Capture mechanism in Palaeotropical pitcher plants ...Capture mechanism in Palaeotropical pitcher...

Capture mechanism in Palaeotropical pitcher plants (Nepenthaceae) isconstrained by climate

Jonathan A. Moran1,*, Laura K. Gray2, Charles Clarke3 and Lijin Chin3

1School of Environment and Sustainability, Royal Roads University, Victoria, British Columbia, V9B 5Y2 Canada, 2Department ofRenewable Resources, University of Alberta, Edmonton, Alberta, T6G 2H1 Canada and 3School of Science, Monash University,

Bandar Sunway, Selangor, 46150 Malaysia* For correspondence. E-mail [email protected]

Received: 6 May 2013 Returned for revision: 3 June 2013 Accepted: 3 July 2013

† Background and Aims Nepenthes (Nepenthaceae, approx. 120 species) are carnivorous pitcher plants with a centreof diversity comprising the Philippines, Borneo, Sumatra and Sulawesi. Nepenthes pitchers use three main mechan-isms for capturing prey: epicuticular waxes inside the pitcher; a wettable peristome (a collar-shaped structure aroundthe opening); and viscoelastic fluid. Previous studies have provided evidence suggesting that the first mechanism maybe more suited to seasonal climates, whereas the latter two might be more suited to perhumid environments. In thisstudy, this idea was tested using climate envelope modelling.† Methods A total of 94 species, comprising 1978 populations, were grouped by prey capture mechanism (large peri-stome, small peristome, waxy, waxless, viscoelastic, non-viscoelastic, ‘wet’ syndrome and ‘dry’ syndrome).Nineteen bioclimatic variables were used to model habitat suitability at approx. 1 km resolution for each group,using Maxent, a presence-only species distribution modelling program.† Key Results Prey capture groups putatively associated with perhumid conditions (large peristome, waxless, visco-elastic and ‘wet’ syndrome) had more restricted areas of probable habitat suitability than those associated putativelywith less humid conditions (small peristome, waxy, non-viscoelastic and‘dry’ syndrome). Overall, the viscoelasticgroup showed the most restricted area of modelled suitable habitat.† Conclusions The current study is the first to demonstrate that the prey capture mechanism in a carnivorous plant isconstrained by climate. Nepenthes species employing peristome-based and viscoelastic fluid-based capture arelargely restricted to perhumid regions; in contrast, the wax-based mechanism allows successful capture in both per-humid and more seasonal areas. Possible reasons for the maintenance of peristome-based and viscoelastic fluid-basedcapture mechanisms in Nepenthes are discussed in relation to the costs and benefits associated with a given preycapture strategy.

Key words: Biogeography, carnivorous plants, climate envelope modelling, Nepenthes, Nepenthaceae,Palaeotropics, pitcher plant, capture mechanism.

INTRODUCTION

The monotypic family Nepenthaceae comprises approx. 120species of carnivorous pitcher plants (McPherson, 2009).Invertebrate prey is attracted to the pitcher via a combinationof nectar, colour patterns and, in some cases, fragrance (e.g.Moran, 1996; Moran et al., 1999, 2012; DiGiusto et al., 2008,2010). Nepenthes benefit from prey-derived nutrients, whichaugment the supply available via the roots (Moran and Moran,1998; Pavlovic et al., 2007). The centre of diversity ofNepenthes comprises the islands of Borneo, Sumatra, Sulawesiand the Philippines; the range extends north to Assam, west toMadagascar and south-east to Australia and New Caledonia(Danser, 1928). Nepenthes pitcher production incurs significantresource costs (Osunkoya et al., 2007, 2008). Therefore, incommon with other carnivorous plants, Nepenthes pitcherplants are restricted to habitats in which the nutritional benefitsfrom carnivory outweigh the costs of producing carnivorousorgans (Givnish et al., 1984). As a consequence, members ofthe genus are somewhat patchily distributed across their range.

For example, they are not typically found in the lowland diptero-carp forests that historically covered much of the region.

It has been suggested that the Nepenthaceae provide anexample of adaptive radiation, based on pitcher specializationfor nutrient capture (e.g. Chin et al., 2010; Pavlovic, 2012).Certainly, there exists a range of specializations. For example,Nepenthes ampullaria derives a significant proportion of its ni-trogen from interception of falling leaf litter (Moran et al.,2003; Pavlovic et al., 2011), while four other species(Nepenthes baramensis, Nepenthes lowii, Nepenthes rajah andNepenthes macrophylla) sequester nutrients from mammalianexcreta (Clarke et al., 2009; Chin et al., 2010; Grafe et al.,2011; Wells et al., 2011). Nepenthes albomarginata is atermite specialist (Moran et al., 2001; Merbach et al., 2002),and Nepenthes bicalcarata has a mutualistic nutritional associ-ation with the ant Camponotus schmitzi (Clarke and Kitching,1995; Merbach et al., 2007; Bazile et al., 2012; Thornhamet al., 2012). However, the majority of Nepenthes speciesstudied thus far capture and digest a range of small arthropods,and many are particularly attractive to ants (e.g. Juniper et al.,

# The Author 2013. Published by Oxford University Press on behalf of the Annals of Botany Company. All rights reserved.

For Permissions, please email: [email protected]

Annals of Botany of 13

doi:10.1093/aob/mct195, available online at www.aob.oxfordjournals.org

at Royal R

oads University L

ibrary on August 26, 2013

http://aob.oxfordjournals.org/D

ownloaded from

mailto:[email protected]

http://aob.oxfordjournals.org/

1989; Moran, 1996; Moran et al., 1999; Di Giusto et al., 2008;Bauer et al., 2009). Although Nepenthes pitchers appear to besimple gravity-fed pitfall traps, there are three distinct mechan-isms by which they effect capture of their invertebrate prey.The first two mechanisms function to trap the prey; the third toretain it (for a recent review, see Benz et al., 2012).

(1) A wettable peristome (function: trapping). The peristome isthe collar-shaped structure that surrounds the pitcheropening, and is the location of the major extrafloral nectariesin many species (Juniper et al., 1989). The surface of theperistome is covered in microscopic radial ridges, arrayedanisotropically to aid traction in the direction of the pitchermouth, but to deny traction in the opposite direction, forsome prey taxa (Bohn and Federle, 2004). Capillary actionbetween the ridges also renders the peristome highly wet-table, and it has been shown that a wetted peristome providesvery little traction for insect feet, effectively causing the preyto ‘aquaplane’ and slip into the pitcher. Conversely, manyinsect visitors have little trouble safely traversing the peri-stome when it is dry (Bohn and Federle, 2004; Bauer et al.,2008, 2009, 2011, 2012a, b). The width of the peristome(and thus its presumed effectiveness as a conveyingsurface) varies considerably among species.

(2) Epicuticular waxes (function: trapping). In many Nepenthesspecies, the upper part of the inner pitcher wall is covered inone or more layers of epicuticular waxes (Riedel et al., 2003,2007; Poppinga et al., 2010). These waxes can interfere withan insect’s foothold in a number of ways. First, wax crystalscan detach from the top layer to clog the claws, thus denyingtraction. Secondly, thewaxysurface is unwettable, denying afoothold to insects that rely on capillary forces for traction(Juniper and Burras, 1962; Gaume et al., 2002, 2004; Gorbet al., 2004; Gorb and Gorb, 2006; Wang et al., 2009;Scholz et al., 2010). Finally, the microscopic roughness ofthe wax crystal layer itself may reduce the effective areaavailable for traction by the prey’s adhesive pads (Scholzet al., 2010). In many Nepenthes species, epicuticularwaxes work in conjunction with semi-lunate cells. Theseare modified stomatal guard cells that line the upper part ofthe inner pitcher wall. They present an anisotropic surfacethat denies traction to the claws of insects attempting toescape (Pant and Bhatnagar, 1977; Owen and Lennon,1999; Gaume et al., 2002; Thornhill et al., 2008; Wanget al., 2009; Gorb and Gorb, 2011). In contrast to theaction of the peristome, the waxy zone appears to operateequally well under both wet and dry conditions.

(3) Viscoelastic pitcher fluid (function: retention). A number ofNepenthes species produce pitcher fluid with viscoelasticproperties. Insects falling into this type of fluid find it impos-sible to extricate themselves, and quickly drown (Gaume andForterre, 2007; Gaume and Di Giusto, 2009; Bonhommeet al., 2011).

