Molecular Phylogenetics and Evolution

Molecular Phylogenetics and Evolution 49 (2008) 92 101 Contents lists available at ScienceDirect Molecular Phylogenetics and Evolution journal homepage: www.elsevier.com/locate/ympev The genus Coleodactylus (Sphaerodactylinae, Gekkota) revisited: A molecular phylogenetic perspective Silvia Rodrigues Geurgas a, *, Miguel Trefaut Rodrigues a, Craig Moritz b a Departamento de Zoologia, Instituto de Biociências, Universidade de São Paulo, 05508-090, São Paulo, SP, Brazil b Museum of Vertebrate Zoology, 3101 Valley Life Sciences Building #3160, UC Berkeley, CA 94720, USA article info abstract Article history: Received 31 January 2008 Revised 28 May 2008 Accepted 30 May 2008 Available online 7 June 2008 Keywords: Coleodactylus Sphaerodactylinae Phylogeny Pleistocene refuges 16S RAG-1 c-mos Nucleotide sequence data from a mitochondrial gene (16S) and two nuclear genes (c-mos, RAG-1) were used to evaluate the monophyly of the genus Coleodactylus, to provide the first phylogenetic hypothesis of relationships among its species in a cladistic framework, and to estimate the relative timing of species divergences. Maximum Parsimony, Maximum Likelihood and Bayesian analyses of the combined data sets retrieved Coleodactylus as a monophyletic genus, although weakly supported. Species were recovered as two genetically and morphological distinct clades, with C. amazonicus populations forming the sister taxon to the meridionalis group (C. brachystoma, C. meridionalis, C. natalensis, and C. septentrionalis). Within this group, C. septentrionalis was placed as the sister taxon to a clade comprising the rest of the species, C. meridionalis was recovered as the sister species to C. brachystoma, and C. natalensis was found nested within C. meridionalis. Divergence time estimates based on penalized likelihood and Bayesian dating methods do not support the previous hypothesis based on the Quaternary rain forest fragmentation model proposed to explain the diversification of the genus. The basal cladogenic event between major lineages of Coleodactylus was estimated to have occurred in the late Cretaceous (72.6 ± 1.77 Mya), approximately at the same point in time than the other genera of Sphaerodactylinae diverged from each other. Within the meridionalis group, the split between C. septentrionalis and C. brachystoma + C. meridionalis was placed in the Eocene (46.4 ± 4.22 Mya), and the divergence between C. brachystoma and C. meridionalis was estimated to have occurred in the Oligocene (29.3 ± 4.33 Mya). Most intraspecific cladogenesis occurred through Miocene to Pliocene, and only for two conspecific samples and for C. natalensis could a Quaternary differentiation be assumed (1.9 ± 1.3 Mya). Ó 2008 Elsevier Inc. All rights reserved. 1. Introduction * Corresponding author. Fax: +55 11 3091 7553. E-mail address: sgeurgas@hotmail.com (S.R. Geurgas). The New World Sphaerodactylinae includes five genera of diurnal geckos found in forested areas in West Indies, Central and northern South America: Coleodactylus, Gonatodes, Lepidobepharis, Pseudogonatodes, and Sphaerodactylus (Underwood, 1954; Kluge, 1967, 1987, 1995; Gamble et al., 2008a). Monophyly of sphaerodactyls is well supported by both morphological (Noble, 1921; Kluge, 1987, 1995) and molecular data (Gamble et al., 2008a). Except for Gonatodes, genera present claws enclosed in an ungual sheath, one of the characters long recognized as diagnostic as diagnostic of the family (Noble, 1921; Kluge, 1967, 1995). Accounting for only five of the about 150 species described for the subfamily, Coleodactylus is the most widespread genus in Brazil, extending its distribution from eastern Amazonian Forest, through Cerrado and Caatinga in central Brazil to the Atlantic Forest. Three species are widely distributed: C. amazonicus is found in eastern and central Amazonian Forest, and through southern Venezuela to southern Guyana (Ávila-Pires, 1995); C. meridionalis is associated mainly with the Atlantic Forest, but also occurs in fragments of forests in Caatinga and Cerrado (Vanzolini, 1980; Freire, 1999; Colli et al., 2002), and C. brachystoma occurs in forested enclaves in the Cerrado of Central Brazil (Vanzolini, 1968a,b, 1970; Colli et al., 2002). The other two species have a more restricted distribution: C. septentrionalis occurs from eastern Venezuela to western Suriname and northern Roraima, (Ávila-Pires, 1995) and C. natalensis is confined to forested areas between dunes at the Parque Nacional das Dunas, Rio Grande do Norte (Freire, 1999). Although all species are restricted to the leaf litter (Vanzolini, 1980; Ávila-Pires, 1995), C. amazonicus is basically found in shaded forest environments (Vitt et al., 2005), whereas the other species can occur in habitats ranging from closed-canopy wet forests to more mesic open formations (Vanzolini, 1980; Ávila-Pires, 1995; Freire, 1999; Vitt et al., 2005). The most remarkable biogeographic characteristic of Coleodactylus is the broad disjunct distribution of C. meridionalis and C. septentrionalis. These species are inferred to be closely 1055-7903/$ - see front matter Ó 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2008.05.043

S.R. Geurgas et al. / Molecular Phylogenetics and Evolution 49 (2008) 92 101 93 related (Vanzolini, 1980), but are separated by the Amazon basin, which is occupied by the congener, C. amazonicus. The attempt to explain this pattern of distribution was fundamental for the development of the diversification model of South American biota based on recent and rapid cycles of forest expansion and retraction caused by climatic alterations proposed by Vanzolini and Williams (1970) for the Anolis nitens (formely Anolis chrysolepis; Ávila-Pires, 1995) species group (Vanzolini, 1980). Considering the Amazon basin as the putative center of Coleodactylus diversification and the limited dispersal capacity of these leaf litter geckos, the disjunct distribution was interpreted as evidence of speciation by geographical isolation caused by successive disruptions of a formerly continuous forest (Vanzolini, 1957, 1968b, 1970, 1980; Vanzolini and Williams, 1970). According to Vanzolini s (1957, 1968b, 1970, 1980) hypothesis, a first episode of forest fragmentation would have led to the differentiation of the widespread ancestral stock in C. brachystoma in a southern refuge and C. meridionalis in a northern refuge. The latter species expanded its range through a continuous forest across the Amazon basin and Atlantic Forest during