared with AM associations in typical mixed forest standsand that EM host species are often found at low abundancesoutside of the monodominant patches (Henkel 2003; Cor-rales et al. 2016).The mechanisms that cause mycorrhizal status to gener-

ate monodominance are not fully understood but likely in-volve positive plant-soil feedback as one form of priority ef-fects. EM monodominant stands are thought to result inpart from positive feedback caused by their EM associa-tion, in which local soil conditions are made more favor-able to EM than to AM host trees (Dickie et al. 2014). Thesechanges in local soil conditionsmay involve the species com-position of mycorrhizal fungi and other soil microbes orplant-induced nutrient depletion (Corrales et al. 2016; Peay2016). Priority effects driven by such self-enforcing habitatmodification—or niche construction (sensu Odling-Smeeet al. 2003)—can lead to alternative stable states, in whichEM trees may form monodominant stands when they es-tablish earlier than other species after local disturbance butare otherwise excluded by AM trees.Intriguingly, however, these potential instances of alterna-

tive stable states with either high-diversity AM assemblagesor low-diversity EM monodominant stands are observed inonly some tropical regions (e.g., South America and Africa)but not in others (e.g. Southeast Asia), where a diverse suiteof AM and EMplant species codominate. To our knowledge,no study has directly addressed why this difference exists,

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Tropical Forest Monodominance S107

yet this geographical contrast presents itself as an opportu-nity to gain a better understanding of how and when mono-dominance may arise as an alternative stable state.

Natural History of Mycorrhizal Associationas a Motivation for This Article

The purpose of this article is to develop a biogeographicalhypothesis to explain why EM monodominance arises insome tropical forest regions and not in others. More gener-ally, our goal is to use monodominance as a case in point todiscuss how community convergence can be influenced bytrait convergence and how the extent of this effect can bemediated by the properties of the species pool.

Our motivation for this goal comes from the idea thatEM monodominance may be associated with convergentvalues of plant traits related to resource economy, such asleaf chemistry and decomposability (Torti et al. 2001; Pehet al. 2011). Traits of EM host plant species (such as specificleaf area, leaf C and N content, and C∶N ratio) have beenfound to differ significantly from those of AMhost plant spe-cies (Read 1991; Phillips et al. 2013), suggesting that severallineages of EM plant species may have converged and, in theprocess, as a functional group diverged from AM plant spe-cies in these traits. Even so, there can still be substantial traitvariation across plant species within the EM group (Koeleet al. 2012), which may affect the strength and direction ofplant-soil feedbacks. For example, interspecific variation inleaf C∶N ratios may increase the variation in the strengthand direction of plant-soil feedbacks among EM species (Keet al. 2015).

In this article, we seek to link the possibility that inter-specific trait variation is associated with variation in plant-soil feedbacks with one aspect of forest communities that ishighly variable among tropical regions: the size of the localpool of tree species that host EM fungi (Peay 2016). Thissize ranges from only a few species in South America (e.g.,Baraloto et al. 2011) to hundreds in Southeast Asia (e.g.,Brearley 2012). We focus on this difference among regionsin species pool size and the associated potential variationin the strength of species interactionsmediated by plant-soilfeedbacks. We hypothesize that a diverse EM species pooland the resulting complexity in local interactions amongEM speciesmakes trait convergence in these species difficultand that this obstacle to trait convergence in turn preventsthe strong divergence of local communities that is necessaryfor the emergence of monodominant patches.

To develop this hypothesis, we use results from computersimulation of plant community assembly. The simulationmodel is not intended to replicate actual community assem-bly to quantitatively predict community patterns, but insteadto explore possible qualitative outcomes arising from a smallset of assumptions that characterize species interactions me-

This content downloaded from 132All use subject to University of Chicago Press Term

diated by plant-soil feedbacks. We also examine whetherempirical leaf functional trait data are consistent with the hy-pothesis developed with our model. Variation in leaf func-tional traits reflects selection for ecological strategies of re-source acquisition and conservation (Donovan et al. 2011;Reich 2014) and is correlated with mycorrhizal status (Read1991; Phillips et al. 2013). The leaf data we use for this pur-pose are from two typical mixed forests in different geo-graphic regions. One is from a Neotropical forest in the Pe-ruvian Amazon, where EM plants are rare and the EM plantspecies pool is small. The other is from a Paleotropical forestin Borneo, where EM plants are common and the EM plantspecies pool is large.

Simulation Methods

Overview

The simulation model we used is a modified version of Fu-kami and Nakajima’s (2013) individual-based, spatially im-plicitmodel of plant community assembly through stochasticsequential immigration of species from an external speciespool. In this model, which is built on traditional plant com-petition models (Chesson 1985; Pacala and Tilman 1994;Mouquet et al. 2002; Fukami and Nakajima 2011), plant in-dividuals compete for local resources during the individ-ual establishment stage. They also affect one another’s com-petitiveness via plant-soil feedback. We assume that plantscan take on one of two mycorrhizal associations, EM orAM. There are other potential mycorrhizal states (e.g., non-mycorrhizal or ericoid mycorrhizal), but EM and AM asso-ciations represent most tree species and individuals in trop-ical forests (McGuire et al. 2008; Brundrett 2009). We alsoassume that plants that host both EM and AM fungi can beconsidered primarily EM associated, because the root sys-tems of these plants tend to become dominated by EM fungi(Egerton-Warburton and Allen 2001).We vary the size of the EM host species pool (the number

of EM host species) while holding constant the number ofAM species in the pool, and we examine how the size of theEM species pool influences the way local plant communitiesassemble, particularly with respect to the emergence of thealternative stable state of EMmonodominance. By replicat-ing community assembly in many local forest patches withthe same species pool, we examine how species abundancedistributions may vary among patches that have the sameset of local abiotic conditions within them but differ in thehistory of stochastic species arrival. In particular, we assessthe conditions under which the local abundance of EM hostsshows strong bimodality, in which a single EM host speciesbecomes monodominant in some local patches but remainsrare or absent in all other patches.As in Fukami and Nakajima’s (2013) model, plant re-

cruitment occurs in sites that are arranged in a patch. Plants

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S108 The American Naturalist

are recruited from both seeds produced by established indi-viduals within the patch and immigration from an externalspecies pool. For immigration, species of plants were ran-domly chosen for immigration each time step from the spe-cies pool. The chosen species immigrated as a small numberof propagules to a local forest patch consisting of 1,000 re-cruitment sites. One thousand recruitment sites roughly cor-respond to the sapling abundance we may expect in a hect-are, the scale at which small monodominant stands areobserved. All recruitment sites were initially empty. Subse-quently, only one plant individual could establish and pro-duce propagules in each recruitment site, even when multi-ple propagules arrived from the species pool or from localdispersal within the patch. Propagules were distributed ran-domly to recruitment sites within the patch each time step.Of the propagules that arrived at a recruitment site, the onethat belonged to the species that best fit the environmentalcondition of the recruitment site could establish and pro-duce propagules (determination of environmental fit is de-scribed in more detail below). All plants within a patch pro-duced propagules once each time step (e.g., each year oreach mast event) until they died. All plants died with a fixedprobability, and when they did, recruitment sites becameempty and available for a new plant individual to establish.Real tree species vary in dispersal ability, fecundity, andmor-tality rates, but we kept these constant across species to focuson the mechanisms of monodominance in relation to ourhypothesis. This process of immigration, arrival, establish-ment, reproduction, and death was repeated for 400 timesteps. All simulations were carried out using Mathematica8.0 (Wolfram Research, Champaign, IL). Code and data areavailable as supplementary material, available online.1

Species Pools

Regional species pools each contained 100 AM plant spe-cies and one to 20 EM plant species. Each species i was de-fined by a value Ri, which can be thought of as a multitraitphenotype that determines how well species perform in lo-cal abiotic conditions during the recruitment stage (see de-tails below). Values of Ri were chosen randomly between 0and 1 from a beta distribution, where the probability den-sity for value x was proportional to xa21(12 x)b21 (Mouquetet al. 2002).We set a p b, which causes the probability den-sity for Ri to have a peak at 0.5, but the value of a was variedin order to examine the effect that the amount of variationin Ri among species had on community assembly. We re-fer to 1=a as interspecific phenotypic variability v. A largerv means that Ri values are more variable among species(fig. 1B). In addition to Ri, species were also characterized

1. Code that appears in the American Naturalist is provided as a convenienceto the readers. It has not necessarily been tested as part of the peer review.

This content downloaded from 132All use subject to University of Chicago Press Term

by a set of values (Sij) that defined the strength of plant-soilfeedback, in which the presence of a plant individual be-longing to species j at a recruitment site during a given timestep changed the competitive ability of a plant belonging tospecies i at that recruitment site during the following timestep (see details below).

Local Patches

Local forest patches each consisted of 1,000 recruitmentsites. Each recruitment site could accommodate only oneadult tree and was characterized by a value that representsabiotic conditions (Hk), chosen randomly between 0 and 1from a beta distribution. For the beta distribution forHk, weset a p b p 2, which—as with the case of Ri—produces a

Arbuscularmycorrhizal

trees

Ecto-mycorrhizal

trees1

1

1.1 if i = j[1, 1.1] if i ≠ j

1.1 if i = j[1, 1.1] if i ≠ j

A

B

v = 1

v = 0.1

v = 0.01

v = 0.0001

Freq

uenc

y

Species value, Ri (solid lines) orabiotic condition, Hk (dashed line)

0.2 0.4 0.6 0.8 1.00

1

2

3

4

Figure 1: Model description. A, Schematic summary of the strengthof plant-soil feedback (Sij) assumed in the simulation model. The val-ues indicated were used for the simulation shown in figures 2–4 butchanged as specified in figure 5 to evaluate the effect of feedbackstrength on community assembly. An Sij of 1 means no feedback. AnSij value that is 11 means positive feedback. All intraspecific feedbacks(i.e., i p j) were assumed to have Sij of 1.1, and Sij values for interspe-cific feedbacks (i.e., i ( j) were assigned a random value between 1 and1.1, so that they are all positive, but their strengths vary with speciesidentity and are never greater than intraspecific feedbacks. B, Probabil-ity distribution for abiotic condition of recruitment sites (Hk; dashedline) and for species phenotypic value (Ri) among ectomycorrhizalspecies in the species pool under different degrees of species variabil-ity (v, which is 1=b for the ß distribution for Ri).