There is considerable variation in the degree to which a givenNepenthes species employs each mechanism. Bauer et al.(2012a) divided the capture strategy of Nepenthes species intotwo broad groups: peristome based and wax based. In addition,Benz et al. (2012) demonstrated an inverse relationshipbetween peristome width and extent of the waxy zone in eight

Nepenthes species, suggesting disruptive selective pressure toadopt one of two capture mechanisms. Similarly, Bonhommeet al. (2011) established an inverse relationship between theviscoelasticity of pitcher fluid and the amount of wax producedin 23 Nepenthes species, once more suggesting disruptive selec-tion. A potential driver of such selection is climate: Bauer et al.(2012a) raised the possibility that a perhumid climate mightfavour the peristome-based strategy, whereas drier or more sea-sonal climates might favour a wax-based strategy (see points 1and 2, above). Bonhomme et al. (2011) suggested that the pro-duction of viscoelastic fluid (as opposed to waxes) favours thecapture of volant prey, which are relatively more common thancrawling prey such as ants, in montane habitats. Of greater rele-vance to the current study, they also demonstrated that the visco-elastic Nepenthes fluid retains its retentive properties, even whendiluted by the heavy rainfall typical of such environments.

Therefore, we decided to test the idea that the employment ofthe various capture mechanisms described above might be drivenby climatic variables. Specifically, we explored the possibilitythat peristome-based and viscoelastic capture strategies arerestricted to perhumid environments, whereas the wax-basedstrategy is favoured in drier or more seasonal environments.We took a climate envelope modelling approach, using Maxent,a widely used species distribution modelling program (Phillipset al., 2004, 2006; Elith et al., 2011; http://www.cs.princeton.edu/~schapire/maxent/).

MATERIALS AND METHODS

Nepenthes species population localities and pitcher characteristics

Geographical co-ordinates were recorded for populations of 94Nepenthes species (Table 1). This represents approx. 80 % ofthe species currently described (McPherson, 2009). A total of1978 populations were recorded from Thailand, Laos, Vietnam,Cambodia, Peninsular Malaysia, Sumatra, Java, Borneo, thePhilippine Archipelago and Sulawesi. The populations of 69species were observed directly in the wild by one of us (C.C.);data on the remaining species were derived from examinationsof herbarium material, cultivated plants and taxonomic descrip-tions (e.g. Danser, 1928). For many Nepenthes species withlimited geographical distributions, the observations representthe entire currently known geographical range. Members of theNepenthaceae are remarkable for high degrees of endemism:many are montane with restricted geographical ranges, often asingle mountain or a small number of adjacent mountains (e.g.N. macrophylla, N. clipeata, N. aristolochioides; Clarke, 1997,2001). Consequently, for 76 of our study species (approx.80 %), only a small number of populations (,10) are currentlyknown to exist. In contrast, a small number of species havevery large geographical ranges, and a consequently largernumber of identified populations. For example, 460 populationsof N. gracilis were identified, spanning Thailand, PeninsularMalaysia, Sumatra, Borneo and Sulawesi; similarly, there were342 populations of N. ampullaria identified, encompassingThailand, Peninsular Malaysia, Sumatra and Borneo (Table 1).Initially, we considered randomly sub-sampling from the set ofpopulations for these wide-ranging species, to limit any potentialtaxonomic bias in our results as a result of one or two species nu-merically dominating a given prey capture mechanism group

Moran et al. — Climate and capture mechanism in Palaeotropical pitcher plants of 13

at Royal R

oads University L



ownloaded from

http://www.cs.princeton.edu/~schapire/maxent/




TABLE 1. Summary of Nepenthes species used in the study, including pitcher characteristics

Species No. of populations Peristome size* Wax Viscoelastic fluid†

Nepenthes adnata 3 Small Present AbsentNepenthes alata 17 Small Present AbsentNepenthes alba 10 Small Present AbsentNepenthes albomarginata 27 Small Present AbsentNepenthes ampullaria 342 Large Absent AbsentNepenthes andamana 5 Small Present AbsentNepenthes argentii 2 Medium Absent AbsentNepenthes aristolochioides 2 Medium Absent PresentNepenthes baramensis 12 Medium Present AbsentNepenthes bellii 5 Medium Absent NA‡

Nepenthes benstonei 9 Small Present AbsentNepenthes bicalcarata 16 Large Absent AbsentNepenthes bokorensis 3 Small Present AbsentNepenthes bongso 13 Large Reduced PresentNepenthes boschiana 3 Large Present AbsentNepenthes burbidgeae 4 Large Absent PresentNepenthes burkei 1 Large Reduced AbsentNepenthes chang 2 Small Present AbsentNepenthes chaniana 5 Medium Present PresentNepenthes clipeata 1 Small Present NANepenthes copelandii 1 Medium Reduced PresentNepenthes densiflora 4 Medium Reduced PresentNepenthes diatas 3 Medium Present PresentNepenthes dubia 3 Small Absent PresentNepenthes edwardsiana 2 Medium Present AbsentNepenthes ephippiata 3 Medium Reduced AbsentNepenthes eustachya 14 Small Present AbsentNepenthes eymae 5 Medium Reduced PresentNepenthes faizaliana 3 Large Present AbsentNepenthes flava 1 Large Absent PresentNepenthes fusca 47 Medium Reduced PresentNepenthes glabrata 8 Small Present PresentNepenthes glandulifera 1 Small Reduced PresentNepenthes gracilis 460 Small Present AbsentNepenthes gracillima 7 Medium Present AbsentNepenthes gymnamphora 86 Small Present AbsentNepenthes hamata 3 Medium Present AbsentNepenthes hirsuta 19 Small Reduced AbsentNepenthes hispida 5 Small Present AbsentNepenthes holdenii 1 Small Present AbsentNepenthes inermis 9 Small Reduced PresentNepenthes izumiae 3 Medium Present AbsentNepenthes jacquelineae 2 Large Absent PresentNepenthes jamban 1 Medium Absent PresentNepenthes kampotiana 3 Small Present AbsentNepenthes kerrii 2 Small Present AbsentNepenthes kongkandana 2 Small Present AbsentNepenthes lavicola 1 Medium Present AbsentNepenthes lingulata 1 Medium Present AbsentNepenthes longifolia 5 Medium Present PresentNepenthes lowii 19 Medium Reduced AbsentNepenthes macfarlanei 15 Medium Present PresentNepenthes macrophylla 3 Medium Present PresentNepenthes macrovulgaris 4 Small Reduced AbsentNepenthes mapuluensis 3 Large Present NANepenthes maxima 10 Medium Present PresentNepenthes merrilliana 9 Medium Absent AbsentNepenthes mikei 3 Small Present AbsentNepenthes mira 1 Medium Absent NANepenthes mirabilis 148 Small Present AbsentNepenthes muluensis 6 Small Present AbsentNepenthes murudensis 4 Small Present AbsentNepenthes northiana 5 Large Reduced PresentNepenthes ovata 4 Large Reduced PresentNepenthes petiolata 2 Medium Present PresentNepenthes philippinensis 4 Small Present Absent

Continued


at Royal R

oads University L



ownloaded from


(see below for details regarding groups). However, to have down-sampled would have created false absences, which would havereduced the accuracy of the models used. It is important tostress that for the purposes of this study, we were focusednot on individual species, but on the capture mechanism.Therefore, if a population of plants using a given mechanismwas known and located, it was included in the analysis. For ourpurposes, the specific status of a plant was less important thanthe fact that, for instance, a population of Nepenthes conformingto a given capture mechanism group or syndrome (see below)was present at a given location. For each species, the followingpitcher attributes were quantified.

(1) Peristome size. We adopted the approach of Bauer et al.(2012a) in assigning peristomewidth classes based on the re-lationship of peristome width to pitcher length. The relativeperistome size of 65 species was derived from fig. 8 of Baueret al. (2012a); values for the remaining 29 species in thecurrent study were derived from field measurements byC.C. Species were then assigned to one of three peristomesize categories (small, medium and large). These correspondto the following categories of peristome width in fig. 8 ofBauer et al. (2012a): small ¼ –20 to –10; medium ¼ –5to +5; large ¼ + 10 to +25. These numerical values referto the magnitude of the residuals from the peristome

width/pitcher length regression relationship: positivevalues denote larger than average peristomes; negativevalues, smaller than average (Table 1). We did not apply acorrection for phylogeny (see Cheverud et al., 1985) to theperistome size measurements, since Bauer et al. (2012a)found no significant difference between phylogeneticallycorrected and uncorrected morphological data for the 65species that they examined in their study (i.e. no detectablephylogenetic signal with regards to morphological traits).Since their 65 species account for 70 % of the totalnumber of species used in the current study, we believe ithighly unlikely that the other species we used woulddepart from this pattern in any significant way.

