[1801 2362014 Sysbio-syu035tex] Page 1 1ndash14

Syst Biol 0(0)1ndash14 2014copy The Author(s) 2014 Published by Oxford University Press on behalf of the Society of Systematic Biologists All rights reservedFor Permissions please email journalspermissionsoupcomDOI101093sysbiosyu035

Insights on the Evolution of Plant Succulence from a Remarkable Radiation in Madagascar(Euphorbia)

MARGARET EVANS123lowast XAVIER AUBRIOT34 DAVID HEARN5 MAXIME LANCIAUX3 SEBASTIEN LAVERGNE6CORINNE CRUAUD7 PORTER P LOWRY II38 and THOMAS HAEVERMANS3

1 Laboratory of Tree-Ring Research 1215 E Lowell Street and 2Department of Ecology and Evolutionary Biology 1041 E Lowell Street University of ArizonaTucson AZ 85721 USA 3ISYEB Institut de Systeacutematique Eacutevolution Biodiversiteacute (UMR 7205 CNRS MNHN EPHE UPMC) Museacuteum national

drsquohistoire naturelle National Herbarium CP 39 57 rue Cuvier 75231 Paris Cedex 05 France 4Department of Life Sciences Natural History MuseumCromwell Road London SW7 5BD UK 5Department of Biological Sciences Towson University 8000 York Rd Baltimore MD 21252 USA 6Laboratoire

drsquoEcologie Alpine UMR CNRS 5553 Universiteacute Joseph Fourier Grenoble I BP 53 38041 Grenoble Cedex 9 France 7Genoscope Centre National deSequenccedilage CP5706 2 rue Gaston Creacutemieux 91057 Evry Cedex France and 8Missouri Botanical Garden PO Box 299 St Louis MO 63166 USA

lowastCorrespondence to be sent to Laboratory of Tree-Ring Research 1215 E Lowell Street University of Arizona Tucson AZ 85721 USAE-mail mekevansuarizonaedu

Margaret Evans and Xavier Aubriot are co-first authors of this article

Received 9 May 2013 reviews returned 3 September 2013 accepted 9 May 2014Associate Editor Luke Harmon

AbstractmdashPatterns of adaptation in response to environmental variation are central to our understanding of biodiversity butpredictions of how and when broad-scale environmental conditions such as climate affect organismal form and functionremain incomplete Succulent plants have evolved in response to arid conditions repeatedly with various plant organssuch as leaves stems and roots physically modified to increase water storage Here we investigate the role played byclimate conditions in shaping the evolution of succulent forms in a plant clade endemic to Madagascar and the surroundingislands part of the hyper-diverse genus Euphorbia (Euphorbiaceae) We used multivariate ordination of 19 climate variablesto identify links between particular climate variables and three major forms of succulencemdashsucculent leaves cactiform stemsucculence and tubers We then tested the relationship between climatic conditions and succulence using comparativemethods that account for shared evolutionary history We confirm that plant water storage is associated with the twocomponents of aridity temperature and precipitation Cactiform stem succulence however is not prevalent in the driestenvironments countering the widely held view of cactiforms as desert icons Instead leaf succulence and tubers aresignificantly associated with the lowest levels of precipitation Our findings provide a clear link between broad-scale climaticconditions and adaptation in land plants and new insights into the climatic conditions favoring different forms of succulenceThis evidence for adaptation to climate raises concern over the evolutionary future of succulent plants as they along withother organisms face anthropogenic climate change [Adaptation climate comparative analysis Euphorbia ordinationphylogeny]

Succulent plants have such diverse downright bizarregrowth forms that they make an irresistible targetfor evolutionary studies of the relationship betweenform and function between organism and environment(Nobel 1988 Edwards and Donoghue 2006 Hearn 2006Ogburn and Edwards 2009) Succulence is thought tobe an adaptation to arid conditions (Futuyma 1997Niklas 1997 Arakaki et al 2011) Indeed the radiation ofsucculent plants in arid regions on two continents in twodistantly related families cacti (Cactaceae) in the NewWorld and spurges (Euphorbiaceae) in the Old Worldis used as a textbook example of convergent adaptiveevolution (Raven et al 1986 Futuyma 1997 Niklas 1997Stearns and Hoekstra 2005) Surprisingly then thereare very few comparative studies of the relationshipbetween succulence and climate (Hearn 2004 2013) Inparticular there are no studies of much less predictionabout how climate influence the tremendous varietyof forms of succulence among land plants rangingfrom leaf succulents such as agaves (Agave) and livingstones (Lithops) to stem succulents such as cacti andbottle trees (Adansonia and Pachypodium) to root orroot-like succulents (see Fig 1ce) Arguably the mostinteresting groups of succulent plants are those wherewater storage occurs in radically different parts of the

plant body among close relatives (Fig 1 Hearn et al2013) suggesting evolutionary lability and begging thequestion of whether there is adaptive value in suchstriking changes in morphology

Here we examine the relationship between climateand different forms of succulence in a subclade of thegiant genus Euphorbia (comprising sections GoniostemaDenisophorbia and Deuterocalli hereafter EuphorbiaGDD) that is fantastically varied in form (Fig 1) andendemic to Madagascar and surrounding islands TheEuphorbia GDD clade thus carries another kind ofsignificance it is representative of the endemism thatplaces Madagascar in the top tier of global biodiversityhotspots (Myers et al 2000 Ganzhorn et al 2001Mittermeier et al 2005 Phillipson et al 2006) We focuson three kinds of succulence (i) leaf succulence (ii)cactiform stem succulence and (iii) below-ground water-storing organs either a caudex (a swollen perennialstem at or near-ground level) or tuber (a more generalterm referring to a storage organ derived from stem orroot tissue) Cactiform stem succulents are considerediconic of deserts but in fact they are not prevalent inplaces subject to very long or unpredictable droughtInstead they store water above ground fully exposedto the evaporative powers of the sun and wind making

1

Systematic Biology Advance Access published June 27 2014 at M

useum N

ational dHistoire N

aturelle on July 1 2014httpsysbiooxfordjournalsorg

Dow

nloaded from

[1801 2362014 Sysbio-syu035tex] Page 2 1ndash14

2 SYSTEMATIC BIOLOGY

FIGURE 1 Growth form diversity in the Euphorbia clade GDD (sections Goniostema Denisophorbia and Deuterocalli) a) E aff pyrifoliaa non-succulent tree b) E capmanambatoensis a cactiform c) E cylindrifolia a dwarf chamaephyte with highly succulent leaves (left leaf) d)E mahafalensis a shrub e) E primulifolia var primulifolia a true geophyte with non-succulent leaves (right leaf) and f) E alluaudii a coraliformtree These represent mature specimens with the exception of E alluaudii which reaches a maximum height of 10 m Drawings by A Haevermans

them relatively vulnerable to water loss they aremost prevalent in areas with seasonal drought buta reliable season of precipitation (Gibson and Nobel1986 Burgess and Shmida 1988 Burgess 1995 Eggli andNyffeler 2009 Ogburn and Edwards 2010) In contrastwe suggest that in very dry climates water storageshould occur below ground or near the soil surfacewhere evaporative potential is lower We argue thatbelow-ground succulents are more narrowly adaptedto extremely dry environments than cactiform speciessince tubers are highly susceptible to rot a fact wellknown to succulent enthusiasts and Irish-Americanimmigrants alike (Hearn 2004) We also expect below-ground or near-ground forms of succulence to beassociated with cooler temperatures both because riskof frost damage selects for these forms in cold climatesand ground-hugging growth forms put photosyntheticstructures in a warmer microenvironment (near the soilsurface) where they can be more efficient Cactiformstem succulents by contrast are known to be limitedby cold conditions (Shreve 1911 Steenbergh and Lowe1977 Nobel 1980 Gibson and Nobel 1986 Piersonand Turner 1998 Godinez-Aacutelvarez et al 2003 Ogburnand Edwards 2010) Thus we propose that thereare two fundamental categories of succulencemdashthosepositioning water-storing tissues below ground or near

the surface versus above groundmdashassociated with verydry and relatively cool conditions versus moderately dryand warm conditions respectively