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Tropical Forest Monodominance S109

peak of Hk values at 0.5 (fig. 1B). For all simulations in arun, the Hk values remained the same.

Competitive Ability

Values of Hk, Ri, and Sij together determined competitiveability (Cijk) of species i at recruitment site k when the re-cruitment site was occupied by species j at the previoustime step:

Cijk p (12 jHk 2 Rij)Sij:

Thus, assuming that Sij p 1 (i.e., no effect of species j andspecies i through plant-soil feedback, as explained below),a species would have a high value of Cijk if it had a value ofRi that is close to that of Hk, as in Mouquet et al.’s (2002)model. In our simulation, a close match between Hk and Ri

is most likely when a species’ Ri value is 0.5, since Hk alsohas a peak at 0.5 (fig. 1B).

Plant-soil feedback moderates competitive ability, withthe value of Sij defining the direction and strength of thefeedback. Specifically, Sij defines the effect of species j occu-pying a given recruitment site during a given time step onthe competitive ability of species i in that microsite duringthe following time step. Thus, Sij affects the competitiveability of species i independently of both recruitment sitecondition (as defined by Hk) and species phenotype (as de-fined by Ri) and represents the amount by which the differ-ence between the abiotic environment and the species’matchto the environment is improved (Sij 1 1), worsened (Sij ! 1),or not affected (Sij p 1) via plant-soil feedback (Fukami andNakajima 2013). That is, if Sij p 1, there is no net effect ofplant-soil feedback, whereas Sij 1 1 and Sij ! 1 represent pos-itive and negative plant-soil feedback, respectively.

We set the values of Sij to attempt to represent the inter-actions within and between AM and EM plants that are ob-served in nature (fig. 1A). There are many ways in whichplant-soil feedback could be structured, but here we assumethat plants facilitated the growth of conspecific individualsvia local accumulation of specific mycorrhizal fungi that arebeneficial to them, via modification of soil nutrient com-position to the species’ own benefit via their specific mycor-rhizal association (Corrales et al. 2016) or via better protec-tion from pathogens through added physical or chemicaldefenses in mycorrhizal roots (Duchesne et al. 1989; News-ham et al. 1995; Bennett et al. 2017). Plants can also nega-tively affect conspecifics by, for example, accumulating soilpathogens (Kulmatiski et al. 2008). However, studies thatexamine net effects of pathogens plusmycorrhizae often findoverall positive effects from soil biota (e.g., fig. 1B inManganet al. 2010; see also Cortois et al. 2016; Bennett et al. 2017;

This content downloaded from 132All use subject to University of Chicago Press Term

Teste et al. 2017). Similarly, in a large analysis of forest plotsin North America and New Zealand, Dickie et al. (2014)found that most tree species exhibited positive density de-pendence. We therefore assumed in our model that positiveeffects outweigh negative ones to result in net positive plant-soil feedback for conspecifics. Accordingly, we set Sij p 1:1when i p j.We also assumed that EM tree species shared similar ni-

trogen economy (Corrales et al. 2016) as well as similar as-sociations with EM fungi. As a result, EM plant species canbenefit each other to some extent, but the strength of in-terspecific plant-soil feedback is variable among EM plantspecies pairs and is never greater than that of intraspecificfeedback. We believe that these assumptions are plausiblebecause evidence suggests that EM fungi tend to be gener-alists with high overlap between co-occurring species (Ken-nedy et al. 2003; Peay et al. 2015; but see Bennett et al. 2017)and that shared EM fungi can facilitate interspecific recruit-ment (Horton et al. 1999; Nara 2006b). In addition, there isevidence that different combinations of EM fungi and plantscan have different effects on plant growth, even among closelyrelated species (Nara 2006a; Fransson et al. 2015). Conse-quently, when i ( j, we assigned Sij values by taking a uni-form randomly drawn value between 1 and 1.1 for each Sij.For simplicity, we assume that this same pattern of intra-and interspecific feedback strengths also holds for AM plantspecies. In other words, EM and AM host species have sim-ilar capacities for positive plant-soil feedback in our model.Finally, we also assumed that EM and AM plant species didnot facilitate each other through plant-soil feedback. Thus,Sij p 1betweenall pairs of anEMspecies and anAMspecies.The direction and strength of plant-soil feedback in trop-

ical forests have not been thoroughly characterized, and it isuncertain whether the assumptions specified above are al-ways realistic. To examine whether model results are sensi-tive to these assumptions, we ran two sets of additional sim-ulation. In the first, we assumed that among EMhost plants,Sij took a uniform randomly drawn value between 1 and 1.1,even when i p j, in order to relax the assumption that allintraspecific feedbacks were more or equally positive com-pared with any interspecific feedbacks (fig. S1B; figs. S1–S5 are available online). In a second set, we used a combinationof alternative assumptions. Specifically, we assumed (1) var-iable positive intra- and interspecific feedbacks for EM hostplants, (2) variable intra- and interspecific feedbacks (bothpositive and negative) for AM host plants, and (3) negativeeffects of EM host plants on AM host plants. For assump-tion 1, we tried two variants. In one (variant 1), we usedthe original assumption as above, that is, that the strengthof interspecific plant-soil feedback is variable among EMplant species pairs and is never greater than that of intra-specific feedback (fig. S1C). The other (variant 2) is that in-terspecific and intraspecific feedbacks are all positive but

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S110 The American Naturalist

equally variable, such that some intraspecific feedbacks areless positive than some interspecific feedbacks (fig. S1D).

Community Assembly

At each time step, each species in the species pool immigratedto the local patch with probability I. For each of the 100 AMspecies in the pool, we used I p 0:05. For each EM species,we used I p 1=(number of EM species in the pool). Forexample, when we had two or 20 EM species in the pool,I was 0.5 or 0.05 for each EM species, respectively. Thisway, we standardized for the total immigration frequencyfor all EMplants across the gradient of the size of theEMspe-cies pool. We also did additional simulation in which I p0:05 for each species, regardless of the size of the EM speciespool used, in order to decouple the effects of regional speciesrichness per se from those of regional relative abundance.

At each recruitment site in the local patch, spe-cies i arrived with probability 12 exp[2(Pi 1 FNi)=(total number of recruitment sites, i.e., 1,000)] at each timestep. Here, Pi is the number of propagules of species i thatimmigrate from the species pool (20 propagules for specieschosen for that time step for immigration from the speciespool, and 0 seed for all other species), F is fecundity (50 forall species), and Ni is the number of plants belonging to spe-cies i in the local patch (0 for all species in the first time step,i.e., at t p 1, which is analogous tomodeling recruitment af-ter a stand-replacing disturbance). The specifications werechosen so as to have values that likely fall within a plausiblerange. When the number of recruitment sites that were as-signed to receive a propagule of species i exceeded Pi 1 FNi

(which rarely happens), Pi 1 FNi recruitment sites wererandomly selected from these recruitment sites, and a prop-agule of the species were assigned only to the selected re-cruitment sites.

Given this probability, there were three possibilities re-garding plant establishment and seed production in each re-cruitment site. First, if the recruitment site were already oc-cupied by a plant, that plant remained there. In other words,we assumed that seeds could not displace established adults.Second, if the recruitment site were empty and no speciesarrived at that recruitment site, it remained empty. Third, ifthe recruitment site were empty and one or more species ar-rived at that recruitment site, of those species that arrived,the one with the greatest value of Cijk (independent of thenumber of propagules of each species) was assumed to occupythe recruitment site and produce propagules starting the fol-lowing time step. The seed-to-adult recruitment processesare important in many of the dynamics modeled here, butprocesses operating during those stages per se were not thefocus of our model, which was intended to identify the min-imal processes required to produce monodominance.

This content downloaded from 132All use subject to University of Chicago Press Term

After plant establishmentwas completed for all recruitmentsites, plants occupying a recruitment site died with probabil-itym, which was 0.4 for all species. Thus, competitive abilityin our model is not related to dispersal, fecundity, or baselinesurvival; it dictates only a species’ ability to win a recruitmentsite.We assembled communities for 400 time steps. From vi-sual inspection of results, 400 time steps seemed long enoughfor most communities to reach an equilibrium state (fig. 2).

Manipulating Species Pool Size and Species Variability

Wemanipulated two factors—the number of EM species inthe species pool and the amount of variation in Ri among

Loca

l abu

ndan

ce

Time step

A

B

100 200 300 400

200

150

100

50

200

100

300

400

Figure 2: Examples of simulated population dynamics for two forestpatches, showing ectomycorrhizal (EM) monodominance (A) andarbuscular mycorrhizal (AM) dominance (B) as alternative stablestates. Both A and B show results for one instance of local commu-nity assembly under the same species spool, and each line indicates aspecies, with warm colors denoting EM tree species and cold colorsdenoting AM tree species. In these examples, there were 100 AM treespecies and 10 EM tree species in the species pool, and species variabil-ity amongEM species (v) was set to be small (v p 0:0001), with all EMspecies having an Ri value very close to the optimal, 0.5 (see fig. 1B).Depending on assembly history, local communities develop either asEMmonodominance (A), in which a single EM species becomes dom-inant (orange line) and all others are AM species, or as an AM-onlycommunity (B), in which a few EM species were able to colonize theforest patch initially but became locally extinct because they werenot common initially and were excluded competitively by AM speciesthrough positive plant-soil feedback among AM species.