(2) Presence/absence of wax. We used species data describingwax production from Bonhomme et al. (2011), Bauer et al.(2012a) and Benz et al. (2012), as well as direct observationof specimens in the field by C.C. (Table 1). For the reasonsoutlined above, we did not apply a phylogenetic correctionto the wax data-set.

(3) Presence/absence of viscoelastic pitcher fluid. For eachspecies, the presence or absence of viscoelastic pitcherfluid was noted. Data for 23 species were taken from thedata set presented by Bonhomme et al. (2011); the remainingdata were from direct field observations by C.C. A potentialsource of error in our approach is that assessment of

TABLE 1. Continued

Species No. of populations Peristome size* Wax Viscoelastic fluid†

Nepenthes pilosa 1 Small Reduced PresentNepenthes rafflesiana 238 Medium Reduced PresentNepenthes rajah 5 Large Absent AbsentNepenthes ramispina 7 Small Present AbsentNepenthes reinwardtiana 39 Small Present AbsentNepenthes rhombicaulis 3 Medium Present AbsentNepenthes sanguinea 68 Small Present AbsentNepenthes sibuyanensis 2 Medium Absent AbsentNepenthes singalana 8 Medium Present AbsentNepenthes smilesii 19 Small Present AbsentNepenthes spathulata 9 Medium Present PresentNepenthes spectabilis 7 Medium Present PresentNepenthes stenophylla 22 Small Present PresentNepenthes sumatrana 6 Medium Reduced PresentNepenthes suratensis 3 Small Present AbsentNepenthes talangensis 3 Small Absent PresentNepenthes tenuis 1 Medium Absent PresentNepenthes tentaculata 56 Small Present AbsentNepenthes thai 1 Small Present AbsentNepenthes thorelli 3 Small Present AbsentNepenthes tobaica 13 Small Present PresentNepenthes tomoriana 3 Small Present NANepenthes truncata 6 Large Reduced AbsentNepenthes veitchii 13 Large Absent AbsentNepenthes ventricosa 2 Large Absent AbsentNepenthes villosa 5 Medium Present AbsentNepenthes vogelli 7 Medium Absent PresentPopulation total 1978Species total 94

* Categories correspond to those presented in fig. 8 of Bauer et al. (2012a) as follows: small ¼ –20 to –10; medium ¼ –5 to +5; large ¼ + 10 to +25 (valuesrefer to the magnitude of the residuals from the peristome width/pitcher length regression relationship).

† Field evaluations of pitcher fluid viscoelasticity were based on direct observations, not rheological measurements. This is a potential source of error.‡ NA ¼ data unavailable.Data sources: Bonhomme et al. (2011); Bauer et al. (2012a); Benz et al. (2012); personal field measurements/observations (C.C.).


at Royal R

oads University L



ownloaded from


viscoelasticity was made based on direct observation ofpitcher fluid, rather than rheological measurement.

Nepenthes undergo ontogenic change in pitcher morphology.Young plants typically produce terrestrial pitchers that are ovoidin shape; as the plant grows, aerial pitchers are produced. Theseare usually more elongate in form. In a number of species, thisontogenic shift in gross morphology is accompanied bychanges in wax production, fluid viscoelasticity and/or relativewidth of the peristome. In such cases, we took the approachthat if, for example, viscoelastic fluid was produced by eitherpitcher form of a given species, then that species was classedas employing viscoelastic fluid (Table 1). The same approachwas taken regarding ontogenic changes in wax production andperistome size.

Nepenthes capture mechanism groupings

Based on the three attributes outlined above, we grouped thespecies into eight putative capture mechanism groups (hereaftercalled groups), summarized in Table 2. Six of the groups focus ona single pitcher characteristic (peristome size, viscoelasticity offluid and waxiness). The final two groups represent combinationsof these characteristics: the ‘wet’ syndrome comprises speciesthat possess large peristomes, and produce no wax; the ‘dry’ syn-drome comprises species with small peristomes, wax and non-viscoelastic fluid. It might have been considered optimal tohave incorporated only species that produce viscoelastic fluidinto the ‘wet’ syndrome, based on the observations of Bonhommeet al. (2011). However, onlyseven species possessed pitchers withlarge peristomes, viscoelastic fluid and no wax, and the totalnumber of populations for this combination of traits was only18, far below the threshold of 80 required for meaningful model-ling using Maxent (Elith et al., 2011). The large and medium peri-stomesize classes were combined in the large peristomegroup and‘wet’ syndrome. This is because there appears to be little differ-ence between the two peristome size categories from a functionalviewpoint. For example, both N. bicalcarata (large peristome cat-egory) and N. rafflesiana (medium peristome category) employperistome aquaplaning as a prey-trapping mechanism (Bohn andFederle, 2004; Bauer et al., 2008).

Climate variables

We considered 19 bioclimatic variables as potential predictorsof Nepenthes group distributions (summarized in Table 3).These variables are freely available at 30 arc-seconds (approx.1 km) resolution from the WorldClim data set Version 1.4,

Release 3 (Hijmans et al., 2005; http://www.worldclim.org/bioclim). This data set comprises gridded precipitation and tem-perature data based on monthly measurements from globally dis-tributed weather stations. The data are averaged from the years1950–2000 to generate a single 50 year ‘climate surface’ foreach month. The resulting surfaces are then utilized to generatebioclimatic variables (e.g. mean temperature of the wettestquarter), which are considered more meaningful biologicallythan monthly measurements for defining the environmental tol-erances of a species (Graham and Hijmans, 2006; Murienneet al., 2009). These variables have been widely used to predictplant species distributions in other habitat modelling studies(e.g. Kumar et al., 2006; Guisan et al., 2007a, b; Murienneet al., 2009; Kumar and Stohlgren, 2009). We did not usemonthly meteorological data due to differences in seasonalweather patterns over the extensive geographical area coveredin the study. For instance, in western Java, the period of highestrainfall occurs between December and February; in contrast,the highest rainfall in northern Thailand occurs during September(source: Meteorological Service of Singapore;http://www.weather.gov.sg). To have used monthly data might have prevented themodel from associating particular climate variables with a highlycorrelated capture group, thus reducing model accuracy.

Climate envelope model development and evaluation

Using the bioclimatic variables discussed above, weemployed Maxent V. 3.3.3k (Phillips et al., 2004) to modelgroup distributions in our geographical area of interest. Thisencompassed Thailand, Laos, Vietnam, Cambodia, PeninsularMalaysia, Sumatra, Java, Borneo, the Philippines and Sulawesi.Maxent is a maximum entropy-based machine learning programthat estimates the probability distribution for the occurrence of aspecies based on constraints of environmental variables onspecies distribution (Phillips, 2006; Phillips et al., 2006; Elithet al., 2011). The current study modelled distributions of capturemechanisms (Table 2) rather than individual species, and gener-ated an estimate of the probability of the presence of a givengroup that varied from 0 to 1, where 0 was the lowest and 1 thehighest probability. The advantage of this modelling approachis that it only requires presence data and environmental variablelayers (continuous or categorical) for the study area (Phillips andDudık, 2008). Moreover, Maxent has been found to produce themost accurate models among many different species distributionmodelling methods (Elith et al., 2006; Hijmans and Graham,2006; Townsend Peterson et al., 2007).