We addressed these predictions about climate anddifferent forms of succulence in the Euphorbia GDD cladeusing a two-step analysis The first is an exploratoryphase Wetdry and coldwarm gradients can bedescribed with a variety of summary statistics (meanmaximum or minimum temperature and precipitation)at various time scales Correlations between thesevariables are often high making it inappropriate toinclude them in a single test (multicollinearity) yetmultiple tests raise the risk of type 1 error A key objectiveof the exploratory stage is thus to identify a suitablesubset of climate variables for statistical testing Towardthis we used multivariate ordination techniques togenerate a reduced-dimension climate space in which weexamined the distribution of succulence variables Thisidentified mean annual temperature (MAT) and meanannual precipitation (MAP) as suitable climate variablesfor statistical analysis Second we tested the relationshipbetween succulence variables and MAT or MAP takinginto account the pseudoreplication that arises due tothe relatedness of species (Felsenstein 1985) The resultsyield new insight into the climatic conditions favoringthe evolution of different types of succulence in plants

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 3 1ndash14

2014 EVANS ET ALmdashEVOLUTION OF SUCCULENCE 3

MATERIALS AND METHODS

PhylogenyEuphorbia GDD forms a well-supported and

morphologically diversified clade of at least 90species (123 taxa including infraspecific entities) withsection Goniostema accounting for the majority of theclade (sim75 species Haevermans et al 2009) Speciesdiscovery circumscription and description are notcomplete in this group but recent progress providesa robust phylogenetic framework for analysis of traitevolution (Steinmann 2001 Steinmann and Porter2002 Haevermans et al 2004 Zimmerman et al 2010Aubriot 2012 Horn et al 2012 Dorsey et al 2013) Ourphylogenetic reconstruction involved a sample of 279individuals corresponding to 82 species of which 30samples represented outgroup taxa for the GDD cladeselected from four recently established subgenera ofEuphorbia (Horn et al 2012) E sections GoniostemaDenisophorbia and Deuterocalli were representedrespectively by 203 34 and 12 samples correspondingto approximately 63 8 and 2 species representing90 of the estimated diversity in the Euphorbia GDDclade (Supplementary Fig S1 and Table S1 availablefrom httpdoiorg105061dryadvq6mp) DNAwas sequenced for six chloroplast markers (atpI-atpHpsbA-trnH ndhA matK trnQ-5rsquorps16 and rbcL) andtwo nuclear regions (ITS and ETS) resulting in a datamatrix of 8507 bp (lt1 missing data) Our approachwas to use all of the available data (sequences) inorder to verify operational taxonomic units (OTUs)and estimate phylogeny then trim the trees to retaina single sample per ingroup OTU (ie removingduplicates) retaining all samples that were associatedwith both an unambiguous taxonomic circumscriptionand reliable locality data (Supplementary Fig S1httpdoiorg105061dryadvq6mp) Rerunningthe phylogenetic analysis with one sample per finalretained OTU (67 taxa) yields essentially the sametopology Sequence matrices were aligned usingMUSCLE (version 38 Edgar 2004) and MAFFT (version6 Katoh et al 2002) the aligned data are availableat doi105061dryadvq6mp The best-fit model ofnucleotide substitution evolution for each regionwas determined using Akaikersquos information criterionimplemented in MrModelTest 22 (Nylander 2004)GTR++I for ITS ETS atpI-atpH GTR+ for psbA-trnH ndhA matK trnQ-5rsquorps16 and GTR+I for rbcLPhylogenetic analysis of the nuclear and chloroplastmarkers separately revealed topological incongruencebetween them chloroplast markers which are lackingin sequence variation do not recover as monophyleticE sections Goniostema and Denisophorbia The evidencesuggests this incongruence is real with unknownpotentially complex causes (eg incomplete lineagesorting andor hybridization) We made the decisionto concatenate all regions into a combined data setbecause this offers a consensus tree most consistent withstrong morphological data supporting monophyleticE sections Goniostema and Denisophorbiamdashspecies of the

former are united by strictly bisexual cyathia uniquemodifications of stipules into spines prickles or comb-like enations and verrucose seeds whereas species ofthe latter are united by uni- or bisexual cyathia a uniquehabit with chandelier-like branching and plagiotropicbranches (as in Fig 1a) and smooth unornamentedseeds Analyses based on the other most extensivesampling efforts to date (though substantially less thanhere Haevermans et al 2004 Zimmerman et al 2010Dorsey et al 2013) come to the same conclusionmdashthatis that E sections Goniostema and Denisophorbia aremonophyletic Topology reconstruction and relativedivergence times were estimated simultaneously usingBEAST v172 (Drummond et al 2012) with the generegions partitioned according to the best-fit models ofevolution and a Yule speciation tree prior (Drummondet al 2007) Note that our analyses do not dependupon the absolute time scale of the Euphorbia GDDradiation instead dating served to produce credibleultrametric trees for visualizing trait evolution andphylogenetically corrected statistical tests of traitndashclimate relationships Because the fossil record forEuphorbia is poor we relied upon two dates estimatedby Bruyns et al 2011 as temporal constraints the ages ofsubgenus Euphorbia (mean = 2994 Myr SD = 453 Myr95 HPD = 2249ndash3956) and our ingroup EuphorbiaGDD (mean = 1265 Myr SD = 30 Myr 95 HPD =7099ndash1962) Uncertainty regarding these dates wasincorporated by assigning normal prior distributionsto these two calibration points (Ho 2007 Couvreuret al 2008 Bergh and Linder 2009 Su and Saunders2009) Substitution models rate heterogeneity and basefrequencies were unlinked across partitions Divergencetimes were estimated under a relaxed uncorrelatedmolecular clock that allows rates to vary independentlyalong branches according to a lognormal distribution(Drummond et al 2007) Three independent MarkovChain Monte Carlo (MCMC) simulations were runon the CIPRES Science Gateway Web server (Milleret al 2010) each 80 million generations sampling every4000 generations MCMC samples were inspected forconvergence and parameter stability (using Tracer15 Rambaut and Drummond 2007) The first 25 ofeach chain was removed as burn-in and chains werecombined using LogCombiner 172 (Drummond et al2012)

Trimming trees yields a 67-tip phylogeny (Fig 2)that accounts for approximately 74 of the speciesdiversity in the Euphorbia GDD clade In the processof trimming we made the following decisions aboutproblematic taxa First we split E pyrifolia a variablemember of E section Denisophorbia occurring on islandssurrounding Madagascar (Mauritius AssumptionIsland Aldabra Atoll and the Seychelles islands)into four distinct OTUs to account for the significantmorphological diversity it comprises Second many ofthe spiny shrubby taxa related to E mahafalensis andE milii are very difficult to identify at the species leveldue to ambiguous descriptions and often fragmentarytype material hindering taxonomic clarity Because

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 4 1ndash14

4 SYSTEMATIC BIOLOGY

FIGURE 2 Trimmed phylogeny of Euphorbia GDD derived from the maximum clade credibility tree Node support is shown for all nodeswith posterior probability greater than 050 support for the backbone and key nodes defining seven subclades (see Fig 6) is highlighted withblack points Vertical bars indicate these seven subclades plus E section Goniostema The asterisk indicates the E primulifolia clade

the entities related to E mahafalensis (E retrospinaE razafindratsirae E croizatii E rubrostriata andE hoffstaetteri) are quite homogeneous in overallmorphology geographic distribution and ecologicalpreferences (climate and substrate) we treated thisgroup of putative species as constituting a single OTUthe ldquoE mahafalensis complexrdquo The E milii assemblagewhich comprises numerous infraspecific concepts ispolyphyletic (Aubriot 2012) As a consequence wereduced our sampling of E milii to two monophyleticgroups for which we had clear morphological andgeographical data E milii var milii from the northernpart of Madagascarrsquos central highlands and E milii varlongifolia from the southern part

Finally we made decisions concerning the polyphylyand paraphyly respectively of two additional speciesEuphorbia primulifolia and E alluaudii The three samplesof E primulifolia in our analysis each from a distinctgeographical area were separated in the consensus

phylogeny by very strongly supported nodes so weconsidered each to constitute a distinct OTU accordingto their geographical provenance (E aff primulifoliavar primulifolia E aff primulifoliamdashIsalo and E affprimulifoliamdashHorombe) In the small section Deuterocallicomposed of just E alluaudii and E cedrorum wediscovered that E alluaudii is composed of twoassemblages one closer to E cedrorum than to theother assemblage of E alluaudii Because all E alluaudiispecimens are very similar morphologically it wasimpossible to associate them with one assemblageversus the other Keeping two distinct OTUs wouldhave drastically reduced the amount of occurrencedata since each of the OTUs would then have to berepresented by the small number of individuals usedfor molecular analyses In order to better reflect the verybroad geographic distribution of what is traditionallyreferenced to as E alluaudii we selected one asrepresentative of E alluaudii and removed the other from

