sign in sign up
<<
Home , Actizera lucida, Acytolepis, Brephidium, Caleta (butterfly), Caleta elna, Catochrysops panormus

Journal of (J. Biogeogr.) (2015)

ORIGINAL One-note samba: the biogeographical ARTICLE history of the relict Brazilian butterfly Elkalyce cogina Gerard Talavera1,2,3*, Lucas A. Kaminski1,4, Andre V. L. Freitas4 and Roger Vila1

1Institut de Biologia Evolutiva (CSIC- ABSTRACT Universitat Pompeu Fabra), 08003 Barcelona, Aim Biogeographically puzzling taxa represent an opportunity to understand Spain, 2Department of Organismic and the processes that have shaped current distributions. The systematic Evolutionary Biology and Museum of Comparative Zoology, Harvard University, placement and biogeographical history of Elkalyce cogina, a small lycaenid but- Cambridge, MA 02138, USA, 3Faculty of terfly endemic to Brazil and neighbouring Argentina, are long-standing puzzles. Biology and Soil Science, St Petersburg State We use molecular tools and novel biogeographical and life history data to clar- University, 199034 St Petersburg, Russia, ify the and distribution of this butterfly. 4 Departamento de Biologia , Instituto Location South America, with emphasis on the Atlantic Rain Forest and Cer- de Biologia, Universidade Estadual de rado biomes (Brazil and Argentina). Campinas, 13083-970 Campinas, SP, Brazil Methods We gathered a data set of 71 () samples, including representatives of all described subtribes and/or sections. Among these, we contributed new sequences for E. cogina and four additional relevant taxa in the target subtribes Everina, Lycaenopsina and Polyommatina. We inferred a molecular phylogeny based on three mitochondrial genes and four nuclear markers to assess the systematic position and time of divergence of E. cogina. Ancestral geographical ranges were estimated with the R package BioGeoBEARS. To investigate heterogeneity in clade diversification rates, we used Bayesian analysis of macroevolutionary mixtures (bamm). Results Our results confirm the hypothesis that E. cogina belongs to the sub- tribe Everina and not Lycaenopsina, but unexpectedly recovered it as the sister group to the rest of Everina, with an estimated divergence time of approxi- mately 10 Ma. Ancestral geographical range reconstruction points to an old colonization from Asia, the centre of diversity for the Everina, to the New World. The Neotropical Polyommatina lineage diversified to produce almost 100 species in multiple genera, whereas the E. cogina lineage did not diversify at all. Such lack of diversification is unique among the seven Everina/ Polyommatina lineages that colonized the New World. We also show that the larvae of E. cogina feed on , supporting the identification of this host- plant family as the ancestral state for the whole group. Main conclusions The age and biogeographical reconstruction of the Elkalyce lineage are similar to those of the Neotropical lineage of Polyommatina and suggest that both travelled via the route proposed by Vladimir Nabokov (Asia– Beringia–North America–South America). This coincidence suggests that the climatic conditions at c. 10 Ma favoured dispersal from Asia to the Neotropics and that later events may have erased traces of these butterfly lineages in North America. *Correspondence: Gerard Talavera, Institut de Biologia Evolutiva (CSIC-Universitat Pompeu Keywords Fabra), Passeig Marıtim de la Barceloneta 37, Beringia, biogeographical disjunction, colonization, dispersal, diversification, 08003 Barcelona, Spain. E-mail: [email protected] Elkalyce cogina, Fabaceae, , New World, relict

ª 2015 John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/jbi 1 doi:10.1111/jbi.12671 G. Talavera et al.

This is just a little samba, cal species Elkalyce cogina (Schaus, 1902), which is restricted to Brazil and Misiones province in northern Argentina. The Built upon a single note. species was originally described within the by Other notes are bound to follow, Schaus (1902), but because no close relatives were attributa- ble based on morphology, the monotypic genus Elkalyce But the root is still that note ... Balint & Johnson, 1996 was erected, thought to be sister to Oreolyce and so belonging to the section (equiva-  Samba de Uma Nota So (‘One-Note Samba’), a song composed lent to the subtribe Lycaenopsina) within by Antonio Carlos Jobim and Newton Mendoncßa. (Balint & Johnson, 1996). More recently, Robbins & Duarte (2006) suggested that Elkalyce cogina could be related to INTRODUCTION in the Everes section (subtribe Everina), also within Polyommatinae, supporting an earlier unpublished hypothe- Intriguing disjunct distributions that apparently formed after sis of Heinz Ebert and John H. Eliot (see comments in the break-up of Gondwana are found within many butterfly Brown, 1993; Robbins & Duarte, 2006). In both hypotheses, families as the result of a complex biogeographical history of the putative sister groups of E. cogina are exclusively dis- colonization and events (de Jong & van Acther- tributed in Asia, and therefore this taxon seems to represent berg, 2007; Pena~ et al., 2010). The Lycaenidae also include a South American endemic lycaenid whose closest relatives notable disjunct distributions that are generally only attribu- occur in the Old World. table to long-distance dispersal events, including examples in In this study, we present a detailed molecular phylogeny the Spalgini, Lycaenini, Heliphorini and Polyommatini (de of Polyommatini including E. cogina and thus provide evi- Jong & van Actherberg, 2007; Vila et al., 2011). The study of dence about its systematic placement within the tribe. By the affinities and origins of these taxa is crucial to under- inferring divergence times and ancestral geographical range, standing the processes that led to the observed patterns of we estimate the time of colonization of the New World by and diversification world-wide. this lineage, and we explore how diversification patterns With over 1200 species distributed world-wide, the tribe within the group can elucidate the unique biogeographical Polyommatini is one of the two richest lycaenid clades, and evolutionary history of E. cogina. alongside the Neotropical/Holarctic-distributed Eumaeini (Eliot, 1973). These two tribes represent the only Lycaenidae groups with important South American radiations, but their MATERIALS AND METHODS current biogeographical patterns are strikingly different: whereas most Eumaeini diversity occurs in the Neotropics, Distribution and life history data the Polyommatini are most species-rich in the Holarctic region. Specimens of E. cogina were studied and collected in the field ~ Within the Polyommatini, a complex scenario of indepen- at three localities in Sao Paulo state (SP), south-eastern Bra- ° 0 ° 0 – dent New World colonization events was hypothesized by zil: (1) Serra do Japi, Jundiaı (23 13 S, 46 58 W, 900 ° 0 ° 0 Nabokov (1944, 1945) in the 1940s and recently corrobo- 1200 m), (2) Grota Funda, Atibaia (23 10 S, 46 31 W, – ~ rated by Vila et al. (2011), who showed that at least five col- 1050 1100 m), and (3) Alto do Capivari, Campos do Jordao ° 0 ° 0 – onization events from the Old World to the New World (22 44 S, 45 31 W, 1600 1800 m). Locality data for occurred between c. 11 and 1 Ma, apparently always through E. cogina were compiled from the list presented in Robbins Beringia (the region surrounding the Bering strait). This sce- & Duarte (2006); additional data were also obtained from nario was postulated and evaluated for the richest subtribe, specimens recently collected by the authors and deposited in Polyommatina, but other subtribes, like Leptotina, Everina, the Zoological Collection of Campinas State University, and ~ Glauchospsychina and Brephidiina, also have representatives from the personal communication of Ezequiel Nunez-Bustos both in the Old World and the New World. (Museo Argentino de Ciencias Naturales ‘Bernardino Riva- Such complex biogeographical histories can only be unrav- davia’, Buenos Aires, Argentina). A map of the updated dis- elled by means of a reliable and abundant record or tribution of E. cogina is presented in Fig. 1. inferred from well-founded phylogenies. For butterflies, although the fossil record is extremely poor (de Jong, 2007), Taxon sampling and molecular data knowledge of their phylogenetic relationships is improving, resulting in novel insights into long-standing biogeographical A large collection of 75 Polyommatini samples was used, questions. Following the seminal taxonomic revision by Eliot including at least one representative of each described sub- (1973), the higher level systematics of Polyommatini have tribe or section sensu Eliot (1973). Most of the data set was recently been improved using molecular phylogenies (Wie- gathered from our previous studies (Vila et al., 2011; Talav- mers et al., 2010; Vila et al., 2011; Talavera et al., 2013). era et al., 2013), from which multilocus sequences were Some enigmatic taxa remain dubiously placed at the tribal available. An additional specimen of Elkalyce cogina was col- and subtribal levels, however, as is the case for the Neotropi- lected from Petropolis (Brazil) and taxon sampling for the

