A Molecular View of MicrobialDiversity and the Biosphere

Norman R. Pace

Over three decades of molecular-phylogenetic studies, researchers have compiled anincreasingly robust map of evolutionary diversification showing that the main diversityof life is microbial, distributed among three primary relatedness groups or domains:Archaea, Bacteria, and Eucarya. The general properties of representatives of the threedomains indicate that the earliest life was based on inorganic nutrition and that pho-tosynthesis and use of organic compounds for carbon and energy metabolism camecomparatively later. The application of molecular-phylogenetic methods to study naturalmicrobial ecosystems without the traditional requirement for cultivation has resulted inthe discovery of many unexpected evolutionary lineages; members of some of theselineages are only distantly related to known organisms but are sufficiently abundant thatthey are likely to have impact on the chemistry of the biosphere.

Microbial organisms occupy a peculiarplace in the human view of life. Microbesreceive little attention in our general textsof biology. They are largely ignored by mostprofessional biologists and are virtually un-known to the public except in the contextsof disease and rot. Yet, the workings of thebiosphere depend absolutely on the activi-ties of the microbial world (1). Our textsarticulate biodiversity in terms of large or-ganisms: insects usually top the count ofspecies. Yet, if we squeeze out any one ofthese insects and examine its contents un-der the microscope, we find hundreds orthousands of distinct microbial species. Ahandful of soil contains billions of microbialorganisms, so many different types that ac-curate numbers remain unknown. At mostonly a few of these microbes would beknown to us; only about 5000 noneukary-otic organisms have been formally described(2) (in contrast to the half-million de-scribed insect species). We know so littleabout microbial biology, despite it being apart of biology that looms so large in thesustenance of life on this planet.

The reason for our poor understandingof the microbial world lies, of course, in thefact that microbes are tiny, individually in-visible to the eye. The mere existence ofmicrobial life was recognized only relativelyrecently in history, about 300 years ago,with Leeuwenhoek’s invention of the mi-croscope. Even under the microscope, how-ever, the simple morphologies of most mi-crobes, usually nondescript rods andspheres, prevented their classification bymorphology, the way that large organismshad always been related to one another. It

was not until the late 19th century and thedevelopment of pure-culture techniquesthat microbial organisms could be studied asindividual types and characterized to someextent, mainly by nutritional criteria. How-ever, the pure-culture approach to the studyof the microbial world seriously constrainedthe view of microbial diversity because mostmicrobes defy cultivation by standard meth-ods. Moreover, the morphological and nu-tritional criteria used to describe microbesfailed to provide a natural taxonomy, or-dered according to evolutionary relation-ships. Molecular tools and perspective basedon gene sequences are now alleviating theseconstraints to some extent. Even the earlyresults are changing our perception of mi-crobial diversity.

A Sequence-Based Mapof Biodiversity

Before the development of sequence-basedmethods, it was impossible to know theevolutionary relationships connecting all oflife and thereby to draw a universal evolu-tionary tree. Whittaker, in 1969, just as themolecular methods began to develop, sum-marized evolutionary thought in the con-text of the “Five Kingdoms” of life: animals,plants, fungi, protists (“protozoa”), andmonera (bacteria) (3). There also was rec-ognized a higher, seemingly more funda-mental taxonomic distinction between eu-karyotes, organisms that contain nuclearmembranes, and prokaryotes, predecessorsof eukaryotes that lack nuclear membranes(4). These two categories of organisms wereconsidered independent and coherent relat-edness groups. The main evolutionary di-versity of life on Earth, four of thefive traditional taxonomic kingdoms, wasthought to lie among the eukaryotes, par-

ticularly the multicellular forms.The breakthrough that called to ques-

tion many previous beliefs and brought or-der to microbial, indeed biological, diversityemerged with the determination of molec-ular sequences and the concept that se-quences could be used to relate organisms(5). The incisive formulation was reachedby Carl Woese who, by comparison of ribo-somal RNA (rRNA) sequences, establisheda molecular sequence–based phylogenetictree that could be used to relate all organ-isms and reconstruct the history of life (6,7). Woese articulated the now-recognizedthree primary lines of evolutionary descent,termed “urkingdoms” or “domains”: Eucarya(eukaryotes), Bacteria (initially called eu-bacteria), and Archaea (initially called ar-chaebacteria) (8).

Figure 1 is a current phylogenetic treebased on small-subunit (SSU) rRNA se-quences of the organisms represented. Theconstruction of such a tree is conceptuallysimple (9). Pairs of rRNA sequences fromdifferent organisms are aligned, and the dif-ferences are counted and considered to besome measure of “evolutionary distance” be-tween the organisms. There is no consider-ation of the passage of time, only of changein nucleotide sequence. Pair-wise differencesbetween many organisms can then be usedto infer phylogenetic trees, maps that repre-sent the evolutionary paths leading to themodern-day sequences. The tree in Fig. 1 islargely congruent with trees made using anymolecule in the nucleic acid–based, infor-mation-processing system of cells. On theother hand, phylogenetic trees based onmetabolic genes, those involved in the ma-nipulation of small molecules and in inter-action with the environment, commonly donot concur with the rRNA-based version[see (10, 11) for reviews and discussions ofphylogenetic results with different mole-cules]. Incongruities in phylogenetic treesmade with different molecules may reflectlateral transfers or even the intermixings ofgenomes in the course of evolution. Somemetabolic archaeal genes, for instance, ap-pear much more highly related to specificbacterial versions than to their eucaryal ho-mologs; other archaeal genes seem decidedlyeukaryotic in nature; still other archaealgenes are unique. Nonetheless, the recentlydetermined sequence of the archaeon Meth-anococcus jannaschii shows that the evolu-tionary lineage Archaea is independent ofboth Eucarya and Bacteria (12).

