Anolis chrysolepis Duméril and Bibron, 1837 (Squamata: Iguanidae), Revisited: Molecular Phylogeny and Taxonomy of the Anolis chrysolepis Species Group

Abstract The Anolis chrysolepis species group is distributed across the entire Amazon basin and currently consists of A. bombiceps and five subspecies of A. chrysolepis. These lizards are characterized by moderate size, relatively narrow digital pads, and a small dewlap that does not reach the axilla. We used the mitochondrial gene ND2 to estimate phylogenetic relationships among putative subspecies of A. chrysolepis and taxa previously hypothesized to be their close relatives. We also assessed the congruence between molecular and morphological datasets to evaluate the taxonomic status of group members. On the basis of the two datasets, we present a new taxonomy, elevating each putative subspecies of A. chrysolepis to species status. We provide new morphological diagnoses and new distributional data for each species.


INTRODUCTION
The Pleistocene Refuge Hypothesis proposed almost simultaneously by Haffer (1969) and Vanzolini and Williams (1970) posits that patches of lowland tropical forest that existed during dry periods in the Pleistocene served as core areas for speciation in birds and in the lizard complex Anolis chrysolepis, respectively. Although the Pleistocene Refuge Hypothesis has been falsified for members of the A. chrysolepis species group because diversification occurred much earlier (15 mya) than the Pleistocene (Glor et al., 2001), relationships among all members of the group have not been worked out and related taxa (e.g., A. meridionalis and A. bombiceps) have not been properly placed with reference to the A. chrysolepis complex, and current names do not accurately reflect the evolutionary history of the group (Glor et al., 2001;Nicholson et al., 2005). Because the A. chrysolepis species group has been and continues to be a model for evolutionary (Nicholson et al., 2006(Nicholson et al., , 2007Schaad and Poe, 2010) and ecological (Vitt and Zani, 1996;Vitt et al., 2001 studies, it is critical that their relationships be properly understood. Here we present a phylogenetic hypothesis for the A. chrysolepis species group using a much larger set of samples than was available previously and provide species names for taxa that can be identified as independent evolutionary lineages. Following de Queiroz (2007), we consider independent evolutionary lineages, here recognized on the basis of gene trees, analogous to species. Results of this study should be directly applicable to phylogeographic and phyloecological studies of the A. chrysolepis species group.
The A. chrysolepis group comprises two species: A. chrysolepis Dumé ril and Bibron, 1837, and A. bombiceps Cope, 1876. Anolis chrysolepis is currently composed of five subspecies: A. chrysolepis chrysolepis, in eastern Guiana (Brazil, French Guiana, Suriname, and southern Guyana); A. chrysolepis planiceps Troschel, 1848, in western Guiana (Brazil, Suriname, northwestern Guyana, Venezuela, and Trinidad); A. chrysolepis scypheus Cope, 1864, in western Amazonia (Colombia, Ecuador, Peru, and northwestern Brazil); A. chrysolepis tandai Avila-Pires, 1995, in southwestern Amazonia (Brazil and Peru); and A. chrysolepis brasiliensis Vanzolini and Williams, 1970, in Brazil, from Maranhão and enclaves of open vegetation in southern Pará south to São Paulo (Vanzolini and Williams, 1970;Avila-Pires, 1995;Icochea et al., 2001;Santos-Jr et al., 2007). Anolis bombiceps occurs in western Amazonia, in Peru, Colombia, and Brazil, at least in partial sympatry with A. c. scypheus and perhaps also with A. c. tandai (Avila-Pires, 1995). Members of the A. chrysolepis group are characterized by their moderate size (up to 83 mm snout-vent length); short heads; supraorbital semicircles usually forming a pronounced ridge; relatively narrow digital pads, with distal lamellae under phalanx ii forming a slightly prominent border; a dewlap that does not reach the axilla and is present in both sexes (but smaller in females); and keeled, imbricate ventral scales that are distinctly larger than dorsals.
The A. chrysolepis species group was examined morphologically by Vanzolini and Williams (1970), who recognized four sub-species of A. chrysolepis and a distinct species, A. bombiceps. Vanzolini and Williams (1970: 13) believed the level of differentiation between the subspecies were ''closest to species difference, and indicative, perhaps, of past and future potential species formation.'' Anolis chrysolepis was later examined by Avila-Pires (1995) under the name A. nitens. She described another subspecies, A. n. tandai, and observed that most specimens occurring in areas of intergradation according to Vanzolini and Williams (1970) could be assigned to one of the recognized subspecies.