2 Journal of Biogeography ª 2015 John Wiley & Sons Ltd Biogeography of the relict butterfly Elkalyce cogina

Figure 1 Recorded distribution of Elkalyce cogina (black dots) in South America. Locality data include records compiled by Robbins & Duarte (2006), new collections and museum data. target subtribes Everina, Lycaenopsina and Polyommatina with the following parameters: b2, 50% + 1 of the sequences; was strengthened with additional taxa. (see Table S1 in b3,3;b4,5;b5, all (Castresana, 2000; Talavera & Castresana, Appendix S1 in Supporting Information for a list of speci- 2007). The final combined alignment consisted of 4850 bp, mens used in this study.) The samples (bodies in ethanol comprising 2172 bp of COI + leu-tRNA + COII, 1172 bp of and wings in glassine envelopes) are stored in the DNA and EF-1a, 811 bp of 28S, 370 bp of Wg, and 328 bp of H3 and Tissues Collection of the Museum of Comparative Zoology is available in treebase at http://www.treebase.org/ (study (Harvard University, Cambridge, MA, USA) and of the Insti- ID 17438). tut de Biologia Evolutiva (Barcelona, Spain). Genomic DNA Maximum likelihood (ML) and Bayesian inference (BI) was extracted from a leg or from a piece of the abdomen were employed to estimate evolutionary relationships. ML using the DNeasy Tissue Kit (Qiagen, Valencia, CA, USA) was run using phyml 3.0 (Guindon et al., 2010), and BI following the manufacturer’s protocols. Fragments from was employed to simultaneously infer phylogenetic relation- three mitochondrial genes – cytochrome oxidase I (COI), ships and divergence times with the software beast 1.8.0 leucine transfer RNA (leu-tRNA) and cytochrome oxidase II (Drummond et al., 2012). jModeltest 2.1.4 (Guindon & (COII) – and from four nuclear markers – elongation factor- Gascuel, 2003; Darriba et al., 2012) was executed to select 1 alpha (EF-1a), 28S ribosome unit (28S), histone H3 (H3) the best-fitting DNA substitution models according to the and wingless (Wg) – were amplified by polymerase chain Akaike information criterion (AIC). For ML, the model reaction and sequenced as described in Talavera et al. GTR+I+G was selected for a single concatenated alignment. (2013). (GenBank codes for the set of sequences used and For BI, the data were partitioned into six markers, consider- obtained specifically for this study are listed in Table S2 of ing COI + leu-tRNA + COII a single evolutionary unit in Appendix S1.) the mitochondrial genome. As a result, the HKY+I+G model was used for H3, the TN+I+G model for Wg and a GTR+I+G model for EF-1a, 28S and the mitochondrial Alignment, phylogenetic inference and dating fragment. The gamma distribution was estimated automati- A molecular matrix was generated for each independent mar- cally from the data using six rate categories. Because of the ker, with the mitochondrial markers treated as a single evo- absence of reliable primary calibration points () for lutionary unit. All sequences were edited and aligned, the Lycaenidae studied, normally distributed time to the together with those obtained in Vila et al. (2011), using gen- most recent common ancestor (TMRCA) priors based on eious 6.1.5 (Biomatters, Auckland, New Zealand; available published age estimates were used as secondary calibrations at: http://www.geneious.com/). Ambiguously aligned regions on six well-supported nodes (posterior probability ≥ 0.95) and regions of the matrix lacking more than 50% of data to estimate divergence times (Fig. 2). The 95% HPD distri- were removed using gblocks 0.96 under a relaxed criterion bution of these TMRCA priors included the maximum and