Interpreting the MolecularTree of Life

“Evolutionary distance” in this type of phy-logenetic tree (Fig. 1), the extent of se-quence change, is read along line segments.

The author is in the Departments of Plant and MicrobialBiology and Molecular and Cell Biology, University ofCalifornia, Berkeley, CA 94720–3102, USA. E-mail:nrpace@nature.berkeley.edu

SCIENCE z VOL. 276 z 2 MAY 1997 z www.sciencemag.org734

The tree can be considered a rough map ofthe evolution of the genetic core of thecellular lineages that led to the modernorganisms (sequences) included in the tree.The time of occurrence of evolutionaryevents cannot be extracted reliably fromphylogenetic trees, despite common at-tempts to do so. Time cannot be accuratelycorrelated with sequence change becausethe evolutionary clock is not constant indifferent lineages (7). This disparity is evi-denced in Fig. 1 by the fact that linesleading to the different reference organismsare not all the same length; these differentlineages have experienced different extentsof sequence change. Nonetheless, the orderof occurrence of branchings in the trees canbe interpreted as a genealogy, and intrigu-ing insights into the evolution of cells areemerging.

A sobering aspect of large-scale phyloge-netic trees such as that shown in Fig. 1 isthe graphical realization that most of ourlegacy in biological science, historicallybased on large organisms, has focused on anarrow slice of biological diversity. Thus,we see that animals (represented in Fig. 1 byHomo), plants (Zea), and fungi (Coprinus)constitute small and peripheral branches ofeven eukaryotic cellular diversity. If theanimals, plants, and fungi are taken to com-prise taxonomic “kingdoms,” then we mustrecognize as kingdoms at least a dozen othereucaryotic groups, all microbial, with asmuch or more independent evolutionaryhistory than that which separates the threetraditional eukaryotic kingdoms (13).

The rRNA and other molecular datasolidly confirm the notion stemming fromthe last century that the major organelles ofeukaryotes—mitochondria and chloro-plasts—are derived from bacterial symbi-onts that have undergone specializationthrough coevolution with the host cell. Se-quence comparisons establish mitochondriaas representatives of Proteobacteria (thegroup in Fig. 1 including Escherichia andAgrobacterium) and chloroplasts as derivedfrom cyanobacteria (Synechococcus andGloeobacter in Fig. 1) (14). Thus, all respi-ratory and photosynthetic capacity of eu-karyotic cells was obtained from bacterialsymbionts; the “endosymbiont hypothesis”for the origin of organelles is no longerhypothesis but well-grounded fact. The nu-clear component of the modern eukaryoticcell did not derive from one of the pro-karoytic lineages, however. The rRNA andother molecular trees show that the eukary-otic nuclear line of descent extends as deep-ly into the history of life as do the bacterialand archaeal lineages. The mitochondrionand chloroplast came in relatively late. Thislate evolution is evidenced by the fact thatmitochondria and chloroplasts diverged

from free-living organisms that branch pe-ripherally in molecular trees. Moreover, themost deeply divergent eukaryotes even lackmitochondria (15). These latter organisms,little studied but sometimes troublesomecreatures such as Giardia, Trichomonas, andVairimorpha, nonetheless contain at least afew bacterial-type genes (16). These genesmay be evidence of an earlier mitochondrialsymbiosis with Eucarya that was lost (11) orperhaps other symbiotic or gene-transferevents between the evolutionary domains.

The root of the universal tree in Fig. 1,the point of origin of the modern lineages,cannot be established using sequences ofonly one type of molecule. However, recentphylogenetic studies of gene families thatoriginated before the last common ancestorof the three domains have positioned theroot of the universal tree deep on the bacte-rial line (10). Therefore, Eucarya and Ar-chaea had a common history that excludedthe descendants of the bacterial line. Thisperiod of evolutionary history shared byEucarya and Archaea was an important timein the evolution of cells, during which therefinement of the primordial information-processing mechanisms occurred. Thus,modern representatives of Eucarya and Ar-chaea share many properties that differ frombacterial cells in fundamental ways. One ex-

ample of similarities and differences is in thenature of the transcription machinery. TheRNA polymerases of Eucarya and Archaearesemble each other in subunit compositionand sequence far more than either resemblesthe bacterial type of polymerase. Moreover,whereas all bacterial cells use sigma factors toregulate the initiation of transcription, euca-ryal and archaeal cells use TATA-bindingproteins (17, 18).

Because of the shared history of Eucaryaand Archaea, we should, perhaps, look tothe Archaea to identify fundamental prop-erties of far more complex cells such as ourown. The eukaryotic nuclear membrane, forinstance, is considered by cell biologists tobe an intrinsic component of the nucleus,somehow responsible for its integrity. Thefact that Archaea remained “prokaryotic,”that is, did not develop a nuclear mem-brane, indicates that a membrane is notrequired for nuclear function, which Ar-chaea certainly achieve (as do Bacteria, forthat matter). Indeed, the archaeal nuclearzone even seems to exclude ribosomes (19),and the genome ofM. jannaschii is sprinkledwith homologs of eucaryal nuclear and nu-cleolar structural genes (12). What consti-tutes a “nucleus?” Certainly the acquisitionof the nuclear membrane was a relativelylate event in the establishment of the eu-

Fig. 1. Universal phylo-genetic tree based onSSU rRNA sequences.Sixty-four rRNA se-quences representativeof all known phyloge-netic domains werealigned, and a tree wasproduced using FASTD-NAML (43, 52). That treewas modified, resultingin the composite oneshown, by trimming lin-eages and adjustingbranch points to incor-porate results of otheranalyses. The scale barcorresponds to 0.1changes per nucleotide.