Very little subsequent taxonomic research has been conducted on the species of the A. chrysolepis group. One molecular phylogenetic study included three of the described A. chrysolepis subspecies and found they formed a weakly supported clade (Glor et al., 2001). Glor et al. (2001Glor et al. ( : 2664 concluded that, ''further study of geographical genetic interactions among these subspecies probably will reveal that they are distinct species.'' Additional molecular phylogenetic research, with broad outgroup sampling, recovered a well-supported clade consisting of A. onca, A. annectans, A. lineatus, A. auratus, A. meridionalis, and A. chrysolepis, although A. chrysolepis was represented by just a single individual from Roraima, Brazil (Nicholson et al., 2005). Members of this clade were included in another phylogenetic analysis (Nicholson et al., 2006), using the same three A. chrysolepis subspecies of Glor et al. (2001), which recovered a paraphyletic A. chrysolepis. Nicholson et al. (2006) found that A. c. tandai was more closely related to A. meridionalis and the A. onca + A. annectans clade, whereas A. c. scypheus and A. c. planiceps formed a clade that was the sister group to the remaining species + A. auratus. Like Glor et al. (2001), Nicholson et al. (2006) stressed the need for additional research into the systematics of A. chrysolepis and the possible existence of cryptic species. Anolis bombiceps has not been included in any molecular studies so far.
The name A. chrysolepis has a long and confusing history, with both A. nitens and A. chrysolepis considered valid names for the species (Hoogmoed, 1973;Avila-Pires, 1995;Myers and Donnelly, 2008). Myers and Donnelly (2008) presented a detailed history of the use of these names, and Myers (2008) requested the International Commission of Zoological Nomenclature (ICZN) to give precedence of A. chrysolepis Dumé ril and Bibron, 1837, over Draconura nitens Wagler, 1830, which was accepted (ICZN, 2010. In the present work, we analyzed mitochondrial DNA from the protein coding gene ND2 and associated tRNA and morphological data from all five described subspecies of A. chrysolepis and related taxa to 1) recover the phylogenetic relationships among subspecies of A. chrysolepis and test previous phylogenetic hypotheses, 2) evaluate the taxonomic status of described subspecies of A. chrysolepis, and 3) present a revised taxonomy that incorporates this phylogenetic information.

Taxon Sampling and DNA Sequencing
We sampled representatives of each of the five subspecies of A. chrysolepis (Table 1, Figure 1). Species previously shown to be closely related to A. chrysolepis were also included either from newly sequenced samples (e.g., A. bombiceps) or from previously published GenBank material (Glor et al., 2001;Nicholson, 2002;Nicholson et al., 2005Nicholson et al., , 2006. Genomic DNA was extracted from muscle, liver, or tail clips using DNeasy Blood and Tissue Kit (Qiagen, Valencia, California). Polymerase chain reaction was used to amplify portions of the mitochondrial protein-coding gene ND2 (NADH dehydrogenase subunit 2) and adjacent tRNAs with primers LVT_Metf.6_AnCr (AAGCTATTGGGCCCATACC) and LVT_5617_AnCr (AAAGTGYTTGAG-TTGCATTCA) (Rodriguez Robles et al., 2007). Polymerase chain reaction cleanup and DNA sequencing was performed by Agencourt Bioscience (Beverly, Massachusetts). Sequences were edited and aligned using SEQUENCHER ver. 4.2 (Gene Codes, Ann Arbor, Michigan). ND2 sequences were translated into amino acids using MacClade ver. 4.08 (Maddison and Maddison, 1992) to confirm alignment and gap placement and ensure there were no premature stop codons.

Phylogenetic Analyses
We analyzed the ND2 data using parsimony in PAUP ver. 4.0b10 (Swofford, 2001). Parsimony analysis was conducted using a heuristic search with 1,000 random taxon additions and tree bisection and reconnection (TBR) branch swapping and all characters equally weighted. We conducted 1,000 bootstrap replicates with 25 random additions per replicate to assess nodal support (Felsenstein, 1985).