Journal of Biogeography 3 ª 2015 John Wiley & Sons Ltd emi,F Germain, 4 2 Figure Talavera G. ure 06 n oymaiaaeidctdwt oordbxs pesd n nesd htgah ftemntpcspecies monotypic the of photographs underside and Upperside boxes. coloured with for indicated cogina hypotheses are Elkalyce taxonomic Polyommatina in and involved 2006) subtribes Duarte, Polyommatini two Vila The on World. based were probability nodes (posterior these nodes probabilities well-supported (posterior six relationships The supported indicate lines Thick 25 aeincrnga o oymaiibsdo ee genes: seven on based Polyommatini for chronogram Bayesian  vir1884 evrier tal. et r hw fml pcmnE.Oberth Ex. specimen (female shown are hts sl B Zsolt Photos: . 20 tal. et 21) rnhsi leidct oymaiaadEeialnae hthv ooie h New the colonized have that lineages Everina and Polyommatina indicate blue in Branches (2011). alint). 15 6 ≥ .5 sdfrcokclbainaemre ihcoksmos g rosfor priors Age symbols. clock with marked are calibration clock for used 0.95) rCl.BiihMsu 973 ihlblindicating: label with 1927-3, Museum British Coll. ur € 10 4 ≥ 5 .5;nd asso netit netmtddvrec times. divergence estimated in uncertainty show bars node 0.95); COI 3 1 5 , 2 leu-tRNA layecogina Elkalyce , COII 0 , (B EF-1 VL01L152 MWT93B024 AS92Z024 AS92Z272 JXM99T709 DL99T242 AS92Z185 JE01C283 VL05Z998 VL05Z995 AS92Z130 AD00P266 AD03B062 KD94R033 AAM98V076 TL96W903 DL02P705 TL97W507 KD94Q064 MWT93A070 TL97W513 MWT93D008 AS92Z186 MWT93D010 AS92Z131 TL96W917 MWT93E012 RV03V095 TQ05M179 AD00P540 MAT99Q954 MH01I001 RV03V234 VL01L462 CCN05I856 AD00P092 NK00P712 VL02X393 AD00P206 AP98W773 NK00P594 KD93C045 AH00T289 KD93C021 RV03V188 NK00P777 VL02X159 MWT93D027 VL02X094 AS92Z312 MWT93A009 MFB00N206 MFB00N227 AS92Z065 NK00P165 AD00P129 AD00P053 NK00P575 KD93Q030 AH98Y715 KD95Z506 MWT93B058 KD93C044 MWT93B038 RV03N585 TL96W908 AH95Y685 MC04Z114 RV05M735 VL02X098 09X164 ln ono,19;Rbis& Robbins 1996; Johnson, & alint Ma a , Wg Maurus vogelii eurypilus trochylus hanno Pamiria chrysopis lucifer lajus ammon Afarsia morgiana Glabroculus cyane Tongeia bisudu Alpherakya sarta ,

Euphilotes enoptes marina Echinargus isola podarce echo lygdamus comyntas icarioides usta hintza alsulus Plebejidea loewii eumedon Leptotes trigemmatus Cupido minimus cuzquenha Neolysandra coelestina agestis semiargus Paralycaeides inconspicua Rueckbeilia fergana idas bellargus amandus Rimisia miris dukei marshalli ª mirza aethiops hippocrates cyara falkensteini hylax cnejus caelius panormus Maculinea arion Pseudochrysops bornoi karsandra miskini Elkalyce cogina Eldoradina cyanea lucida Plebulina emigdionis nyseus 28S Itylos titicaca Pseudolucia chilensis Oraidium barberae rosimon alecto boeticus Edales pandava angusta Lycaenopsis haraldus malaya puspa elna 05Jh ie osLtd Sons & Wiley John 2015 Br and  sl oaFiug,P. Friburgo, Nova esil, ora fBiogeography of Journal H3 45 bp). (4850

Lycaenopsina Everina Polyommatina Biogeography of the relict butterfly Elkalyce cogina minimum ages inferred in Vila et al. (2011) for the selected clade to be combined. for the genera was nodes. These inferences were based on the application of a assigned according to Talavera et al. (2013). wide range of substitution rates described for mitochondrial Ancestral geographical ranges were estimated in order to regions in invertebrates. For COI, a slow substitution rate infer the most likely biogeographical origin for the Everina of 6.5 9 10 9 and a fast rate of 9.5 9 10 9 (Quek et al., subtribe and so know the directions of the intercontinental 2004) were used, and for COI + leu-tRNA + COII, a substi- dispersals that resulted in the current distributions. The like- tution rate of 11.5 9 10 9 substitutions site 1 yr 1 (Brower, lihoods of the two hypotheses were assessed with the R pack- 1994) was applied. The resulting 95% HPD intervals were age biogeobears 0.2.1 (Matzke, 2013a,b) (see Table S3 in 3.26–6.06 Ma for node 1, 2.97–5.61 Ma for node 2, 3.43– Appendix S1). For all analyses, 10 large-scale biogeographical 6.35 Ma for node 3, 6.76–12.2 Ma for node 4, 4.78– regions and a dispersal multiplier matrix were coded as in 9.32 Ma for node 5 and 9.95–17.2 Ma for node 6. An Vila et al. (2011). In order to avoid sampling effects, termi- uncorrelated relaxed clock (Drummond et al., 2006) and a nals were considered placeholders for the genera they repre- constant population size under a coalescent model were sented and the distribution range of the genus was used (see established as priors. Two independent chains were run for Table S4 in Appendix S1). A test was performed for a disper- 50 million generations each, sampling values every 1000 sal–extinction–cladogenesis (DEC) model (Ree & Smith, steps. A conservative burn-in of 500,000 generations was 2008), a maximum-likelihood version of the dispersal–vicari- applied for each run after checking Markov chain Monte ance model (DIVALIKE) (Ronquist, 1997) and a Bayesian Carlo (MCMC) convergence through graphically monitoring biogeographical inference (BAYAREALIKE) (Landis et al., likelihood values in Tracer 1.5 (available at: http://beast. 2013). The three models were also tested when allowing for bio.ed.ac.uk/Tracer). Independent runs were combined in founder-effect (j) (Matzke, 2014). The resulting logcombiner 1.8.0, as implemented in beast, and all models recovered the common ancestor of Everina and Poly- parameters were analysed using Tracer 1.5 to determine ommatina in several regions simultaneously, including both whether they had also reached stationarity. Tree topologies the Old World and the New World, a scenario that is biolog- were assessed using treeannotator 1.8.0 in the beast ically implausible. AIC scores and weights were used to package to generate a maximum-clade-credibility tree of all detect the best-performing model (BAYAREALIKE), which sampled trees with median node heights. figtree 1.4.0 was subsequently used to independently test each of the six (available at: http://beast.bio. ed.ac.uk/FigTree) was used to favoured areas as a potential origin by fixing the root. visualize the consensus tree along with node ages, age devia- tions and node posterior probabilities. RESULTS