ARTICLES

www.sciencemag.org z SCIENCE z VOL. 276 z 2 MAY 1997 735

caryal line of descent, occurring only afterthe separation from Archaea. Perhaps thenuclear membrane is after all not funda-mental to the function of the nucleus butrather is a relatively late-arriving embellish-ment. One hypothesis would be that thenuclear membrane was an invention de-rived from the Golgi apparatus to serve as agathering basket for nuclear products, fordistribution by the Golgi throughout thecell. The properties of nuclear pores wouldbe consistent with this hypothesis; they arelarge orifices, typically .10 nm in diameter,unlikely to gate anything except large mol-ecules (20). The evolutionary record sug-gests, then, that we look to something morefundamental than the nuclear membranefor the integrity of the nucleus and bywhich to define the essential quality of theeukaryotic cell. The shared evolutionaryhistory of Eucarya and Archaea suggeststhat we may be able to recognize the mostfundamental elements of our own nucleusthrough study of the archaeal version.

The Metabolic Diversity of Life

The molecular-phylogenetic perspective(Fig. 1) is a reference framework withinwhich to describe microbial diversity; thesequences of genes can be used to identifyorganisms. This capability is an importantconcept for microbial biology. It is not pos-sible to describe microorganisms as tradi-tionally done with large organisms, throughtheir morphological properties. To be sure,some microbes are intricate and beautiful inthe microscope, but they are mainly rela-tively unfeatured at the resolution of rou-tine microscopy. Therefore, in order to dis-tinguish different types of microbes, micro-biologists early turned to metabolic proper-ties such as utilizable sources of nutrition, forinstance, sources of carbon, nitrogen, andenergy. Microbial taxonomy accumulated asanecdotal descriptions of metabolically andmorphologically distinct types of organismsthat were essentially unrelatable. Molecularphylogeny now provides a framework withinwhich we can relate organisms objectively,and also through which we can interpret theevolutionary flow of the metabolic machin-eries that constitute microbial diversity.

Laboratory studies of microbial metabo-lism have focused mainly on organisms suchas Escherichia coli and Bacillus subtilis. In thebroad sense, such organisms metabolizemuch as animals do; we are all “organo-trophs,” using reduced organic compoundsfor energy and carbon. Organotrophy is notthe prevalent form of metabolism in theenvironment, however. Autotrophic metab-olism, fixation of CO2 to reduced organiccompounds, must necessarily contribute to agreater biomass than organotrophic metabo-

lism, which it supports (a principle longappreciated by ecologists). Energy for fixingCO2 is gathered in two ways: by phototrophy(photosynthesis) or lithotrophy (couplingthe oxidation of reduced inorganic com-pounds such as hydrogen, hydrogen sulfide,or ferrous iron to the reduction of a chemicaloxidant, a terminal electron acceptor such asoxygen, nitrate, sulfate, sulfur, or carbon di-oxide). Thus, metabolic diversity can begeneralized in terms of organotroph or au-totroph, phototroph or lithotroph, and thenature of the electron donor and acceptor.

The phylogenetic distributions of differ-ent types of carbon and energy metabolismamong different organisms do not necessar-ily follow the evolutionary pattern of rRNA(Fig. 1). Presumably, this lack of correspon-dence is because of past lateral transfers ofthose metabolic genes and larger scale sym-biotic fusions. Nonetheless, there are do-main-level tendencies that may speak tothe ancestral nature of the three domains oflife (21). The perspective here is currentlylimited mainly to Archaea and Bacteria.Such broad generalities cannot yet be as-sessed for the Eucarya because so little isknown about the metabolic breadth of thedomain, the properties of the most deeplydivergent lineages. There is considerableinformation about one pole of eukaryoticdiversity, that represented by animals,plants, and fungi. We know little about theother pole, the amitochondriate organismsthat spun off of the main eucaryal line earlyin evolution (22). The known instances ofsuch lineages, represented by Trichomonas,Giardia, and Vairimorpha in Fig. 1, are pri-marily pathogens. Pathogenicity to humansis a rare trait among the rest of eucaryotesand bacteria, and no archaeal pathogen isknown. This correlation may indicate thatnonpathogenic, deeply divergent eucaryotesare abundant in the environment but notyet detected. They should be sought inanaerobic ecosystems, possibly coupledmetabolically to other organisms. A driv-ing theme of the eucaryal line seems to bethe establishment of physical symbiosiswith other organisms. Beyond that, thegeneral metabolism of the rudimentary eu-karyotic cell seems simple, based on fer-mentative organotrophy. By virtue of sym-biotic partners, however, eukaryotes areable to take on phototrophic or lithotro-phic life-styles and to use the electron-acceptor oxygen (23).

Symbiotic microbes commonly conferthe lithotrophic way of life even on ani-mals, although this was only recently rec-ognized. The 2-m-long tubeworm Riftiapachyptila, for instance, lives in the vicinityof sea-floor hydrothermal vents and metab-olizes hydrogen sulfide and carbon dioxideby means of sulfide-oxidizing, carbon diox-

ide–fixing bacterial symbionts (24). Thisinvertebrate and metabolically similar onesmay contribute significantly to primary pro-ductivity in the ocean (25). It is not nec-essary to go to unusual (from our perspec-tive) places such as ocean-floor vents toencounter other equally fascinating hydro-gen sulfide–dependent eukaryotes (26).Underfoot at the ocean beach, for example,microbial respiration of seawater sulfate cre-ates a hydrogen sulfide–rich ecosystem pop-ulated by little-known creatures such asKentrophoros, a flat, gulletless ciliate thatunder the microscope appears fuzzy becauseit cultivates on its outer surface a crop ofsulfide-oxidizing bacteria (27). These bac-teria are ingested by endocytosis and there-by provide nutrition for Kentrophoros. Inother anaerobic environments, methano-gens, members of Archaea, live intracellu-larly with eukaryotes and serve as metabolichydrogen sinks (28). Still other symbiosesbased on inorganic energy sources are allaround us and are little explored for theirdiversity of microbial life (26).