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 5 1ndash14

2014 EVANS ET ALmdashEVOLUTION OF SUCCULENCE 5

our data set We regard our final selection of samplesto be representative of the diversity of the EuphorbiaGDD clade especially since we included representativesfrom problematic taxa as well as some of the ldquohiddenrdquodiversity within the clade through the inclusion ofsamples from OTUs whose historical delimitation wasshown to be paraphyletic or polyphyletic

Succulence and Life Form DataSome definitions of succulence exclude plants with

a caudex or tuber under the expectation that thesestructures primarily serve the function of storing carbonrather than water (Eggli and Nyffeler 2009) It is our viewthat the water and carbon storage value of tubers maybe intimately intertwined in seasonal environments (seethe ldquoDiscussionrdquo section) and in at least some species itis clear that tubers do store water (Pate and Dixon 1981Olson and Carlquist 2001 Eggli and Nyffeler 2009 Hearn2009a 2009b 2013) For these reasons we include tubersin our analysis of succulence and climate The GDD cladeincludes a diversity of forms of succulence trees andshrubs with stems and leaves exhibiting varying degreesof succulence some with tuberous roots ldquocactiformrdquospecies (single-stemmed or sparsely branched plantswith highly succulent stems bearing leaves only duringthe wet season) geophytes and dwarf chamaephyteswith a highly developed underground caudex somewith succulent leaves and thorny semisucculent shrubssuch as Euphorbia milii the famous ldquocrown-of-thornsrdquo(Fig 1 Haevermans et al 2004 Aubriot 2012) We tackledthis diversity of degrees and forms of succulence bygathering data on three charactersmdashleaf succulencethe presence of a tuber (including a caudex) andgrowth formmdashtreated as categorical The characters andcharacter states were defined as follows

1 Leaf succulence (0) leaves weakly succulent (1)leaves rigid ldquosemisucculentrdquo and (2) leaves highlysucculent (Fig 1)

Definition and coding of character states were basedon visual inspection of the thickness of the leaf bladevisibility of veins and leaf texture (viz leaves pliable ornot) Weakly succulent leaves (0) have a thin blade withclearly visible veins and are easily bent Semisucculentleaves (1) have a slightly thick rigid blade typically withveins that are not prominent they tend to break ratherthan bend Highly succulent leaves (2) have a very thickrigid and inflexible blade with inconspicuous veins

2 Tuber (0) nontuberous roots and (1) tuberous roots(Fig 1ce)

The absence (0) or presence (1) of a primarilysubterranean caudex or swollen root system wascaptured using a binary scheme

3 Growth form (0) geophyte (1) dwarfchamaephyte (2) sparsely branched ldquocactiformrdquo(3) shrub and (4) tree (Fig 1)

Geophytes (0) are mostly subterranean with a largecaudex and underground stems only leaves andinflorescences are exposed above the surface Dwarfchamaephytes (1) have a subterranean caudex that bearsshort unbranched stems not exceeding 60 cm above thesurface Cactiform species (2) are unbranched or sparselybranched with highly succulent stems Shrubs (3) lack atrue trunk and have numerous basal branches Trees (4)have a trunk a central stem unbranched at the base

These five categories were selected to capture themajor growth forms in the GDD clade while avoidingan excess of sparsely populated categories For examplethe two species of E section Deuterocalli (E alluaudii andE cedrorum) in our sample are classified as succulenttrees they have a growth form described as ldquocoraliformrdquoowing to the thickness of their stems and densebranching pattern (Fig 1f) We did not attempt to modelthe evolution of this growth form because it has a singleorigin in the GDD clade (convergent evolution of thecoraliform habit has occurred elsewhere in Euphorbia)

Some species were difficult to categorize For exampleEuphorbia didiereoides E duranii and E guillauminianahave growth forms intermediate between the shruband cactiform categories We treated them as shrubspreferring to err on the conservative side in ourdefinition of cactiform Ideally continuous data onsucculence should be collected (Ogburn and Edwards2012) but this was not practical in a group that includesso many microendemics including several exampleswhere an accepted species was described long ago buthas not been recorded since

Coding for these three characters drew upon multiplesources of information observations in the fieldexamination of living collections in greenhouses (at theMNHN arboretum in Chegravevreloup France HeidelbergBotanical Garden in Germany and WageningenBotanical Garden in the Netherlands) and outdoorcollections in Madagascar (Tsimbazaza BotanicalGarden in Antananarivo Antsokay Arboretum inToliara) as well as herbarium specimens (mostly fromMNHN and MBG) When living or dry specimens werenot available we used the primary literature as well asthe expertise and personal experience of the authorsand colleagues

Locality and Climate DataWe assembled an initial database of 653 occurrences

of members of the GDD clade approximately 75from museum collections mainly the MNHN andMBG herbaria (available on SONNERAT [httpsciencemnhnfrinstitutionmnhnsearch lastaccessed May 28 2014] and TROPICOS [httpwwwtropicosorg last accessed May 28 2014]) Theremaining localities were derived from monographsunpublished records and expert knowledge (AubriotX Haevermans T unpublished data) All locality datawere carefully evaluated some records were discardedbecause of imprecision species misidentification or

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 6 1ndash14

6 SYSTEMATIC BIOLOGY

a) b)

c)

FIGURE 3 a) Euphorbia GDD localities used in this study (crosses) and major bioclimatic regions of Madagascar as defined by Cornet (1974) b)Loading of the 19 Bioclim variables in the OMI 1 versus 2 climate space with precipitation variables colored blue and temperature variables coloredyellow The variables are MAT mean diurnal temperature range (DiurTRange) isothermality (Isotherm) temperature seasonality (TSeason)maximum temperature of the warmest month (MaxTMo) minimum temperature of the coldest month (MinTMo) temperature annual range(TRange) mean temperature of the wettest quarter (TWetQ) mean temperature of the driest quarter (TDriQ) mean temperature of the warmestquarter (TWarmQ) mean temperature of the coldest quarter (TColdQ) MAP precipitation of the wettest month (PWetMo) precipitation of thedriest month (PDryMo) precipitation seasonality (PSeason) precipitation of the wettest quarter (PWetQ) precipitation of the driest quarter(PDryQ) precipitation of the warmest quarter (PWarmQ) precipitation of the coldest quarter (PColdQ) c) Ordination diagram of OMI axis 1versus 2 with centroids for each species (points) Colored circles overlain here were placed to encompass speciesrsquo geographic distributions withcolors corresponding to panel a

duplication The final data set contained 529 occurrences(Fig 3a) which were used for climatic data extractionClimate data were the 19 ldquoBioclimrdquo variables availablefrom worldclim (httpwwwworldclimorg lastaccessed May 28 2014 Hijmans et al 2005) downloadedat the highest possible resolution (sim1 km2) and extractedusing ArcGIS version 93 These include 11 temperature-related variables (MAT minimum temperature inthe coldest month mean temperature of the wettestand driest quarters etc) and 8 precipitation-relatedvariables (MAP precipitation of the driest and wettestquarters precipitation seasonality etc)

OrdinationWe compared two different multivariate ordination

techniques for reducing the dimensionality of theBioclim data and examining the distribution ofsucculence characters principal components analysis(PCA) and the outlying mean index (OMI Doleacutedecet al 2000) OMI is a type of discriminant analysis sounlike PCA it is designed to maximize separation ofspeciesrsquo niches in the reduced-dimension niche space

In particular an OMI analysis produces a transformationof the original variables (here the 19 Bioclim variables)that maximizes the distance between the centroid of ahypothetical species that occurs evenly across the studylandscape (ie the mean environmental conditionsamong the sampled sites) and the centroids of the studyspecies (Doleacutedec et al 2000 Thuiller et al 1994) Thisanalysis was implemented using the R package ade4(Thioulouse et al 1997)