Diversification rates and ancestral geographical Distribution and life history range inference Published geographical distribution data and field observa- To investigate possible diversification-rate heterogeneity tions show that E. cogina is associated with montane habi- affecting the Polyommatina + Everina clade, and therefore tats, usually above 800 m (with the exception of the the role of Elkalyce cogina, we used Bayesian analysis of southernmost records in Argentina) (Fig. 1). The species has macroevolutionary mixtures (bamm; Rabosky et al., 2013, usually been observed associated with open habitats, espe- 2014a). The software was run on the beast maximum-cred- cially partly flooded montane swamps (as noted by Robbins ibility tree and non-random missing data were accounted for & Duarte, 2006), where other polyommatines such as Lep- by assigning the proportion of the species sampled per totes cassius, Hemiargus hanno and are also com- genus. Four MCMC chains were run for 10 million genera- mon (although these three species have much broader tions with a burn-in of 10%. Convergence of chains was geographical and elevational distributions). Adults are gener- checked in coda (Plummer et al., 2006) and bamm results ally only found in particular localities, sometimes at great were analysed and visualized with the package bammtools densities. For example, we observed about 20 individuals in (Rabosky et al., 2014b) in R 3.1.1 (R Core Team, 2014). a small swamp area of c. 100 m2 in Serra do Japi (Jundiaı, A lineages-through-time (LTT) plot was calculated for the SP), and more than 50 individuals in a 30-m sector of a trail Polyommatina + Everina clade and contrasted with the through a swamp area in Campos do Jord~ao (SP). Adults of diversification patterns of a data set of simulated trees. The E. cogina are present all year round, although they are more R package treesim (Stadler, 2011) was used to simulate 100 common from January to April. They have a low, fluttering trees by fixing the extant number of species per genus and flight, usually 10–50 cm above ground, usually flying inside the time since the most recent common ancestor (function the vegetation or just above it. Adults were observed feeding ‘sim.bd.taxa.age’). The speciation and extinction rates used from flowers of small Asteraceae, Fabaceae and Lamiaceae in the simulations (k = 0.80, l = 0.40) were calculated that were growing in the marshy areas. In Serra do Japi, one through the extended version of the birth–death models larva has been reared on inflorescences of Desmodium unci- (function ‘bd.ext’) in the ape package (Paradis et al., 2004), natum (Fabaceae), but no oviposition behaviour has ever which allows phylogenetic and taxonomic data for a given been observed. This represents the first host-plant record for

Journal of Biogeography 5 ª 2015 John Wiley & Sons Ltd G. Talavera et al.

E. cogina. Because no ants were recorded tending larvae at 275.4) (see Table S3 in Appendix S1). Among the six any of the studied sites, a low degree of myrmecophily or a potential ancestral regions for the root of the tree, the Orien- non-myrmecophilous condition is assumed. tal region was strongly favoured (LnL, 142.96; AIC, 291.92), followed by the East Palaearctic region (LnL, 144.30; AIC, 294.60) (see Table S3 in Appendix S1). Thus, Phylogenetic systematics a dispersal event from Asia to the New World is recovered ML and BI recovered phylogenetic trees displaying generally as the most likely hypothesis, although the exact route is concordant topologies that are very similar to those previ- uncertain. ously published for Polyommatini (Vila et al., 2011; Talavera et al., 2013) for the supported nodes (Fig. 2, Fig. S1 in DISCUSSION Appendix S1). Elkalyce cogina was recovered basally in the Everina subtribe with strong support (posterior probability, The systematic position of Elkalyce cogina 1.0; bootstrap support, 100), and is thus most closely related to the Old World genera Tongeia, Talicada and Cupido. The Our results contradict the hypothesis of Balint & Johnson inferred divergence times also generally matched previously (1996) that Elkalyce cogina belongs to the subtribe Lycae- published probability distributions, and the estimated diver- nopsina (Lycaenopsis section). Rather, the present data and gence between E. cogina and the common ancestor of Ever- analyses support the close relationship to the Everina ina was 10.17 Ma (95% highest posterior density interval, (Everes section) proposed by Robbins & Duarte (2006) and HPD, 7.69–12.86 Ma). confirm the ideas of H. Ebert, as mentioned by Brown (1993). Talavera et al. (2013) suggest 15 Myr as a mini- mum age for subtribes in Polyommatini. Accordingly, the Diversification monotypic genus Elkalyce (10.17 Myr; HPD, 7.69– Net diversification rates, as estimated by the bamm model, 12.86 Myr) may be included within Everina. (Note that accelerated noticeably during the Palaearctic Polyommatina Everina is the correct name for this subtribe, rather than radiation (Fig. 3a). The 95% HPD sampled by bamm Cupidina as used by Talavera et al., 2013;.) Concerning the comprised the ten most probable distinct shift configura- position within the subtribe, our phylogeny shows Elkalyce tions. The single best shift configuration in bamm inferred to be the sister group to all the other Everina sampled one increasing rate shift for the most diverse Palaearctic (Fig. 2), and thus this taxon is not ‘sister to, or congeneric Polyommatina clade, with a probability of 0.52 (Fig. 3b). with, Tongeia’ as suggested by Eliot (1988, unpublished let- The second-best configuration (f = 0.14) suggested an ters to Robbins) as described in Robbins & Duarte (2006). additional minor second shift in the form of a diversifica- Interestingly, Robbins & Duarte (2006) expressed doubts tion decrease for the E. cogina lineage(Fig.3c).Thediver- regarding the homology of the male genital ‘false alulae’ sification-through-time curve showed that diversification because of slight differences in their position, which led rates peaked between 7 and 4 Ma (Fig. 3d). This strong them to be cautious about synonymizing Elkalyce with signal can be attributed to the shift detected for Palaearc- Tongeia. Although monotypic genera are sometimes consid- tic Polyommatina; the peak disappeared when diversifica- ered of dubious classificatory value (Farris, 1976), the phy- tion rates in this clade were not considered (Fig. 3e). The logenetic position and age (cf. discussion of generic ages in LTT plot showed an overall accumulation of lineages Talavera et al., 2013) of Elkalyce support the acceptance of increasing fairly smoothly (Fig. 3f). The curve displayed this genus. minor fluctuations in the cladogenesis through time that might result from clade diversification heterogeneity, but Historical biogeography and diversification generally fitted within the margins of the simulated trees under a constant-rate model (Fig. 3f). No strong shifts in Ancestral geographical range reconstruction indicates the diversification rate, either increases or decreases, were most likely scenario to consist of the common ancestor of detected in the lineages colonizing the New World, and Everina + Polyommatina originating in the Old World – only Elkalyce displayed a marginally significant rate probably in the Oriental region or in the Eastern Palaearctic decrease. Indeed, Elkalyce was recovered as the oldest lin- – followed by dispersal to the New World by the Elkalyce lin- eage with a single extant taxon in the Everina + Poly- eage (see Table S3 in Appendix S1). The centre of diversity ommatina tree (Fig. 3g). for Everina is clearly , as is the case for several other lycaenid groups. Indeed, all the genera except Elkalyce occur there, some of them (Bothrinia, Shijimia and Talicada) Ancestral geographical range exclusively. The remarkable disjunction between Elkalyce and The BAYAREALIKE model (log-likelihood, LnL, 137.28; the other Everina taxa does not, however, represent the only AIC, 278.4) was by far the best among all models tested with example of a New World lycaenid whose closest relatives are biogeobears, and the addition of founder-event speciation in the Old World. Other notorious biogeographical disjunc- (j) parameters improved the results (LnL, 134.71; AIC, tions exist, including those of two species of ,