Many lithotrophic, but comparatively feworganotrophic, representatives of Archaeahave been obtained in pure culture (29).There are primarily two metabolic themes,both relying on hydrogen as a main source ofenergy. Among the known Euryarchaeota—one of the two archaeal kingdoms knownthrough cultivated organisms—the mainelectron acceptor is carbon dioxide, and theproduct, methane—“natural gas.” Most ofthe methane encountered in the outer fewkilometers of Earth’s crust or on the surface isdetermined by isotopic analysis to be theproduct of methanogenic Archaea, past andpresent. Such organisms probably constitutea large component of global biomass. Theycertainly offer an inexhaustible source ofrenewable energy to humankind.

The general metabolic theme of the oth-er established kingdom of Archaea, Crenar-chaeota, is also the oxidation of molecularhydrogen, but with a sulfur compound as theterminal electron acceptor. All of the culti-vated representatives of Crenarchaeota alsoare thermophiles. Consequently, such or-ganisms have been referred to as “ther-moacidophilic” or “hyperthermophilic” Ar-chaea; some grow at the highest knowntemperatures for life, up to 113°C in the caseof Pyrolobus fumaris (30). These crenarchae-otes might seem bizarre, capable of thrivingat temperatures above the usual boilingpoint of water on a diet of H2, CO2, andelemental sulfur and exhaling hydrogen sul-fide. Yet, in terms of the molecular struc-tures of the basic cellular machineries, thesecreatures resemble eukaryotes far moreclosely than either resembles the bacteriumE. coli (17).

The metabolic diversity of microbes is

SCIENCE z VOL. 276 z 2 MAY 1997 z www.sciencemag.org736

usually couched in terms of the utilizationof complex organic compounds. From thatstandpoint, metabolic diversity seems, onthe basis of cultivated instances of organ-isms, to have flowered mainly among theBacteria. Even here, however, reliance onorganic nutrients probably was not ances-tral. The most deeply branching of the cul-tured bacterial lineages, represented byAquifex and Thermotoga in Fig. 1, are basi-cally lithotrophs that use hydrogen as anenergy source and electron acceptors suchas sulfur compounds (Thermotoga) or lowlevels of oxygen (Aquifex) (31). Cultivatedinstances of these deeply branching bacte-rial lineages also are all thermophilic andthus share two important physiological at-tributes with the deeply branching andslowly evolving Archaea: a hydrogen-basedenergy source and growth at high tempera-tures. This coincidence suggests that thelast common ancestor of all life also metab-olized hydrogen for energy at high temper-atures. This inference is consistent withcurrent notions regarding the origin of life,that it came to be in a geothermal setting athigh temperature (32).

Chlorophyll-based photosynthesis was abacterial invention. It seems to have ap-peared well after the establishment of thebacterial line of descent, at or before thedivergence of the line in Fig. 1 leading toChloroflexus, a photosynthetic genus (33),and after the deeper divergences such asthose leading to Aquifex and Thermotoga,genera that are not known to have photo-synthetic representatives. Most bacterialphotosynthesis is anaerobic, however. Oxy-genic photosynthesis, the water-based pho-tosynthetic mechanism that produces thepowerful electron acceptor oxygen, aroseonly in the kingdom-level lineage of cya-nobacteria. This invention changed the sur-face of Earth profoundly and conventional-ly is thought to be the basis, directly orindirectly, of most present-day biomass.

Anaerobic photosynthesis is widely dis-tributed in the late-branching bacterialkingdoms. The more ancient theme oflithotrophy, metabolism of inorganic com-pounds, is also widely distributed phyloge-netically, intermixed with organotrophicorganisms. The pattern suggests that organ-otrophy arose many times from otherwisephotosynthetic or lithotrophic organisms.Indeed, many instances of Bacteria canswitch between these modes of nutrition,carrying out photosynthesis in the light andlithotrophy or organotrophy in the dark.Particularly among Bacteria, this type ofenergy metabolism seems highly volatile inevolution: Bacteria that are closely relatedby molecular criteria can display strikinglydifferent phenotypes when assessed in thelaboratory through the nature of their car-

bon and energy metabolism. In the relative-ly closely related “gamma subgroup” of thekingdom of Proteobacteria (delineated bythe genus Escherichia in Fig. 1), for instance,we find the phenotypically disparate orga-nisms E. coli (organotroph), Chromatiumvinosum (hydrogen sulfide–based pho-totroph), and the symbiont of the tube-worm R. pachyptila (hydrogen sulfide–basedsymbiont). The superficial metabolic diver-sity of these types of Bacteria belies theirunderlying close evolutionary relatedness,giving no hint of the close similarities oftheir basic machineries. The versatility ofBacteria makes the metabolic machineriesof Archaea and Eucarya seem comparative-ly more monotonous. As the sequences ofdiverse genomes are compared, it will bepossible to map the flow of metabolic genesonto the rRNA-based tree and thereby seehow metabolic diversity has been moldedthrough evolution.

The molecular perspective gives us morethan just a glimpse of the evolutionary past;it also brings a new future to the disciplineof microbial biology. Because the molecu-lar-phylogenetic identifications are basedon sequence, as opposed to metabolic prop-erties, microbes can be identified withoutbeing cultivated. Consequently, all the se-quence-based techniques of molecular biol-ogy can be applied to the study of naturalmicrobial ecosystems, heretofore littleknown with regard to organismal makeup.