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Tropical Forest Monodominance S111

EM host species (i.e., species variability v, as defined in“Species Pools”)—to examine their effects on EM abun-dance in assembled local communities. To this end, undereach of the species pools we used (i.e., 16 pools in fig. 3),we ran the simulation 100 times, with each run consideredone instance of community assembly in a local forest patch.Under each species pool, the replicated 100 patches had thesame set of recruitment sites, which allowed us to evaluatewhether plant communities that develop in different patchesthat share the same set ofHk values in them diverged in spe-cies composition as a result of random variation in the his-tory of species arrival from the species pool. If they did di-verge, that would be evidence for alternative stable states.We were particularly interested in assessing the conditionsunder which EM monodominance arose. We defined EMmonodominance as the case in which only one dominantEM species occupied one-fourth (25%) or more of the avail-able recruitment sites in a patch, with the other sites beingoccupied by AM species.

Simulation Results and Discussion

Simulation results verified that communities with (fig. 2A)and without (fig. 2B) monodominant EM species could de-velop as alternative stable states in our model. Under thesame species pool, communities were either dominated byAM species without any EM species present or had at least25% EM individuals, representing two alternative stablestates, similar to those seen in some Neotropical rainforests.

By varying species pool size and phenotypic variability(fig. 3), we found that EM monodominance arose as an al-ternative state whenever trait variability among EM specieswas low. In these cases, all EM species in the species poolhad an Ri value that was nearly optimal for the abiotic con-ditions available in the forest patch. Under this condition, asingle dominant EM species was competitive enough to formmonodominantstandswhenitarrivedearlyduringlocalcom-munity assembly but not when it arrived late, as indicated bythe strongly bimodal pattern in the top row in figure 3. Oneof the twomodes had no EMspecies, and the othermode hadonly one EM species, even when the species pool containedmultiple EM species.

Regardless of species pool size, as Ri values became morevariable among EM species in the local species pool (i.e.,moving fromtop tobottomrows infig. 3), theEMtotal abun-dance became more variable among forest patches, and thebimodal pattern became increasingly obscure. The frequencyof monodominance declined with greater phenotypic vari-ability when there was more than one EM species in the poolbecause, although thenumberof EMtrees in apatchmight behigh, these patches consisted of multiple EM species.

The additional simulation indicated that these results gen-erally held under the alternative assumptions examined, but

This content downloaded from 132All use subject to University of Chicago Press Term

only when all intraspecific feedbacks were more or equallypositive compared with interspecific feedbacks (fig. S1C).Otherwise, multiple EMplant species often coexisted as a re-sult of mutual facilitation among them, with only a smallnumber of monodominance patches developing (fig. S1B,S1D). Therefore, one condition for monodominance forma-tion in our model is stronger positive intraspecific relativeto interspecific feedback among EM trees, which is consis-tent with some recent empirical data (Bennett et al. 2017).Strong intraspecific feedback is possible if, for example, atree species cultivates an assemblage of highly beneficial EMfungi. The additional simulation in which immigration rateI was 0.05, regardless of the size of the EM species pool, in-dicated that the number of monodominant patches is posi-tively correlated with the immigration rate per EM species(fig. S2A, S2C).Together, these results suggest that there is no effect of

species pool size per se on whether EM monodominancewill emerge or on how strong the bimodal pattern reflectingmonodominance formation will be. Rather, it is the amountof phenotypic variability among EM species that shapes thepattern of EM abundance in local patches. Specifically, re-duced variability tends to cause greater frequency of mono-dominance (fig. 3). In order to directly test for the relation-ship between species pool size and phenotypic variability ofEM species, we ran additional simulations to examine howspecies pool size influences the relationship between spe-cies phenotypic value (Ri) and the abundance of EM spe-cies when EM species variability in the species pool is large(v p 1). We found that when species pool size was small,there was a relatively good correspondence between Ri valueand EM species abundance (fig. 4A). We call this corre-spondence trait-environment matching, meaning that thecloser species are in their phenotype Ri to the optimal (i.e.,0.5 in our simulation), the more abundant they tend to be-come. In this case, the abundances of species follow what isexpected on the basis of their match to the environment(i.e., jRi 2Hkj).As species pool size was increased, this correspondence

became weaker (fig. 4), such that species with an optimalRi value (closer to 0.5) were not necessarily more abundantthan those with a suboptimal value (away from 0.5). Whenthe species pool was diverse, two additional abundance peaksappeared (atRi of around 0.2 and 0.8), so that species that didnot have an Ri close to 0.5 could nevertheless be as abundantas those with an optimal Ri, as most clearly seen in figure 4D.In short, for EM species in our model, it is adaptive to havetraits that best match the underlying abiotic environmentalconditions if there are not many EM species in the speciespool. However, if the species pool has many EM species, ourresults suggest that there can be disruptive natural selectionthat keeps some species away from having the abioticallyoptimal Ri value if the species pool has many species. As a

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

12

1020

Num

ber o

f ect

omyc

orrh

izal

tree

s in

a lo

cal p

atch

(num

ber o

ut o

f 100

0)

Num

ber o

f ect

omyc

orrh

izal

tree

spe

cies

in th

e sp

ecie

s po

ol

Number of local patches (number out of 100)Variability, v, in species values, Ri, among

ectomycorrhizal tree species in the species pool

10.10.010.0001

200

400

600

800

1000

50 40 30 20 1060 50 40 30 20 10

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

200

400

600

800

1000

50 40 30 20 1060 50 40 30 20 10

70 60 50 40 30 20 10 50 40 30 20 10 60 50 40 30 20 10 60 50 40 30 20 10

60 50 40 30 20 10

60 50 40 30 20 10

50 40 30 20 10

50 40 30 20 1050 40 30 20 10

70 60 50 40 30 20 1070 60 50 40 30 20 10

50 40 30 20 10

0

1.0

1.0 1.0

0

1.0

1.0

1.5

0

1.0

1.0

0

1.0

1.0

1.0

0

1.0

1.044

3029

44

5245

13

561

204

3013

73

0

1.0

1.0

1.5

0

1.0

1.1

2.5

0

1.01.8

2.63.2

0

2.0

2.8

1.0

0

1.01.0

1.9

0

1.02.0

2.0

1.6

0

1.02.0

3.02.

3 2.83.0

3.0

0

2.05.0

2.0

3.53.0

4.3

0

1.43.5

1.3

1.72.8

4.0

0

1.0

1.7

1.83.0

3.3

0

1.02.0

1.0

1.02.0

2.0

Figu

re3:

Histogram

sshow

ingectomycorrhizaltree

abun

dancein

alocalforestpatchacross

agradient

ofthenu

mberof

ectomycorrhizaltree

speciesin

thespeciespo

olandthespecies

phenotypicvariability(v)am

ongectomycorrhizaltree

speciesin

thespeciesspool(seefig.1B

).The

S ijvaluesused

areas

specified

infigure

1A.T

henu

mberaboveeach

barindicatesthemean

numberof

ecotom

ycorrhizaltree

speciesthatwerepresentin

thepatch,averaged

acrossallp

atches

belongingto

thebar.The

numberin

thetoprightcorner

ofeach

histogram

indicatesthe

percentage

ofthepatchesthatshow

edmon

odom

inance

byou

rcriteria(onlyon

eectomycorrhizaltree

speciesin

thepatchandmorethan

25%of

thestem

sbelong

tothatspecies).E

achhis-

togram

show

sresults

foron

especiespo

ol,but

wereplicated

speciespo

olsforeach

histogram

andfoun

dthat

results

wereconsistent

acrossreplicated

pools.

This content downloaded from 132.236.027.111 on July 22, 2017 05:14:57 AMAll use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-

c).

Tropical Forest Monodominance S113

result, convergent trait evolution that results in low interspe-cific trait variability and tight trait-environment matchingin the species pool may be more likely when the EM speciespool is small, whereas divergent trait evolution may arisewhen the EM species pool is large.

Why do EM species with suboptimal phenotypes obtaincomparable densities to those with optimal phenotypeswhen the species pool is large (fig. 4D)? We suspected thatit was because the abundance of each EM species is affectedso strongly by the plant-soil feedback—which is mademorecomplex by the larger number of EM species that each havedifferent strengths of plant-soil feedback—that the under-lying influence of abiotic conditions is overwhelmed andobscured. To test this idea, we reran the simulation, but thistime with the strength of plant-soil feedback among EMand/or AM species (Sij) all set to 1.1 instead of being drawnuniform-randomly from [1, 1.1]. Setting all Sij values to 1.1makes interspecific feedback strengths all the same andequal to intraspecific feedback strength, eliminating thecomplexity of feedback strength. In this additional simula-tion, we found good trait-environment matching—such asthe one seen under small species pools (fig. 4A)—if all Sijvalues among EM species were set to 1.1 (fig. 5B). Poortrait-environment matching was not, however, eliminatedwhen Sij values among AM species were set to 1.1 (fig. 5C,

This content downloaded from 132All use subject to University of Chicago Press Term

5D). These results support our hypothesis that it is the var-iation in interspecific feedback strengths among EM speciesthat makes trait-environment matching poor and mono-dominance unlikely when the species pool is large.In summary, our results suggest that two interdependent

processes influence the likelihood of trait and communityconvergence. First, trait-environment matching that leadsto trait convergence among EM speciesmay be less likely un-der a larger EM species pool because of local plant-soil feed-backs that vary in strength. Second, weak trait convergencemay in turn impede the development of EMmonodominance,which represents an extreme case of community divergence.