Mitochondrial DNA (mtDNA) has been widely used to recover phylogenetic relationships among species and to delimit species (Avise et al., 1998;Grau et al., 2005;Gamble et al., 2008;Fenwick et al., 2009), and because of its shorter coalescent times, it is considered a good indicator of population history and species limits (Avise et al., 2000;Wiens and Hollingsworth, 2000;Wiens and Penkrot, 2002;Zink and Barrowclough, 2008;Barrowclough and Zink, 2009). However, the high substitution rate of mitochondrial DNA makes saturation, especially at third codon positions, a possible problem for accurate phylogenetic reconstruction (Jukes, 1987;Yoder et al., 1996, Glor et al., 2001, Hudson and Turelli, 2003. One way to minimize the effects of saturation is to use model-based phylogenetic methods like maximum likelihood (ML) and Bayesian analyses (Felsenstein, 1978;Jukes, 1987;, Lartillot et al., 2007. Additionally, the use of partitioned model-based analyses, with separate models of molecular evolution for each gene or codon, can minimize phylogenetic error (Bull et al., 1993;Lemmon and Moriarty, 2004;Nylander et al., 2004;Brandley et al., 2005). We conducted Bayesian analyses using MrBayes 3.1.2  ) on both the partitioned and unpartitioned datasets. Data were partitioned by codon with a fourth partition for tRNAs and the optimal partitioning strategy selected using Bayes Factors calculated from the harmonic mean likelihood values (Nylander et al., 2004;Brandley et al., 2005). We estimated the best fit model of sequence evolution for the data as a whole and for each partition separately using AIC scores in Modeltest (Posada, 2008). Bayesian analyses were initialized with a neighbor-joining tree and two separate analyses conducted for each partitioning strategy. Each analysis consisted of seven heated chains and one cold chain run for 2 million generations, with sampling every 1,000 generations. Postburnin convergence was checked by visual inspection of likelihood values by generation using Tracer 1.5 (Rambaut and Drummond, 2009) and visual inspection of split frequencies using AWTY (Nylander et al., 2008). We also conducted partitioned Maximum Likelihood analysis, with data partitioned as above using RAXML ver. 7.0.4 (Stamatakis, 2006) using the GTR+GAMMA model for all partitions. We conducted 1,000 ''fast bootstrap'' replicates and 10 separate maximum likelihood searches. Bootstrap values $70 were considered as indicating strong support for both parsimony and ML analyses. We calculated net among group distances (Nei and Li, 1979) between major lineages of the ''A. chrysolepis species group'' using MEGA 4 (Kumar et al., 2008). We calculated both uncorrected p-distances and corrected distances using the GTR model.
On the basis of our best ML tree, we compared alternative phylogenetic hypotheses using the Shimodaira-Hasegawa (SH) test (Shimodaira and Hasegawa, 1999) and the Approximately Unbiased (AU) test (Shimodaira, 2002). Three alternative hypotheses were considered: 1) monophyly of A. chrysolepis subspecies, excluding A. bombiceps and A. meridionalis; 2) monophyly of the A. chrysolepis subspecies + A. bombiceps, excluding only A. meridionalis; and 3) monophyly of all A. c. tandai specimens, as identified by morphological data. We used RAxML7.0.4 (Stamatakis, 2006) to compute per-site log likelihoods that were input into CONSEL (Shimodaira and Hasegawa, 2001) to calculate P values. We also tested alternative phylogenetic hypotheses in a Bayesian framework and calculated the Posterior Probabilities of alternative hypotheses using the tree filter option in PAUP*.

Morphological Analyses
We collected morphological and morphometric data from 403 specimens (Appendix 1) from the following zoological collections: MZUSP, Measurements were recorded with digital calipers to the nearest 0.1 mm on the right side of the body, except when specimens were damaged (in this case, the left side was used). Scale and measurement terminology follows Avila-Pires (1995).
We recorded the following morphometric data: snout-vent length (SVL), tail length (from posterior edge of precloacal plate), head width, head height, mouth length (from tip of snout to posterior margin of mouth), distance between orbits (minimum), ear-opening diameter, distance between nostrils (minimum), distance from mouth to ear (from anterior margin of earopening to posterior margin of mouth), snout length (from tip of snout to anterior margin of orbit), interparietal length, tibia length, foot length (from toe IV base to the heel), fourth toe length (from toe IV nail to toe base), and fourth toe maximum width. Additionally, we recorded the following meristic characters: scales around midbody, postrostrals, supralabials, infralabials, loreals (under second canthal), canthals, scales between second canthals, scales between supraorbital semicircles (minimum), scales between interparietal and supraorbital semicircles (minimum), postmentals, fourth finger lamellae, and fourth toe lamellae.