A Sequence-Based Glimpse ofBiodiversity in the Environment

Knowledge of microorganisms in the envi-ronment has depended in the past mainlyon studies of pure cultures in the laboratory.Rarely are microbes so captured, however.Studies of several types of environmentsestimate that more than 99% of organismsseen microscopically are not cultivated byroutine techniques (34). With the se-quence-based taxonomic framework of mo-lecular trees, only a gene sequence, not afunctioning cell, is required to identify theorganism in terms of its phylogenetic type.The occurrence of phylogenetic types oforganisms, “phylotypes,” and their distribu-tions in natural communities can be sur-veyed by sequencing rRNA genes obtainedfrom DNA isolated directly from the envi-ronment. Analysis of microbial ecosystemsin this way is more than a taxonomic exer-cise because the sequences provide experi-mental tools—for instance, molecular hy-bridization probes—that can be used toidentify, monitor, and study the microbialinhabitants of natural ecosystems (35).

Ribosomal RNA genes are obtained bycloning DNA isolated directly from the en-vironment. “Shotgun libraries” of random

DNA fragments are a source of rRNA, aswell as other genes, but require sorting ofrRNA genes from the others. The quickestway to survey the constituents of microbialecosystems is through the use of the poly-merase chain reaction (PCR) (36). Thehighly conserved nature of rRNA allows forthe synthesis of “universal” PCR primersthat can anneal to sequences conserved inthe rRNA genes from all three phylogeneticdomains. In principle, PCR carried out withthese primers amplifies the rRNA genes ofall types of organisms present in an envi-ronmental sample. Individual types of genesin the mixture are separated by a cloningstep and then sequenced.

A molecular-phylogenetic assessment ofan uncultivated organism can provide in-sight into many of the properties of theorganism through comparison with its stud-ied relatives. One example of the perspec-tive that phylogeny can offer on an other-wise unknown organism is seen with thesulfur-oxidizing microorganisms that pro-vide nutrition to symbiotic invertebratessuch as the vent tubeworm R. pachyptila(24). Although many attempts to cultivatethe symbionts for phenotypic characteriza-tion failed, rRNA analyses revealed thatmany of the basic cellular properties of thesymbionts were already familiar to us. TheRiftia symbiont and a number of other sul-fur-oxidizing symbionts associated with in-vertebrate animals all proved to be fairlyclosely related to one another, close rela-tives also to the intensively studied organ-isms E. coli and Pseudomonas aeruginosa(37). Because of their phylogenetic proxim-ity, many of the properties of the symbiontscan be inferred from those of the well-studied organisms. For instance, we can pre-dict with good confidence the nature of theribosome and antibiotic-susceptibility pat-terns, the nature of the DNA-replicativemachinery, the character of the RNA poly-merase complex, the character of biosyn-thetic pathways and their regulatory mech-anisms, the nature of the cell envelope andenergy transduction schemes, and manyother cellular properties of the symbionts.On the other hand, because E. coli and P.aeruginosa do not oxidize sulfur, these rela-tions cannot provide insight into the sulfur-oxidative pathways of the symbionts. TherRNA sequence does, however, identifyfree-living and cultivated (but less-studied)close relatives of the symbionts—for exam-ple Thiomicrospira sp. L-12 ( a hydrother-mal-vent isolate) and Thiothrix sp.—thatalso rely on sulfur oxidation and so arelikely to provide good models for this pro-cess in the symbionts.

Every nucleic acids–based study of nat-ural microbial ecosystems so far performedhas uncovered novel types of rRNA se-

ARTICLES

www.sciencemag.org z SCIENCE z VOL. 276 z 2 MAY 1997 737

quences, often representing major new lin-eages only distantly related to known ones.The discovery of rRNA sequences in theenvironment that diverge more deeply inphylogenetic trees than those of cultivatedorganisms is particularly noteworthy. Itmeans that the divergent organisms recog-nized by rRNA sequence are potentiallymore different from known organisms in thelineage than the known organisms are fromone another. The deepest divergences inboth the Bacteria and Archaea were firstdiscovered in rRNA-based surveys of hotspring–associated communities in Yellow-stone National Park.

The geothermal features of Yellow-stone National Park have been favoritehaunts of high-temperature biologists fordecades (38). Currently, rRNA-basedmethods are being used to survey phylo-types present in a number of Yellowstonehot springs with disparate chemical set-tings. One of these, Octopus Spring (Fig.2A)—a near-boiling, slightly alkaline, ex-tremely low-nutrient flow near Old Faith-ful geyser containing an abundant commu-nity of pink filaments—yielded the firstevidence for the lineage currently thoughtto be the most deeply divergent in theBacteria. When this lineage, representedby Aquifex and EM17 (pink filamentclone) (39) in Fig. 1, was first encounteredby 5S rRNA sequence (40), little could beinferred about the physiology of the asso-ciated organism because no cultivated spe-cific relative had yet been described. Sub-sequent clues to the nature of the pinkfilaments came with the discovery of A.pyrophilus, cultured from an Icelandic hotspring with a chemical character similar tothat of Octopus Spring (41), and determi-nation of the 16S rRNA sequence for thepink filaments (39). The A. pyrophilus andpink filament sequences are sufficientlyclosely related (Fig. 1) that many of theirproperties are likely to be shared. Themode of nutrition of the pink filaments,for instance, is predicted to be that ofAquifex, consumption of hydrogen withlow levels of oxygen and fixation of car-bon dioxide. Many other representativesof the Aquifex-EM17 relatedness group(Aquificales) have now been cultured,mainly from high-temperature settings,and all are thermophilic hydrogen oxidiz-ers (31).