Empirical Trait Data Analysis

Study Systems and Methods

To begin to assess whether empirical data are consistentwith the hypothesis developed with the model above, we an-alyzed leaf trait data from trees in two geographic regions:Southeast Asia (Borneo), where EM trees are common inmost lowland forests, and South America (Peruvian Ama-zon), where EM trees are generally rare but can becomemonodominant. Both regions contain large areas of low-land tropical rainforests with extraordinary levels of tree di-

Number of ectomycorrhizal tr

2 10 201

Species value, Ri

ee species in the species pool

Spe

cies

abu

ndan

ce, s

umm

edac

ross

40

loca

l pat

ches

DCBA

HGFE

0.2 0.4 0.6 0.8 1.00.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0

0.2 0.4 0.6 0.8 1.00.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0

8000600040002000

7000

6000

5000

4000

3000

2000

1000 5001500

35002500

5000

3000

1000 500

150020002500

1000

100

300

500

700 350

250

150

50

200015001000500

2500

Figure 4: Relationship between species value (Ri) and the abundance of ectomycorrhizal tree species across a gradient of the number ofectomycorrhizal tree species in the species pool. A–D, Results for a total of 2,400 community assembly outcomes comprising 100 replicatedspecies pools, under each of which 40 replicated local communities were assembled. The Y-axis is the mean abundance of a species in a localcommunity, averaged across the 40 replicated local communities, which were then averaged across species having similar Ri values (witherror bars showing standard errors). Solid lines are the simulation results, and dotted lines show the relationship that would be expectedif abundance were determined solely by availability of recruitment sites (i.e., the frequency of recruitment sites having specific values ofHk, abiotic condition of microsite). E–H, Results based on the same data as for A–D, but instead of showing mean abundances and standarderrors, the abundance of each species is indicated (each circle represents a species).

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S114 The American Naturalist

versity. The Peruvian Amazon has greater regional tree di-versity (approximately 5,000 species; Pennington et al. 2004)than Borneo (approximately 3,000 species) probably be-cause it is connected to the rest of the Amazon basin, theworld’s largest contiguous area covered by lowland rainfor-est. Thus, at the scale of a local patch of forest (e.g., LambirHills in Borneo or Allpahuayo-Mishana in Peru), we expectthe regional tree species pool to be reasonably similar, withPeru being slightly larger. However, the pool of EM trees ismuch larger in Borneo than in Peru.

For our analysis, trait data were taken from two separatestudies of plant functional ecology inPeruvianAmazon (For-tunel et al. 2012, 2014) and Borneo (Russo et al. 2005, 2013).Detailed methods on data collection can be found fromFortunel et al. (2012, 2014) and in the appendix, respectively.Data in Peru were collected from multiple 0.5-ha Gentryplots, whereas those in Borneo were collected from withina single 52-ha forest dynamics plot in Lambir Hills NationalPark.We used data on four leaf traits that are involved in thefast-slow plant economics spectrum (Reich 2014) and thatare thought to be strongly filtered by environmental condi-tions (Fortunel et al. 2014): specific leaf area (SLA), leaf car-bon and nitrogen concentrations by mass, and leaf C∶N ra-tio. Leaf C∶N ratio is considered a particularly importanttrait in determining the strength of plant-soil feedbacks (DeDeyn et al. 2008; Ke et al. 2015). Data from individual plantswere averaged within species. Species’mycorrhizal status was

This content downloaded from 132All use subject to University of Chicago Press Term

classified at the genus or family level on the basis of Brun-drett (2009) and Peay et al. (2013). Data are deposited in theDryad Digital Repository: http://dx.doi.org/10.5061/dryad.c0kr7 (Fukami et al. 2017).For each region, we analyzed trait data from lower fertil-

ity (sand) and higher fertility (clay) soil habitats separatelybecause these habitats differ in species pool compositionwithin each region. The trait data set from Peru included365 AM and 13 EM species from clay soils and 140 AMand 3 EM species from sandy soils. The trait data set fromBorneo included 98 AM and 17 EM species from clay soilsand 111 AM and 35 EM species from sandy soils. We tookthree approaches to look for differences in EM traits be-tween the two species pools (Peru and Borneo) within soiltypes. First, we compared mean trait values for EM versusAM host species within each species pool, using a two-sample t-test, to evaluate our prediction that EM speciesare functionally distinct from AM species. Second, we com-pared the variance of trait values for EM versus AM specieswithin each species pool, using a one-sided variance ratioF-test, to evaluate our prediction that EM species have largertrait variability than AM species. Third, we compared thevariance in trait values between EM species pools in Borneoand Peru, using a one-sided variance ratio F-test, to evalu-ate our prediction that there should be greater variation intrait values among EM host species in Borneo than in Peru,where EM monodominance is found. We do not approach

Species value, Ri

Spe

cies

abu

ndan

ce, s

umm

ed

acro

ss 4

0 lo

cal p

atch

es

Strengths of inter-specific feedback in EM trees 1.1 llA]1.1 ,1[

Strengths of inter-specific feedback in A

M trees

[1, 1.1]A

ll 1.1A

DC

B

0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0

0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0

20406080

100120 200

150

100

50

400

300

200

100

500

400300

200100

Figure 5: Effect of changing Sij values on the relationship between species value (Ri) and species abundance. Symbols are as in figure 4, exceptthat data for each panel are based on 60 (not 100) replicate species pools. EM, ectomycorrhizal; AM, arbuscular mycorrhizal.

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Tropical Forest Monodominance S115

this question in an explicit phylogenetic framework, butour test is conservative because we predict larger trait varia-tion in the narrower phylogenetic pool, that is, in Borneo,which had only one EM family (Dipterocarpaceae).

Trait Analysis Results and Discussion

We found significant differences in some mean foliar traitvalues between AM and EM species in both Peru and Bor-neo that were consistent with our model prediction (table 1;fig. 6). In Peru, foliar C∶N ratio was significantly differentbetween AM and EM species on both soil habitats, and fo-liar nitrogen concentration was different only on clay soil.In Borneo, foliar carbon concentration differed significantlybetween AM and EM species on both habitats, but no othersignificant differences were observed. The difference foundin Peru between AM and EM host species in foliar C∶N wasparticularly large, in contrast to virtually identical means forthis trait in Borneo (fig. 6).

One possible explanation for these inconsistencies lies inthe relative differences in the sizes of EM and AM host spe-cies pools in each forest. To the extent that greater richnessof the species pool causes more complex plant-soil feed-backs within each mycorrhizal group, in Peru (with fewerEM host species), their traits may track environmental con-ditions closely, causing a large divergence from those of AMhost species. Since the pool of AM species is larger, theymaybe influenced strongly by the complex plant-soil feedbacks,reducing their match to underlying abiotic conditions. Incontrast, in Borneo, where bothmycorrhizal groups are spe-cies rich, complex plant-soil interactions might have pre-vented species traits from tracking the environment in bothEM and AM host species.

Another explanation lies in the difference in the relativesoil fertility of these regions. Averaging over soil types andmycorrhizal status, foliar N was significantly lower (F1, 781 p133:7, P ! :001) and C∶N significantly higher (F1, 781 p109:8, P ! :001) in Borneo than in Peru, and both soils atthe Bornean site are particularly low in soil N, P, K, and

This content downloaded from 132All use subject to University of Chicago Press Term

other nutrients (Baillie et al. 2006). Low soil fertility isstrongly associated with overall low foliar nutrient concen-trations and higher C∶N ratios (Chapin et al. 1993), whichcould give less latitude for differences between AM and EMhost species in Borneo compared with more fertile sites.Trait differences between AM and EM host species in Bor-neo may be more tightly linked to leaf structure and physicaldefense, as suggested by the significant differences betweenmycorrhizal groups in foliar C, which is likely to affect leaflitter decomposition and, in turn, the strength of plant-soilfeedbacks and the extent of trait-environment matching.There were significant differences in the variance of leaf

traits between EM and AM host species in Peru and Bor-neo. However, in contrast to our expectations, variance intraits was often larger for AM than EM host species in Bor-neo (three of four traits), whereas no significant differenceswere observed in trait variances between AM and EM hostspecies in Peru (table 2). Furthermore, results for trait var-iance of EM species pools did not support our model pre-diction either. The variance in species traits was not largeramong EM host species in Borneo than in Peru for any ofthe traits measured for clay (SLA F16, 12 p 0:56, P p :86;leaf carbon concentration [LCC]F16, 12 p 0:52,P p :89; leafnitrogen concentration [LNC] F16, 12 p 0:05, P p 1; C∶NF16, 12 p 0:87, P p :61) or sandy (SLA F34, 2 p 2:32, P p:35; LCC F34, 2 p 5:13, P p :18; LNC F34, 2 p 0:03, P p 1;C∶N F34, 2 p 0:49, P p :85) habitats.How could these apparent discrepancies between model

predictions and empirical patterns be reconciled? Our fo-cus in this article has been on contrasting alternative stablestates between diverseAMandmonodominant EMpatches,but as a family, the Dipterocarpaceae dominate this andmany other forests in Southeast Asia in terms of basal area.One potential explanation for their trait convergence maythen be that phylogenetic relatedness of dipterocarp specieshave caused them to act as if they belonged to one species,thereby allowing for trait-environment matching in theabsence of complex interspecific plant-soil feedback. As forthe apparent lack of trait convergence among EM relative

Table 1: Difference in species’ mean values of leaf traits between arbuscular mycorrhizal (AM) and ectomycorrhizal (EM)tree species across two edaphic habitats in Peru and Borneo

No. species

SLA LCC

.236.027.111 on July 22, s and Conditions (http://w

LNC

2017 05:14:57 AMww.journals.uchicago.edu

C∶N

Region and habitat

AM EM t P t P t P t

/t-and-c).