Tests of TraitndashClimate RelationshipsExamination of the distribution of traits in climate

space led us to select MAT and MAP for analysesof traitndashclimate relationships (Results) We usedphylogenetic logistic regression to evaluate theability of these dimensions of climate to predictspeciesrsquo character state Each trait was modeled asa binomial response converting the two traits withmore than two categories (leaf succulence and growthform) into binary variables Semisucculent leaveswere grouped with nonsucculent leaves yielding avariable that contrasts species with highly succulent

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 7 1ndash14

2014 EVANS ET ALmdashEVOLUTION OF SUCCULENCE 7

leaves against those without highly succulent leavesGrowth form was recoded into two variables onecontrasting species that are cactiform versus notand another contrasting taxa that are geophytes ordwarf chamaephytes versus not highlighting the mostconspicuous forms of above-ground versus below-ground succulence respectively Our tests accountedfor the relatedness of species by including a correlationstructure derived from phylogenies (implemented usingthe MCMCglmm package in R Hadfield 2010 Hadfieldand Nakagawa 2010) Prior distributions for the climateeffects were normally distributed centered on zerowith variance proportional to the estimated residualvariance (Hadfield 2012) Based on convergence andautocorrelation statistics from initial simulations onthe trimmed maximum clade credibility tree we ransimulations of 15 times 106 iterations eliminating the first5 times 105 as burn-in and thinning the chains to 11000samples yielding 1000 samples per chain To accountfor phylogenetic uncertainty including uncertaintyabout the monophyly of Euphorbia sections Goniostemaand Denisophorbia we repeated MCMCglmm tests (onesimulation per tree) using a random sample of 10 000trees taken from the post burn-in phase of the threeBEAST runs described above A climate effect wasconsidered significant if at least 95 of its posteriordistribution fell on one side of zero

RESULTS

PhylogenyThe aligned data matrix comprised 8507 characters

of which 2478 were variable and 1919 parsimonyinformative Bayesian analysis conducted with BEASTyielded a well-resolved phylogeny (Fig 2) in whichposterior probability support was generally strongexcept for some relationships in the deeper nodesThe analysis adds support for the monophyly of theEuphorbia GDD and all three of its constituent clades(E sections Goniostema Denisophorbia and Deuterocalli)corroborating earlier results (Haevermans et al 2004Zimmerman et al 2010 Dorsey et al 2013) andquestioning the suggestion made by Bruyns et al (2006)and Horn et al (2012) that E sections Goniostema andDenisophorbia may be paraphyletic These differences intopology likely result from differences in taxon samplingand in DNA regions examined (Aubriot 2012) Manyof the relationships within E section Goniostema arestrongly supported and are consistent with previousstudies (Haevermans et al 2004 Zimmerman et al2010) These relationships considered along withecomorphological features allowed us to identify fivewell-supported subclades within E section Goniostema(Aubriot 2012)

Ordination of ClimatePCA and OMI eigenvectors were oriented similarly

but differed in magnitude (Fig 3c vs Supplementary

Fig S2b httpdoiorg105061dryadvq6mp) Wepreferred the output from the OMI analysis over PCAbecause it nicely captured the four major bioclimaticregions defined by Cornet (1974)mdashthe humid warmeastern forest the cool central highlands the subaridsouth and southwest and the climatically heterogeneouswestmdasheach one falling into one-quadrant of the OMIaxis 1 versus 2 climate space (Fig 3b) Furtherspecies were better separated from one another in OMIaxis 1 versus 2 space compared with PCA axis 1versus 2 space (Fig 3b vs Supplementary Fig S2ahttpdoiorg105061dryadvq6mp) MAT is highestin the upper left and lowest in the lower right of theOMI ordination diagram whereas MAP is highest inthe lower left and lowest in the upper right (Fig 3b) Thedriest region of the ordination diagram has the strongestprecipitation seasonality and the coolest region hasthe greatest temperature seasonality (Fig 3b) thuswe take MAT and MAP as proxies for seasonalityof temperature and precipitation respectively Speciesof Euphorbia belonging to the GDD clade are foundthroughout Madagascar with multiple representativesin each of Cornet (1974) bioclimatic regions (Fig 3ac)

Distribution of Succulence Characters in Climate SpaceExamination of the distribution of character states

in OMI axis 1 versus 2 climate space allowed us toselect two climate variables for traitndashclimate tests MATand MAP The prevalence of highly succulent leaves(orange Fig 4a) and tubers (red Fig 4b) varies along thediagonal of the OMI ordination diagram stretching fromthe upper right to lower left Prevalence of the cactiformgrowth form varies along the upper left to lowerright diagonal (blue Fig 4c) Prevalence of geophytesand dwarf chamaephytes varies along both of thesediagonals (green and orange Fig 4c) either individuallyor taken together The upper right to lower left diagonalis a precipitation gradient from dry and seasonalprecipitation to mesic and even precipitation (Fig 3b)whereas the upper left to lower right diagonal is atemperature gradient from warm and even temperatureto cooler and seasonal temperature variation (Fig 3b)MAT and MAP which load most directly on thesediagonals (Fig 3b) were selected for tests of the abilityof climate variation to predict speciesrsquo character state

Leaf SucculenceThe majority of species have nonsucculent leaves

(47 out of 67 total) with fewer having semisucculentleaves (8 species) or strongly succulent leaves (12species) Those with nonsucculent leaves occur in allfour quadrants of the climate diagram (green Fig 4a)whereas most with highly succulent leaves (10 out of 12)are found in the upper right quadrant (orange Fig 4a)corresponding to the subarid south of MadagascarOne species with highly succulent leaves (E francoisii)and half of the species with semisucculent leaves

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 8 1ndash14

8 SYSTEMATIC BIOLOGY

a)

b)

c)

FIGURE 4 Distribution of traits a) leaf succulence b) tuber andc) growth formmdashin the climate space defined by OMI axis 1 versus 2(similar to PCA) Inset illustrates the interpretation of this climate space(see Fig 3b)mdashwarm versus cool dry versus wet as well as precipitationand temperature seasonality (P seas and T seas respectively) Eachtaxon is represented by an ellipse of the 95 density of that taxonrsquoslocalities ellipses are colored according to the speciesrsquo trait state (listedin Table 1) Full taxon names as well as abbreviations used here arelisted in Table 1

(E thuarsiana E tardieuana E elliotii and E mangelsdorfii)are found in places that are surprisingly wet at themacroscale (eastern rainforest Fig 4a) Phylogeneticlogistic regression indicates that MAP is significantlypredictive of highly succulent leaves which have atendency to evolve in places with the least precipitation(Fig 5a) This result is robust to phylogenetic uncertaintyMAP had a significant effect on the presence of highlysucculent leaves on more than 99 of the 10 000trees examined At the lowest MAP observed amongEuphorbia GDD species (376 mm per year) the predictedprobability of having highly succulent leaves is 064(Fig 5a)

TubersNearly half of the Euphorbia GDD species in our sample

have a tuber (31 out of 67 red Fig 4b) They are especiallyconcentrated in the subarid south just 2 of 21 taxa in theupper right quadrant of the ordination diagram lack atuber (Fig 4b) Phylogenetic logistic regression indicatesthat MAP is significantly related to the presence of atuber taxa occurring in places with less precipitationare more likely to have a tuber (Fig 5b) This pattern wassignificant on 98 of 10 000 trees At the lowest MAPobserved among Euphorbia GDD species (376 mm peryear) the predicted probability of having a tuber is 086(Fig 5b)

Life FormThe 14 ldquocactiformrdquo species tend to occur in the

warmest region of the climate space (Fig 4c andSupplementary Fig S3a httpdoiorg105061dryadvq6mp) with the notable exceptions ofEuphorbia horombensis E lophogona and E boissieri(MAT = 205C 229C and 237C respectively)The effect of MAT on the cactiform life form is notstatistically significant on any of 10 000 trees considerednor is the effect of MAP But the evolution of two growthforms with a majority of their biomass below groundmdashgeophytes and dwarf chamaephytesmdashis significantlyassociated with low average annual precipitationandor low average annual temperature Evolution ofgeophytism is significantly associated with low MAT(on 100 of 10 000 trees) whereas evolution of the dwarfchamaephyte form is significantly associated withlow MAP (on 88 of 10 000 trees) Evolution of thesetwo growth forms considered together is significantlyassociated with low MAT (on 99 of 10 000 treesFig 5c) At the lowest MAT observed among EuphorbiaGDD species (166C) the predicted probability of beingeither a geophyte or dwarf chamaephyte is 073 (Fig 5c)Two other growth form categories trees and shrubs arefound in all four quadrants of the climate space