6 Journal of Biogeography ª 2015 John Wiley & Sons Ltd Biogeography of the relict butterfly Elkalyce cogina

(a) (b) f = 0.52 (c) f = 0.14 Chilades (12) 2 Edales (9) Echinargus (1) Cyclargus (7) 0.2 1.00 Hemiargus (5) Eldoradina (2) 0.75

Diversification rate Diversification Nabokovia (3) -0.08 0.50 Pseudolucia (46) Pseudochrysops (1) 0.25 Itylos (24) Paralycaeides (3) Marginal shift probability Plebulina (1) Icaricia (7) (d) 2.0 (e) 2.0 Rueckbeilia (2) Afarsia (9) 1.5 1.5 Kretania (17) Alpherakya (5) Glabroculus (2) 1.0 1.0 Aricia (15) Maurus (1) 0.5 0.5 Plebejidea (2) Net diversification rate Net diversification Eumedonia (3) rate Net diversification

Agriades (19) 0.0 0.0 Rimisia (1) 15 10 5 0 15 10 5 0 Cyaniris (2) (f) Time before present (Ma) (g) Time before present (Ma) Lysandra (15) 500 200

Polyommatus (183) 180

Neolysandra (6) 160 100 Pamiria (7) 140

Patricius (7) 50 120

Plebejus (40) 100

Freyeria (3) 80 10 Tongeia (16) 60 Talicada (3) 5

Species number 40

Cupido (8) Species number (log) 2 20 Cupido (11) 0 1 Elkalyce (1) 15 10 5 0 15 10 5 0 Time before present (Ma) Time before present (Ma)

Figure 3 Diversification for the Polyommatina + Everina clade. (a) bamm phylorate plot showing average net diversification rates along each branch of the Polyommatina + Everina clade. Warmer colours denote faster diversification rates (in lineages per Myr). Number of species per genera is indicated in parentheses. (b,c) The two shift configurations with the highest posterior probability. Circles denote locations of rate shifts (red, rate increase; blue, rate decrease) and are proportional to the overall marginal probability of a shift on the branch. Trajectories of diversification rate through time (RTT) for (d) the whole Polyommatina + Everina clade and (e) after excluding the hyperdiverse Palaearctic Polyommatina subclade. Colour density shading illustrates the relative probability of a rate at any point in time. (f) Lineages-through-time (LTT) plot for the Polyommatina + Everina clade (red) and LTT plots for 100 simulated trees (black). (g) Species richness of each terminal lineage (representing genera) against their estimated ages (Myr). The red dot corresponds to Elkalyce cogina. apparently related to South African taxa in the same genus were warmer than at present and a land bridge existed in and in Oraidium, and Zizula cyna – the only other species of Beringia, which undoubtedly facilitated this dispersal route Zizula occurring in , Southeast Asia and for presumably warm-adapted butterflies. The discovery that (Robbins & Duarte, 2006; de Jong & van Actherberg, 2007). E. cogina feeds on Fabaceae, as was presumably also the case Although debate about a Gondwanan vicariant or dispersal for the ancestors of the Neotropical Polyommatina and of origin has been generated for other higher level butterfly taxa the Icaricia + Plebulina clade (Vila et al., 2011), helps (e.g. de Jong, 2003; Hall et al., 2004; Braby et al., 2005; explain their parallel biogeographical histories. In addition, Kodandaramaiah & Wahlberg, 2007; Pena~ et al., 2010), most this result strengthens the hypothesis that Fabaceae was the cases in Lycaenidae are arguably too young to consider a ancestral host-plant for both Polyommatina and Everina Gondwanan vicariant-origin hypothesis. subtribes (see Fig. S7 in Vila et al., 2011). Within the Ever- Vila et al. (2011) proposed that at least five colonization ina, the genus Cupido includes two apparent sister species events from the Old World to the New World through Ber- in the Nearctic – Cupido comyntas and Cupido amyntula – ingia occurred within the Polyommatina, the sister subtribe that are estimated to have colonized the New World less to Everina. The ancestor of Elkalyce might represent another than 2 Ma, the estimated time of divergence between C. case that occurred approximately 10 Ma, at a similar time comyntas and the Palaearctic Cupido alcetas (Fig. 2). Overall, to the ancestor of the current Neotropical Polyommatina at least seven Old World-to-New World colonization events (c. 10.7 Ma) and the ancestor of the Icaricia + Plebulina have occurred within the Polyommatina + Everina clade clade (c. 9.3 Ma). During this period, climatic conditions (Fig. 4, Table 1).