the subsequent wetter period. A second fragmentation episode would have separated eastern and western populations of C. meridionalis, giving rise to C. septentrionalis and promoting the speciation of C. amazonicus in a non-specified refuge. Competitive exclusion with C. amazonicus (Vanzolini, 1968b, 1970) or ecological adaptations of C. meridionalis and C. septentrionalis to drier formations could have precluded these species from recolonizing the Amazon Basin (Vanzolini, 1980). Despite studies focused on intraspecific morphological variation (Vanzolini, 1957; Ávila-Pires, 1995; Freire, 1999), ecology (Ramos, 1981; Vitt et al., 2005), and karyotype description (Santos et al., 2003), the knowledge about Coleodactylus is extremely limited, and even the monophyly of the genus has yet to be established. Since the species were intuitively grouped based on the overall conservative morphology and structure and asymmetry of the ungual sheath (Vanzolini, 1957), the monophyly of Coleodactylus have been accepted, yet so far not rigorously tested. Although having an asymmetrical digit, C. amazonicus has four scales forming the ungual sheath instead of the five scales described for the other species of the genus. The biological meaning of this variation is unknown, but since the relevance of terminal toe pad scalation in delimiting sphaerodactyl genera has long been recognized (Noble, 1921; Parker, 1926), this issue deserves a further investigation. In addition, the diversification hypothesis proposed by Vanzolini (1957, 1968b, 1970, 1980) rests on some other premises that also remain untested, one of them concerning the assumed close relationship between species displaying the disjunction. No formal phylogenetic analysis had been done for the genus, and the affinity between C. meridionalis and C. septentrionalis was based entirely on external morphological similarities (Vanzolini, 1980). Therefore, phylogenetic relationships among species remain unknown, making any biogeographic interpretation speculative. Finally, for the diversification of Coleodactylus to be consistent with the scenario of recent speciation proposed, species should exhibit a divergence time frame roughly corresponding to the Pleistocene. A recent molecular phylogenetic study, however, indicated a late Cretaceous origin for the New World sphaerodactyl genera, and placed the divergence between C. brachystoma and C. septentrionalis during Oligocene (Gamble et al., 2008a), suggesting a much deeper history for Coleodactylus diversification. In order to improve the understanding about the evolutionary history of Coleodactylus, nucleotide sequence data from a mitochondrial gene, 16S rrna, and two nuclear genes, c-mos and RAG-1, were used to (1) evaluate the monophyly of the genus, (2) provide the first phylogenetic hypothesis of relationships among its species in a cladistic framework, and (3) estimate the relative timing of species divergences. The results were compared with the previous hypothesis based on the Quaternary rain forest fragmentation model proposed to explain the diversification of the genus. 1.1. Systematic background Barbour (1921), in his review of the genus Sphaerodactylus, noticed that the species described as S. meridionalis Boulenger, 1888, collected on the northeastern region of the Atlantic coast, and S. amazonicus Andersson, 1918, collected in the Amazon Basin near Manaus, although having an asymmetrical ungual sheath, did not present the supraciliary spine characteristic of all species of Sphaerodactylus, and suggested that they should not belong to the genus. Parker (1926) erected the genus Coleodactylus for S. meridionalis, based on the lack of the supraciliary spine, clavicle not perforated and differences in composition and asymmetry of the five scales enclosing the claws. Subsequently, Wettstein (1928) described a new species, C. zernyi, from the lower Tapajós River in Amazon basin, which was distinguishable from C. meridionalis by having keeled dorsal scales and one post-nasal, and from S. amazonicus by the number of post-nasals and absence of a pattern of longitudinal bands on the head. The differences in number, shape, and asymmetry of the ungual sheath scales in relation to that of C. meridionalis were described. Despite of the lack of the supraciliary spine, the justification for incorporing the new species into the genus Coleodactylus instead of in Sphaerodactylus was the absence of a second outer scale on the ungual sheath. In the only review of the genus Coleodactylus, Vanzolini (1957) synonymized S. amazonicus and C. zernyi within C. amazonicus and placed Homonota brachystoma Amaral, 1935, and S. pfrimeri Miranda-Ribeiro, 1937, in the synonymy of C. brachystoma. A new species, C. guimaraesi, was described from the upper Madeira River in Amazon Basin, differing from C. amazonicus in having smooth dorsal scales and presenting some differences on the ungual sheath, based on a scheme of putative fusions of different scales (Vanzolini, 1968a). Additionally, the author confirmed the disjunct distribution of over 2000 km between populations of C. meridionalis from northwestern Amazonia and northeastern Brazil, already related by Parker (1935). Based on general morphological features and geographical distribution in relation to the Amazon basin, Vanzolini (1957) identified three evolutionary branches in Coleodactylus, represented by C. amazonicus, C. meridionalis, and C. brachystoma + C. guimaraesi, in a pre-cladist scheme that can be considered the first phylogenetic hypotheses for the genus. No sister-group relationships among branches were explicitly indicated, and apparently all lineages were considered to be derived directly from the ancestral stock. A more careful evaluation led Vanzolini (1968a,b) to correct the description of dorsal scales of C. guimaraesi from smooth to keeled, and to reinterpret relationships among species in a pre-cladistic evaluation of shared plesiomorphic (smooth dorsal scales and less asymmetrical digits) and apomorfic (keeled dorsal scales and more asymmetrical digits) character states. The author mentioned (1968a) that Coleodactylus might be represented by two pairs of subspecies, C. meridionalis + C. brachystoma and C. amazonicus + C. guimaraesi, but admitting that the last two species could be synonyms (which was later formally stated by Ávila-Pires, 1995). No direct ancestor-descendant relationship between pairs of subspecies was stated, and the possibility of potential subspecies was never mentioned again.