Given that we only considered continuous environmental vari-ables, and the lowest number of samples within a single group

TABLE 2. Putative Nepenthes prey capture mechanism groups modelled in the study

Group Peristome Wax Viscoelastic fluid No. of species No. of populations

Small peristome Small Any Any 41 1098Large peristome Large or medium Any Any 53 881Waxy Any Yes Any 56 1170Waxless Any No Any 20 422Viscoelastic Any Any Yes 33 460Non-viscoelastic Any Any No 56 1505‘Dry’ syndrome Small Yes No 29 1010‘Wet’ syndrome Large or medium No Any 18 416


at Royal R

oads University L



ownloaded from

http://www.worldclim.org/bioclim



http://www.weather.gov.sg




TABLE 3. Results of Maxent models for putative prey capture mechanism groups in Nepenthes

Permutation importance (%) by group

Bioclimatic variable* Large peristome Small peristome Waxy Waxless Viscoelastic Non-viscoelastic ‘Wet’ syndrome ‘Dry’ syndrome

BIO1 (annual mean temperature) 0 4.90 0 0 0.85 0 0 10.72BIO2 (mean diurnal temperature range) 10.66 13.03 9.92 8.11 4.85 8.91 2.51 10.77BIO3 (isothermality)† 8.51 10.37 7.95 4.96 0 7.05 0 7.66BIO4 (temperature seasonality)‡ 0.40 0.23 0.20 0.21 0.81 0.27 0.39 0.40BIO5 (maximum temperature of warmest month) 0 0 0 0 0.01 0 0.90 0BIO6 (minimum temperature of coldest month) 9.10 0.69 1.40 9.46 6.50 5.11 14.37 0.31BIO7 (annual temperature range)§ 16.49 11.72 7.88 12.72 6.45 10.97 0.77 8.24BIO8 (mean temperature of wettest quarter) 16.51 17.12 22.50 9.59 27.73 18.58 10.98 16.27BIO9 (mean temperature of driest quarter) 0 0 0 0 0 0 0 0BIO10 (mean temperature of warmest quarter) 1.31 7.55 10.11 11.10 0.17 13.86 12.95 12.22BIO11 (mean temperature of coldest quarter) 0 0 0 0 0 0 0 0BIO12 (annual precipitation) 9.56 9.37 12.14 5.53 18.32 7.13 9.42 8.52BIO13 (precipitation of wettest month) 1.12 1.09 0.76 1.60 0.88 1.08 2.81 0.62BIO14 (precipitation of driest month) 8.18 2.54 3.49 13.29 4.36 7.10 18.39 5.00BIO15 (precipitation seasonality)} 0.60 1.70 0.94 0.09 5.62 0.05 0.72 0.10BIO16 (precipitation of wettest quarter) 16.43 14.24 16.99 17.04 21.17 15.37 21.60 12.33BIO17 (precipitation of driest quarter) 0.07 3.79 2.51 3.52 0.62 2.21 2.42 3.18BIO18 (precipitation of warmest quarter) 1.06 1.55 3.10 2.75 1.27 2.19 1.77 3.32BIO19 (precipitation of coldest quarter) 0.00 0.11 0.10 0.03 0.38 0.12 0 0.31Test AUC** 0.81 0.77 0.77 0.82 0.86 0.77 0.83 0.80

* Bioclimatic variables are in bold if permutation importance is ≥5 % for any group.† Isothermality is a measure of diurnal temperature range compared with annual temperature range, calculated as: (BIO2/BIO7) × 100.‡ (Standard deviation) × 100.§ (BIO5–BIO6).} Coefficient of variation of precipitation.** AUC: area under curve, a measure of model performance. A value of 0.5 represents a model that performs no better than random; a value of 1 represents a perfect fit.

Mora

net

al.—

Clim

ate

an

dca

ptu

rem

echa

nism

inP

ala

eotro

pica

lp

itcher

pla

nts

Pag

e6

of

13

at Royal Roads University Library on August 26, 2013 http://aob.oxfordjournals.org/ Downloaded from


(n ¼ 378) was almost 5-fold greater than the acceptable minimumof 80 observations (Elith et al., 2011), we were able to use theproduct feature type for our model execution. This feature usesthe product of pairs of continuous environmental variables, to-gether with the linear feature of each variable, to constrain the co-variance of the respective variables to the empirical values, therebyincorporating interactionsbetweenpredictorvariables into thesuit-able habitat projections (Phillips et al., 2006). Additionally, thedefault Maxent model parameters (regularization multiplier¼ 1and convergence threshold¼ 1025) were used. However, to de-crease the chance of the model over- or underpredicting suitablehabitat for each group, we increased the maximum number ofiterations from 500 (the default value) to 5000, to allow themodel a sufficient number of iterations for convergence.

When defining the environmental tolerances of the groupsexamined in this study, it was important to consider potential geo-graphic sampling bias within the occurrence data, given differingdegrees of accessibility between sampling locations. Forexample,a potential source of bias in species distribution modelling basedon presence-only data is the tendency to over-represent observa-tions close to roads and other means of access to a given area.Phillips et al. (2009) suggested a method to correct for samplingbias, by selecting background data with similar sampling bias tothe occurrence data. Following the steps outlined in Phillips(2006) and Young et al. (2011), a bias file was generated usingthe smallest administrative unit (recorded as district, county ormunicipality) for all countries in our study area. The geographicspace of an administrative unit was included in the bias file if anobservation fell within the unit. By including the bias file,Maxent sampled the default 10 000pseudo-absences fromthe col-lection of small-scale geographic units where observations wererecorded, rather than from the entire prediction area.

Finally, for model evaluation, the random test percentage wasset at 30, so that 70 % of the observations for a given group wererandomly selected as a training data set and the remaining 30 %were used for model testing. Model fit was evaluated by thewidely used area under the curve (AUC) of the receiver operatingcharacteristic (ROC), which represents the probability that themodel classifier will correctly identify a randomly chosen truepresence (Fielding and Bell, 1997; Fawcett, 2006). The AUCis a threshold-independent measure of model performance thatranges from 0 to 1; avalue of 0.5 represents a model that performsno better than random, whilst 1 is maximally predictive. Toevaluate model performance further, the replicates option wasset at 10 so that the selection of testing data and calculation ofAUC values were repeated. Therefore, final model performancewas evaluated with the mean AUC for each group over the tenreplicates. Similarly suitable habitat (in terms of the climate en-velope) for each group was displayed as the average projectedprobability of presence across the ten replicates. Although theMaxent parameter settings listed above were constant for allgroups, each was modelled and evaluated individually.Graphical outputs of the model for each group were producedusing the ArcMap V.10 software package (Esri Inc., Redlands,CA, USA).

Evaluation of bioclimatic variables as predictors of occurrence

Maxent provides several methods for quantifying the rela-tive contribution of each variable to the model. We used

‘permutation importance’ to investigate which bioclimaticvariables were most important for predicting geographical dis-tribution of a given group. This is a measure of decrease in train-ing AUC (expressed as a percentage) resulting from randomlypermuting values of a given variable during the training phaseof model development. The decrease in training AUC is in pro-portion to the importance of that variable to the model (Phillips,2006). In addition, we compared the relative magnitude of im-portant variables between group pairs (large peristome vs.small peristome, waxy vs. waxless, viscoelastic vs. non-viscoelastic, and ‘wet’ syndrome vs. dry’ syndrome), to allowidentification of variables that might be important in delimitingsuitable habitat for a given group. For each bioclimatic vari-able, permutation importance values were summed for eachpair of groups. The relative contribution from each memberof the pair to the total was then calculated as a percentage, asfollows:

% importance of bioclimatic variable to group A

= [PIA/(PIA + PIB)] × 100

Where PIA and PIB denote permutation importance of agiven bioclimatic variable to groups A and B, respectively.We considered variables with permutation importance values≥5 % as important; the comparisons included only those vari-ables with values at or above this threshold for at least one groupin each pair.

RESULTS

Model evaluation

Model evaluation statistics for group predictions are presented inTable 3. Mean AUC values were consistently high, ranging from0.77 to 0.86. A high AUC value indicates a high level of modelaccuracy: a value of 0.75 means that 75 % of the time arandom sample from presence predictions will have a scoregreater than a random selection from absence predictionsacross all available probability thresholds. Therefore, an AUCvalue of 0.5 indicates a random predictor, and values between0.5 and 0.6 are generally considered indicative of a failedmodel (Fawcett, 2006). Overall, groups with capture mechan-isms predicted to be associated with wet environments (largeperistome, waxless, viscoelastic and ‘wet’ syndrome) haveslightly higher AUC values than groups predicted to occur indrier or more seasonal environments (small peristome, waxy,non-viscoelastic and ‘dry’ syndrome; Table 3).