A few measurements and scale counts could not be assessed for all specimens analyzed. In multivariate analysis, cases with missing observations will be dropped, weakening the analysis because of loss of information and degrees of freedom. To avoid simply deleting entire rows of data, missing observations can be estimated using a variety of methods, including mean substitution, regression, expectation maximization, maximum likelihood and multiple imputation (Tabachnick and Fidell, 2001;Quinn and Keough, 2002). Among these approaches for imputing values to missing observations, multiple imputation is the most robust and also makes fewer assumptions about the pattern of missing observations (Rubin, 1996;Van Buuren et al., 2006). Therefore, we imputed missing data using multivariate imputations by chained equations (Van Buuren et al., 2006), as implemented by package mice in R v. 2.12.0 (R Development Core Team, 2009).
To partition the total morphometric variation between size and shape variation, we defined body size as an isometric size variable (Rohlf and Bookstein, 1987) following Somers (1986): we calculated an isometric eigenvector, defined a priori with values equal to p 20.5 , where p is the number of variables (Jolicoeur, 1963), and obtained scores from this eigenvector, hereafter called body size, by postmultiplying the n 3 p matrix of log-transformed data, where n is the number of observations, by the p 3 1 isometric eigenvector. To remove the effects of body size from the log-transformed data, we used Burnaby's method (Burnaby, 1966): we postmultiplied the n 3 p matrix of the log-transformed data by a p 3 p symmetric matrix, L, defined as: where I p is a p 3 p identity matrix, V is the isometric size eigenvector defined above, and V T is the transpose of matrix V (Rohlf and Bookstein, 1987). Hereafter, we refer to the resulting size-adjusted variables as shape variables.
To identify morphometric and meristic variables that best discriminate among species, we used a stepwise discriminant analysis coupled with 100-fold cross-validation to measure classification performance (Quinn and Keough, 2002) using the package klaR in R v. 2.12.0 (R Development Core Team, 2009).

Phylogenetic Analyses
We sequenced 1,088 base pairs of the mitochondrial ND2 gene and adjacent tRNAs, which contained 82 variable sites and 633 parsimony-informative characters. Thirty-nine new mtDNA sequences from 34 localities ( Fig. 1) are reported and aligned with 14 previously published sequences.
A comparison of the partitioned Bayesian analyses to the unpartitioned analyses strongly favored the partitioned strategy (Bayes Factors . 860). We observed convergence among multiple Bayesian runs and utilized post-burnin samples (burnin 5 1,000) to estimate model parameters and tree topology (Fig. 2). The partitioned ML analysis produced a single tree (Fig. 3, ln L 5 216,649.1489) that had a similar topology to the partitioned Bayesian consensus tree at well-supported nodes. The Parsimony analysis produced 54 equally most parsimonious trees (TL 5 3,832, CI 5 0.337683, RI 5 0.682552, RC 5 0.230486, HI 5 0.662317; Fig. 4). Subspecies formed strongly supported monophyletic groups in all analyses, with the exception of specimens of A. c. tandai from Acre. All analyses also recovered a paraphyletic A. chrysolepis with regard to A. bombiceps and A. meridionalis (Figs. 2-4). Sampled individuals of A. chrysolepis, A. bombiceps, and A. meridionalis were members of one of two clades; one (Clade A) composed of A. c. chrysolepis, A. c. tandai, and Anolis meridionalis and another (Clade B) composed of A. c. brasiliensis, A. c. planiceps, A. c. scypheus, and A. bombiceps. Relationships among taxa in clade B were similar across all trees, with an A. bombiceps + A. c. scypheus clade and an A. c. planiceps + A. c. brasiliensis clade that are sister taxa. Relationships within Clade A varied depending on the analysis. Parsimony analysis recovered A. c. tandai from Acre (LSUMZ H13599) as the sister taxon of the A. c. tandai + A. c. chrysolepis clade. The ML and Bayesian trees, on the other hand, recovered the Acre A. c. tandai as the sister taxon of A. c. chrysolepis, but with low bootstrap support. The A. c. chrysolepis + A. c. tandai clade was well supported in all analyses, whereas the A. c. chrysolepis + A. c. tandai + A. meridionalis clade received poor nodal supported.