94 S.R. Geurgas et al. / Molecular Phylogenetics and Evolution 49 (2008) 92 101 In 1980, the northwestern populations of C. meridionalis were raised to species status, C. septentrionalis, based on the lower number of ventral scales but mainly on the presence of dorsal paired dark bordered light spots (Vanzolini, 1980). Freire (1999) recognized a close relationship between C. natalensis and the former two species, but no explicit suggestion regarding its phylogenetic placement was given. C. natalensis (Freire, 1999), resembles C. septentrionalis in presenting a similar pattern of dorsal light spots, but differs from it by the number of ventral and mid-body scales, and is distinguished from C. meridionalis, its geographically closest species, by the color pattern, body size, lower number of post-rostrals and higher number of toe lamellae (Freire, 1999). 2. Materials and methods 2.1. Taxon sampling The five currently recognized species of Coleodactylus were represented by 41 individuals corresponding to 21 localities, chosen to cover as much of the distribution of the species as possible (Table Table 1 List of taxa used in this study Family Species Locality Voucher GenBank 16S c-mos RAG-1 Sphaerodactylidae Coleodactylus amazonicus Acajatuba, AM (1) MTR10278 DQ110455 EU435219 EU435174 MTR10280 DQ104109 EU435220 EU435175 Altér do Chão, PA (2) MTR09744 DQ110418 EU435221 EU435176 MTR09746 DQ104102 EU435222 EU435177 Apiaú, RR (3) MTR09818 EU435266 EU435223 EU435178 MTR09819 EU435267 EU435224 EU435179 igarapé Camaipí, AP (4) JM104 DQ110435 EU435225 EU435180 MTR6247 DQ104104 EU435226 EU435181 Rio Preto da Eva, AM (5) MTR09898 EU435264 EU435227 EU435182 MTR09899 EU435265 EU435228 EU435183 São José das Pombas, AM (6) MTR10182 DQ110485 EU435229 EU435184 MTR10199 DQ104110 EU435230 EU435185 Serra do Kukoinhokren, PA (7) MTR36308 DQ104103 EU435231 EU435186 C. brachystoma Serra da Mesa, GO (8) MTRCB200 DQ104113 EU435232 EU435187 MTR09756 DQ110556 EU435233 EU435188 Uruçuí-Una, PI (9) MTR5225 DQ110558 EU435234 EU435189 MTR1701 DQ104116 EU435235 EU435190 Paranã, TO (10) MTR4318 DQ104115 EU435236 EU435191 MTR4129 DQ110563 EU435237 EU435192 Serra do Amolar, MS (11) IAH020 EU435268 EU435238 EU435193 IAH637 EU435269 EU435239 EU435194 C. meridionalis Carolina, MA (12) ESTR0973 EU435274 EU435240 EU435195 ESTR0112 DQ104126 EU435241 EU435196 Central, BA (13) MTR10360 EU435272 EU435242 EU435197 MTR09882 EU435273 EU435243 EU435198 Mamanguape, PB (14) MTR09762 DQ104118 EU435244 EU435199 MTR09768 DQ110493 EU435245 EU435200 Murici, AL (15) MTR10368 EU435270 EU435246 EU435201 MTR10369 EU435271 EU435247 EU435202 Pacoti, CE (16) MTR094 DQ110491 EU435248 EU435203 MTR4538 DQ104117 EU435249 EU435204 Una, BA (17) MD1721 DQ104121 EU435250 EU435205 MD2613 DQ110505 EU435251 EU435206 C. natalensis Natal, RN (18) MTR09906 DQ104127 EU435252 EU435207 MTR09907 DQ110564 EU435253 EU435208 C. septentrionalis Boa Vista, RR (19) MTR09795 DQ104131 EU435254 EU435209 MTR09796 DQ110548 EU435255 EU435210 Maracá Island, RR (20) MTR09782 DQ110542 EU435256 EU435211 MTR09789 DQ104130 EU435257 EU435212 Maú River, RR (21) MTR09808 DQ104129 EU435258 EU435213 MTR09809 DQ110516 EU435259 EU435214 Gonatodes humeralis Aripuanã, MT LG1177 EU435278 EU435263 EU435218 Lepidoblepharis xanthostigma Costa Rica no voucher EU435277 EU435262 EU435217 Pseudogonatodes guianensis Rio Preto da Eva, AM MTR09893 EU435275 EU435260 EU435215 MTR09894 EU435276 EU435261 EU435216 Sphaerodactylus leucaster X86056 S. shrevei AY662570 AY662623 Teratoscincus keyserlingii AY753545 T. przewalskii AY662569 AY662624 Gekkonidae Gekko gecko AY282753 AY172929 AY662625 Eublepharidae Eublepharis macularius AB028762 AF039470 E. turcmenicus AY662622 Diplodactylidae Pseudothecadactylus lindneri AF215247 AF090846 AY662626 Locality numbers correspond to those in Figs. 1 3. Brazilian States abbreviations (under Localitiy ) are as follows: AM, Amazonas; AL, Alagoas; BA, Bahia; CE, Ceará; GO, Goiás; MA, Maranhão; MT, Mato Grosso; MS, Mato Grosso do Sul; PA. Pará; PB, Pernambuco; PI, Piauí; RN, Rio Grande do Norte; RR, Roraima; TO, Tocantins.