No statistically significant relationships were foundbetween climate and the cactiform growth form Thisincluded models with a linear term for MAP alinear plus quadratic term the preceding two models

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 9 1ndash14

2014 EVANS ET ALmdashEVOLUTION OF SUCCULENCE 9

a) b) c)

FIGURE 5 Lower panels stochastic character mapping of three traits related to succulence on the trimmed maximum clade credibilitytree with the tree plotted in climate space Climate data for the extant taxa are shown in Table 1 Climate at interior nodes was reconstructedusing the function phenogram (package phytools R Revell 2013) stochastic character mapping was performed using makesimmap (also in thepackage phytools) A single stochastic realization of character mapping is shown for each trait Upper panels predicted probability of havingthe succulent character state as a function of climate from a logistic regression not taking into account phylogeny a) Leaf succulence convertedto a binary trait phylogeny plotted with respect to MAP b) As in panel a but the tuber character c) Growth form converted to a binary traitcontrasting geophytes and dwarf chamaephytes against all other growth forms phylogeny plotted with respect to MAT

with a linear term for MAT and the precedingmodels without two outliers with respect to MAP(Euphorbia lophogona and E boissieri) This is in spiteof trends in the data (Supplementary Fig S3abhttpdoiorg105061dryadvq6mp) We suspect thelack of statistical significance is a consequence of lowstatistical power due to a small number of transitions inand out of the cactiform state (Supplementary Fig S3chttpdoiorg105061dryadvq6mp)

DISCUSSION

Although lingering phylogenetic uncertainty suggestssome caution our analyses confirm that succulence inthe Euphorbia GDD clademdashincluding leaf succulencecactiform stem succulence and tuber formationmdashisassociated with aridity (warm dry conditions Figs4 and 5) providing new support for the notion thatsucculence is an adaptive response to aridity Furtherour results support the notion of two fundamentalcategories of succulencemdashthose positioning water-storing tissues below ground or near the surface versusabove groundmdashwhich are associated respectivelywith very dry and relatively cool conditions versusmoderately dry and warm conditions The associationof tubers and dwarf chamaephytes (all of which have atuber) with low MAP confirms the prediction that below-ground water storage should be favored in very dryenvironments Hearn (2013) found a remarkably parallelresult in Adenia (Passifloraceae) a clade centered inMadagascar and Africa with tremendous morphologicaldiversity and expression of succulence in different parts

of the plant body the evolution of tubers is significantlyrelated to lower precipitation The association betweenhighly succulent leaves and low MAP in the EuphorbiaGDD clade also confirms this prediction 9 of the12 species with highly succulent leaves are dwarfchamaephytes thus they store water below ground ina tuber and their water-storing leaves are deployednear the soil surface (Fig 1c) We also found theevolution of below-ground or near-ground succulenceto be associated with cooler temperatures geophytes aswell as geophytes plus dwarf chamaephytes combinedare more prevalent in cooler environments

Contrary to a laypersonrsquos view of cactiforms theyare not prevalent in the driest places occupied bythe Euphorbia GDD clade (upper right quadrantFig 4c) instead this growth form is most prevalent atintermediate levels of MAP (Supplementary Fig S4bhttpdoiorg105061dryadvq6mp) Cactiform stemsucculence is also associated with warm temperatures 11of the 14 cactiform species are found in places with MATbetween 24C and 26C (Table 1 and SupplementaryFig S3a httpdoiorg105061dryadvq6mp) Thesepatterns are consistent with the literature (Gibsonand Nobel 1986 Burgess and Shmida 1988 Burgess1995 Eggli and Nyffeler 2009) but they were notstatistically significant probably because of a lackof power (Supplementary Fig S3c httpdoiorg105061dryadvq6mp) and the presence of ldquoclimateoutliersrdquo that is Euphorbia GDD species that are foundin surprisingly mesic places given their morphologicalcharacteristics For example E lophogona is cactiform yetit occurs in the eastern mesic forest biome where MAPis 2172 mm per year However many of these ldquoclimate

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 10 1ndash14

10 SYSTEMATIC BIOLOGY

TABLE 1 Complete listing of the Euphorbia taxa included in this study abbreviations used in figures the average climatic conditions amonglocalitiesmdashMAT (in C) MAP (in mm) and scores on OMI axes 1 and 2mdashalong with character states

Taxon Abbreviation MAT MAP OMI1 OMI2 Tuber Leaves Growth form Localities

E alfredii alfR 260 1674 minus309 132 Absent Not Cactiform 1E alluaudii all 225 780 210 044 Absent Strong Tree 73E ambarivatoensis ambaR 260 1635 minus313 133 Absent Not Shrub 9E ambovombensis ambo 230 559 253 082 Present Strong Cham 5E analavelonensis ana 205 760 356 minus004 Absent Not Shrub 5E ankaranae ankeR 256 1486 minus286 109 Absent Not Tree 15E ankarensis anks 259 1502 minus301 129 Absent Not Cactiform 10E banae ban 240 484 235 200 Present Not Shrub 6E beharensis beh 240 637 192 131 Present Strong Shrub 12E berevoensis bere 265 1106 minus016 310 Present Not Shrub 5E berorohae bero 259 791 168 305 Present Not Shrub 1E boissieri boisL 237 2894 minus432 minus539 Absent Not Cactiform 4E bongolavensis bong 266 1500 minus141 243 Absent Not Tree 6E bulbispina bulb 253 1193 minus264 109 Absent Not Shrub 4E capmanambatoensis capmR 255 1370 minus299 minus017 Absent Semi Cactiform 5E capsaintemariensis capsR 235 422 211 110 Present Strong Cham 10E cedrorum ced 239 396 221 222 Absent Strong Tree 4E cylindrifolia cyl 236 874 119 021 Present Strong Cham 7E decaryi dec 239 730 157 063 Present Strong Cham 4E didiereoides did 210 856 278 minus040 Absent Not Shrub 9E duranii dur 196 1010 295 minus184 Absent Not Shrub 7E elliotii ellS 227 1821 minus154 minus405 Present Semi Tree 6E erythrocucullata ery 247 749 137 225 Present Not Shrub 5E francoisii fran 229 1607 minus110 minus329 Present Strong Cham 9E geroldii gerL 244 1396 minus221 minus012 Absent Semi Shrub 4E gottlebei got 241 674 269 269 Present Not Shrub 1E guillauminiana gui 265 1565 minus199 280 Absent Not Shrub 6E hedyotoides hed 238 708 182 108 Present Not Tree 13E hermanschwartzii hermR 256 1603 minus291 118 Absent Not Cactiform 6E horombensis hor 205 840 314 minus042 Absent Not Cactiform 6E itremensis itrS 171 1322 392 minus341 Present Not Geophyte 3E kondoi kon 241 376 207 232 Present Not Shrub 8E labatii labL 260 1655 minus315 129 Present Not Cham 3E leuconeura leu 246 1278 minus219 054 Absent Not Cactiform 10E lophogona lopS 229 2172 minus219 minus457 Absent Not Cactiform 12E mahabobokensis mahb 226 757 275 108 Present Not Tree 3E mahafalensis mahf 240 493 236 175 Present Not Shrub 6E mangelsdorffii mangL 216 1842 minus092 minus346 Absent Semi Shrub 1E martinae marR 259 1610 minus304 130 Absent Not Tree 9E milii var longifolia mill 198 918 330 minus113 Absent Not Shrub 4E milii var milii milm 189 1183 299 minus181 Absent Not Shrub 2E millotii millS 252 1420 minus292 minus080 Absent Semi Cactiform 6E moratii morL 255 1320 minus108 222 Present Not Cham 8E neohumbertii neoR 259 1434 minus297 135 Absent Semi Cactiform 12E pachypodioides pachR 259 1627 minus302 126 Absent Not Cactiform 10E parvicyathophora parv 242 477 235 252 Present Strong Cham 5E paulianii pau 251 1293 minus028 204 Absent Not Cactiform 6E pedilanthoides ped 265 1540 minus208 305 Present Not Shrub 7E perrieri per 246 1563 minus148 138 Absent Not Cactiform 10E aff primulifolia var primulifolia primL 183 1376 300 minus265 Present Not Geophyte 9E aff primulifoliamdashHorombe primh 206 847 314 minus053 Present Not Geophyte 4E aff primulifoliamdashIsalo primi 212 792 289 022 Present Not Geophyte 3E aff pyrifoliamdashAldabra pyrald 260 1077 minus288 091 Absent Not Tree 11E aff pyrifoliamdashAssumption pyras 261 1085 minus294 096 Absent Not Tree 7E aff pyrifoliamdashMaurice pyrm 225 1852 minus158 minus389 Absent Not Tree 7E aff pyrifoliamdashSeychelles pyrsR 260 2213 minus605 minus210 Absent Not Tree 11E quartziticola quaS 166 1332 410 minus373 Present Not Geophyte 20E rangovalensis rangL 199 1804 059 minus401 Present Not Tree 18E robivelonae rob 253 1487 minus302 minus100 Absent Not Shrub 2E rossii ros 252 717 132 268 Present Not Shrub 7E sakarahaensis sak 241 666 278 259 Present Not Shrub 4E suzannaendashmarnierae suz 233 875 233 125 Present Strong Cham 1E tardieuana tarR 231 2412 minus271 minus409 Absent Semi Shrub 5E thuarsiana thuS 236 2903 minus441 minus564 Absent Semi Tree 7E tulearensis tul 240 387 211 227 Present Strong Cham 6E viguieri vig 261 1324 minus200 219 Absent Not Cactiform 23E waringiae war 219 886 222 minus027 Present Strong Cham 1