Uncorrected pairwise distances among lineages in the A. chrysolepis species group ranged from 5.0% between A. c. tandai and A. c. chrysolepis to 22.1% between A. c. scypheus and A. meridionalis (Table 2).
Both the SH and AU tests (Table 3) found that the alternative hypothesis of a monophyletic A. chrysolepis, excluding both A. bombiceps and A. meridionalis, resulted in a significantly worse tree than the ML tree. The ML tree constrained to exclude just A. meridionalis was not significantly worse than our best ML tree. Similarly, both tests found no significant difference between a tree constraining a monophyletic A. c. tandai and our best ML tree. Bayesian Posterior Probabilities of alternative hypotheses showed little to no support (e.g., Posterior Probabilities , 0.05) for a monophyletic A. chrysolepis excluding A. bombiceps and A. meridionalis, as well as a monophyletic A. c. tandai. The Bayesian Posterior Probability of a monophyletic A. chrysolepis + A. bombiceps, excluding A. meridionalis, received moderate support.

Morphological Analyses
The stepwise discriminant analysis applied on body size and all shape variables selected tibia length, interparietal length, and snout-vent length (all size-adjusted) as the most powerful discriminators of A. chrysolepis spp., A. bombiceps, and A. meridionalis, with a classification accuracy of 0.67 based on cross-validation. The first two linear discriminant functions based on these three variables explained about 99% of the total variation, the first function mainly representing a contrast between relative tibia length (2) versus relative SVL (+), and the second function repre-senting primarily the variation in interparietal length (  followed by A. c. brasiliensis. Nevertheless, classification accuracy based on morphology was moderate.
The stepwise discriminant analysis applied on meristic counts selected canthals, fourth toe lamellae, and scales between second canthals as the most powerful discriminators of the species, with a classification accuracy of 0.83 based on cross-validation (Fig. 6). The first two linear discriminant functions based on these three variables explained about 93% of the total variation. The first function mainly represented a contrast between canthals and scales between second canthals (2) versus fourth toe lamellae (+), whereas the second function primarily represented the variation in fourth toe lamellae and canthals (

DISCUSSION
The molecular phylogenetic analyses recovered six species-level taxa as part of the A. chrysolepis species group. These taxa can also be morphologically distinguished on the basis of morphometric and meristic characters. Even though we cannot infer relationships among these taxa on the basis of the meristic discriminant analysis, the results of this analysis are consistent with the existence of two clades: one containing A. c. tandai, A. c. chrysolepis, and A. bombiceps and another clade containing A. c. brasiliensis and A. c. planiceps. Meristic characters in A. c. scypheus appear to be intermediate between these two groups, which is also consistent with it being (together with A. bombiceps) the sister clade to A. c. brasiliensis + A. c. planiceps. Anolis meridionalis was quite distinct from other members of the A. chrysolepis species group on the basis of meristic characters.
We define the A. chrysolepis species group as the clade originating with the most recent common ancestor of A. c. chrysolepis and A. c. brasiliensis. Anolis meridionalis has not historically been allied with the A. chrysolepis species group because of its unique morphology. In particular, A. meridionalis differs from other members of the A. chrysolepis species group by having digital dilatations on phalanx ii and iii continuous with scales under phalanx i, instead of forming the prominent border observed in the A. chrysolepis subspecies and A. bombiceps. Although the node leading to the A. chrysolepis species group, including A. meridionalis, was well supported in the ML and Bayesian analyses, the presence of A. meridionalis in clade A received poor support in all phylogenetic analyses. For this reason, we could not reject the alternative hypothesis of a monophyletic A. chrysolepis group exclusive of A. meridionalis. This means that inclusion of A. meridionalis in the A. chrysolepis species group is still uncertain. Future phylogenetic analyses that include additional A. meridio- nalis samples and data from nuclear loci may help resolve this issue. All described taxa in the molecular analyses formed well-supported, monophyletic groups, with the exception of A. c. tandai. The A. c. tandai individual from Acre fit the morphological diagnosis we present in this study but was either the sister taxon to A. c. chrysolepis (ML and Bayesian analyses) or the sister taxon to the A. c. chrysolepis + A. c. tandai clade (parsimony analysis). The apparent paraphyly of A. c. tandai may be due to several phenomena, none of which are mutually exclusive. One possibility is phylogenetic error due to incomplete taxonomic sampling or lack of data (Graybeal, 1998;Mitchell et al., 2000). It is also possible that individuals from the Acre population represent an as yet undescribed, morphologically cryptic  species. Incomplete lineage sorting can also result in discordance between individual gene trees and the species tree because of the retention and/or sorting of ancestral polymorphisms, particularly when populations have diverged recently, have a large effective population size, or both (Maddison, 1997; Ballard and Whitlock, 2004;Maddison and Knowles, 2006). Additional phylogenetic analyses incorporating nuclear genes and additional taxa, as well as using methods that incorporate coalescent processes and incomplete lineage sorting, would be useful in clarifying relationships among A. c. tandai populations.