Journal of Biogeography 7 ª 2015 John Wiley & Sons Ltd G. Talavera et al.

(a) (b) (c) (d) (e)

(h) a

Figure 4 (f) Representative specimen photographs (dorsal view) and maps c showing the current distribution of the seven lineages (a–g) of Polyommatinae that d f colonized the New World. (h) Correlation e between number of species and estimated time since colonization of New World b g (g) (Ma). The monotypic genus Elkalyce (in red) can be considered an outlier breaking up any correlation.

Nevertheless, our bamm results show that the decrease in the Table 1 Inferred timing of colonization (mean SD) and diversification rate of the Elkalyce lineage is only marginally approximate number of species and genera for New World significant. This could be due to the generally low species lineages of Polyommatinae. Values for non-inferred nodes were extracted from Vila et al. (2011). See also Fig. 4. richness of Everina and of the early divergent branches in Polyommatina. Moreover, other examples of old singletons Approximate exist in the Polyommatina, including the Neotropical Pseu- Date of number of Number of dochrysops bornoi [9.89 (7.97–11.88) Myr old]. Pseudochrysops Ancestor colonization (Ma) species genera is only found on Caribbean islands, and islands are well Neotropical 11.74 (9.74–13.86) 91 9 known for harbouring relict lineages. Elkalyce cogina is one Polyommatina of the few polyommatines that occur in eastern South Amer- Elkalyce 10.17 (7.69–12.86) 1 1 ica. Regardless of any factor that explains the low diversity in – – Icaricia Plebulina 8.49 (6.94 10.04) 8 2 Elkalyce, the Neotropical Polyommatina, which are most Plebejus 2.40 (1.50–4.70) 4 1 diverse in western South America, may exhibit high species Cupido 1.65 (1.02–2.45) 2 1 1.10 (0.15–2.40) 3 1 richness due to habitat heterogeneity and Andean orogenesis, group as found for other Neotropical butterflies (see Wahlberg & Agriades optilete 1.00 (0.49–2.00) 1 1 Freitas, 2007; Elias et al., 2009). A significant increase in the diversification rates in Poly- ommatina was detected by bamm for the hyperdiverse It is generally understood that older lineages have had Palaearctic genera. This clade is generally associated with more time to accumulate species (McPeek & Brown, 2007) mountainous habitats and arid conditions, similar to the and an exponential correlation between time and number of Polyommatina in South America, which also display high species is expected if we assume a constant diversification relative species numbers. Polyommatina and Everina are rate. The number of species versus time of colonization both typical of open vegetation, with very few species occur- approximates an exponential curve across the New World ring in forested habitats. This could in part explain why lin- lineages, with the exception of Elkalyce (Fig. 4). There is a eages such as Elkalyce or Pseudochrysops have not diversified: particularly strong contrast between the successful Neotropi- both are endemics of areas with a low diversity of open habi- cal Polyommatina and the monotypic Elkalyce which, despite tats, which could prevent diversification by vicariance in having similar ages, have resulted in c. 91 and 1 species in these two lineages. Therefore, even if the relict distribution the Neotropics respectively. We cannot be certain that no of Elkalyce suggests extinction in the intervening areas after undescribed species exist within Elkalyce; because there is no colonization and dispersal through the New World, the lim- apparent bias in the intensity of research carried out in the ited diversification observed for this lineage might also be a different lineages of South American polyommatines, how- product of a low speciation rate. Although Elkalyce main- ever, it seems unlikely that the documented contrast in rich- tained the apparently ancestral host-plant family (Fabaceae), ness is an artefact. This pattern shows that Elkalyce has New World Polyommatina genera including Pseudolucia, diversified much less than most extant New World lineages. Icaricia, Plebejus and Agriades displayed host-plant shifts to