Substrate is indicated with a superscript above the taxon abbreviation (R = rock outcropping S = sand L = laterite) Character states in bold arethose contrasted against the other state(s) in phylogenetic logistic regression

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 11 1ndash14

2014 EVANS ET ALmdashEVOLUTION OF SUCCULENCE 11

outliersrdquo are found on edaphically arid substrates(Table 1) E lophogona grows on sand and thus ata microenvironmental scale is subjected to ariditya phenomenon arising from the interaction betweenclimate and substrate (Nobel 1983 McAuliffe 1994Schwinning and Ehleringer 2001 Schwinning 2010) Wereturn to this point below regarding niche conservatismversus lability

There is still much to be learned about the resourcestrategies of succulent plants In particular water andcarbon dynamics may be intimately intertwined inshaping growth form variation in environments withseasonal precipitation In Euphorbia GDD tubers areassociated with the longest drought season and thus theshortest wet seasonmdashthat is they are associated withthe need to mobilize both water and carbon rapidlyto deploy temporary photosynthetic organs (drought-deciduous leaves) and upregulate the physiologicalprocesses necessary to take advantage of a short windowof opportunity for resource acquisition growth andreproduction That is water stored in tubers maysupport photosynthesis on a seasonal time scale as hasrecently been shown for stem-stored water in baobabs(Chapotin et al 2006a 2006b) rather than the diurnaltime scale at which stem-stored water in cacti supportsphotosynthesis

Herbivory and fire in addition to climate are thoughtto interact to shape growth form variation in aridand semiarid environments (Burgess 1995) Howeverherbivory seems an unlikely explanation for shiftsfrom above-ground to below-ground water storagesince cactiform stem succulents have evolved successfuldefenses against herbivory both physical and chemical(eg spines and toxic sap) Depending on fire intensityand frequency cactiform stem succulents can be quitevulnerable to burning (Thomas 1991 Thomas andGoodson 1992 Pfab and Witkowski 1999) whereasbelow-ground forms of succulence are better designedto survive fire (Lesica 1999 Tyler and Borchert 2002Proches et al 2006) However the prehuman fire regimeof Madagascar is quite contentious (Lowry et al 1997)making it difficult to conclude whether fire played a rolein shaping growth form in Euphorbia GDD

Climatic Niche Evolution versus ConservatismAt first glance climatic niches seem quite labile in

Euphorbia GDD Although most species are found inplaces that are relatively warm (22ndash26C MAT) and dry(50ndash150 cm MAP Fig 6) there are several exceptionscertain taxa live in cooler places such as E quartziticolaE itremensis E primulifolia var primulifolia E milii varmilii and E milii var longifolia (of the central highlands)and others occur in wetter places such as E boissieriE thuarsiana E tardieuana E aff pyrifolia Seychellesand E lophogona (of Madagascarrsquos eastern forests andthe Seychelles Islands) Across the Euphorbia GDD taxathat we sampled MAP ranges from 38 cm per year to

a)

b)

FIGURE 6 a) Trimmed phylogeny of Euphorbia GDD (as in Fig 3)plotted with respect to MAT MAT data for the extant taxa are shownin Table 1 ancestral reconstruction of MAT used the ldquopicrdquo methodin traitgram (package picante R) Colors indicate the seven stronglysupported subclades illustrated in Figure 3 Horizontal bars indicatethe range of values for each subclade b) As in a but MAP

more than 290 cm per year nearly a 10-fold differenceand MAT from 17C to 27C (Table 1)

Multiple independent colonizations of mesicenvironments are suggested by the fact that the speciesoccupying these areas are derived from five separatewell-supported subclades (Figs 2 and 6b) sectionDeuterocalli (red) section Denisophorbia (pink) a pair ofsizeable clades with species in section Goniostema thatare found in northern Madagascar which we refer toas the ldquothuarsianardquo clade and the ldquoboissierirdquo clade (afterthe first-named species in each yellow and orangerespectively) E primulifolia var primulifolia (gray) andanother pair of clades in section Goniostema whosemembers are mostly found in southern Madagascarhereafter the ldquomiliirdquo and ldquolophogonardquo clades (blue andgreen respectively also after the oldest named speciesin each Fig 6a) Multiple independent colonizations

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 12 1ndash14

12 SYSTEMATIC BIOLOGY

are also suggested to a lesser degree for the invasion ofcooler environments (Fig 6a)

This appears to paint a picture of niche labilityuntil one takes into consideration substrate and scaleMany of the Euphorbia GDD taxa occur on substratesmdashrock outcrops sands and laterite soils (Table 1)mdashthat can be expected to yield limited pulses ofavailable moisture even in the face of considerableprecipitation given the shallow root profile of the plantsFurther temperature and precipitation estimated at the1 km2 scale (macroenvironmental conditions) can beprofoundly affected not just by substrate but also slopeaspect and local topography (Kumar et al 1997 AustinVan Niel 2011) We conclude that niche lability amongmembers of Euphorbia GDD may not be as great asit first seems Instead we suspect microenvironmentalspecialization allows for a degree of niche conservatism(toward a warm dry niche) in the face of considerablemacroenvironmental variation Conclusions about nichelability derived from macroscale habitat data should beviewed with caution

Testing for TraitndashClimate RelationshipsmdashA MethodologicalNote

OMI axes 1 and 2 were not significantly relatedto variation in succulence characters (results availablefrom the authors) This raises an important pointspecies do not evolve in response to principal axes ofclimate variation constructed by multivariate ordinationbut rather they respond to physiologically importantclimatic variation which may or may not load parallelto such statistically constructed principal axes of climatevariation In some cases traits may vary with principalaxes of climate variation (eg Boucher et al 2012) inother cases they may not Here MAP and MAT loadon the diagonals of the ordination diagram of OMI axis1 versus 2 (Fig 3b) and it is along these diagonalsthat traits related to succulence vary (Fig 4) Testsof traitndashclimate relationships must navigate betweenthe statistical rock of multicollinearity among climatevariables and the statistical hard place of elevatedtype 1 error with multiple tests but the solution ofusing principle axes of climate variation for traitndashclimatetests is not without pitfalls Interesting and importanttraitndashclimate patterns can be overlooked if traits areonly related to principal components We suggest thatexamination of the distribution of traits in an ordinationdiagram is necessary for a robust analysis of traitndashclimaterelationships and that ordination should ultimately beused as an exploratory tool to refine hypotheses abouttraitndashclimate relationships