P

Peru:

Clay 362 13 .06 .95 1.59 .14 24.18 .001* 6.15 !.001*

Sand

140 3 .97 .41 2.15 .89 22.87 .10 4.68 .03*

Borneo:

Clay 98 17 .98 .34 24.43 !.001* 1.06 .29 21.34 .19 Sand 111 35 1.29 .20 24.55 !.001* .58 .56 2.56 .58

Note: SLA, specific leaf area; LCC, leaf carbon concentration; LNC, leaf nitrogen concentration; C∶N, leaf C∶N ratio.* P ! :05.

S116 The American Naturalist

to AMhost species in Peru, this result may reflect low statis-tical power due to small sample size. The number of EM spe-cies was much smaller in Peru (13 and three species) than inBorneo (17 and 35 species).

Discussion

Taken together, our simulation results suggest one poten-tial reason why EM monodominance is favored when the

This content downloaded from 132All use subject to University of Chicago Press Term

species pool has a small number of EM species. Specifically,results summarized in figure 3 indicate that monodomi-nance can emerge only when EM species phenotypes (Ri)closely match the environmental conditions available in lo-cal forest patches (Hk). Results presented in figure 4 indi-cate in turn that such optimal trait-environment matchingis more likely under small species pools because EM spe-cies can thenmore easily track the environment. As a result,trait evolution of EM species converges toward abiotically

0 20 40 60 80

0.00

0.02

0.04

0.06

0.08

0.10

Clay soil in Peru

Den

sity

0 20 40 60 80

0.00

0.02

0.04

0.06

0.08

0.10

Sandy soil in Peru

0 20 40 60 80

0.00

0.02

0.04

0.06

0.08

0.10

Clay soil in Borneo

Foliar C:N

Den

sity

0 20 40 60 80

0.00

0.02

0.04

0.06

0.08

0.10

Sandy soil in Borneo

Foliar C:N

C

A B

D

Figure 6: Probability density plots showing the frequency of tree species having particular foliar C∶N ratios under two soil conditions (lessfertile sand or more fertile clay) in Borneo and Peru. Dark gray indicates ectomycorrhizal (EM) host species. Light gray indicates all otherspecies, most of which are arbuscular mycorrhizal (AM) hosts. Dashed lines indicate means.

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Tropical Forest Monodominance S117

optimal conditions. Finally, results in figure 5 show that thereason why species do not track the environment under di-verse species pools is because the complex plant-soil feedbackamong the many EM species obscures the effect of environ-mental heterogeneity on local species abundances. In otherwords, environmental filtering (Kraft et al. 2015a) is weakerthan priority effects.

Although we have focused on mycorrhizal associations,monodominance may arise through other pathways (Pehet al. 2011). For example, a defensive ant-plant mutualismgenerates monodominant stands of Duroia hirstuta knownas devil’s gardens in the Amazon (Frederickson et al. 2005).Still, many tree species that form monodominant stands inthe tropics are EM. Studies have uncovered evidence sup-porting a range of ways in which EM associations may gen-erate positive plant-soil feedbacks that lead to monodomi-nance, including commonmycorrhizal networks (McGuire2007), alteration of local nutrient economies (Corrales et al.2016), variation in leaf chemistry (Torti et al. 2001), and EMmining of organic nitrogen (Orwin et al. 2011). Our modeldoes not distinguish between these different mechanisms.Rather, our work is complementary to these studies in thatwe ask why geographic differences in species pools may leadto some tropical regions containing EM monodominantstands and other regions showing local codominance ofmul-tiple tree species of both mycorrhizal types.

We chose one simple way to represent plant-soil feed-backs. Namely, we assumed that plant-soil feedbacks werenet positive and most beneficial to conspecifics. Choosingappropriate values for plant-soil feedbacks is not trivial, be-cause in nature these feedbacks are hard to estimate. Em-pirical measurement of their strength should take into ac-count multiple contributing factors, such as abiotic effects(Waring et al. 2015) and the combined effects of pathogensandmutualists (Klironomos2002) andnatural enemies (Bag-chi et al. 2014). In addition, whether net feedbacks are pos-itive or negative also depends on the choice of other het-erospecific comparisons and sterilization methods used togenerate the reference point. A number of other factors

This content downloaded from 132All use subject to University of Chicago Press Term

may influence feedback strength, such as phylogenetic relat-edness (Gilbert and Webb 2007; Liu et al. 2012), local den-sity (Comita et al. 2010; Mangan et al. 2010; Liu et al. 2012),and mycorrhizal type (Johnson et al. 2012; Dickie et al.2014; Bennett et al. 2017).In this study, we attempted to keep the model focused

on differences in species pool size by keeping the nature ofplant-soil feedback identical between AM and EM species.Through additional simulation (figs. S1, S2), we have madean initial effort to examine the robustness of our model pre-dictions to assumptions regarding plant-soil feedback. Fu-ture iterations of this model could explore more complexplant-soil feedback structure, such as more negative (or lesspositive) intraspecific feedbacks and stronger positive feed-backs in EM compared with AM plants. It is also possiblethat EM plants are better able to make the environment lesssuitable for AM plants, owing to the abilities of EM fungito access organic forms of nutrients and preempt nutrientuptake by AM fungi. This scenario can be evaluated furtherwith our model. Finally, we assigned Ri and Sij values inde-pendently, but in nature, species with similar traits—as re-presented bymore similar values of Ri—might have the ten-dency to share fungal symbionts and therefore have morepositive Sij. Phylogenetic relationships of plant species, whichwe did not consider in this article, may prove useful in somecases as a proxy for estimating Ri and Sij values.It is well established in plant-soil feedback theory that pos-

itive feedback causes alternative stable states, whereas nega-tive feedback facilitates species coexistence (Bever et al. 2012).Our simulation results are consistent with this prior theory.However, most previous plant-soil feedback theories haveconsidered interactions between only two plant species. Here,we have studied interactions among many plant species (seealso Fukami and Nakajima 2011, 2013), which is the onlyway to directly address the effect of the number of species inthe regional speciespoolon local communityassemblydrivenby plant-soil feedback.One benefit of our model is that it makes specific pre-

dictions about trait variation of plant species and about

Table 2: Difference in leaf trait variance between arbuscular mycorrhizal (AM) and ectomycorrhizal (EM) tree speciesacross two edaphic habitats in Peru and Borneo

No. species

SLA LCC

.236.027.111 on July 22s and Conditions (http://

LNC

, 2017 05:14:57 AMwww.journals.uchicago.edu

C∶N

Region and habitat

AM EM F ratio P F ratio P F ratio P F ratio

/t-and-c).

P

Peru:

Clay 362 13 1.11 .45 1.03 .52 .34 1.00 1.65 .16 Sand 140 3 5.71 .16 4.24 .21 .79 .71 3.95 .22

Borneo:

Clay 98 17 1.15 .40 3.17 .006* 4.00 .001* 2.55 .02*

Sand

111 35 1.42 .12 4.07 !.001* 7.05 !.001* 5.44 !.001*

Note: SLA, specific leaf area; LCC, leaf carbon concentration; LNC, leaf nitrogen concentration; C∶N, leaf C∶N ratio.* P ! :05.

S118 The American Naturalist

mycorrhiza-driven plant-soil feedbacks under different spe-cies pool sizes, specifically that divergence in the values oftraits related to local environmental fit should be greaterin more diverse species pools. Our trait analysis did not yieldstrong support for this prediction, perhaps because these aredifficult effects to measure empirically, since many factorsnot included in our model influence trait variation in naturalsettings. Moreover, uncertainties regarding the mycorrhizalstatus (e.g., EMorAM) of plant species in different geograph-ical locations may also have caused the poor correspondenceof model predictions and empirical patterns (Forrestel et al.2017). In addition, detecting patterns in traits across a gra-dient of species pool size is made difficult by the inherentcorrelation between species pool size and statistical powerfor any analysis that uses species as a unit of replication.

Our trait analysis is only a first step to begin to evaluatethe empirical relevance of the hypothesis we have devel-oped here through the simulation modeling. Research onthe links between plant traits, mycorrhizal associations, andthe strength of plant-soil feedback is still at an early stage ofdevelopment (Laughlin et al. 2015; Cortois et al. 2016; Ben-nett et al. 2017), and the dearth of relevant data may in partexplain thepoor correspondenceof our empirical resultswiththe model predictions. We list three future directions ofempirical research. First, with respect to the determinants ofmonodominance, we need to better understand which traitsare important for tree species’ fit to the local environment(Kraft et al. 2008; Fortunel et al. 2014; Laughlin et al. 2015;Forrestel et al. 2017) and how correlated different traits arewith competitive abilities and the strength of plant-soil feed-backs (Uriarte et al. 2010; Fortunel et al. 2016). For example,some traits may bemore closely related to competitive abilityand others to niche differences (Mayfield and Levine 2010;Kraft et al. 2015b; Kunstler et al. 2016). Likewise, the pat-terns observed in the relationship between trait variabilityand monodominance may depend on which traits are in-volved inwhich kinds of interactions in the system. This pos-sibility is supported by the fact that in our analyses of traitvariation, some traits showed patterns consistent with ourpredictions, whereas others did not. To wit, plant-soil feed-backs involving EMspecies in Borneomay bemediatedmorestrongly by the effects of trait variation in leaf structure andphysical defense on litter decomposability than by the effectsof foliar N because of the generally low fertility of soils in thisBornean forest. Thus, whether patterns of variation consis-tent with trait-environment matching are observed may notonly depend on how correlated the traits examined are withplant-soil feedbacks but also be constrained by the local envi-ronment. Second, to complement the indirect inference ofspecies performance and interactions through trait analysis,more direct measurements of theCijk values could be helpful.One approach may be leveraging trait values in neighbor-hood models (Uriarte et al. 2010; Fortunel et al. 2016) as es-

This content downloaded from 132All use subject to University of Chicago Press Term

timates of the Cijk in our model to test the effects of AM andEM trees on each other’s growth and survival. Third, we com-pared just two regions here, but it will be more informativeto analyze data frommore regions—including Africa, wheremonodominance also arises—to assess the potential appli-cability of our hypothesis across geographical regions withdifferent evolutionary histories.If species pool properties are key determinants of com-

munity assembly (Ricklefs and Schluter 1993; Zobel 2016),as we have considered in this article, a fundamental ques-tion is what causes differences in species pools among trop-ical regions in the first place. For example, what might ex-plain the unusually high prevalence of EM lineages in theAsian tropics? Any understanding of why the EM pool is solarge in Asia must reckon with the radiation of the Diptero-carpoideae (Dipterocarpaceae), which numbers more than470 species (Ashton 2002). More generally, many hypothe-ses have been put forward to explain the disparities in treespecies diversity among different tropical regions. However,there is no consensus explanation (Couvreur 2014). Someauthors emphasize the relative importance of extrinsic fac-tors, such as climate or climatic stability (Couvreur 2014),whereas others highlight intrinsic factors, such as historyof diversification of different lineages in different regions,which have resulted in, for example, a diverse understory treeflora only in South America (Gentry 1993; Terborgh et al.2016). Another potentially important intrinsic factor is EMassociation and its influence on the origin and maintenanceof tropical tree diversity, as outlined here. Clearly, more re-search is needed to understand EM influences on patternsof tree monodominance and how this relates to the commu-nity assembly and diversity of regional tree floras.