Our results show broad congruence among molecular and morphological data sets that are consistent with independent evolutionary lineages. Most importantly, each of these lineages is morphologically diagnosable. Genetic distances among sister taxa in the A. chrysolepis group were also comparable to ND2 distances among sister species in other squamate taxa (Macey et al., 1998(Macey et al., , 1999Glor et al., 2001;Oliver et al., 2009). Therefore, we elevate each subspecies to species status under the general lineage species concept (de Queiroz, 1998(de Queiroz, , 1999(de Queiroz, , 2005(de Queiroz, , 2005a(de Queiroz, , 2005b(de Queiroz, , 2007. To facilitate future studies, each species, including A. bombiceps and A. meridonalis, is diagnosed below and an identification key is provided, considering morphological data collected for this study as well as data from the literature. Table 6 compares the main meristic and morphometric characters.

Taxonomy/Species Accounts
All descriptions of color pattern are based on literature, photographs of live animals, and preserved specimens. Abbreviated Description. Maximum SVL 74 mm. Vertebral region with distinctly enlarged scales, middorsal row largest; number of rows of enlarged scales increases posteriorly. Scales on upper arms smaller than, to subequal to, vertebral scales. Supraorbital semicircles with scarcely enlarged scales. Supraocular scales keeled, slightly larger than or subequal to scales on snout, grading into granules laterally and posteriorly. Interparietal subequal to or slightly larger than adjacent scales (Fig. 7A, B). Color in Preservative. Color pattern sexually dimorphic. Male dorsal color pale or grayish-brown with or without a wide light vertebral band bordered by a grayishbrown irregular band laterally. Paired triangular spots may be present along back; most specimens with paired triangular spots on sacral region. Female dorsal pattern less variable. A thin dark brown line begins at posterior corner of each eye at each side, converging toward neck and continuing along the body, where they delimit a lighter or plumbeous vertebral band that darkens and expands laterally on tail.
Distribution. Southern Guyana, Suriname, French Guiana and northern Brazil, in the states of Amapá and Pará.
Color in Preservative. Color pattern sexually dimorphic. In males, vertebral region usually distinct from flanks, with unclear limits between these areas. A pair of subtriangular dark spots present on sacral region. Some specimens may present sinuous lines, assuming subtriangular shapes along dorsum. Females usually with a welldelimited vertebral band, similar to Anolis chrysolepis females; occasionally dorsal pattern similar to males.
Dewlap Color in Life and Preservative. In preservative, male dewlap royal blue or blackish-blue with light scales. Female dewlap with central blue spot surrounded by a cream area; scales usually light colored. In life, male dewlap skin frequently blue or blackish, with light scales. Avila-Pires (1995) mentioned the dewlap in MPEG 15986 as ''ultramarine with cream-color scales on rim.'' Dewlap in females, when extended, presents a large and central blue spot, surrounded by a cream area. Scales are frequently cream to orange. When not extended, dewlap presents a light rim and is blue laterally. Avila-Pires (1995) described the holotype MPEG 15850 female dewlap color as ''sulphur-yellow with a large indigo-blue spot.'' Comparison with Other Species from the A. chrysolepis Species Group. As already mentioned by Avila-Pires (1995), this species has the longest tibia in relation to SVL (0.30-0.43). For differences with A. chrysolepis, see above. Avila-Pires (1995) also mentioned the possible sympatry with A. bombiceps, which also has a blue or blackish blue dewlap (with no sexual dimorphism), but they can be distinguished by female dewlap color (a central blue spot, surrounded by a pale area in A. tandai), by the minimum number of scales between supraorbital semicircles (1-4 in A. tandai and 1-2 in A. bombiceps) and by the number of postmentals (4-8 in A. tandai and 6-8 in A. bombiceps).