8 Journal of Biogeography ª 2015 John Wiley & Sons Ltd Biogeography of the relict butterfly Elkalyce cogina new families of plants, which is a well-known driver of but- REFERENCES terfly diversification (Nylin & Wahlberg, 2008; Fordyce, 2010; Nylin et al., 2014). Balint, Z. & Johnson, K. (1996) Systematic position and Larval association with ants (myrmecophily) is another generic status of Lycaena cogina Schaus, 1902: an ende- factor that might increase diversification in Lycaenidae (Pel- mic Neotropical Lycaenopsina (Lepidoptera, Lycaenidae). lissier et al., 2012). This may be important when comparing Acta Zoologica Academiae Scientiarum Hungaricae, 41, the diverse and generally myrmecophilous Polyommatina 343–347. with the Everina (Cupido section in Fiedler, 1991; Benyamini, Benyamini, D. (1995) Synopsis of biological studies of the Chi- 1995), but it is not particularly relevant in explaining the lean Polyommatini (Lepidoptera, Lycaenidae), Vol. 52. uniqueness of Elkalyce, because its larvae apparently display a Reports of the Museum of Natural History, University of low degree of myrmecophily, similar to that observed in the Wisconsin, Stevens Point, 1–51. other Everina taxa. Braby, M.F., Trueman, J.W.H. & Eastwood, R. (2005) When and where did troidine butterflies (Lepidoptera: Papilion- idae) evolve? Phylogenetic and biogeographic evidence CONCLUSIONS suggests an origin in remnant Gondwana in the Late Cre- We provide the first molecular assessment of the taxon taceous. Invertebrate Systematics, 19, 113–143. Elkalyce cogina, a species of Polyommatini, endemic to Brower, A.V. (1994) Rapid morphological radiation and con- Brazil and neighbouring Argentina, of unclear systematic vergence among races of the butterfly Heliconius erato placement and biogeographical history. Our results show inferred from patterns of mitochondrial DNA evolution. that E. cogina belongs to the Everina subtribe, which Proceedings of the National Academy of Sciences USA, 91, agrees with one of the published hypotheses based on 6491–6495. morphology. The supported position of E. cogina as sister Brown, K.S., Jr (1993) Neotropical Lycaenidae: an overview. to the other Everina taxa studied and the estimated diver- of Lycaenidae (butterflies) (ed. by T.R. gence time of about 10 Ma are, however, unexpected. We New), pp. 45–61. IUCN, Gland, Switzerland. observed the larvae of E. cogina feeding on Desmodium Castresana, J. (2000) Selection of conserved blocks from uncinatum (Fabaceae), which suggests that the common multiple alignments for their use in phylogenetic analysis. ancestor of both the sister subtribes Everina and Molecular Biology and Evolution, 17, 540–552. Polyommatina fed on Fabaceae. A biogeographical history Darriba, D., Taboada, G.L., Doallo, R. & Posada, D. (2012) parallel to that of the Neotropical Polyommatina (Asia– jModelTest 2: more models, new heuristics and parallel Beringia–North America–South America) is proposed. The computing. Nature Methods, 9, 772. lack of diversification in Elkalyce is unique among lineages de Jong, R. (2003) Are there butterflies with Gondwanan of New World colonists. ancestry in the Australian region? Invertebrate Systematics, 17, 143–156. de Jong, R. (2007) Estimating time and space in the evolu- ACKNOWLEDGEMENTS tion of the Lepidoptera. Tijdschrift voor Entomologie, 150, We thank Ezequiel Nunez-Bustos~ and Alfred Moser for sup- 319–346. plying additional distribution data. Maice S. Franco collected de Jong, R. & van Actherberg, C. (2007) Global disjunctions the larva in Serra do Japi. Nicholas Matzke helped with bio- and flying . Biogeography, time, and place: distribu- geobears. Gabriel Miller reviewed the English language. Zolt tions, barriers and islands (ed. by W. Renema), Topics in Balint kindly provided photos of museum specimens of Elka- Geobiology, 29,5–44. Springer, Berlin. lyce cogina. Comments from Bob Robbins and two anony- Drummond, A.J., Ho, S.Y.W., Phillips, M.J. & Rambaut, A. mous referees greatly improved this manuscript. Funding (2006) Relaxed phylogenetics and dating with confidence. was provided by the Spanish Ministerio de Economıa y PLoS Biology, 4, e88. Competitividad (project CGL2013-48277-P). G.T. is sup- Drummond, A.J., Suchard, M.A., Xie, D. & Rambaut, A. ported by the grant BP-A00275 (AGAUR-Generalitat de Cat- (2012) Bayesian phylogenetics with BEAUti and the alunya), Marie Curie Actions FP7-PEOPLE-2013-IOF BEAST 1.7. Molecular Biology and Evolution, 29, 1969– (project 622716) and the grant 1.50.1617.2013 at Saint 1973. Petersburg State University; L.A.K. was supported by CAPES Elias, M., Joron, M., Willmott, K., Silva-Brand~ao, K.L., Kai- (3200-14-0); A.V.L.F. thanks ICMBio for research permits ser, V., Arias, C.F., Gomez Pinerez,~ G.L., Uribe, S., Brower, (SISBIO no. 10802-5), CNPq (fellowship 302585/2011-7 and A.V.Z., Freitas, A.V.L. & Jiggins, C.D. (2009) Out of the grant 564954/2010-1), RedeLep-SISBIOTA-Brasil/CNPq Andes: patterns of diversification in clearwing butterflies. (563332/2010-7), the National Science Foundation (DEB- Molecular Ecology, 18, 1716–1729. 1256742), BR-BoL (MCT/CNPq/FNDCT 50/2010) and Eliot, J.N. (1973) The higher classification of the Lycaenidae FAPESP (grant 2012/50260-6 and BIOTA-FAPESP Programs (Lepidoptera): a tentative arrangement. Bulletin of the Bri- 2011/50225-3 and 2013/50297-0). tish Museum of Natural History (Entomology), 28, 373–506.

Journal of Biogeography 9 ª 2015 John Wiley & Sons Ltd G. Talavera et al.