For each group, graphical representations of modelled habitatsuitability are presented in Fig. 1A–H. Probability of habitatsuitability is plotted on a colour scale, where grey representsvalues of 0–0.1 and dark blue represents values of 0.9–1. Ingeneral, the groups putatively associated with perhumid condi-tions (large peristome, waxless, viscoelastic and ‘wet’ syn-drome; Fig. 1B, D, F, H) show less extensive areas of probablehabitat suitability than do those associated putatively with lesshumid conditions (small peristome, waxy, non-viscoelastic and‘dry’ syndrome; Fig. 1A, C, E, G). The most restricted area ofsuitable habitat is associated with the viscoelastic group(Fig. 1F).


at Royal R

oads University L



ownloaded from


0–0·1N

N

0·1–0·20·2–0·30·3–0·40·4–0·50·5–0·60·6–0·70·7–0·80·8–0·90·9–1

Probability ofhabitat suitabilityA Small peristome B Large peristome C Waxy D Waxless

E Non-viscoelastic F Viscoelastic G Dry syndrome H Wet syndrome

FI G. 1. Probability of habitat suitability foreight putative Nepenthes prey capture mechanism groups, modelled from 19 bioclimatic variables using Maxent. Probability is mapped by colour, from grey (0–0.1) todark blue (0.9–1). Scale bar ¼ 1000 km.

Mora

net

al.—

Clim

ate

an

dca

ptu

rem

echa

nism

inP

ala

eotro

pica

lp

itcher

pla

nts

Pag

e8

of

13

at Royal Roads University Library on August 26, 2013 http://aob.oxfordjournals.org/ Downloaded from


Bioclimatic variable importance

Of the 19 bioclimatic variables used in the models, two pro-duced permutation importance values of zero for all models, i.e.they contributed no useful information to account for the observeddistributions: mean temperature of the driest quarter (BIO9) andmean temperature of thecoldest quarter (BIO11;Table 3).Anadd-itional six bioclimatic variables produced values of ,5 % for allgroups, essentially contributing no useful information to themodels: temperature seasonality (BIO4), maximum temperatureof the warmest month (BIO5), precipitation of the wettest month(BIO13), precipitation of the driest quarter (BIO17), precipitationof the warmest quarter (BIO18) and precipitation of the coldestquarter (BIO19; Table 3). Conversely, three variables were im-portant in informing the model across all groups: mean tempera-ture of the wettest quarter (BIO8), annual precipitation (BIO12)and precipitation of the wettest quarter (BIO16), which producedhigh permutation importance values (≥5 %), ranging from 5.5 %to as high as 27.7 %. The remaining eight bioclimatic variablesproduced values that varied across groups and, therefore,provide insight into the differential influence of climatic factorsbetween them (see below).

Comparisons of bioclimatic variable importance for habitatsuitability between groups

There are discernible patterns of differences in permutationimportance for habitat suitability between the groups putativelyassociated with perhumid conditions and those putatively asso-ciated with drier environments (Fig. 2). Compared with thesmall peristome group, the large peristome group is influencedto a higher degree by minimum temperature of the coldestmonth (BIO6), precipitation of the driest month (BIO14) andannual temperature range (BIO7). Conversely, the small peri-stome group is more strongly influenced by the mean temperatureof the warmest quarter (BIO10; Fig. 2A). The presence of wax ishighly influenced by mean temperature of the wettest quarter(BIO8), annual precipitation (BIO12) and isothermality(BIO3) compared with the absence of wax. Conversely, thewaxless group is influenced more by annual temperature range(BIO7), precipitation of the driest month (BIO14) andminimum temperature of the coldest month (BIO6; Fig. 2B).Habitat suitability for the viscoelastic group is highly influencedby seasonality of precipitation (BIO15), annual precipitation(BIO12), mean temperature of the wettest quarter (BIO8), pre-cipitation of the wettest quarter (BIO16) and minimum tempera-ture of the coldest month (BIO6) compared with thenon-viscoelastic group. Conversely, the latter is more highlyinfluenced by isothermality (BIO3), mean temperature of thewettest quarter (BIO10), mean diurnal temperature range(BIO2), annual temperature range (BIO7) and precipitation ofthe driest month (BIO14; Fig. 2C). Comparing the ‘wet’ and‘dry’ syndromes, the former is influenced more by theminimum temperature of the coldest month (BIO6), precipita-tion of the driest month (BIO14) and precipitation of thewettest quarter (BIO16). Conversely, annual mean temperature(BIO1), isothermality (BIO3), annual temperature range(BIO2), mean diurnal temperature range (BIO2) and mean tem-perature of the wettest quarter (BIO8) are more important for the‘dry’ syndrome (Fig. 2D).

DISCUSSION

With a geographical range extending from Madagascar to NewCaledonia, Nepenthaceae is a successful and diverse carnivorousfamily. Nepenthes can be found from sea level to .3000 m

Min temp. coldest month (BIO6)A

B

C

D

Min temp. coldest month (BIO6)

Precip. seasonality (BIO15)



Precip. driest month (BIO14)




Annual temp. range (BIO7)




Precip. wettest quarter (BIO16)




Annual precip. (BIO12)




Mean temp. wettest quarter (BIO8)




Isothermality (BIO3)




Mean diurnal temp. range (BIO2)




Mean temp. warmest quarter (BIO10)



Annual mean temp. (BIO1)


100 50 0 50 100% Small peristome% Large peristome

100 50 0 50 100% Wax absent% Wax present

100 50 0 50 100% Non-viscoelastic fluid% Viscoelastic fluid

100 50 0 50 100% Dry syndrome% Wet syndrome

FI G. 2. Comparison of relative magnitude of permutation importance (%) ofbioclimatic variables to the Maxent model between putative Nepenthes preycapture mechanism groups. Only those variables with permutation importancevalues ≥5 % for at least one group in each pair were evaluated. (A) Large peri-stome vs. small peristome. (B) Wax present vs. wax absent. (C) Viscoelastic

fluid vs. non-viscoelastic fluid. (D) ‘Wet’ syndrome vs. ‘dry’ syndrome.


at Royal R

oads University L



ownloaded from


elevation, on a variety of substrates (Clarke, 1997). Nepenthesoccur in both perhumid and seasonal tropical environments,and the range of pitcher morphologies is striking, even to thecasual observer. This diversityof pitcher morphology is mirroredby the range of nitrogen sources exploited, including bat excreta,leaf litter and arthropods (e.g. Moran et al., 2003; Di Giusto et al.,2008; Grafe et al., 2011; Pavlovic et al., 2011). Yet, despite thiswide range of pitcher structure and function, we detected a con-sistent, large-scale pattern defining the relationship between bio-climatic variables and prey capture mechanism.

Peristome

Bohn and Federle (2004) and Bauer et al. (2008) demonstratedthe profound reliance of peristome function on ambient moisturein Nepenthes species that rely on this structure as the primarytrapping mechanism. The current study confirms that Nepenthesspecies employing peristome-based prey capture are generallyrestricted toperhumid regions.AsFig.1 demonstrates,habitat suit-able for the large peristome group is considerably less extensivethanthatforthesmallperistomegroup.Itisconfinedlargelytoperhu-mid regions, notably the eastern Philippines, western Borneo,western Sumatra and eastern Peninsular Malaysia. In contrast,habitat suitable for the small peristome group is much less geo-graphically constrained, and extends into regions with more sea-sonal climates, both north and south of the Equator.

The relative importance of individual bioclimatic factors tothe Maxent models also provides clues to climatic constraintson peristome use. Compared with the small peristome group,habitat suitable for the large peristome group is influenced moreby minimum temperature of the coldest month, annual tempera-ture range and precipitation of the driest month. The first twovariables could be considered pertinent with regards to montaneclimates (a large proportion of species in the large peristomegroup are montane/sub-montane in habit); the last variable sug-gests a constraint with regards to periods of low moisture avail-ability, further supporting the idea that peristome-based captureis centred on perhumid regions.