S.R. Geurgas et al. / Molecular Phylogenetics and Evolution 49 (2008) 92 101 95 Savannah 20 1 3 6 19 21 Amazon Forest 11 1, Fig. 1). Ingroup taxa also included members of Gonatodes, Pseudogonatodes, Lepidoblepharis, and Sphaerodactylus, so that the monophyly of Coleodactylus in relation to the New World sphaerodactyls could be assessed. In the absence of the closest genera of Sphaerodactylini (Gamble et al., 2008a), monophyly of the group was established by including Gekko and Teratoscincus as outgroup taxa. Phylogenies were rooted with members of Eublepharinae and Diplodactylinae, in accordance with previous molecular phylogenetic analyses (Townsend et al., 2004; Gamble et al., 2008a). Voucher specimens from which sequence data were obtained in this study are deposited at the Museu de Zoologia, Universidade de São Paulo (MZUSP) and Coleção Zoológica do Departamento de Biologia e Zoologia da Universidade Federal de Mato Grosso (UFMT), Brazil. 2.2. Laboratory procedures 5 500 km 2 4 7 Cerrado Caatinga Total genomic DNA was extracted from liver or tail tissues, stored either frozen or ethanol fixed, by the standard proteinase K protocol (Sambrook et al., 1989). Approximately 500 bp of 16S, 572 bp of RAG-1 and 528 bp of c-mos genes were amplified and sequenced in both directions with the primers and conditions presented in Table 2. All PCR products were enzymatically purified 8 12 10 9 13 C. amazonicus C. brachystoma C. meridionalis C. natalensis C. septentrionalis 16 17 18 14 15 Atlantic Forest Fig. 1. Sampling localities of Coleodactylus. Locality numbers correspond to Table 1, Figs. 2 and 3. with Exonuclease I and Shrimp Alkaline Phosphatase (USB or Fermentas). Automated sequencing was performed using BigDye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems), followed by analysis on ABI Prism 310, 3700 or 3170 Genetic Analyzer Sequencers (Applied Biosystems) according to the manufacturer s instructions. Sequences were edited in Sequence Navigator (PE Applied Biosystems) or Sequencher v. 4.1.2 (Gene Codes Corporation) and initially aligned using the default parameters of ClustalW (Thompson et al., 1994). Primary homology between bases (sensu de Pinna, 1991) of the 16S gene were then hypothesized by comparison with the secondary structure model proposed for mammals (Burk et al., 2002). The absence of well-conserved motifs into the length-variable loops regions between stems 40/41 and 42/45 prevented the establishment of provisional homology between bases or positions, and these segments were excluded from analyses. Indels of nuclear genes were inserted based on conservation of the amino acid reading frame, and those sharing 5 0 and 3 0 termini could be confidentially considered homologous (Simmons and Ochoterena, 2000). Sequences of the nuclear gene c-mos obtained in this study have been combined with previously published data, and only 369 bp were used in the phylogenetic analyses. GenBank accession numbers for all sequences are indicated in Table 1. 2.3. Phylogenetic analyses Prior to the phylogenetic analyses, each gene was separately tested for homogeneity of base composition among taxa using the base frequencies option implemented in PAUP 4.0b10 (Swofford, 2003), in order to avoid the potential effects that nucleotide composition differences among sequences could cause in the resulting tree topologies and nodal support recovered (e.g., Harris, 2003; Gruber et al., 2007). The effect of multiple substitutions in each dataset was also evaluated by plotting uncorrected codon-based (RAG-1 and c-mos) and total (16S) transition and transversion distances against the corresponding corrected pairwise distances using the appropriate model of evolution identified by MrModeltest v.2.2 (Nylander, 2004). In addition, congruence between different gene partitions was tested between all pairwise combinations of partitions using the incongruence length difference test (ILD; Farris et al., 1994), with the null distributions generated by 1000 replications, under a heuristic search with 20 random addition sequences per replicate. Although a high probability of type I error had been detected in different simulations (Dolphin et al., 2000; Barker and Lutzoni, 2002; Darlu and Lecoin- Table 2 List of primer sequences used in this study Gene Primer Sequence (5 0-3 0 ) PCR conditions 16S 16S F.1 a TGTTTACCAAAAACATAGCCTTTAGC 94 C (40 s), 45 to 51 C (40 s), 72 C (40 s) 35 16SF.st31 0d AGGTAACGCCTGCCCAGTGA 94 C (40 s), 50 C (40 s), 72 C (40 s) 35 16S R.0 a TAGATAGAAACCGACCTGGATT c-mos LSCH1 b CTCTGGKGGCTTTGGKKCTGTSTACAAGG 94 C (40 s), 50 to 55 C (40 s), 72 C (40 s) 35 LSCH2 b GGTGATGGCAAARGAGTAGATGTCTGC RAG-1 F94 c TGGAARTTCAARCTGTTCAAAGT 94 C (40 s), 49 to 51 C (40 s), 72 C (40 s) 35 F104 c CAAAGTGAGATCNCTTGAAAA 94 C (40 s), 49 to 51 C (40 s), 72 C (40 s) 35 R387 c GTNTCATCATCTACTGGTCCA R522 d AAATTAGTTGGATGGATTGTGTCCA F2568a c GGATGAATGGRAATTTTGCCAGA 94 C (40 s), 48 to 52 C (40 s), 72 C (40 s) 35 F2568 b c GGATGAATGGAAAYTTTGCTMGA 94 C (40 s), 48 to 52 C (40 s), 72 C (40 s) 35 R2876 c TTTGTTCCCAGATTCATTTCC R2901 d TTTATTTCCGGACTCATTTCC PCR cycles included a initial denaturation step of 94 C for 5 min, and a final elongation step of 72 C for 7 min. a Whiting et al., 2003. b Godinho et al., 2005. c Townsend et al., 2004. d This study.