Conclusions

To our knowledge, this is the first study to develop a hypoth-esis to explain why alternative stable states of either local EMmonodominance or relative rarity arise in some tropical re-gions and not others. Specifically, our hypothesis is that anincrease in the size of the EM tree species pool results in anincrease in the complexity of biotic interactions, which inturn prevents monodominance by a single EM species. Wehave proposed that under a diverse EM host species pool,tree speciesmaintain divergence in traits because of the com-plex plant-soil feedbacks among the many different tree spe-cies, with no species achieving a monodominant status. Incontrast, under a depauperate EMhost species pool, tree spe-cies experience strong selection for evolution toward theoptimal abiotic conditions, allowing them to become com-petitive enough to form monodominant stands via positiveplant-soil feedback that causes priority effects. Overall, thisstudy highlights the importance of geographical variation inspecies pools in understanding the conditions under which

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Tropical Forest Monodominance S119

community divergence results in vastly different alternativestable states.

Returning to the general topic of convergence, our goal inthis article has been to point out the utility of studying thelink between trait convergence and community convergence.To this end, we have explored how trait convergence mayaffect community convergence, using the case of tropicalmonodominanceasan illustrative example.Althoughwehavedeveloped only one hypothesis, one thing that is clear fromour work is that the natural historical knowledge of how con-vergent traits influence species interactions is essential to un-derstanding community convergence.

Acknowledgments

We thank A. A. Agrawal for the invitation to contribute thisarticle to the special issue arising from the Vice PresidentialSymposium of the American Society of Naturalists held inAustin, Texas, in June 2016. Comments from A. A. Ag-rawal, P.-J. Ke, and three anonymous reviewers improvedthe article. T.F. thanks the National Science Foundation(NSF; DEB 1555786) and the Terman Fellowship at Stan-ford University for financial support as well as the Centerfor Macroecology, Evolution, and Climate and the Sectionof Microbiology at the University of Copenhagen for sab-batical support. K.G.P. was supported in part by the NSF(DEB 1249342 and RAPID 1361171) and a Department ofEnergy Early Career Grant (DE-SC0016097). S.E.R. was sup-ported in part by the NSF (RAPID 1361171). Author partici-pation: T.F. and K.G.P. designed the study; M.N. conductedand analyzed simulation in collaboration with T.F.; C.B.,P.V.A.F., and C.F. collected data from Peru; S.E.R. collecteddata fromBorneo; K.G.P. and S.E.R. analyzed data fromPeruand Borneo; T.F. wrote the first draft of the manuscript; andall authors contributed to writing the manuscript.

APPENDIX

Methods for Quantifying Leaf Functional Traitsin Bornean Rain Forest

Study System and Species Selection

LambirHills National Park (Lambir) is located in the north-western part of Borneo in the Malaysian state of Sarawak(47200N, 1137500E). Lambir receives approximately 3,000 mmof rainfall annually, with all months averaging 1100 mm(Watson 1985). The region has the highest tree species rich-ness recorded in the Paleotropics (Ashton 2005), with speciesin the Dipterocarpaceae dominating the forest (Lee et al.2002). The soils and geomorphology of Lambir have beenpreviously described (Baillie et al. 2006; Tan et al. 2009). Thesoils range from coarse loams that are sandstone derived,leached, nutrient depleted, and well drained with substantial

This content downloaded from 132All use subject to University of Chicago Press Term

raw humus, to clays that are shale derived, less nutrient de-pleted, and less well drained, with little raw humus. Tree com-munity composition and demography vary across soil types,with most species exhibiting soil habitat specialization andtree demography varying across soil types (Davies et al. 2005;Russo et al. 2005). Lambir is the site of a 52-ha research plotthat was established in 1991 as part of the Center for TropicalForest Science Forest Global Earth Observatory plot network(Anderson-Teixeira et al. 2015) tomonitor woody plants. Alltrees ≥ 1 cm in diameter at breast height are tagged, mapped,and dentified, and their diameters are measured to the near-est 1 mm.

Tree species were sampled for the quantification of leaffunctional traits fromwithin and near the Lambir plot (Russoet al. 2013). The species identity of individuals sampled out-side of the plot was verified using an on-site herbarium andby consultation with local botanists (S. Tan). Species wereselected to encompass a wide range of families—targetingtaxa contributing substantially to forest basal area in the Lam-bir plot—to target species-rich genera, such as Shorea (Dip-terocarpaceae) and Diospyros (Ebenaceae), and to includespecies with a range of shade tolerance niches. For each spe-cies, juvenile (1–5 cm in diameter) and adult (110 cm in di-ameter) trees were sampled as much as possible, with one to22 individuals sampled per species.

Quantification of Leaf Functional Traits

From each tree, three to five mature, sunlit, minimallydamaged leaves were harvested. The petiole was cut fromthe lamina, which was gently cleaned of debris. Fresh leaflaminas were scanned (Canon LiDE 110), and the imageswere analyzed with ImageJ software (Schneider et al. 2012)to estimate the area of each. After oven drying at 607C for3 days, the dry weight of each lamina was recorded. TheSLA (cm2/g) was calculated as fresh area divided by dryweight, and SLA was averaged across leaves of each indi-vidual. After drying, the midvein was removed from eachlamina. Laminas from each individual were ground to-gether to a fine, uniform powder and analyzed by elemen-tal combustion for carbon and nitrogen content (CostechAnalytical Elemental Combustion System 4010). For eachindividual, percent carbon (C) and nitrogen (N) were cal-culated as the mass of C (or N) in the sample/dry mass ofthe sample # 100, and C∶N ratio was calculated as per-cent C/percent N. Trait values were averaged across indi-viduals to obtain mean values for each tree species.

Literature Cited

Ashton, P. S. 2002. Dipterocarpaceae. Pages 182–197 in K. Kubitzki,ed. The families and genera of vascular plants. V. Flowering plants:

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S120 The American Naturalist

Dicotyledons. Malvales, Capparales and non-betalain Caryophyl-lales. Springer, Berlin.

Bagchi, R., R. E. Gallery, S. Gripenberg, S. J. Gurr, L. Narayan, C. E.Addis, R. P. Freckleton, and O. T. Lewis. 2014. Pathogens and in-sect herbivores drive rainforest plant diversity and composition.Nature 506:85–88.

Baillie, I. C., P. S. Ashton, S. P. Chin, S. J. Davies, P. A. Palmiotto, S. E.Russo, and S. Tan. 2006. Spatial associations of humus, nutrientsand soils in mixed dipterocarp forest at Lambir, Sarawak, Ma-laysian Borneo. Journal of Tropical Ecology 22:543–553.

Baraloto, C., S. Rabaud, Q. Molto, L. Blanc, C. Fortunel, B. Hérault,N. Dávila, I. Mesones, M. Rios, E. Valderrama, and P. V. A. Fine.2011. Disentangling stand and environmental correlates of above-ground biomass in Amazonian forests. Global Change Biology17:2677–2688.

Bennett, J. A., H. Maherali, K. O. Reinhart, Y. Lekberg, M. M. Hart,and J. Klironomos. 2017. Plant-soil feedbacks and mycorrhizal typeinfluence temperate forest population dynamics. Science 355:181–184.

Bever, J. D., T. G. Platt, and E. R. Morton. 2012. Microbial popula-tion and community dynamics on plant roots and their feedbackson plant communities. Annual Review of Microbiology 66:265–283.

Brearley, F. 2012. Ectomycorrhizal associations of the Dipterocarpa-ceae. Biotropica 44:637–648.

Brundrett, M. C. 2002. Coevolution of roots and mycorrhizas of landplants. New Phytologist 154:275–304.

———. 2009. Mycorrhizal associations and other means of nutri-tion of vascular plants: understanding the global diversity of hostplants by resolving conflicting information and developing reli-able means of diagnosis. Plant and Soil 320:37–77.

Cavender-Bares, J., K. H. Kozak, P. V. A. Fine, and S. W. Kembel.2009. The merging of community ecology and phylogenetic biol-ogy. Ecology Letters 12:693–715.

Chapin, F. S., III, K. Autumn, and F. Pugnaire. 1993. Evolution ofsuites of traits in response to environmental stress. American Nat-uralist 142:78–92.

Chesson, P. L. 1985. Coexistence of competitors in spatially and tem-porally varying environments: a look at the combined effects ofdifferent sorts of variability. Theoretical Population Biology 28:263–287.

Clements, F. E. 1936. Nature and structure of the climax. Journal ofEcology 24:252–284.