Distribution. South of the Amazon River and west of the Tapajó s River, in Brazil (states of Pará, Amazonas, Rondô nia, Acre, and north of Mato Grosso), and in Peru.
Color in Preservative. No sexual dimorphism in color pattern. Specimens usually have many chevrons along back, with tips directed posteriorly, sometimes forming the posterior border of rhomboid figures. A pair of triangular spots commonly present on sacral region. Myers and Donnelly (2008) described the color pattern of two adult males and one adult female as ''orange with white or grayish white scales in basal rows, scales darker gray or blackish gray in distal rows.'' Dewlap Color in Life and Preservative. Dewlap red, fading rapidly in preserved specimens, appearing cream-white, with light scales. A lateral lavender area may be present as mentioned by Avila-Pires (1995). In life, dewlap skin orange to reddish with grayish to cream scales. Myers and Donnelly (2008) found variation in the dewlap of four juveniles, including a female that had ''a large bluish black basal spot on the dewlap, which had a bright orange periphery and mostly white scales (only a few dark scales).'' Comparison with the Other Species from the A. chrysolepis Species Group. This species has the proportionately largest interparietal scale. It differs from its sister taxon A. brasiliensis mainly by dewlap color (red in A. planiceps and blue or grayish/ blackish blue in A. brasiliensis) and body size (A. planiceps reaches 76 mm, whereas A. brasiliensis reaches 69 mm).
Distribution. Venezuela, Trinidad, Guyana, and the states of Roraima and Amazonas on the northern part of Brazil.
Color in Preservative. No sexual dimorphism in color pattern. Dorsal color grayishbrown or pale white, either uniform or not. A light vertebral band may be present, either narrow with undefined margins or wide; in both cases surrounded by darker area. A pair of triangular spots on sacral region commonly present, may be accompanied by second pair at the base of tail. Ventral region usually pale-white, may be marbled with brown spots.
Dewlap Color in Life and Preservative. Dewlap blue or grayish-blue, with light or grayish scales. In life, dewlap usually grayish blue or blackish blue, with dark scales varying from light-cream to dark gray. Some specimens from Tocantins state show the dewlap skin grayish-green tending to yellowish-beige along rim, with scales grayishbrown or pale-cream tending to brownish along rim. Some irregular light-blue lines may be present (Fig. 8D). Vanzolini and Williams (1970) do not describe the dewlap color but refer to the frontispiece plate representing the dewlap color in life of a male as green with a brown edge along rim. Avila-Pires (1995) observed in specimens from Carajás, Southern Pará, ''a blue dewlap, lighter in females, with scales varying from light to dark gray or cream and the surrounding area may be chrome-orange.'' Comparison with the Other Species from the A. chrysolepis Species Group. Anolis brasiliensis, along with A. bombiceps, has the largest toe IV among the other species of the A. chrysolepis group. Anolis brasiliensis differs from A. planiceps, mainly by dewlap color (red in A. planiceps and blue or grayish/blackish blue in A. brasiliensis) and body size (A. planiceps reaches 76 mm, whereas A. brasiliensis reaches 69 mm). Anolis brasiliensis is broadly sympatric with A. meridionalis, although they occur in different habitats (see A. meridionalis description below) and they differ mainly by digital dilatations under phalanx ii and iii that form a prominent border in A. brasiliensis. Distribution. Brazil, in southern Pará, Tocantins, Piauí, Maranhão, Ceará, Goiás, Mato Grosso, Minas Gerais, São Paulo, and Distrito Federal. Two individuals housed in MCZ under the numbers R-60580 and R-60581 are labeled as paratypes of A. c. brasiliensis but have as localities Rio Juruá, Brazil, and Loreto, Peru, respectively. Vanzolini and Williams (1970) did not mention those individuals, which mean that they are not paratypes. Moreover, even though they have typical A. brasiliensis characteristics, it is extremely unlikely that the species occurs in these localities. Given their questionable data, we did not consider these individuals in the morphological analyses or elsewhere in this study.