Hot springs on the northern flank ofthe Yellowstone caldera usually have highconcentrations of iron (II), hydrogen sul-fide, hydrogen, and carbon dioxide—awealth of foodstuffs for compatible physi-ologies. Ongoing sequence-based studiesof the microbial inhabitants of one ofthese springs, Obsidian Pool (Fig. 2B),have radically revised our view of the phy-

logenetic diversity of Archaea. All culti-vated Crenarchaeota branch in the clusterbracketed by Pyrodictium and Thermofilumin Fig. 1. Discovery of a rich abundance ofdiverse crenarchaeal rRNA genes in Ob-sidian Pool sediment (for example, pSLsequences in Fig. 1), scores of new genera,expanded the known phylogenetic diver-sity (estimated by specific line-segmentlengths) of Crenarchaeota severalfold(42). More surprising, other sequencesfrom Obsidian Pool (pJP27 and pJP28 inFig. 1) seem to branch so deeply in theoverall archaeal tree that they constitute anew kingdom-level branch of Archaea,recognized provisionally as “Korarcha-eota” (43). It now will be interesting tostudy other genes from these novel organ-isms. These genes, as well as informationon the physiology and other properties ofthe organisms, will be obtained mostreadily if they can be cultured. Even with-out cultivation, however, cloning largefragments of environmental DNA andthen “chromosome walking” to assemblecontiguous clones offers access to the ge-nomes of these or other uncultivated or-ganisms (44).

Continuing study of Obsidian Pool isexpanding the known extent of bacterial,as well as archaeal, diversity. ObsidianPool, judged extremely inhospitable fromthe human standpoint, contains a richdiversity of sequence types representingmost of the known bacterial kingdoms, aswell as kingdom-level divergences neverdescribed by cultivation (45). Phyloge-netic studies of cultured and environmen-tal sequences have expanded substantiallyour appreciation of the scope of bacterialdiversity: In 1987, only about 12 phyloge-netic kingdoms (main phyla) of Bacteriawere recognized (Fig. 3, inset) (7), butnow, at least 25 to 30 distinct, kingdom-level phylogenetic divergences are re-solved (Fig. 3). The topology of the bac-terial tree is remarkable. Bacterial diversi-ty seems to have arisen mainly from an

explosive radiation of lineages, ratherthan from the sequential divergence ofmain lines seen, for instance, in the euca-ryal domain (Fig. 1). Preliminary resultsfrom Obsidian Pool also call into questionanother supposition based on culture stud-ies, that Archaea dominate high-tempera-ture environments. Quantitative hybrid-ization of domain-specific oligonucleotideprobes to rRNA genes obtained by PCRindicates that bacterial genes outnumberarchaeal genes by 50 :1 in this environ-ment. Such conclusions, of course, arecompromised to an unknown extent byconsiderations such as nonuniform ampli-fication of different rRNA genes, but thetrend seems to indicate that bacteria dom-inate this environment.

It is not necessary to go to extremeenvironments to encounter exotic diversi-ty; it is all around us. Phylotypes that,because of their abundance, must be sig-nificant contributors to the biospherehave escaped detection until the se-quence-based methods developed. Oneexample of an arena for research openedby the molecular methods involves therecently discovered mesophilic (low-tem-perature) Crenarchaeota (represented inFig. 1 by pGrf and marSBAR). On thebasis of culture-studies, crenarchaeoteshad been thought to be restricted to high-temperature environments. Cloned rRNAgene analysis shows, however, that low-temperature versions of Crenarchaeota areabundant globally in marine (19, 46) andterrestrial (47) environments, in typically30 to 50% of planktonic rRNA genes inlimited samplings of Atlantic, Pacific, andAntarctic waters (48). The physiologies ofthe low-temperature crenarchaeotes areunknown; none has yet been cultivated.The properties of their remote relatives—the cultivated, high-temperature Crenar-chaeota—hint that the mesophilic typesmight also engage in hydrogen metabo-lism, perhaps using some oxidation state ofsulfur as an electron acceptor.

A B

Fig. 2. Yellowstone National Park hot springs rich in microbial diversity. (A) Octopus Spring. The sourcepool of this hot spring is 90° to 93°C and extremely low in nutrients but contains abundant biomass andthe deepest known evolutionary divergences in the domain Bacteria. (B) Obsidian Pool. Molecularstudies find that the inhospitable environment of this hot spring, 75° to 95°C in temperature andcontaining high concentrations of iron (II) and hydrogen sulfide, supports an extensive diversity ofpreviously unknown microbial life, both archaeal and bacterial.

SCIENCE z VOL. 276 z 2 MAY 1997 z www.sciencemag.org738

Microbial Diversity and the Limitsof the Biosphere

Textbooks generally portray only a part ofthe global distribution of life, the part thatis immediately dependent on either theharvesting of sunlight or the metabolismof the decay products of photosynthesis.The molecular phylogenetic record shows,however, that lithotrophic metabolismpreceded and is more widespread phyloge-netically and geographically than is eitherphototrophy or organotrophy. The litho-trophic biosphere potentially extends ki-lometers into the crust of Earth, an essen-tially unknown realm (49). These consid-erations may indicate that lithotrophycontributes far more to the biomass ofEarth than currently thought.