CONCLUSION

Although it has long been understood that plantsare shaped by their environment (Raunkiaeligr 1934Daubenmire 1974 Niklas 1997) we still have much to

learn about how temperature precipitation and theirseasonality interact with nonclimatic factorsmdashsuch asfire herbivory and substratemdashto shape plant form andfunction Here a study integrating trait climate andphylogenetic data for a remarkable radiation in oneof the worldrsquos most exciting laboratories of evolution(Madagascar) yields insight into the fascinating formsof succulent plants and the climatic conditions favoringdifferent forms of succulence This has implicationsfor the conservation of succulent species many ofwhich are threatened if different forms of succulenceare favored by different climatic regimesmdashif theirmorphology indeed is the result of adaptation toenvironmentmdashwe can expect these iconic organismsto be affected by climate change This is particularlysignificant in Madagascar one of the worldrsquos mostimportant biodiversity hotspots where succulent andspiny Euphorbia are a dominant component of manyvegetation types In fact climate effects on succulentEuphorbia seem to have already taken hold in SouthAfrica where there is widespread mortality of Euphorbiaingens trees (Van der Linde et al 2012)

SUPPLEMENTARY MATERIAL

Data available from the Dryad Digital Repositoryhttpdxdoiorg105061dryadvq6mp

FUNDING

Financial support for field and laboratory workwas provided by the MNHN ATMs ldquoTaxonomiemoleacuteculaire DNA barcode et gestion durable descollectionsrdquo and ldquoBiodiversiteacute actuelle et fossile Crisesstress restaurations et panchronisme le messagesysteacutematiquerdquo This work was also supported by theANR-funded project EVORANGE (ANR-09-PEXT-011)

ACKNOWLEDGEMENTS

The authors thank A Soulebeau for assistance with theEuphorbia section Denisophorbia species and the BoEMmolecular laboratory (Museacuteum National drsquoHistoireNaturelle Paris [MNHN]) for assistance with molecularwork Some of the sequencing was performed underagreement No 200567 between Genoscope and theMNHN under the project ldquoMacrophylogeny of liferdquodirected by Guillaume Lecointre The authors aregrateful to J Hadfield and K Holsinger for theirassistance with MCMCglmm Thanks go to the JCVIInc for support access and use of the computationalgrid employed for comparative analyses Commentsfrom M Ogburn KL Prudic B Boyle and membersof the Sanderson laboratory significantly improved themanuscript

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 13 1ndash14

2014 EVANS ET ALmdashEVOLUTION OF SUCCULENCE 13

REFERENCES

Arakaki M Christin P-A Nyffeler R Lendel A Eggli U OgburnRM Spriggs E Moore MJ Edwards EJ 2011 Contemporaneousand recent radiations of the worldrsquos major succulent plant lineagesProc Natl Acad Sci USA 1088379ndash8384

Aubriot X 2012 Radiations eacutevolutives ldquoinnovations cleacutesrdquo et notionsdrsquoespegraveces dans le genre Euphorbia L agrave Madagascar [PhDdissertation] Paris (France) Museacuteum National drsquoHistoire Naturelle

Austin MP Van Niel KP 2011 Improving species distribution modelsfor climate change studies variable selection and scale J Biogeogr381ndash8

Bergh NG Linder HP 2009 Cape diversification and repeatedout-of-southern-Africa dispersal in paper daisies (AsteraceaendashGnaphalieae) Mol Phylogenet Evol 515ndash18

Boucher FC Thuiller W Roquet C Douzet R Aubert S AlvarezN Lavergne S 2012 Reconstructing the origin of high-alpineniches and cushion life form in the genus Androsace (Primulaceae)Evolution 66 1255ndash1268

Bruyns PV Mapaya RJ Hedderson TJ 2006 A new subgenericclassification for Euphorbia (Euphorbiaceae) in southern Africabased on ITS and psbAndashtrnH sequence data Taxon 55397ndash420

Bruyns PV Klak C Hanacek P 2011 Age and diversity in Old Worldsucculent species of Euphorbia (Euphorbiaceae) Taxon 601717ndash1733

Burgess TL 1995 Desert grassland mixed shrub savanna shrubsteppe or semidesert scrub The dilemma of coexisting growthforms In McClaran MP Van Devender TR editors The DesertGrassland Tucson (Arizona) The University of Arizona Pressp 31ndash67

Burgess TL Shmida A 1988 Succulent growth-forms in aridenvironments In Whitehead EE Hutchinson CF TimmermannBN Verity RG editors Arid lands today and tomorrow Boulder(CO) Westview Press p 383ndash395

Chapotin SM Razanameharizaka JH Holbrook NM 2006a Waterrelations of baobab trees (Adansonia spp L) during the rainy seasondoes stem water buffer daily water deficits Plant Cell Environ291021ndash1032

Chapotin SM Razanameharizaka JH Holbrook NM 2006b Baobabtrees (Adansonia) in Madagascar use stored water to flush new leavesbut not to support stomatal opening before the rainy season NewPhytol 169549ndash559

Cornet A 1974 Essai de cartographie bioclimatique agrave MadagascarNotice explicative No 55 Paris ORSTOM

Couvreur TLP Chatrou LW Sosef MSM Richardson JE 2008Molecular phylogenetics reveal multiple tertiary vicariance originsof the African rain forest trees BMC Biol 654

Daubenmire RF 1974 Plants and environment 3rd ed New YorkWiley

Doleacutedec S Chessel D Gimaret-Carpentier C 2000 Niche separationin community analysis a new method Ecology 812914ndash2927

Dorsey BL Haevermans T Aubriot X Morawetz JJ Riina RSteinmann VW Berry PE 2013 Phylogenetics morphologicalevolution and classification of Euphorbia subgenus Euphorbia Taxon62291ndash315

Drummond AJ Ho SYW Rawlence N Rambaut A 2007 A RoughGuide to BEAST 14 [Internet] Available from URL httpbeastmcmcgooglecodecomfilesBEAST14_Manual_6July2007pdf(last accessed April 2013)

Drummond AJ Suchard MA Xie D Rambaut A 2012 Bayesianphylogenetics with BEAUti and the BEAST 17 Mol Biol Evol291969ndash1973

Edgar RC 2004 MUSCLE multiple sequence alignment withhigh accuracy and high throughput Nucleic Acids Res321792ndash1797

Edwards EJ Donoghue MJ 2006 Pereskia and the origin of the cactuslife-form Am Nat 167777ndash793

Eggli U Nyffeler R 2009 Living under temporarily arid conditionsmdashsucculence as an adaptive strategy Bradleya 2713ndash36

Felsenstein J 1985 Phylogenies and the comparative method Am Nat1251ndash15

Futuyma DJ 1997 Evolutionary biology 3rd ed Sunderland (MA)Sinauer Associates

Ganzhorn JU Lowry PP II Schatz GE Sommer S 2001 Thebiodiversity of Madagascar one of the worldrsquos hottest hotspots onits way out Oryx 35346ndash348

Gibson AC Nobel PS 1986 The cactus primer Cambridge (UK)Harvard University Press

Godinez-Aacutelvarez H Valverde T Ortega-Baes P 2003 Demographictrends in the Cactaceae Bot Rev 69173ndash203

Hadfield JD 2010 MCMC methods for multi-response generalisedlinear mixed models the MCMCglmm R Package J Stat Softw331ndash22

Hadfield JD 2012 MCMCglmm course notes [Internet]Available from URL httpcranr-projectorgwebpackagesMCMCglmmvignettesCourseNotespdf

Hadfield JD Nakagawa S 2010 General quantitative genetic methodsfor comparative biology phylogenies taxonomies meta-analysisand multi-trait models for continuous and categorical charactersJ Evol Biol 23494ndash508

Haevermans T Hoffmann P Lowry PP II Labat J-N RandrianjohanyE 2004 Phylogenetic analysis of the Madagascan Euphorbiasubgenus Lacanthis based on ITS sequence data Ann Missouri BotGard 91247ndash259

Haevermans T Rouhan G Hetterscheid W Teissier M Belarbi KAubriot X Labat J-N 2009 Chaos revisited nomenclature andtypification of the Malagasy endemic Euphorbia subgenus Lacanthis(Raf) MG Gilbert Adansonia 31279ndash299

Hearn DJ 2004 Growth form evolution in Adenia (Passifloraceae) anda model of the evolution of succulence [PhD dissertation] Tucson(Arizona) The University of Arizona

Hearn DJ 2006 Adenia (Passifloraceae) and its adaptiveradiation phylogeny and growth form diversification Syst Bot31805ndash821