96 S.R. Geurgas et al. / Molecular Phylogenetics and Evolution 49 (2008) 92 101 tre, 2002; Dowton and Austin, 2002), the ILD is the best understood test of phylogenetic incongruence of all tests available and can be considered as a conservative first test for identifying potential incongruence among data partitions (Hipp et al., 2004; Planet, 2006). In the absence of conflicting results, data were combined to perform phylogenetic analyses using Maximum Parsimony (MP) and Maximum Likelihood (ML) implemented in PAUP 4.0b10 (Swofford, 2003) and Bayesian analysis (BA) implemented in MrBayes 3.0b4 (Ronquist and Huelsenbeck, 2003). MP searches were conducted with equal character weighting and gaps treated as missing data under the heuristic search with tree bisection reconnection (TBR) branch swapping and 1000 random-addition sequence replicates. To investigate the contribution of the indels to the phylogenetic reconstruction, an additional parsimony analysis was performed coding gaps of the nuclear genes as a presence/absence characters matrix (Simmons and Ochoterena, 2000). Nodal support was estimated using non-parametric bootstrapping (Felsenstein, 1985) with 10,000 replicates with five random addition sequence replicates each and TBR branch swapping. Consensus trees were obtained following the 50% majority rule, and nodes with bootstrap P70% were considered strongly supported. The ML tree was obtained using a heuristic search, with five random-addition sequence replicates and TBR branch swapping using GTR + I + C nucleotide substitution model selected by MrModeltest v2.2 (Nylander, 2004). Nodal support was estimated using non-parametric bootstrapping (Felsenstein, 1985) with 100 replicates with one random addition sequence replicate and TBR branch swapping. For Bayesian analyses, the best-fit model of nucleotide substitution for each data partition was selected using the hierarchical likelihood ratio (hlrt) criterion implemented in MrModeltest v.2.2 (Nylander, 2004). Because genes sampled evolve at different rates, and base frequencies and substitution rates may vary with codon positions, three different partition schemes were used for phylogenetic analyses: (1) one-partition, using a single model for the whole dataset (GTR + I + C); (2) four-partitions, using a separate model for each gene, considering the two regions of RAG-1 separately; and (3) ten-partitions, using a different model for each codon position of the three protein coding fragments and for the mitochondrial gene (Table 3). Two independent Bayesian analyses were performed for each partition, with a random starting tree, four incrementally heated Markov chains, and 4,000,000 generations, with trees sampled every 100 generations to estimate likelihood and sequence evolution parameters. Stationarity for each run was detected by plotting the likelihood scores of the trees against generation time, and the topology, posterior probability values, and branch lengths inferences were estimated after discarding 0.25% of the initial trees of each run as burn-in samples. Nodes with posterior probability P95% on a 50% majority rule consensus tree from both runs were considered significant support for a given clade. 2.4. Comparing alternative topologies A recent molecular phylogeny based on five nuclear genes (Gamble et al., 2008a) had suggested that Sphaerodactylus, Pseudogonatodes, and Coleodactylus constitute a monophyletic sister group to Gonatodes + Lepidoblepharis. In that study, C. brachystoma + C. septentrionalis were placed as the sister genus of Pseudogonatodes, in agreement with the relationship based on morphology proposed by Kluge (1995). These studies, however, were not designed to test the monophyly of Coleodactylus, and the genus was implicitly assumed to be monophyletic. In order to evaluate the previous hypothesis relative to the phylogenetic position of Coleodactylus within Sphaerodactylinii against the topology obtained here and to explore the possibility of nonmonophyly of the genus, two different approaches were used to test the following alternative topologies: (H1) Coleodactylus constrained to be a monophyletic sister genus of Pseudogonadotes; and (H2) C. brachystoma + C. meridionalis + C. natalensis + C. septentrionalis constrained to be the sister group of Pseudogonadotes and C. amazonicus as the sister clade of all Sphaerodactilini. In the first approach, likelihood scores of unconstrained and constrained topologies were compared using the non-parametric Shimodaira Hasegawa test (Shimodaira and Hasegawa, 1999) implemented in PAUP 4.0b10 (Swofford, 2003), using RELL bootstrapping (1000 replicates) and the same likelihood parameters used in the ML analyses. In the second approach, the presence of the alternative topologies was detected within the set of topologies contained in the 95% credible set of Bayesian trees from both runs of the ten-partitions scheme, sampled after burn-in using the option filter in PAUP 4.0b10 (Swofford, 2003). Alternative topologies were considered statistically reject if they were absent in credible set of trees (Fessler and Westneat, 2007; Zaldivar-Riverón et al., 2007). 2.5. Divergence times The topology obtained from the Bayesian analyses under the ten-partition scheme was used to estimate the branch lengths under a GTR + I + C model of evolution in a maximum likelihood approach (PAUP 4.0b10), and the assumption of rate constancy of DNA substitution through time among taxon was tested by comparing the log-likelihood scores from trees constructed with and without a molecular clock constraint (Felsenstein, 1981). As the Table 3 Summary of character variation for nuclear protein coding genes and mitochondrial ribosomal gene used in this study Gene Characters Nucleotide Composition (%) hlrt Model Total V PI A C G T { 2 P RAG-1 (1) 281 152 78 0.37 0.20 0.20 0.23 14.06 1.00 HKY 1 st positon 94 43 28 0.33 0.23 0.29 0.15 12.69 1.00 HKY 2 nd position 94 44 22 0.43 0.19 0.15 0.23 10.28 1.00 HKY 3 rd position 93 62 29 0.32 0.19 0.16 0.33 12.05 1.00 HKY RAG-1 (2) 302 77 49 0.34 0.20 0.22 0.24 8.15 1.00 K80+C 1 st position 100 16 10 0.29 0.21 0.29 0.21 1.57 1.00 K80 2 nd position 101 8 3 0.36 0.20 0.19 0.25 1.62 1.00 JC 3 rd position 101 53 36 0.36 0.19 0.19 0.26 18.50 1.00 K80+C c-mos 369 136 90 0.28 0.21 0.23 0.28 9.34 1.00 HKY+I 1 st position 123 39 24 0.29 0.21 0.30 0.20 5.65 1.00 K80+ C 2 nd position 123 28 17 0.29 0.20 0.21 0.30 3.94 1.00 SYM+C 3 rd position 123 69 49 0.25 0.21 0.18 0.36 16.12 1.00 HKY 16S 371 127 93 0.31 0.25 0.22 0.22 18.24 1.00 GTR+I+C The best-fit model of nucleotide evolution used in Bayesian analyses under four-partition and ten-partition schemes are indicated. V: number of variable sites; PI: number of parsimoniously informative sites.