Part of that lithotrophic biomass is inmicrohabitats all around us, usually awayfrom light and oxygen. It is not necessaryto look far to find such environments: therumens of cattle and the guts of termitesand humans, for example, are significantsources of methane, a signature of hydro-gen metabolism. Most life that depends oninorganic energy metabolism, however,probably is in little-known environments,based on poorly understood geochemis-tries. The oceans, for instance, cover 70%of Earth’s surface to an average depth of 4km. Most life in the ocean is microbial,and the metabolic patterns of such organ-isms are not understood: Large standingcrops of low-temperature crenarchaeotes,potential hydrogen oxidizers, may indicatean unsuspected, lithotrophy-based foodchain in the oceans. Another little-stud-ied environment with global significanceis the deep subsurface (50). There is in-creasing evidence that the crust of Earth is

shot through with biomass, wherever thephysical conditions permit. Metabolism ofhydrogen is a dominant theme among or-ganisms isolated from geothermal settingsor deep aquifers (51). Hydrogen is gener-ated readily by abiotic mechanisms such asinteraction of water with iron-bearing ba-salt, the main stuff of Earth’s crust; conse-quently, a food source is unlikely to belimiting in most subterranean environ-ments. Rather, it is likely to be the oxi-dant, the terminal electron acceptor, thatlimits growth. Nonetheless, it seems pos-sible that much, perhaps most, of the bio-mass on Earth is subterranean, a biologicalworld based on lithotrophy. Although themetabolic rate of this subterranean bio-sphere is likely to be far slower than in themore dynamic, photic environment, life islikely to be as pervasive in occurrence, andperhaps in cellular diversity, as we experi-ence on the surface.

Opportunity for an EnvironmentalGenome Survey

It is clear from even the small number ofenvironments so far studied with the mo-lecular methods that our understanding ofthe makeup of the natural microbial worldis rudimentary. The sequence-based meth-ods, however, now provide a way to surveybiodiversity rapidly and comprehensively.Ribosomal RNA genes gathered from theenvironment are snapshots of organisms,representatives of different types of ge-nomes, targets for further characterization ifthey seem interesting or useful. If we wantto understand the biosphere, I think it im-portant, even essential, that we undertake arepresentative survey of microbial diversityin the environment. A complete cataloging

of Earth’s microbial biota is needless and, ofcourse, impossible. A representative survey,however, is worthwhile. A representativesurvey could be achieved with modest ef-fort, with the use of automated sequencingtechnology. Analysis of 1000 clones (todetect the most abundant genome types)from each of 100 chemically different envi-ronments would be comparable to the effortto sequence a single microbial genome. Thequestions are large and many: What kindsof organisms do we share this planet withand depend on? What model systems shouldwe choose for laboratory studies of environ-mental processes? How extensive is thefund of biodiversity from which we candraw useful lessons and resources? Can weuse the distribution of microbes as a biosen-sor array to map and monitor the chemis-tries of Planet Earth? Are there deeperbranchings in the tree of life than the lin-eages we know?

The opportunities for the discovery ofnew organisms and the development of re-sources based on microbial diversity aregreater than ever before. Molecular se-quences have finally given microbial biolo-gists a way to define their subjects, throughmolecular phylogeny. The sequences alsoare the basis of the tools that will allowmicrobial biologists to explore the distribu-tion and roles of the organisms in the en-vironment. Microbial biology can now be awhole science; the organism can be studiedin the ecosystem.

REFERENCES______________

1. M. T. Madigan, J. M. Martinko, J. Parker, Brock Bi-ology of Microorganisms (Prentice-Hall, Upper Sad-dle River, NJ, ed. 8, 1996).

2. A. T. Bull, M. Goodfellow, J. H. Slater, Annu. Rev.Microbiol. 46, 219 (1992).

Fig. 3. Diagrammatic repre-sentation of the known phy-logenetic span of Bacteria in1987 (inset) and today. Phy-logenetic trees containingsequences from the indicat-ed organisms or groups oforganisms, chosen to repre-sent the broadest diversityof Bacteria, were used asthe basis for this diagram(compiled with P. Hugen-holtz). Filled sectors indicatethat several representative

sequences fall within the indicated depth of branching. Linesdesignated by OP represent one or more phylotypes that wereidentified in Obsidian Pool by means of molecular methods buthave not been not cultivated. The inset is an outline of thebacterial tree compiled by Woese in 1987 (7 ).

ARTICLES

www.sciencemag.org z SCIENCE z VOL. 276 z 2 MAY 1997 739

3. R. H. Whittaker, Science 163, 150 (1969).4. E. Chatton, Titres et Travoux Scientifiques (Sete,

Sottano, Italy, 1937).5. E. Zuckerkandl and L. Pauling, J. Theor. Biol. 8, 357

(1965); R. M. Schwartz and M. O. Dayhoff, Science199, 395 (1978).

6. C. R. Woese and G. E. Fox, Proc. Natl. Acad. Sci.U.S.A. 74, 5088 (1977).

7. C. R. Woese, Microbiol. Rev. 51, 221 (1987).8. iiii, O. Kandler, M. L. Wheelis, Proc. Natl. Acad.

Sci. U.S.A. 87, 4576 (1990).9. D. L. Swofford, G. J. Olsen, P. J. Waddell, D. M.

Hillis, in Molecular Systematics, D. M. Hillis, C.Moritz, B. K. Mable, Eds. (Sinauer, Sunderland, MA,1996), pp. 407–514.

10. W. F. Doolittle and J. R. Brown, Proc. Natl. Acad.Sci. U.S.A. 91, 6721 (1994).

11. J. D. Palmer, Science 275, 790 (1997).12. C. J. Bult et al., ibid. 273, 1058 (1996).13. The taxonomic term “kingdom” has no molecular

definition. I use the term to indicate main lines ofradiation in the particular domain; 14 such “king-dom-level” lines are associated with the eucaryal lineof descent in Fig. 1 [see also (22)].