Hearn DJ 2009a Descriptive anatomy and evolutionary patterns ofanatomical diversification in Adenia (Passifloraceae) Aliso 2713ndash38

Hearn DJ 2009b Developmental patterns in anatomy are sharedamong separate evolutionary origins of stem succulent and storageroot-bearing growth habits in Adenia (Passifloraceae) Am J Bot961941ndash1956

Hearn DJ 2013 Dissection of evolutionary networks to assess theirrole in the evolution of robustness function and diversificationEvolution 672273ndash2283

Hearn DJ Poulsen T Spicer R 2013 The evolution of growth formswith expanded root and shoot parenchymatous storage is correlatedacross the eudicots Int J Plant Sci 1741049ndash1061

Hijmans RJ Cameron SE Parra JL Jones PG Jarvis A 2005 Veryhigh resolution interpolated climate surfaces for global land areasInt J Climatol 251965ndash1978

Ho SYW 2007 Calibrating molecular estimates of substitution ratesand divergence times in birds J Avian Biol 38409ndash414

Horn JW Van Ee BW Morawetz JJ Riina R Steinmann VWBerry PE Wurdack KJ 2012 Phylogenetics and the evolutionof major structural characters in the giant genus Euphorbia L(Euphorbiaceae) Mol Phylogenet Evol 63305ndash326

Katoh K Misawa K Kuma K Miyata T 2002 MAFFT a novelmethod for rapid multiple sequence alignment based on fast Fouriertransform Nucleic Acids Res 303059ndash3066

Kumar L Skidmore AK Knowles E 1997 Modelling topographicvariation in solar radiation in a GIS environment Int J GeogrInform Sci 11475ndash497

Lesica P 1999 Effects of fire on the endangered geophytic herb Silenespaldingii (Caryophyllaceae) Am J Bot 86996ndash1002

Lowry PP II Schatz GE Phillipson PB 1997 The classification ofnatural and anthropogenic vegetation in Madagascar In GoodmanSM Patterson BD editors Natural change and human impactin Madagascar Washington (DC) Smithsonian Institution Pressp 93ndash123

McAuliffe JR 1994 Landscape evolution soil formation and ecologicalpatterns and processes in Sonoran Desert bajadas Ecol Monogr64111ndash148

Miller MA Pfeiffer W Schwartz T 2010 Creating the CIPRES sciencegateway for inference of large phylogenetic trees Proceedings of theGateway Computing Environments Workshop (GCE) 2010 Nov 14New Orleans LA p 1ndash8

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from

[1801 2362014 Sysbio-syu035tex] Page 14 1ndash14

14 SYSTEMATIC BIOLOGY

Mittermeier RA Robles GP Hoffman M Pilgrim J BrooksT Mittermeier CG Lamoreux J da Fonseca GAB 2005Hotspots revisited Earthrsquos biologically richest and most endangeredterrestrial ecoregions Chicago (IL) The University of ChicagoPress

Myers N Mittermeier RA Mittermeier CG Da Fonseca GAB KentJ 2000 Biodiversity hotspots for conservation priorities Nature403853ndash858

Niklas KJ 1997 The evolutionary biology of plants Chicago (IL) TheUniversity of Chicago Press

Nobel PS 1980 Morphology surface temperatures and northernlimits of columnar cacti in the Sonoran Desert Ecology 611ndash7

Nobel PS 1983 Biophysical plant physiology and ecology SanFrancisco (CA) WH Freeman and Co

Nobel PS 1988 Environmental biology of Agaves and Cacti New YorkCambridge University Press

Nylander JAA 2004 MrModeltest v2 Program distributed by theauthor Uppsala (Sweden) Evolutionary Biology Centre UppsalaUniversity

Ogburn RM Edwards EJ 2009 Anatomical variation in Cactaceaeand relatives trait lability and evolutionary innovation Am J Bot96391ndash408

Ogburn RM Edwards EJ 2010 The ecological water-use strategies ofsucculent plants Adv Bot Res 55179ndash225

Ogburn RM Edwards EJ 2012 Quantifying succulence a rapidphysiologically meaningful metric of plant water storage Plant CellEnviron 351533ndash1542

Olson ME Carlquist S 2001 Stem and root anatomical correlationswith life form diversity ecology and systematics in Moringa(Moringaceae) Bot J Linn Soc 135315ndash348

Pate JS Dixon KW 1981 Plants with fleshy underground storageorgansmdasha Western Australian survey In Pate JS McComb AJeditors The biology of Australian plants Nedlands (Australia)University of Western Australia Press p 181ndash215

Pfab MF Witkowski ETF 1999 Fire survival of a criticallyendangered succulent Euphorbia clivicola R A Dyermdashfire-avoideror fire-tolerant Afr J Ecol 37249ndash257

Phillipson PB Schatz G Lowry PP II Labat J-N 2006A catalogue of the vascular plants of Madagascar In GhazanfarSA Beentje H editors African plants biodiversity ecologyphytogeography and taxonomy Kew (UK) Royal Botanic Gardensp 613ndash627

Pierson EA Turner RM 1998 An 85-year study of saguaro (Carnegieagigantea) demography Ecology 792676ndash2693

Proches S Cowling RM Goldblatt P Manning JC Snijman DA2006 An overview of Cape the geophytes Biol J Linn Soc 8727ndash43

Rambaut A Drummond AJ 2007 Tracer v14 [Internet] Availablefrom URL httpbeastbioedacukTracer (last accessed April2013)

Raunkiaeligr C 1934 The life forms of plants and statistical plantgeography Oxford (UK) University Press

Raven PH Evert RF Eichhorn SE 1986 Biology of plants 4th edNew York Worth Publishers

Revell LJ 2013 Two new graphical methods for mapping traitevolution on phylogenies Method Ecol Evol 4754ndash759

Schwinning S 2010 The ecohydrology of roots in rocks Ecohydrology3238ndash245

Schwinning S Ehleringer JR 2001 Water use trade-offs andoptimal adaptations to pulse-driven arid ecosystems J Ecol 89464ndash480

Shreve F 1911 The influence of low temperatures on the distributionof the giant cactus Plant World 14136ndash146

Stearns SC Hoekstra RF 2005 Evolution an introduction 2nd edOxford (UK) Oxford University Press

Steenbergh WF Lowe CH 1977 Ecology of the saguaro IIReproduction germination establishment growth and survival ofthe young plant Scientific Monograph Series 8 Washington (DC)National Park Service

Steinmann VW 2001 The evolution of succulence in the new worldspecies of Euphorbia (Euphorbiaceae) [PhD dissertation] Claremont(CA) Claremont Graduate University

Steinmann VW Porter M 2002 Phylogenetic relationships inEuphorbieae (Euphorbiaceae) based on ITS and ndhF sequence dataAnn Missouri Bot Gard 89453ndash490

Su YCF Saunders RMK 2009 Evolutionary divergence times in theAnnonaceae evidence of a late Miocene origin of Pseuduvaria inSundaland with subsequent diversification in New Guinea BMCEvol Biol 9153

Thioulouse J Chessel D Doleacutedec S Olivier JM 1997 ADE4 amultivariate analysis and graphical display software Stat Comput775ndash83

Thomas PA 1991 Response of succulents to fire a review Int JWildland Fire 111ndash22

Thomas PA Goodson P 1992 Conservation of succulents in desertgrasslands managed by fire Biol Conserv 6091ndash100

Thuiller W Lavorel S Midgley G Lavergne S Rebelo T 1994 Relatingplant traits and species distributions along bioclimatic gradients for88 Leucadendron taxa Ecology 851688ndash1699

Tyler C Borchert M 2002 Reproduction and growth of the chaparralgeophyte Zigadensis fremontii (Liliaceae) in relation to fire PlantEcol 16511ndash20

Van der Linde JA Roux J Wingfield MJ Six DL 2012 Die-off ofgiant Euphorbia trees in South Africa symptoms and relationship toclimate South Afr J Bot 83172ndash185

Zimmerman NFA Ritz CM Hellwig FH 2010 Further support forthe phylogenetic relationships within Euphorbia L (Euphorbiaceae)from nrITS and trnL-trnF IGS sequence data Plant Syst Evol28639ndash58

at Museum

National dH

istoire Naturelle on July 1 2014

httpsysbiooxfordjournalsorgD

ownloaded from