Wax

It is evident from Fig. 2A and B that there is a reciprocal patternof relative importance of bioclimatic variables between peristome-based capture and wax-based capture. The three bioclimatic vari-ables identified above as important in the distribution of the largeperistome group (minimum temperature of the coldest month,annual temperature range and precipitation of the driest month)arealso the three most influential in the distribution of thewaxlessgroup. Similarly, the geographical range of suitable habitat ismore restricted for the waxless group than for the waxy group.The maps for the small peristome and waxy groups are almostidentical, as are those for the large peristome and waxless groups.Given the previously noted inverse relationship between waxproduction and peristome size (Bauer et al., 2012a; Benz et al.,2012), this finding is perhaps not surprising. The graphicaloutputs from the models also confirm the prediction that speciesusing wax as a capture mechanism would be less confined to per-humid conditions than those relying on a peristome-based strat-egy. From a functional viewpoint, this also makes intrinsic sense,

since the wax-based capture mechanism does not require criticallevels of ambient moisture, in contrast to peristome-based capture.

Viscoelastic fluid

Of the eight habitat models derived here, that for the viscoelas-tic group appears to be the most tightly constrained by climate, asevidenced by the highly limited geographical extent of suitablehabitat. Bonhomme et al. (2011) noted the tendency for specieswith viscoelastic pitcher fluid to be montane in habit, and this iscorroborated by the distribution of suitable habitat in the currentstudy. Theyalso suggested that the abundance of volant prey rela-tive to crawling prey in tropical montane environments favoursthe deployment of viscoelastic fluid rather than waxes. Their as-sertion was based on experimental testing of the effects of waxand viscoelastic fluid on insect traction. They demonstratedthat wax is a more effective mechanism for preventing escapeof ants, whereas viscoelastic fluid is effective for both ants andflies. Further, the authors determined that an inverse relationship(similar to that between peristome and wax) exists between de-ployment of viscoelastic fluid and deployment of wax. This rela-tionship would account for the almost identical areas of suitablehabitat modelled for the waxy and non-viscoelastic groups. It hasbeen shown that even when highly diluted with water (as mightbe the case with rainwater in montane and other perhumid envir-onments), Nepenthes viscoelastic fluid retains its prey-retentiveproperties (Gaume and Forterre, 2007; Bonhomme et al., 2011).Thus, we might expect the production of viscoelastic fluid to beassociated with wetter habitats, and the findings of the currentstudy appear to support this idea. The results of the paired groupcomparison show that three of the five important bioclimaticvariables for the viscoelastic group pertain to precipitation. Incontrast, only one of the five important variables for the non-viscoelastic group is based on precipitation; the rest are alltemperature related.

‘Wet’ and ‘dry’ syndromes

Given that the putative ‘wet’ and ‘dry’ syndromes are based oncombinations of the individual prey capture mechanisms dealtwith above, it is perhaps not surprising that the modelledhabitat suitability patterns closely match them, with the formerbeing restricted to perhumid regions and the latter extendinginto more seasonal regions. Of the three bioclimatic variablesshown to be important to the ‘wet’ syndrome model in thepaired comparison, two are precipitation related; in contrast, allfive of the variables shown to be important to the ‘dry’ syndromemodel are temperature related.

No evidence for edaphic influences on group distributions

Climate is not the only determinant of the geographical distri-butions of organisms. For plants, edaphic factors can play a crit-ical role in determining where a given species is likely to occur.As with other carnivorous plant taxa, Nepenthes are restricted tonutrient-poor substrates (Givnish et al., 1984). There is a varietyof such substrates available across the geographical range ofNepenthes, and many species are restricted to one or more ofthem. For example, N. rafflesiana is typically found growingon silica-derived podzols, N. bicalcarata occurs predominantly


at Royal R

oads University L



ownloaded from


on peat-derived soils, Nepenthes faizaliana is restricted to lime-stone outcrops, N. rajah grows on serpentine soils derived fromultramafic bedrock, and Nepenthes sumatrana occurs on sand-stone outcrops (Clarke, 1997, 2001). However, although soiltype may play an important role in the distribution of individualNepenthes species, it does not appear to be a factor in the distri-bution of capture mechanisms examined in the current study.Evidence for this is found in the five species mentioned above:although each is restricted to a particular soil type, all aremembers of the large peristome group. Similar examples canbe found for the other groups. Therefore, we suggest that al-though the distribution of Nepenthes as a genus (as well as indi-vidual species, as highlighted above) may be constrained byclimate and soil type, the geographical distribution of the threemajor capture mechanisms can be explained largely by climaticvariables alone. Nepenthes species employing peristome-basedor viscoelastic mechanisms are more restricted to perhumidareas (including montane environments); in contrast, thoseemploying wax-based capture are less constrained by moistureregime, and can successfully inhabit more seasonal environ-ments.

Reasons for maintenance of multiple prey capture mechanismsin Nepenthaceae

The current study has established that the wax-based capturestrategy in Nepenthes facilitates carnivory in both perhumidand more seasonal environments. If the wax-based strategy is ef-fective over such a large geographical range, why are the othermechanisms (large peristome and viscoelastic fluid) maintainedin Nepenthes at all? There is a consensus that the ‘ancestral’ stateof Nepenthes pitchers is represented by a wax-based capturestrategy, and that this strategy has been lost independently inseveral clades (Meimberg and Heubl, 2006; Gaume and DiGiusto, 2009; Bauer et al., 2012a). What might be the evolution-ary drivers of wax loss and the consequent adoption of alternativestrategies? One possibility is differential effectiveness of thetrapping mechanisms on different prey taxa: Bonhomme et al.(2011) demonstrated that viscoelastic fluid was more effectivethan wax in retaining volant prey. They also suggested thatvolant prey are more prevalent than crawling prey in montanehabitats, thus favouring deployment of viscoelastic fluid.However, a discussion of possible prey specialization withregards to bioclimatic variables is beyond the scope of thecurrent study.

Another possibility is that there exists a cost differentialbetween wax-based and alternative strategies. Regardless oftrapping mechanism, the production of pitchers incurs a cost tothe plant. There is a quantifiable ‘opportunity cost’ associatedwith producing a foliar structure that delivers sub-optimal photo-synthetic performance. For example, pitchers of Nepenthestalangensis exhibit significantly reduced rates of photosynthesis(AN) and quantum yield of photosystem II (FPSII) comparedwith laminae (Pavlovic et al., 2009). Additionally, some compo-nents of the pitcher are more expensive than others to produce.The construction costs of a plant organ such as a leaf can bedefined as the amount of photosynthate required to produce aunit weight of biomass (Poorter and De Jong, 1999; Villaret al., 2006). Simple compounds such as cellulose are relativelyinexpensive to produce; however, synthesis of secondary

compounds (e.g. lignin and tannins) incurs higher constructioncosts (Poorter and De Jong, 1999; Villar et al., 2006; Osunkoyaet al., 2007, 2008). The peristome is a rigid, structurally rein-forced region of the pitcher, and this rigidity is due to the incorp-oration of these more expensive secondary compounds. Theviscoelasticity of pitcher fluid results from the presence of long-chain polymers (Bonhomme et al., 2011) the relative construc-tion costs of which have yet to be determined. The waxes pro-duced by Nepenthes pitchers are rich in primary alcohols andvery long chain aldehydes (Riedel et al., 2003, 2007). Waxesincur significantly higher construction costs than secondarycompounds such as tannins and lignin (Poorter and De Jong,1999; Martınez et al., 2002). The inner walls ofN. albomarginata pitchers produce .35 mg cm22 of wax(Riedel et al., 2007), which would, other factors being equal,incur considerably higher construction costs than pitchers ofwaxless congeners. Therefore, we suggest that the reason thatwax has been lost independently from several clades over timeis probably due to the costs of production outweighing the bene-fits gained by its deployment in a particular environment. If theless resource-costly peristome is as effective as wax in perhumidregions, it would be to a plant’s advantage to reduce or cease ex-pensive wax production to maximize net benefit. This idea is inagreement with the findings of Bonhomme et al. (2011), Benzet al. (2012) and Bauer et al. (2012a): in Nepenthes, there is astrong reciprocal relationship between the use of wax and the de-ployment of alternative capture mechanisms. It is also in agree-ment with the findings of the current study, with regards to thegeographical distributions of wax-based vs. alternative preycapture mechanisms. To conclude, we suggest that climate is astrong candidate as the agent of the observed disruptive selectionbetween wax-based and either peristome-based or viscoelasticfluid-based capture syndromes in Nepenthes.