14. J. Sapp, Evolution by Association: A History of Sym-biosis (Oxford Univ. Press, New York, 1994).

15. T. Cavalier-Smith, Microbiol. Rev. 57, 953 (1993).16. E. T. N. Bui, P. J. Bradley, P. J. Johnson, Proc. Natl.

Acad. Sci. U.S.A. 93, 9651 (1996); A. Germot, H.Philippe, H. le Guyader, ibid., p. 14614; A. J. Roger,C. G. Clark, W. F. Doolittle, ibid., p. 14618.

17. T. L. Marsh, C. I. Reich, R. B. Whitelock, G. J. Olsen,ibid. 91, 4180 (1994).

18. T. Rowlands, P. Baumann, S. P. Jackson, Science264, 1326 (1994).

19. E. F. DeLong, Proc. Natl. Acad. Sci. U.S.A. 89, 5685(1992).

20. L. I. Davis, Annu. Rev. Biochem. 64, 865 (1995).

21. O. Kandler, in Early Life on Earth, S. Bengtson, Ed.(Nobel Symp. 84, Columbia Univ. Press, New York,1993), pp. 152–160.

22. M. L. Sogin, ibid., pp. 181–192.23. D. C. Smith and A. E. Douglas, The Biology of Sym-

biosis (Edward Arnold, London, 1987).24. V. Tunnicliffe, Am. Sci. 80, 336 (1992).25. R. A. Lutz et al., Nature 371, 663 (1994).26. T. Fenchel and B. J. Finlay, Ecology and Evolution in

Anoxic Worlds, R. M. May and P. H. Harvey, Eds.(Oxford Ser. Ecol. Evol., Oxford Univ. Press, Oxford,1995).

27. iiii, Ophelia 30, 75 (1989).28. T. M. Embley and B. J. Finlay,Microbiology 140, 225

(1994).29. M. Kates, D. J. Kushner, A. T. Matheson, Eds., The

Biochemistry of Archaea (Archaebacteria) (Elsevier,Amsterdam, 1993).

30. K. O. Stetter, Am. Soc. Microbiol. News 61, 285(1995).

31. C. Pitulle et al., Int. J. Syst. Bacteriol. 44, 620 (1994).32. N. R. Pace, Cell 65, 531 (1991).33. B. K. Pierson, in Early Life on Earth (Nobel Symp. 84,

Columbia Univ. Press, New York, 1993), pp. 161–180.

34. R. I. Amann, W. Ludwig, K. H. Schleifer, Microbiol.Rev. 59, 143 (1995).

35. N. R. Pace, D. A. Stahl, D. J. Lane, G. J. Olsen, Am.Soc. Microbiol. News 51, 4 (1985); P. Hugenholtzand N. R. Pace, Trends Biotechnol. 14, 190 (1996).

36. R. K. Saiki et al., Science 239, 487 (1988).37. D. L. Distel et al., J. Bacteriol. 170, 2506 (1988).38. T. D. Brock, Thermophilic Microorganisms and Life

at High Temperatures (Springer-Verlag, New York,1978).

39. A.-L. Reysenbach, G. S.Wickham, N. R. Pace, Appl.Environ. Microbiol. 60, 2113 (1994).

40. D. A. Stahl, D. J. Lane, G. J. Olsen, N. R. Pace, ibid.

49, 1379 (1985).41. R. Huber et al., Syst. Appl. Microbiol. 15, 340 (1992).42. S. M. Barns, R. E. Fundyga, M. W. Jeffries, N. R.

Pace, Proc. Natl. Acad. Sci. U.S.A. 91, 1609 (1994).43. S. M. Barns, C. F. Delwiche, J. D. Palmer, N. R.

Pace, ibid. 93, 9188 (1996).44. J. L. Stein, T. L. Marsh, K. Y. Wu, H. Shizuya, E. F.

DeLong, J. Bacteriol. 178, 591 (1996).45. P. Hugenholtz, C. Pitulle, K. L. Hershberger, N. R.

Pace, in preparation.46. J. A. Fuhrman, K. McCallum, A. A. Davis, Nature

356, 148 (1992).47. T. Ueda, Y. Suga, T. Matsuguchi, Eur. J. Soil Sci. 46,

415 (1995); K. L. Hershberger, S. M. Barns, A.-L.Reysenbach, S. C. Dawson, N. R. Pace,Nature 384,420 (1996); S. B. Bintrim, T. J. Donohue, J. Handels-man, G. P. Roberts, R. M. Goodman, Proc. Natl.Acad. Sci. U.S.A. 94, 277 (1997); G. Jurgens, K.Lindstrom, A. Saano, Appl. Environ. Microbiol. 63,803 (1997).

48. E. F. DeLong, K. Y. Wu, B. B. Prezelin, R. V. M.Jovine, Nature 371, 695 (1994).

49. W. C. Ghiorse, Science 275, 789 (1997).50. T. Gold, Proc. Natl. Acad. Sci. U.S.A. 89, 6045

(1992); J. K. Fredrickson and T. C. Onstott, Sci. Am.275, 68 (October 1996); D. Lovely and F. Chapelle,Rev. Geophys. 33, 365 (1995).

51. T. O. Stevens and J. P. McKinley, Science 270, 450(1995); K. Pedersen, Earth Sci. Rev. 34, 243 (1993).

52. B. L. Maidak et al., Nucleic Acids Res. 25, 109(1997).

53. I thank S. Kustu, G. Olsen, and C. Woese for helpfulcomments on the manuscript, and S. Barns, S.Dawson, C. Delwiche, and P. Hugenholtz for assist-ance with the figures. Research in my laboratory issupported by grants from the U.S. Department ofEnergy and NIH.

SCIENCE z VOL. 276 z 2 MAY 1997 z www.sciencemag.org740