S.R. Geurgas et al. / Molecular Phylogenetics and Evolution 49 (2008) 92 101 97 null hypothesis of clocklike evolution was rejected by a Chi square likelihood ratio test ( Ln H0 = 6851.68426; Ln H1 = 6882.58311), divergence times were estimated using the penalized likelihood method (Sanderson, 2002) with the TN algorithm and optimal value of smoothing determined by cross-validation as implemented in r8s (Sanderson, 2003). Outgroups were excluded using the prune command. Since the fossil record for the New World sphaerodactyls is scarce and restricted to some fossils of Sphaerodactylus (Böhme, 1984; Grimaldi, 1995), two secondary calibration points derived from analysis of independent molecular data were used to estimate absolute divergence times within Coleodactylus. Based on the estimates of Gamble et al. (2008a), the origin of the New World sphaerodactyl lineages was fixed to 75.5 Mya, and the divergence of Gonatodes and Lepidoblepharis at 68.2 Mya. Standard deviations of divergence times were estimated using the profile command in r8s (Sanderson and Doyle 2001) from a subset of 718 phylograms with identical topology screened among the last 1,000,000 trees from the ten-partition Bayesian analysis. 3. Results 3.1. Nuclear indels In addition to the 4-codon deletion already described for the 5 region of the gene RAG-1 in the Eublepharinae/Sphaerodactylinae/ Gekkoninae group (Townsend et al., 2004), a total of three additional deletions were evident in sequences of C. brachystoma, C. meridionalis, C. natalensis and C. septentrionalis. Two deletions, corresponding to six and 18 nucleotides, were shared by these four species, and a third deletion, corresponding to six nucleotides, was observed only in C. septentrionalis. A single codon deletion was shared by C. brachystoma, C. meridionalis, and C. natalensis at the 3 0 end of c-mos gene. The indel events in both genes were also detected by Gamble et al. (2008a,b) for C. brachystoma, but were not reported for C. septentrionalis. 3.2. Properties of the dataset The final dataset consisted of a total of 1323 base pairs (bp), being 371 nucleotides from 16S gene, 281 and 302 from the 5 0 one-third and 3 0 two-thirds of RAG-1, respectively, and 369 from c-mos. Of these, 493 were variable and 311 were parsimoniously informative (Table 3). No evidence for differences in base composition among taxa was observed for all genes, although the average nucleotide frequencies among sites showed some variation. The mitochondrial gene had a slight bias in adenine frequency, and the nuclear genes showed a greater content of adenine and thymine. Only the first positions of codons were characterized by higher frequencies of adenine and guanine, in agreement with that reported for Squamata (Harris, 2003; Townsend et al., 2004). Among the nuclear genes, the 5 0 region of RAG-1 was the most variable data, with 54% of variable sites, of which 51% were parsimony-informative. The 3 0 region of RAG-1 and c-mos had similar percentages of variable (25% and 37%, respectively) and informative sites (64% and 66%), indicating similar overall rates of evolution. The relationship between uncorrected p-distances and corrected distances appeared to be linear for the 16S gene and all codon positions of the nuclear genes, suggesting that substitutions have not reached saturation. 3.3. Phylogenetic analyses The partition homogeneity test was not significant for any pairwise combination of datasets (16S vs. RAG-1, 5 0 region: P = 0.19; 16S vs. RAG-1, 3 0 region: P = 0.79; 16S vs. c-mos: P = 0.21; RAG-1, 5 0 region vs. RAG-1, 3 0 region: P = 0.87; RAG-1, 5 0 region vs. c-mos: P = 0.30; RAG-1, 3 0 region vs. c-mos: P = 0.64), and all analyses were based on the combined data. The equally weighted parsimony analysis of the concatenated sequences yielded six equally parsimonious trees, the consensus of which contained 87% well-supported nodes (Fig. 2; TL = 959 steps, CI = 0.64, RI = 0.78). Inclusion of nuclear indels as a presence/absence matrix in the phylogenetic analysis also yielded six equally parsimonious trees (not shown, TL = 963 steps, CI = 0.65, RI = 0.79). Although not improving the topology or bootstrap values of the consensus tree recovered, the congruent phylogenetic signal between codon indels and base substitutions gives additional support to the resulting phylogenetic hypothesis. The ML analysis produced a single most likely tree ( Ln = 6851.33963), with 82% well-supported nodes (Fig. 2). The consensus trees from concatenated, four-partioned and ten-partioned Bayesian analyses were derived from 79802 sample trees each, after discarding the first 10,000 generations from each analysis as burn-in. In general, topology and estimated nodal posterior probabilities from the partitioned analyses were very similar to those derived from the unpartitioned analyses, with 72% well-supported nodes. The discrepancy observed among Bayesian trees involved one weakly supported clade recovered only by the ten-partition scheme (Fig. 2). Pseudothecadactylus Eublepharis Gekko Teratoscincus 0.51/0.96 0.90 Gonatodes Lepidoblepharis * Pseudogonatodes 0.51 71/ 97/ 0.98 Sphaerodactylus C. amazonicus 4 C. amazonicus 2 86/ 92/ 1.0 C. amazonicus 7 C. amazonicus 6 81/ 82/ 1.0 C. amazonicus 3 C. amazonicus 1 79/ 62/ 0.75 C. amazonicus 5 C. brachystoma 10 51/ 50/ 0.51 c-mos indel (3 pb) C. brachystoma 9 C. brachystoma 8 69/ 79/ 1.0 C. brachystoma 11 C. meridionalis 17 C. meridionalis 15 C. meridionalis 16 C. meridionalis 14 RAG-1 85/ 65/ 0.97 C. meridionalis 12 indel C. natalensis 18 RAG-1 indels (6 pb) 62/ 66/ 0.57 C. meridionalis 13 (6 and 18 bp) C. septentrionalis 21 C. septentrionalis 20 0.1 substitutions/site C. septentrionalis 19 Fig. 2. Bayesian tree topology obtained from the molecular data set combined (16S, c-mos, RAG-1). The 50% majority-rule consensus phylogram and posterior probabilities were estimated from 79802 trees derived from analyses under the tenpartition model (see Table 3 for model definition). Stippled lines indicate branches not recovered in the 50% majority-rule consensus of six equally maximum parsimonious trees (length: 959; consistency index: 0.64, retention index: 0.78), the triangle indicates branch recovered by the ML and Bayesian analysis, and the asterisk indicates branch not recovered in ML tree and Bayesian analysis using onepartition and four-partition schemes. Nodes labelled by open circles were highly supported by all three methods of phylogenetic inference (bootstrap P 95% and posterior probability P0.99), The values assigned to the internodes indicate MP bootstrap, ML boostrap and posterior probabilities values, respectively.