Abbreviated Description. Maximum SVL 80 mm. Vertebral scales forming double row of enlarged dorsals along back. Scales of upper arms small relative to other species, larger than vertebral scales. Scales on snout relatively small, with raised surface. Supraorbital semicircles with enlarged scales, generally forming pronounced ridge. A small group of enlarged supraocular scales, grading into granules posteriorly and laterally, anteriorly surrounded by smaller scales. Interparietal distinctly larger than surrounded scales (Fig. 8D-F).
Color in Preservative. No sexual dimorphism in color pattern. Dorsum usually with caudally directed chevrons that may form the posterior border of rhomboid figures, similar to the pattern described for A. planiceps; or, it may show a broad band with lateral expansions (narrower at nape, extending caudally). A pair of subtriangular dark spots may be present on sacral region.
Dewlap Color in Life and Preservative. Dewlap color in preservative usually pale along rim with pale-cream or blackish scales, and blue with light scales in the center (Fig. 7). In life, dewlap skin red along rim (the red color vanishes very easily in preserved specimens) with red or blackish scales, and blue with pale-cream or lightbrown scales laterally. Avila-Pires (1995) described the dewlap color of RMNH 24653 as ''cobalt-blue with red rim, scales white with orange center.'' Comparison with the Other Species from the A. chrysolepis Species Group. Anolis c. scypheus presents the proportionally maximum values of head width, head height, ear diameter, minimum distance between nostrils and SVL among all species of the A. chrysolepis species group. For differences with A. bombiceps, see A. bombiceps diagnosis below.
Distribution. Amazonian Colombia, Ecuador, Peru, and the northwestern part of Amazonas state in Brazil.
Color in Preservative. No sexual dimorphism in color pattern. Dorsal color usually brown, with irregular dark spots between hind limbs and irregular figures across limbs. V-shaped lines along back, with apex directed posteriorly, may be present.
Dewlap Color in Life and Preservative. Dewlap deep blue or blackish in preservative, with light or dark scales. In life, dewlap skin deep blue with light or dark scales.
Color in Preservative. No sexual dimorphism in color pattern observed. Dorsal color usually grayish-brown, with a light cream, sometimes tending to reddish or light-orange, vertebral band. Dark brown V-shaped lines (apex directed posteriorly) may be present, as well as dark-brown irregular figures and spots across the limbs and in paravertebral region. A pair of subtriangular dark spots on sacral region frequently present.
Dewlap Color in Life and Preservative. When extended, dewlap skin in preservative blue or grayish-blue in center and pale or cream along rim, with light or dark scales. When not extended, grayish-blue laterally, with the center cream or beige. In life, dewlap skin deep blue in center and orange to pale yellow along rim. Scales may be darker on the center, tending to cream or beige on the border, changing to orange or pale yellow on anterior base of dewlap, along rim (Fig. 8F). Langstroth (2006) presented a dewlap photo of a male A. meridionalis from near the Zapocó s Reservoir in Bolivia showing a deep blue skin with irregular grayish lines and light or dark scales, tending to grayish-green along rim.
Comparisons with the Other Species from the A. chrysolepis Species Group. Anolis meridionalis is sympatric with A. brasiliensis, of which it can be distinguished mainly by the digital dilatations on phalanx ii and iii, which are continuous with scales under phalanx i and do not form the prominent border observed in A. brasiliensis. Anolis meridionalis can also be distinguished from A. brasiliensis by smaller body size (A. meridionalis reaches 56 mm, whereas A. brasiliensis reaches 69 mm), by the smaller number of scales between supraorbital semicircles (0-1 in A. meridionalis and 0-3 in A. brasiliensis), by the smaller number of scales between supraorbital semicircles and interparietal (1-4 in A. brasiliensis and 0-2 in A. meridionalis), and by the smaller number of fourth finger and toe lamellae (15-20 and 25-32, respectively, in A. brasiliensis, 10-12 and 18-24 in A. meridionalis). Besides these morphological differences, these species do not occur in the same habitat: A. meridonalis is commonly found in open areas densely covered by grass in Brazilian Cerrado (Vanzolini and Williams, 1970), whereas A. brasiliensis is a typical inhabitant of gallery forests in the same biome . Vitt et al. (2008: 146) found ''only two specimens of A. brasiliensis outside of forested habitat in typical Cerrado, and both were inside termite nests and inactive.'' Distribution.