Farris, J.S. (1976) Phylogenetic classification of fossils with Paradis, E., Claude, J. & Strimmer, K. (2004) APE: analyses recent species. Systematic Biology, 25, 271–282. of phylogenetics and evolution in R language. Bioinformat- Fiedler, K. (1991) Systematic, evolutionary, and ecological ics, 22, 289–290. implications of myrmecophily within the Lycaenidae Pellissier, L., Litsios, G., Guisan, A. & Alvarez, N. (2012) (Insecta: Lepidoptera: ). Bonner Zoologische Molecular substitution rate increases in myrmecophilous Monographien, 31,1–210. lycaenid butterflies (Lepidoptera). Zoologica Scripta, 41, Fordyce, J.A. (2010) Host shifts and evolutionary radiations 651–658. of butterflies. Proceedings of the Royal Society B: Biological Pena,~ C., Nylin, S., Freitas, A.V.L. & Wahlberg, N. (2010) Sciences, 277, 3735–3743. Biogeographic history of the butterfly subtribe Euptychiina Guindon, S. & Gascuel, O. (2003) A simple, fast and accurate (Lepidoptera, Nymphalidae, Satyrinae). Zoologica Scripta, algorithm to estimate large phylogenies by maximum like- 39, 243–258. lihood. Systematic Biology, 52, 696–704. Plummer, M., Best, N., Cowles, K. & Vines, K. (2006) Guindon, S., Dufayard, J.-F., Lefort, V., Anisimova, M., Hor- CODA: convergence diagnosis and output analysis for dijk, W. & Gascuel, O. (2010) New algorithms and meth- MCMC. R News, 6,7–11. ods to estimate maximum-likelihood phylogenies: Quek, S.-P., Davies, S.J., Itino, T. & Pierce, N.E. (2004) assessing the performance of PhyML 3.0. Systematic Biol- Codiversification in an ant-plant mutualism: stem texture ogy, 59, 307–321. and the evolution of host use in Crematogaster (Formici- Hall, J.W., Robbins, R.K. & Harvey, D.J. (2004) Extinction dae: Myrmicinae) inhabitants of Macaranga (Euphor- and biogeography in the Caribbean: new evidence from a biaceae). Evolution, 58, 554–570. fossil riodinid butterfly in Dominican amber. Proceedings R Core Team (2014) R: a language and environment for sta- of the Royal Society B: Biological Sciences, 271, 797–801. tistical computing. R Foundation for Statistical Comput- Kodandaramaiah, U. & Wahlberg, N. (2007) Out-of-Africa ing. Vienna, Austria. Available at: http://www.r- origin and dispersal-mediated diversification of the butter- project.org/. fly genus Junonia (Nymphalidae: Nymphalinae). Journal of Rabosky, D.L., Santini, F., Eastman, J., Smith, S.A., Sid- Evolutionary Biology, 20, 2181–2191. lauskas, B., Chang, J. & Alfaro, M.E. (2013) Rates of speci- Landis, M.J., Matzke, N.J., Moore, B.R. & Huelsenbeck, J.P. ation and morphological evolution are correlated across (2013) Bayesian analysis of biogeography when the num- the largest vertebrate radiation. Nature Communications, 4, ber of areas is large. Systematic Biology, 62, 789–804. 1958. Matzke, N.J. (2013a) BioGeoBEARS: biogeography with Rabosky, D.L., Donnellan, S.C., Grundler, M. & Lovette, I.J. Bayesian (and likelihood) evolutionary analysis in R (2014a) Analysis and visualization of complex macroevolu- scripts. R package version 0.2.1. Available at: http:// tionary dynamics: an example from Australian scincid cran.r-project.org/package=BioGeoBEARS (accessed 10 lizards. Systematic Biology, 63, 610–627. March 2015) Rabosky, D.L., Grundler, M., Anderson, C., Title, P., Shi, J.J., Matzke, N.J. (2013b) Probabilistic historical biogeography: Brown, J.W., Huang, H. & Larson, J.G. (2014b) BAMM- new models for founder-event speciation, imperfect detec- tools: an R package for the analysis of evolutionary tion, and fossils allow improved accuracy and model-test- dynamics on phylogenetic trees. Methods in Ecology and ing. Frontiers of Biogeography, 5, 242–248. Evolution, 5, 701–707. Matzke, N.J. (2014) Model selection in historical biogeogra- Ree, R.H. & Smith, S.A. (2008) Maximum likelihood phy reveals that founder-event speciation is a crucial pro- inference of geographic range evolution by dispersal, cess in island clades. Systematic Biology, 63, 951–970. local extinction, and cladogenesis. Systematic Biology, McPeek, M.A. & Brown, J.M. (2007) Clade age and not 57,4–14. diversification rate explains species richness among animal Robbins, R.K. & Duarte, M. (2006) Systematic placement of taxa. The American Naturalist, 169, E97–E106. Lycaena cogina Schaus (Lepidoptera: Lycaenidae: Poly- Nabokov, V. (1944) Notes on the morphology of the genus ommatinae), a biogeographically disjunct New World spe- Lycæides (Lycænidæ, Lepidoptera). Psyche, 50, 104–138. cies. Proceedings of the Entomological Society of Washington, Nabokov, V. (1945) Notes on Neotropical Plebejinæ (Lycae- 108, 226–236. nidæ). Psyche, 51,1–61. Ronquist, F. (1997) Dispersal-vicariance analysis: a new Nylin, S. & Wahlberg, N. (2008) Does plasticity drive spe- approach to the quantification of historical biogeography. ciation? Host-plant shifts and diversification in nympha- Systematic Biology, 46, 195–203. line butterflies (Lepidoptera: Nymphalidae) during the Schaus, W. (1902) Descriptions of new American butterflies. tertiary. Biological Journal of the Linnean Society, 94, Proceedings of the United States National Museum, 24, 115–130. 383–460. Nylin, S., Slove, J. & Janz, N. (2014) Host plant utilization, Stadler, T. (2011) Simulating trees with a fixed number of host range oscillations and diversification in nymphalid extant species. Systematic Biology, 60, 676–684. butterflies: a phylogenetic investigation. Evolution, 68, Talavera, G. & Castresana, J. (2007) Improvement of phylo- 105–124. genies after removing divergent and ambiguously aligned

10 Journal of Biogeography ª 2015 John Wiley & Sons Ltd Biogeography of the relict butterfly Elkalyce cogina

blocks from protein sequence alignments. Systematic Biol- SUPPORTING INFORMATION ogy, 56, 564–577. Talavera, G., Lukhtanov, V.A., Pierce, N.E. & Vila, R. Additional Supporting Information may be found in the (2013) Establishing criteria for higher-level classification online version of this article: using molecular data: the systematics of Polyommatus Appendix S1 Supporting tables (Tables S1–S4) and figure blue butterflies (Lepidoptera, Lycaenidae). Cladistics, 29, (Fig. S1). 166–192. Vila, R., Bell, C.D., Macniven, R., Goldman-Huertas, B., BIOSKETCH Ree, R.H., Marshall, C.R., Balint, Z., Johnson, K., Benya- mini, D. & Pierce, N.E. (2011) Phylogeny and palaeoe- Members of the research team are actively engaged in the cology of Polyommatus blue butterflies show Beringia study of the biodiversity, biogeography and evolution of was a climate-regulated gateway to the New World. Pro- Lepidoptera, with specific interest in unravelling the histori- ceedings of the Royal Society B: Biological Sciences, 278, cal and present-day factors responsible for current species 2737–2744. distributions. Wahlberg, N. & Freitas, A.V.L. (2007) Colonization of and radiation in South America by butterflies in the subtribe Author contributions: G.T. and R.V. conceived the study; Phyciodina (Lepidoptera: Nymphalidae). Molecular Phylo- R.V., L.A.K. and A.V.L.F. collected the data; R.V. obtained genetics and Evolution, 44, 1257–1272. molecular data and G.T. performed phylogenetic and diversi- Wiemers, M., Stradomsky, B.V. & Vodolazhsky, D.I. (2010) fication analyses. G.T. led the writing. All authors con- A molecular phylogeny of Polyommatus s. str. and Plebic- tributed in the form of discussions and suggestions, and ula based on mitochondrial COI and nuclear ITS2 approved the final manuscript. sequences (Lepidoptera: Lycaenidae). European Journal of Entomology, 107, 325–336. Editor: Michelle Gaither

Journal of Biogeography 11 ª 2015 John Wiley & Sons Ltd