98 S.R. Geurgas et al. / Molecular Phylogenetics and Evolution 49 (2008) 92 101 The topologies recovered by MP, ML and BA were highly congruent, and all resolved nodes that received moderate to high bootstrap support (BS P 70%) had also received high posterior probabilities (PP P 0.95). Consensus trees differed from each other only in the degree of resolution of relationships among sphaerodactylini genera. Maximum parsimony analysis identified these relationships as an unresolved polytomy, and maximum likelihood analysis supported only a sister relationship between Gonatodes and Lepidoblepharis. Bayesian trees favored, even though weakly supported, a sister group relationship between Coleodactylus and a clade that includes all other genera of Sphaerodactylini. Within this group, Gonatodes was weakly (concatenated and four-partitioned schemes) to strongly (ten-partitioned scheme) supported as the sister taxon of Lepidoblepharis, with this clade forming a trichotomy with Pseudogonatodes and Sphaerodactylus (concatenated and four-partitioned schemes) or weakly supported as the sister group to Pseudogonatodes + Sphaerodactylus (ten-partitioned scheme). Coleodactylus was recovered as a monophyletic genus, although weakly supported regardless of reconstruction method used. C. amazonicus was placed as a highly divergent sister taxon of a large clade comprising C. brachystoma + C. meridionalis + C. natalensis + C. septentrionalis, hereafter called the meridionalis group. Within the meridionalis group, C. meridionalis is more closely related to C. brachystoma than to C. septentrionalis, and C. natalensis was recovered nested in C. meridionalis, rendering to this last species paraphyletic. All other species were strongly supported as monophyletic, and with the exception of C. natalensis, were themselves composed of highly distinctive clades, which can be structured to geographical subgroups into some extent (Fig. 1). The most evident case is the subdivision between samples of C. amazonicus from the western and eastern Amazonian Forest, a pattern already described for other vertebrate groups (e.g., da Silva and Patton, 1993; Ávila-Pires, 1995; Symula et al., 2003; Gamble et al., 2008b). A geographical structuring seems possible for C. meridionalis populations, for which the geographical distributions of successive branches of the tree are roughly placed increasingly northwards along the Atlantic coast. Interestingly, C. natalensis groups with C. meridionalis populations from Caatinga (Central) and Cerrado (Carolina) in Central Brazil to the closer ones of coastal region. C. septentrionalis also showed a subdivision among populations of northern vs. southern regions of Roraima state. The exception is C. brachystoma, for which no obvious geographical trend could be detected for the significantly differentiated clades of populations. 3.4. Comparison of alternative topologies The relationships among Sphaerodactylini genera presented in Fig. 2 were also retrieved by Gamble et al. (2008a), excepting the position of Coleodactylus, which they recovered as the sister genus of Pseudogonatodes. Among the topologies tested, the unconstrained one (H0) yielded the best tree ( Ln H0 = 6851.33963), in which Coleodatylus was recovered in 42% of the ML bootstrap replicates as the sister group of all Sphaerodactylini. The alternative topology placing Coleodactylus as a monophyletic genus sister to Pseudogonatodes ( Ln H1 = 6853.24352) was recovered in 28% of the bootstrap replicates, and the topology considering only the meridionalis group as the sister group to Pseudogonatodes ( Ln H2 = 6855.12393) was not detected in any of the 100 bootstrap replicates. Nevertheless, alternative topologies were statistically indistinguishable from one another at P 6 0.05 in the Shimodaira Hasegawa test. Similarly, alternative topologies could not be rejected by the Bayesian approach since they were observed within the combined 95% credible set of both runs derived from the tenpartition scheme. From a total of 75718 trees, the unconstrained topology (H0) corresponded to 21,606 trees (28.53%), alternative topology H1 corresponded to 41 trees (0.05%), and alternative topology H2 corresponded to 587 trees (0.77%). 3.5. Divergence Times The result of the molecular dating analysis is shown in Fig. 3. Assuming the New World sphaerodactyls to have last shared the most recent common ancestor at 75.5 Mya and the divergence of Gonatodes and Lepidoblepharis to have occurred at 68.2 Mya (Gamble et al., 2008a), the basal cladogenic event between major lineages of Coleodactylus was estimated to have occurred in the late Cretaceous (72.6 ± 1.77 Mya), approximately at the same point in Gonatodes Lepidoblepharis Pseudogonatode Sphaerodactylus C. amazonicus 2 C. amazonicus 7 C. amazonicus 4 C. amazonicus 3 C. amazonicus 5 C. amazonicus 1 C. amazonicus 6 C. brachystoma 8 C. brachystoma 11 C. brachystoma 9 C. brachystoma 10 C. meridionalis 16 C. meridionalis 14 C. natalensis 18 C. meridionalis 13 C. meridionalis 12 C. meridionalis 15 C. meridionalis 17 C. septentrionalis 20 C. septentrionalis 19 C. septentrionalis 21 Mya 13 12 16 15 14 10 8 9 1.9 ± 1.3 (0.3-7.3) 11 7 1.8 Pleistocene 5.3 Pliocene Miocene Node Age SD Min Max 1 71.7 1.1 68.5 74.6 2 69.4 1.6 65.9 72.6 3 28.4 3.7 19.4 40.1 4 23.1 5.2 8.7 38.6 5 22.1 4.2 11.5 37.9 6 21.4 4.1 13.0 37.8 7 9.9 2.9 4.1 23.9 8 9.7 2.6 4.4 19.3 9 8.9 2.7 3.7 19.4 10 8.5 2.6 2.5 18.0 11 7.9 2.5 3.0 18.4 12 4.1 1.8 0.9 13.0 13 3.7 1.4 0.9 11.3 14 3.5 1.9 0.5 12.1 15 2.9 1.6 0.7 12.5 3 6 5 29.3 ± 4.3 (17.8-44.4) 4 23.0 33.9 Oligocene Eocene 46.4 ± 4.2 (31.8-58.0) 55.8 65.5 Paleocene 68.2 2 1 75.5 72.6 ± 1.8 (64.2-75.2) Late Cretaceous Fig. 3. Chronogram based on penalized likelihood transformation of the ten-partition Bayesian consensus tree (Fig. 2). Divergence time estimates (in million years) and node profile information were obtained from the r8s molecular dating analyses of a subset of 718 phylograms with identical topology screened among the the last 1000000 trees from the ten-partition Bayesian analysis. Black circles represent fixed age nodes, and open circles represent speciation events in Coleodactylus. Locality numbers correspond to Table 1 and Figs. 1 and 2.