ACKNOWLEDGEMENTS

We thank Royal Roads University fora grant to support the study,B. Hawkins for material help, and A. Moran for reviewing themanuscript. Authorcontributions: J.M., C.C. and L.C. conceivedthe study; C.C. collected the data; L.G. and J.M. analysed thedata; J.M. led the writing with contributions from L.G., C.C.and L.C. L.G. constructed the figures. Two anonymous reviewershelped to improve the manuscript considerably.

LITERATURE CITED

Bauer U, Bohn HF, Federle W. 2008. Harmless nectar source or deadly trap:Nepenthes pitchers are activated by rain, condensation and nectar.Proceedings of the Royal Society B: Biological Sciences 275: 259–265.

Bauer U, Willmes C, Federle W. 2009. Effect of pitcher age on trapping effi-ciency and natural prey capture in carnivorous Nepenthes rafflesianaplants. Annals of Botany 103: 1219–1226.

Bauer U, Grafe TU, Federle W. 2011. Evidence for alternative trapping strat-egies in two forms of the pitcher plant, Nepenthes rafflesiana. Journal ofExperimental Botany 62: 3683–3692.

Bauer U, Clemente CJ, Renner T, Federle W. 2012a. Form follows function:morphological diversification and alternative trapping strategies in carniv-orous Nepenthes pitcher plants. Journal of Evolutionary Biology 25:90–102.

Bauer U, Di Giusto B, Skepper J, Grafe TU, Federle W. 2012b. With a flick ofthe lid: a novel trapping mechanism in Nepenthes gracilis pitcher plants.PLoS One 7: e38951.


at Royal R

oads University L



ownloaded from


Bazile V, Moran JA, Le Moguedec G, Marshall DJ, Gaume L. 2012. A carniv-orous plant fed by its ant symbiont: a unique multi-faceted nutritional mu-tualism. PLoS One 7: e36179.

Benz MJ, Gorb EV, Gorb SN. 2012. Diversity of the slippery zone microstruc-ture in pitchers of nine carnivorous Nepenthes taxa. Arthropod-PlantInteractions 6: 147–158.

Bohn HF, Federle W. 2004. Insect aquaplaning: Nepenthes pitcher plantscapture prey with the peristome, a fully wettable water-lubricated anisotrop-ic surface. Proceedings of the National Academy of Sciences, USA 101:14138–14143.

Bonhomme V, Pelloux-Prayer H, Jousselin E, Forterre Y, Labat J-J, GaumeL. 2011. Slippery or sticky? Functional diversity in the trapping strategy ofNepenthes carnivorous plants. New Phytologist 191: 545–554.

Cheverud JM, Dow MM, Leutenegger W. 1985. The quantitative assessmentof phylogenetic constraints in comparative analyses: sexual dimorphismin body weight among primates. Evolution 39: 1335–1351.

Chin L, Moran JA, Clarke C. 2010. Trap geometry in three giant montanepitcher plant species from Borneo is a function of tree shrew body size.New Phytologist 186: 461–470.

Clarke C. 1997. Nepenthes of Borneo. Kota Kinabalu: Natural HistoryPublications.

Clarke C. 2001. Nepenthes of Sumatra and Peninsular Malaysia. Kota Kinabalu:Natural History Publications.

Clarke CM, Kitching RL. 1995. Swimming ants and pitcher plants: a uniqueant–plant interaction from Borneo. Journal of Tropical Ecology 11:589–602.

Clarke CM, Bauer U, Lee CC, Tuen AA, Rembold K, Moran JA. 2009. Treeshrew lavatories: a novel nitrogen sequestration strategy in a tropical pitcherplant. Biology Letters 5: 632–635.

Danser BH. 1928. The Nepenthaceae of the Netherlands Indies. Bulletin duJardin Botanique Buitenzorg, Serie III 9: 249–438.

Di Giusto B, Grosbois V, Fargeas E, Marshall DJ, Gaume L. 2008.Contribution of pitcher fragrance and fluid viscosity to high prey diversityin a Nepenthes carnivorous plant from Borneo. Journal of Biosciences 33:121–136.

Di Giusto B, Bessiere J-M, Gueroult M, et al. 2010. Flower-scent mimicrymasks a deadly trap in the carnivorous plant Nepenthes rafflesiana.Journal of Ecology 98: 845–856.

Elith J, Graham CH, Anderson RP, et al. 2006. Novel methods improve predic-tion of species’ distributions from occurrence data. Ecography 29:129–151.

Elith J, Phillips SJ, Hastie T, Dudık M, Chee YE, Yates CJ. 2011. A statisticalexplanation of MaxEnt for ecologists. Diversity and Distributions 17:43–57.

Fawcett T. 2006. An introduction to ROC analysis. Pattern Recognition Letters27: 861–874.

Fielding AH, Bell JF. 1997.A review of methods for the assessmentof predictionerrors in conservation presence/absence models. Environmental Conservation24: 38–49.

Gaume L, Di Giusto B. 2009. Adaptive significance and ontogenetic variabilityof the waxy zone in Nepenthes rafflesiana. Annals of Botany 104:1281–1291.

Gaume L, Forterre Y. 2007. A viscoelastic deadly fluid in carnivorous pitcherplants. PLoS One 2: e1185.

Gaume L, Gorb S, Rowe N. 2002. Function of epidermal surfaces in the trappingefficiency of Nepenthes alata pitchers. New Phytologist 156: 479–489.

Gaume L, Perret P, Gorb E, Gorb S, Labat J-J, Rowe N. 2004. How do plantwaxes cause flies to slide? Experimental tests of wax-based trappingmechanisms in three pitfall carnivorous plants. Arthropod Structure andDevelopment 33: 103–111.

Givnish TJ, Burkhardt EL, Happel RE, Weintraub JD. 1984. Carnivory in thebromeliad Brocchinia reducta with a cost/benefit model for the general re-striction of carnivorous plants to sunny, moist, nutrient poor habitats.American Naturalist 124: 479–497.

Gorb EV, Gorb SN. 2006. Physicochemical properties of functional surfaces inpitchers of the carnivorous plant Nepenthes alata Blanco (Nepenthaceae).Plant Biology 8: 841–848.

Gorb EV, Gorb SN. 2011. The effect of surface anisotropy in the slippery zone ofNepenthes alata pitchers on beetle attachment. Beilstein Journal ofNanotechnology 2: 302–310.

Gorb E, Kastner V, Peressadko A, et al. 2004. Structure and properties of theglandular surface in the digestive zone of the pitcher in the carnivorous

plant Nepenthes ventrata and its role in insect trapping and retention.Journal of Experimental Botany 207: 2947–2963.

Grafe TU, Schoner CR, Kerth G, Junaidi A, Schoner MG. 2011. A novel re-source–service mutualism between bats and pitcher plants. Biology Letters7: 436–439.

Graham CH, Hijmans RJ. 2006. A comparison of methods for mapping speciesranges and species richness. Global Ecology and Biogeography 15:578–587.

Guisan A, Graham CH, Elith J, Huettmann F. 2007a. Sensitivity of predictivespecies distribution models to change in grain size. Diversity andDistribution 13: 332–340.

Guisan A, Zimmermann NE, Elith J, Graham CH., Phillips S, Peterson AT.2007b. What matters for predicting the occurrences of trees: techniques, dataor species’ characteristics? Ecological Monographs 77: 615–630.

Hijmans RJ, Graham CH. 2006. The ability of climate envelope models topredict the effect of climate change on species distributions. GlobalChange Biology 12: 2272–2281.

Hijmans RJ, Cameron SE, Parra JL, Jones PG, Jarvis A. 2005. Very highresolution interpolated climate surfaces for global land areas.International Journal of Climatology 25: 1965–1978.

Juniper BE, Burras J. 1962. How pitcher plants trap insects. New Scientist 13:75–77.

Juniper BE, Robins RJ, Joel DM. 1989. The carnivorous plants. London:Academic Press.

Kumar S, Stohlgren TJ. 2009. Maxent modeling for predicting suitable habitatfor threatened and endangered tree Canacomyrica monticola in NewCaledonia. Journal of Ecology and Natural Environment 1: 94–98.

Kumar S, Stohlgren TJ, Chong GW. 2006. Spatial heterogeneity influencesnative and non-native plant species richness. Ecology 87: 3186–3199.

Martınez F, Lazo YO, Fernandez-Galiano RM, Merino JA. 2002. Chemicalcomposition and construction costs for roots of Mediterranean trees, shrubspecies and grassland communities. Plant, Cell and Environment 25:601–608.