Comita, L. S., H. C. Muller-Landau, S. Aguilar, and S. P. Hubbell.2010. Asymmetric density dependence shapes species abundancesin a tropical tree community. Science 329:330–332.

Connell, J. H., and M. D. Lowman. 1989. Low-diversity tropical rainforests: some possible mechanisms for their existence. AmericanNaturalist 134:88–119.

Conte, G. L., M. E. Arnegard, C. L. Peichel, and D. Schluter. 2012.The probability of genetic parallelism and convergence in naturalpopulations. Proceedings of the Royal Society B 279:5039–5047.

Cornelissen, J., R. Aerts, B. Cerabolini, M. Werger, and M. van derHeijden. 2001. Carbon cycling traits of plant species are linkedwith mycorrhizal strategy. Oecologia (Berlin) 129:611–619.

Corrales, A., S. A. Mangan, B. L. Turner, and J. W. Dalling. 2016. Anectomycorrhizal nitrogen economy facilitates monodominance ina Neotropical forest. Ecology Letters 19:383–392.

Cortois, R., G. B. De Deyn, T. Schröder-Georgi, and W. H. van derPutten. 2016. Plant-soil feedbacks: role of plant functional groupand plant traits. Journal of Ecology 104:1608–1617.

This content downloaded from 132All use subject to University of Chicago Press Term

Couvreur, T. L. P. 2014. Odd man out: why are there fewer plantspecies in African rain forests? Plant Systematics and Evolution301:1299–1313.

De Deyn, G. B., J. H. C. Cornelissen, and R. D. Bardgett. 2008. Plantfunctional traits and soil carbon sequestration in contrasting bi-omes. Ecology Letters 11:516–531.

Dickie, I. A., N. Koele, J. D. Blum, J. D. Gleason, and M. S. McGlone.2014. Mycorrhizas in changing ecosystems. Botany 92:149–160.

Donovan, L. A., H. Maherali, C. M. Caruso, H. Huber, and H. deKroon. 2011. The evolution of the worldwide leaf economics spec-trum. Trends in Ecology and Evolution 26:88–95.

Drake, J. A. 1991. Community-assembly mechanics and the structureof an experimental species ensemble. American Naturalist 137:1–26.

Duchesne, L. C., B. E. Ellis, and R. L. Peterson. 1989. Disease sup-pression by the ectomycorrhizal fungus Paxillus involutus: contri-bution of oxalic acid. Canadian Journal of Botany 67:2726–2730.

Egerton-Warburton, L., and M. F. Allen. 2001. Endo- and ectomy-corrhizas in Quercus agrifoliaNee. (Fagaceae): patterns of root col-onization and effects on seedling growth. Mycorrhiza 11:283–290.

Fargione, J., C. S. Brown, and D. Tilman. 2003. Community assem-bly and invasion: an experimental test of neutral versus niche pro-cesses. Proceedings of the National Academy of Sciences of theUSA 100:8916–8920.

Forrestel, E. J., M. J. Donoghue, E. J. Edwards, W. Jetz, J. C. O. duToit, and M. D. Smith. 2017. Different clades and traits yield sim-ilar grassland functional responses. Proceedings of the NationalAcademy of Sciences of the USA 114:705–710.

Fortunel, C., P. V. A. Fine, and C. Baraloto. 2012. Leaf, stem and roottissue strategies across 758 Neotropical tree species. FunctionalEcology 26:1153–1161.

Fortunel, C., C. E. T. Paine, P. V. A. Fine, N. J. B. Kraft, and C.Baraloto. 2014. Environmental factors predict community func-tional composition in Amazonian forests. Journal of Ecology102:145–155.

Fortunel, C., R. Valencia, S. J. Wright, N. C. Garwood, and N. J. B.Kraft. 2016. Functional trait differences influence neighbourhoodinteractions in a hyperdiverse Amazonian forest. Ecology Letters19:1062–1070.

Fox, B. J. 1987. Species assembly and the evolution of communitystructure. Evolutionary Ecology 1:201–213.

Fransson, P. M. A., Y. K. Toljander, C. Baum, and M. Weih. 2015.Host plant-ectomycorrhizal fungus combination drives resource al-location in willow: evidence for complex species interaction from asimple experiment. Ecoscience 20:112–121.

Frederickson, M. E., M. J. Greene, and D. M. Gordon. 2005. ‘Devil’sgardens’ bedevilled by ants. Nature 437:495–496.

Fukami, T. 2009. Convergence. Pages 188–191 in R. G. Gillespie andD. A. Clague, eds. Encyclopedia of islands. University of Cali-fornia Press, Berkeley.

———. 2015. Historical contingency in community assembly: inte-grating niches, species pools, and priority effects. Annual Reviewof Ecology, Evolution, and Systematics 46:1–23.

Fukami, T., and M. Nakajima. 2011. Community assembly: alterna-tive stable states or alternative transient states? Ecology Letters14:973–984.

———. 2013. Complex plant-soil interactions enhance plant speciesdiversity by delaying community convergence. Journal of Ecology101:316–324.

Fukami, T., M. Nakajima, C. Fortunel, P. V. A. Fine, C. Baraloto, S. E.Russo, and K. G. Peay. 2017. Data from: Geographical variation in

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

Tropical Forest Monodominance S121

community divergence: insights from tropical forestmonodominanceby ectomycorrhizal trees. American Naturalist, Dryad Digital Repos-itory, http://datadryad.org/resource/doi:10.5061/dryad.c0kr7.

Gentry, A. H. 1993. Diversity and floristic composition of lowlandtropical forests in Africa and South America. Pages 500–547 inP. Goldblatt, ed. Biological relationships between Africa and SouthAmerica. Missouri Botanical Garden Press, St. Louis.

Gilbert, G. S., and C. O. Webb. 2007. Phylogenetic signal in plantpathogen-host range. Proceedings of the National Academy of Sci-ences of the USA 104:4979–4983.

Gleason, H. A. 1927. Further views on the succession concept. Ecol-ogy 8:299–326.

Hart, T. B., J. A. Hart, and P. G. Murphy. 1989. Monodominant andspecies-rich forests of the humid tropics: causes for their co-occurrence. American Naturalist 133:613–633.

Henkel, T. W. 2003. Monodominance in the ectomycorrhizal Dicymbecorymbosa (Caesalpiniaceae) fromGuyana. Journal of Tropical Ecol-ogy 19:417–437.

Horton, T. R., T. D. Bruns, and V. T. Parker. 1999. Ectomycorrhizalfungi associated with Arctostaphylos contribute to Pseudotsugamenziesii establishment. Canadian Journal of Botany 77:93–102.

Janzen, D. H. 1974. Tropical blackwater rivers, animals, and mastfruiting by the Dipterocarpaceae. Biotropica 6:69–103.

Johnson, D. J., W. T. Beaulieu, J. D. Bever, and K. Clay. 2012. Con-specific negative density dependence and forest diversity. Science336:904–907.

Ke, P.-J., T. Miki, and T.-S. Ding. 2015. The soil microbial commu-nity predicts the importance of plant traits in plant-soil feedback.New Phytologist 206:329–341.

Kennedy, P. G., A. D. Izzo, and T. D. Bruns. 2003. There is high po-tential for the formation of common mycorrhizal networks be-tween understorey and canopy trees in a mixed evergreen forest.Journal of Ecology 91:1071–1080.

Klironomos, J. N. 2002. Feedback with soil biota contributes to plantrarity and invasiveness in communities. Nature 417:67–70.

Koele, N., I. A. Dickie, J. Oleksyn, S. J. Richardson, and P. B. Reich.2012. No globally consistent effect of ectomycorrhizal status onfoliar traits. New Phytologist 196:845–852.

Kraft, N. J. B., P. B. Adler, O. Godoy, E. C. James, S. Fuller, and J. M.Levine. 2015a. Community assembly, coexistence and the envi-ronmental filtering metaphor. Functional Ecology 29:592–599.

Kraft, N. J. B., O. Godoy, and J. M. Levine. 2015b. Plant functional traitsand the multidimensional nature of species coexistence. Proceed-ings of the National Academy of Sciences of the USA 112:797–802.

Kraft, N. J. B., R. Valencia, and D. D. Ackerly. 2008. Functional traitsand niche-based tree community assembly in an Amazonian for-est. Science 322:580–582.

Kulmatiski, A., K.H. Beard, J. R. Stevens, and S.M.Cobbold. 2008. Plant-soil feedbacks: a meta-analytical review. Ecology Letters 11:980–992.

Kunstler, G., D. Falster, D. A. Coomes, F. Hui, R. M. Kooyman, D. C.Laughlin, L. Poorter, et al. 2016. Plant functional traits have glob-ally consistent effects on competition. Nature 529:204–207.

Laughlin, D. C., S. J. Richardson, E. Wright, and P. J. Bellingham.2015. Environmental filtering and positive plant litter feedback si-multaneously explain correlations between leaf traits and soil fer-tility. Ecosystems 18:1269–1280.

Li, S. P., M. W. Cadotte, S. J. Meiners, Z. Pu, T. Fukami, and L. Jiang.2016. Convergence and divergence in a long-term old-field succes-sion: the importance of spatial scale and species abundance. Ecol-ogy Letters 19:1101–1109.

This content downloaded from 132All use subject to University of Chicago Press Term

Liu, X., M. Ligan, R. S. Etienne, Y. Wang, C. Staehlin, and S. Yu.2012. Experimental evidence for a phylogenetic Janzen-Connelleffect in a subtropical forest. Ecology Letters 15:111–118.

Mangan, S. A., S. A. Schnitzer, E. A. Herre, K. M. L. Mack, M. C.Valencia, E. I. Sanchez, and J. D. Bever. 2010. Negative plant-soilfeedback predicts tree-species relative abundance in a tropical for-est. Nature 466:752–755.