S.R. Geurgas et al. / Molecular Phylogenetics and Evolution 49 (2008) 92 101 99 time than the other genera diverged from each other. Within the meridionalis group, the split between C. septentrionalis and C. brachystoma + C. meridionalis was placed in the Eocene (46.4 ± 4.22 Mya), and the divergence between C. brachystoma and C. meridionalis was estimated to have occurred in the Oligocene (29.3 ± 4.33 Mya). Most intraspecific cladogenesis occurred through Miocene to Pliocene, and only for two conspecific samples and for C. natalensis could a Quaternary differentiation be assumed (1.9 ± 1.3 Mya). 4. Discussion 4.1. Phylogeny of Coleodactylus The results of phylogenetic analyses of sequence data from one mitochondrial and two nuclear genes presented here retrieved Coleodactylus as a monophyletic group composed of two genetically distinct and well supported clades, one represented by C. amazonicus populations, and one comprising C. brachystoma, C. meridionalis, C. natalensis and C. septentrionalis, referred here as the meridionalis group (Fig. 2). These clades correlate with the variation of the morphological characteristics concerning to dorsal scales and ungual sheath included by Vanzolini (1957) in the original diagnosis of the genus (Parker, 1935) to accommodate C. amazonicus. Species of the meridionalis group are characterized by having claws enclosed by an ungual sheath composed of five asymmetrical scales and smooth dorsal scales, whereas C. amazonicus presents four asymmetrical scales in the ungual sheath and keeled dorsal scales. These differences were used by Vanzolini (1957, 1968b, 1970, 1980) as an evidence to recognize C. amazonicus as a separate branch derived directly from the ancestral stock of Coleodactylus, and to suggest a close relationship among species of the meridionalis group. Within this group, the more explicit statement was the suggestion of Vanzolini (1980) that C. septentrionalis would have originated from C. meridionalis populations from the northwestern Amazonia, isolated after an episode of forest contraction. The molecular analysis contradict this hypothesis and placed C. septentrionalis as the sister taxon to a clade comprising the rest of the meridionalis group. C. meridionalis was recovered as the sister species to the parapatrically distributed C. brachystoma, and C. natalensis was found nested within C. meridionalis. In addition to the morphological autapomorphies, the major clades of Coleodactylus can also be distinguished by the two deletions of 18 and 6 pb in the RAG-1 gene shared by species from the meridionalis group. Within the group, phylogenetic relationships among species were corroborated by two additional indels: a 3 pb deletion in c-mos is shared by C. brachystoma, C. meridionalis, and C. natalensis, whereas a 6 pb deletion in RAG-1 is observed only in C. septentrionalis. Indels in protein-coding DNA sequences are considered rare (Rokas and Holland, 2000), and the value of these mutational events as independent phylogenetic markers to diagnose clades and to resolve phylogenetic relationships have been demonstrated for various groups of organisms (e.g., Venkatesh et al, 2001; Vidal and Hedges, 2002; de Jong et al., 2003; Townsend et al., 2004; van Rheede et al., 2006). The congruence between molecular and morphological data is remarkable, and the results indicate that C. amazonicus is unequivocally distinct from the species of the meridionalis group, Considering, however, that Coleodactylus was recovered as monophyletic genus in all three analysis, even though with a weak support, the taxonomic significance of the variation of the ungual sheath for the genus do not extend beyond the species level. The dataset also did not provide enough resolution regarding the phylogenetic position of the genus within Sphaerodactylini. Recently, a close relationship between ((C. brachystoma, C. septentrionalis) + Pseudogonatodes) was supported by molecular data, in agreement with the morphological (Gamble et al., 2008a). In the current study, excepting the sister taxa relationship between Gonatodes and Lepidoblepharis already recovered by previous molecular data analysis (Gamble et al., 2008a), generic relationships were poorly defined. Coleodactylus was not specifically related to any other genera, being, instead, placed as sister group of all other sphaerodactyls by Bayesian analysis. Some caution, however, must be used in interpreting the results concerning the monophyly of Coleodactylus or the phylogenetic position of the genus within Sphaerodactylini. Molecular data did not find strong support for the null hypothesis, but also did not falsify the alternative hypotheses. This difficulty to recover deep phylogenetic relationships with strong branch support and to statistically reject alternative hypotheses is a characteristic of topologies with weakly supported short interior branches leading to long terminal branches (Weisrock et al., 2005), as the molecular phylogeny obtained here (Fig. 3). This result might be intepreted as a soft polytomy (Maddison, 1989), related to the restrict ability of the available data to resolve dicotomic relationships due to a limited set of synapomorphies accumulated during the short period of time corresponding to the internal branches and the loss of signal due to multiple substitutions along the terminal branches (Weisrock et al., 2005). An alternative explanation for this result is the simultaneous or nearly simultaneous diversification of multiple lineages, characterizing a hard polytomy (Weisrock et al., 2005). Thus, further analyses incorporating additional molecular data and inclusion of more representative taxa from sphaerodactyls is required in order to appropriately ascertain the monophyly of Coleodactylus and to determine the placement of the genus within Sphaerodactylini. Nevertheless, irrespective of the monophyly of the genus or its phylogenetic placement, the molecular data have also evidenced that the previous morphological studies tended to be conservative and to underestimate the diversity of Coleodactylus species. Several recent studies have revealed deep genetic divisions and/or cryptic species within widespread neotropical reptiles (e.g., Glor et al., 2001; Pellegrino et al., 2005; Kronauer et al., 2005; Gamble et al., 2008b). In the present case, it is possible that species might actually represent complexes of species, which could explain the absence of a clear geographical component in the wide range of variation reported for meristic characters and color pattern (Ávila-Pires, 1995; Freire, 1999). For instance, the phylogenetic grouping of C. natalensis with C. meridionalis populations from Caatinga and Cerrado in Central Brazil rather than geographically closer samples from the Atlantic Forest might indicate that C. meridionalis is a single highly structured species with a history of episodes of expansion/colonization, and that C. natalensis originated as a peripheral isolate during one of these events. However, there is the possibility that each clade of C. meridionalis represents a much more widespread species, which might have overlapping distributions. Although incomplete lineage sorting can be hypothesized as a possible cause of the observed paraphyly, imperfect taxonomy has been identified as the main cause of paraphyly in poorly known and undersampled species (e.g., Funk and Omland, 2003; Morando et al., 2003; Avila et al., 2006). Given the broad geographical distribution of species, a denser sampling, including populations from areas not sampled in this study, will probably increase the number of possible clades. The taxonomic status of these clades is currently the focus of an ongoing phylogeographical analysis. 4.2. Evolutionary history Some important outcomes related to the proposed evolutionary history of Coleodactylus can be drawn from the phylogenetic recon-