McPherson S. 2009.Pitcher plants of the Old World. Poole, UK: Redfern NaturalHistory Productions.

Meimberg H, Heubl G. 2006. Introduction of a nuclear marker for phylogeneticanalysis of Nepenthaceae. Plant Biology 8: 831–840.

Merbach MA, Merbach DJ, MaschwitzU, BoothWE,Fiala B, Zizka G.2002.Mass march of termites into the deadly trap. Nature 415: 37.

Merbach MA, Zizka G, Fiala B, Merbach D, Booth WE, Maschwitz U. 2007.Why a carnivorous plant cooperates with an ant-selective defense againstpitcher-destroying weevils in the myrmecophytic pitcher plant Nepenthesbicalcarata Hook.f. Ecotropica 13: 45–56.

Moran JA. 1996. Pitcher dimorphism, prey composition and the mechanisms ofprey attraction in the pitcher plant Nepenthes rafflesiana in Borneo. Journalof Ecology 84: 515–525.

Moran JA, Moran AJ. 1998. Foliar reflectance and vector analysis reveal nutri-ent stress in prey-deprived pitcher plants (Nepenthes rafflesiana).International Journal of Plant Sciences 159: 996–1001.

Moran JA, Booth WE, Charles JK. 1999. Aspects of pitcher morphology andspectral characteristics of six Bornean Nepenthes pitcher plant species:implications for prey capture. Annals of Botany 83: 521–528.

Moran JA, Merbach MA, Livingston NJ, Clarke CM, Booth WE. 2001.Termite prey specialization in the pitcher plant Nepenthes albomarginata– evidence from stable isotope analysis. Annals of Botany 88: 307–311.

Moran JA, Clarke CM, Hawkins BJ. 2003. From carnivore to detritivore?Isotopic evidence for leaf litter utilization by the tropical pitcher plantNepenthes ampullaria. International Journal of Plant Sciences 164:635–639.

Moran JA, Clarke C, Gowen BE. 2012. The use of light in prey capture by thetropical pitcher plant Nepenthes aristolochioides. Plant Signaling andBehavior 8: 957–960.

Murienne J, Guilbert E, Grandcolas P. 2009. Species’ diversity in the NewCaledonian endemic genera Cephalidiosus and Nobarnus (Insecta:Heteroptera: Tingidae), and approach using phylogeny and species’ distri-bution modelling. Biological Journal of the Linnean Society 97: 177–184.

Osunkoya OO, Daud SD, Di Giusto B, Wimmer FL, Holige TM. 2007.Construction costs and physico-chemical properties of the assimilatoryorgans of Nepenthes species in northern Borneo. Annals of Botany 99:895–906.

Osunkoya OO, Daud SD, Wimmer FL. 2008. Longevity, lignin content andconstruction costs of the assimilatory organs of Nepenthes species. Annalsof Botany 102: 845–853.


at Royal R

oads University L



ownloaded from


Owen TP, Lennon KA. 1999. Structure and development of the pitchers from thecarnivorous plant Nepenthes alata (Nepenthaceae). American Journal ofBotany 86: 1382–1390.

Pant DD, Bhatnagar S. 1977. Morphological studies in Nepenthes(Nepenthaceae). Phytomorphology 27: 13–34.

Pavlovic A. 2012. Adaptive radiation with regard to nutrient sequestration strat-egies in the carnivorous plants of the genus Nepenthes. Plant Signaling andBehavior 7: 295–297.

Pavlovic A, Masarovicova E, Hudak J. 2007. Carnivorous syndrome in Asianpitcher plants of the genus Nepenthes. Annals of Botany 100: 527–536.

Pavlovic A, Singerova L, Demko V, Hudak J. 2009. Feeding enhances photo-synthetic efficiency in the carnivorous pitcher plant Nepenthes talangensis.Annals of Botany 104: 307–314.

Pavlovic A, Slovakova L, Santrucek J. 2011. Nutritional benefit from leaf litterutilization in the pitcher plant Nepenthes ampullaria. Plant, Cell andEnvironment 34: 1865–1873.

Phillips S. 2006. A brief tutorial on Maxent. AT&T Research. http://www.cs.princeton.edu/~schapire/maxent/tutorial/tutorial.doc.

Phillips SJ, Dudık M. 2008. Modeling of species distributions with Maxent: newextensions and a comprehensive evaluation. Ecography 31: 161–175.

Phillips SJ, Dudık M, Schapire RE. 2004. A maximum entropy approach tospecies distribution modeling. Proceedings of the Twenty-FirstInternational Conference on Machine Learning 655–662.

Phillips SJ, Anderson RP, Schapire RE. 2006. Maximum entropy modeling ofspecies geographic distributions. Ecological Modelling 190: 231–259.

Phillips SJ, Dudık M, Elith J, et al. 2009. Sample selection bias andpresence-only distribution models: implications for background andpseudo-absence data. Ecological Applications 19: 181–197.

Poppinga S, Koch K, Bohn HF, Barthlott W. 2010. Comparative and functionalmorphology of hierarchically structured anti-adhesive surfaces in carnivor-ous plants and kettle trap flowers. Functional Plant Biology 37: 952–961.

Poorter H, De Jong R. 1999. A comparison of specific leaf area, chemical com-position and leaf construction costs of field plants from 15 habitats differingin productivity. New Phytologist 143: 163–176.

Riedel M, Eichner A, Jetter R. 2003. Slippery surfaces of carnivorous plants:composition of epicuticular wax crystals in Nepenthes alata Blanco pitch-ers. Planta 218: 87–97.

Riedel M, Eichner A, Meimberg H, Jetter R. 2007. Chemical composition ofepicuticular wax crystals on the slippery zone in pitchers of five Nepenthesspecies and hybrids. Planta 225: 1517–1534.

Scholz I, Bueckins M, Dolge L, et al. 2010. Slippery surfaces of pitcher plants:Nepenthes wax crystals minimize insect attachment via microscopic surfaceroughness. Journal of Experimental Biology 213: 1115–1125.

Thornham DG, Smith JM, Grafe TU, Federle W. 2012. Setting the trap: clean-ing behaviour of Camponotus schmitzi ants increases long-term capture ef-ficiency of their pitcher plant host, Nepenthes bicalcarata. FunctionalEcology 26: 11–19.

Thornhill AH, Harper IS, Hallam ND. 2008. The development of the digestiveglands and enzymes in the pitchers of three Nepenthes species: N. alata,N. tobaica and N. ventricosa (Nepenthaceae). International Journal ofPlant Sciences 169: 615–624.

Townsend Peterson A, Papes M, Eaton M. 2007. Transferability and modelevaluation in ecological niche modeling: a comparison of GARP andMaxent. Ecography 30: 550–560.

Villar R, Ruiz-Robleto J, De Jong Y, Poorter H. 2006. Differences in construc-tion costs and chemical composition between deciduous and evergreenwoody species are small as compared to differences among families.Plant, Cell and Environment 29: 1629–1643.

Wang L, Zhou Q, Zheng Y, Xu S. 2009. Composite structure and properties ofthe pitcher surface of the carnivorous plant Nepenthes and its influence onthe insect attachment system. Progress in Natural Science 19: 1657–1664.

Wells K, Lakim MB, Schulz S, Ayasse M. 2011. Pitchers of Nepenthes rajahcollect faecal droppings from both diurnal and nocturnal small mammalsand emit fruity odour. Journal of Tropical Ecology 27: 347–353.

Young N, Carter L, Evangelista P. 2011. A MaxEnt Model v3.3.3e Tutorial(ArcGIS v10). http://ibis.colostate.edu/WebContent/WS/ColoradoView/TutorialsDownloads/A_Maxent_Model_v7.pdf.


at Royal R

oads University L



ownloaded from

http://www.cs.princeton.edu/~schapire/maxent/tutorial/tutorial.doc



http://ibis.colostate.edu/WebContent/WS/ColoradoView/TutorialsDownloads/A_Maxent_Model_v7.pdf




Capture mechanism in Palaeotropical pitcher plants ...Capture mechanism in Palaeotropical pitcher...

Documents

Transcript of Capture mechanism in Palaeotropical pitcher plants ...Capture mechanism in Palaeotropical pitcher...