Mayfield, M. M., and J. M. Levine. 2010. Opposing effects of compet-itive exclusion on the phylogenetic structure of communities. Ecol-ogy Letters 13:1085–1093.

McGuire, K. L. 2007. Common ectomycorrhizal networks may main-tain monodominance in a tropical rain forest. Ecology 88:567–574.

McGuire, K. L., T. W. Henkel, I. G. de la Cerda, G. Villa, F. Edmund,and C. Andrew. 2008. Dual mycorrhizal colonization of forest-dominating tropical trees and the mycorrhizal status of non-dominant tree and liana species. Mycorrhiza 18:217–222.

Melville, J., L. J. Harmon, and J. B. Losos. 2006. Intercontinentalcommunity convergence of ecology and morphology in desertlizards. Proceedings of the Royal Society B 273:557–563.

Mouquet, N., J. L. Moore, and M. Loreau. 2002. Plant species rich-ness and community productivity: why the mechanism that pro-motes coexistence matters. Ecology Letters 5:56–65.

Nara, K. 2006a. Ectomycorrhizal networks and seedling establishmentduring early primary succession. New Phytologist 169:169–178.

———. 2006b. Pioneer dwarf willow may facilitate tree successionby providing late colonizers with compatible ectomycorrhizal fungiin a primary successional volcanic desert. New Phytologist 171:187–198.

Newsham, K. K., A. H. Fitter, and A. R. Watkinson. 1995. Ar-buscular mycorrhiza protect an annual grass from root pathogenicfungi in the field. Journal of Ecology 83:991–1000.

Odling-Smee, J. F., K. N. Laland, and M. W. Feldman. 2003. Nicheconstruction: the neglected process in evolution. Princeton Uni-versity Press, Princeton, NJ.

Orwin, K. H., M. U. Kirschbaum, M. G. St John, and I. A. Dickie.2011. Organic nutrient uptake by mycorrhizal fungi enhances eco-system carbon storage: a model-based assessment. Ecology Letters14:493–502.

Pacala, S. W., and D. Tilman. 1994. Limiting similarity in mechanis-tic and spatial models of plant competition in heterogeneous en-vironments. American Naturalist 143:222–257.

Palmgren, A. 1926. Chance as an element in plant geography. Pages 591–602 in B. M. Duggar, ed. Proceedings of the International Congress ofPlant Sciences, Ithaca, NY.

Peay, K. G. 2016. The mutualistic niche: mycorrhizal symbiosis andcommunity dynamics. Annual Review of Ecology, Evolution, andSystematics 47:143–164.

Peay, K. G., C. Baraloto, P. V. A. Fine. 2013. Strong coupling of plantand fungal community structure across western Amazonian rain-forests. ISME Journal 7:1852–1861.

Peay, K. G., S. E. Russo, K. L. McGuire, Z. Lim, J. P. Chan, S. Tan,and S. J. Davies. 2015. Lack of host specificity leads to indepen-dent assortment of dipterocarps and ectomycorrhizal fungi acrossa soil fertility gradient. Ecology Letters 18:807–816.

Peh, K. S. H., S. L. Lewis, and J. Lloyd. 2011. Mechanisms of mono-dominance in diverse tropical tree-dominated systems. Journal ofEcology 99:891–898.

Pennington, T. D., C. Reynel, and A. Daza. 2004. Illustrated guide tothe trees of Peru. Hunt, Sherborne.

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).

S122 The American Naturalist

Petraitis, P. 2013. Multiple stable states in natural ecosystems. Ox-ford University Press, Oxford.

Phillips, R. P., E. Brzostek, andM. G. Midgley. 2013. The mycorrhizal-associated nutrient economy: a new framework for predictingcarbon-nutrient couplings in temperate forests. New Phytologist199:41–51.

Read, D. J. 1991. Mycorrhizas in ecosystems. Experientia 47:376–391.Reich, P. B. 2014. The world-wide ‘fast-slow’ plant economics spec-

trum: a traits manifesto. Journal of Ecology 102:275–301.Reich, P. B., M. B. Walters, and D. S. Ellsworth. 1997. From tropicsto tundra: global convergence in plant functioning. Proceedings ofthe National Academy of Sciences of the USA 94:13730–13734.

Richards, P. W. 1952. The tropical rain forest, an ecological study.Cambridge University Press, Cambridge.

Ricklefs, R. E., and D. Schluter, eds. 1993. Species diversity in ecolog-ical communities. University of Chicago Press, Chicago.

Russo, S. E., S. J. Davies, D. A. King, and S. Tan. 2005. Soil-relatedperformance variation and distributions of tree species in a Bor-nean rain forest. Journal of Ecology 93:879–889.

Russo, S. E., A. Kochsiek, J. Olney, L. Thompson, A. E. Miller, and S.Tan. 2013. Nitrogen uptake strategies of edaphically specializedBornean tree species. Plant Ecology 214:1405–1416.

Samuels, C. L., and J. A. Drake. 1997. Divergent perspectives on com-munity convergence. Trends in Ecology and Evolution 12:427–432.

Schneider, C. A., W. S. Rasband, and K. W. Eliceiri. 2012. NIH Im-age to ImageJ: 25 years of image analysis. Nature Methods 9:671–675.

Schröder, A., L. Persson, and A. M. de Roos. 2005. Direct experimen-tal evidence for alternative stable states: a review. Oikos 110:3–19.

Smith, M. E., T. W. Henkel, J. K. Uehling, A. K. Fremier, H. D.Clarke, and R. Vilgalys. 2013. The ectomycorrhizal fungal com-munity in a Neotropical forest dominated by the endemic diptero-carp Pakaraimaea dipterocarpacea. PLoS ONE 8:e55160.

Stern, D. L. 2013. The genetic causes of convergent evolution. NatureReviews Genetics 14:751–764.

Sutherland, J. P. 1974. Multiple stable points in natural communities.American Naturalist 108:859–873.

Tedersoo, L., T. W. May, and M. E. Smith. 2010. Ectomycorrhizal life-style in fungi: global diversity, distribution, and evolution of phylo-genetic lineages. Mycorrhiza 20:217–263.

Terborgh, J., L. C. Davenport, R. Niangadouma, E. Dimoto, J. C.Mouandza, O. Schultz, and M. R. Jaen. 2016. The African rain-forest: odd man out or megafaunal landscape? African and Ama-zonian forests compared. Ecography 39:187–193.

Teste, F. P., P. Kardol, B. L. Turner, D. A. Wardle, G. Zemunik, M.Renton, and E. Laliberté. 2017. Plant-soil feedback and the main-tenance of diversity in Mediterranean-climate shrublands. Science355:173–176.

This content downloaded from 132All use subject to University of Chicago Press Term

Torti, S. D., P. D. Coley, and T. A. Kursar. 2001. Causes and conse-quences of monodominance in tropical lowland forests. AmericanNaturalist 2:141–153.

Uriarte, M., N. G. Swenson, R. L. Chazdon, L. S. Comita, W. J. Kress,D. Erickson, J. Forero-Montana, J. K. Zimmerman, and J. Thomp-son. 2010. Trait similarity, shared ancestry and the structure ofneighbourhood interactions in a subtropical wet forest: implica-tions for community assembly. Ecology Letters 13:1503–1514.

Van Nes, E. H., and M. Scheffer. 2004. Large species shifts triggeredby small forces. American Naturalist 164:255–266.

Waring, B. G., L. Alvarez-Cansino, K. E. Barry, K. K. Becklund, S.Dale, M. G. Gei, A. B. Keller, et al. 2015. Pervasive and strong ef-fects of plants on soil chemistry: a meta-analysis of individualplant ‘Zinke’ effects. Proceedings of the Royal Society B 282:20151001.

Zobel, M. 2016. The species pool concept as a framework for study-ing patterns of plant diversity. Journal of Vegetation Science 27:8–18.

References Cited Only in the Online Appendixes

Anderson-Teixeira, K. J., S. J. Davies, A. C. Bennett, E. B. Gonzalez-Akre, H. C. Muller-Landau, S. J. Wright, K. Abu Salim, et al. 2015.CTFS-ForestGEO: a worldwide network monitoring forests in anera of global change. Global Change Biology 21:528–549.

Ashton, P. S. 2005. Lambir’s forest: the world’s most diverse knowntree assemblage? Pages 191–216 inD.W. Roubik, S. Sakai, and A. A.Hamid, eds. Pollination ecology and rain forest diversity, SarawakStudies. Springer, New York.

Davies, S. J., S. Tan, J. V. LaFrankie, and M. D. Potts. 2005. Soil-related floristic variation in the hyperdiverse dipterocarp forest inLambir Hills, Sarawak. Pages 22–34 in D. W. Roubik, S. Sakai,and A. A. Hamid, eds. Pollination ecology and rain forest diversity,Sarawak Studies. Springer, New York.

Lee, H. S., S. J. Davies, J. V. LaFrankie, S. Tan, T. Yamakura, A.Itoh, T. Ohkubo, and P. J. Ashton. 2002. Floristic and structuraldiversity of mixed dipterocarp forests in Lambir Hills NationalPark, Sarawak, Malaysia. Journal of Tropical Forest Science 14:379–400.

Tan, S., T. Yamakura, M. Tani, P. Palmiotto, J. D. Mamit, C. S. Pin,S. Davies, P. Ashton, and I. Baillie. 2009. Review of soils on the52-ha long term ecological research plot in mixed dipterocarp for-est at Lambir, Sarawak, Malaysian Borneo. Tropics 18:61–86.

Watson, H. 1985. Lambir Hills National Park: resource inventorywith management recommendations. National Parks and WildlifeOffice, Kuching.

Symposium Editor: Anurag A. Agrawal

.236.027.111 on July 22, 2017 05:14:57 AMs and Conditions (http://www.journals.uchicago.edu/t-and-c).