INTRODUCTION

Tropidurus Wied, 1825, is one of the most ubiquitous lizard genera occupying open land-scapes in tropical and subtropical South America (Carvalho, 2013). However, the numerous field observations and large number of specimens in scientific collections hardly translate into a complete understanding of its phylogenetic relationships, taxonomic diversity, biogeography, and evolutionary history (Carvalho, 2013; Carvalho et al., 2013). The monophyly and internal relationships in Tropidurus were not rigorously established until Frost et al. (2001) employed molecular (mtDNA) and morphological characters—combining novel data with those from Frost (1992) and Harvey and Gutberlet (2000)—to build a comprehensive phylogeny. That study placed Uranoscodon as the sister taxon of all other tropidurines; recovered Plica, Uracentron, and Strobilurus as the sister clade of Tropidurus; erected a new genus Eurolophosaurus for the former Tropidurus nanuzae group; and restricted Tropidurus to a monophyletic group that predominantly occupies the open-dry South American diagonal, Amazonian savanna enclaves, and large area of the Brazilian Atlantic coast (Rodrigues, 1987, 1988; Frost et al., 2001; Ávila-Pires, 1995; Harvey and Gutberlet, 1998; Carvalho, 2013; Carvalho et al., 2013).

Currently, the genus comprises 25 nominal species in four species groups as per Frost et al. (2001): (1) Tropidurus bogerti group, monotypic: T. bogerti Roze, 1958; (2) Tropidurus spinulosus group: T. callathelys Harvey and Gutberlet, 1998, T. guarani Alvarez et al., 1994, T. melanopleurus Boulenger, 1902, T. spinulosus (Cope, 1862), and T. xanthochilus Harvey and Gutberlet, 1998; (3) Tropidurus semitaeniatus group: T. helenae (Manzani and Abe, 1990), T. jaguaribanus Passos et al., 2011, T. pinima (Rodrigues, 1984), and T. semitaeniatus (Spix, 1825); (4) Tropidurus torquatus group: T. catalanensis Gudynas and Skuk, 1983, T. chromatops Harvey and Gutberlet, 1998, T. cocorobensis Rodrigues, 1987, T. erythrocephalus Rodrigues, 1987, T. etheridgei Cei, 1982, T. hispidus (Spix, 1825), T. hygomi Reinhardt and Lütken, 1861, T. imbituba Kunz and Borges-Martins, 2013, T. insulanus Rodrigues, 1987, T. itambere Rodrigues, 1987, T. montanus Rodrigues, 1987, T. mucujensis Rodrigues, 1987, T. oreadicus Rodrigues, 1987, T. psammonastes Rodrigues et al., 1988, and T. torquatus (Wied, 1820). Over 70% of these taxa were named in the last 35 years, and this ascending, nonasymptotic taxonomic curve suggests a promising future with respect to the number of Tropidurus species still to be uncovered (fig. 1). Morphological variation indicates the existence of species complexes under several nominal species in the genus (e.g., Vanzolini and Gomes, 1979; Vanzolini et al., 1980; Vanzolini, 1986; Rodrigues, 1987; Vitt and Caldwell, 1993; Vitt et al., 1996; Vitt et al., 1997; Frost et al., 1998; Gainsbury and Colli, 2003; Valdujo et al., 2009; Werneck and Colli, 2006; Werneck et al., 2015), although species delimitation based solely on morphology has been proven insufficient to elucidate diversity in several cases (Rodrigues, 1987; Frost et al., 1998; Werneck et al., 2015).

FIGURE 1.

Taxonomic curve showing a steep, nonasymptotic increment in the number of species of the lizard genus Tropidurus described since 1820. Specimen of T. sertanejo, n. sp., MZUSP 104274 (allotype).

During an expedition undertaken in July 2009 for collection of amphibians and reptiles in several sites in the State of Bahia, Brazil, some of us (A.L.G.C., H.R.S., and R.M.) collected a few juvenile Tropidurus at Reserva Particular do Patrimônio Natural (RPPN) Fazenda Pé da Serra, Serra do Arame, Municipality of Ibotirama. These specimens possessed characters unusual among Tropidurus such as the absence of mite pockets on the lateral neck, leading us to suspect that they represented an unnamed species. We conducted a second expedition to RPPN Fazenda Pé da Serra, which was focused on sampling additional specimens of the suspected new taxon. This second expedition was carried out in July 2013 by A.L.G.C., P.L.V.P., and R.M., and resulted in the collection of mature individuals of three sympatric Tropidurus species (fig. 2). Tropidurus hispidus (Spix, 1825) was found around the margins of an artificial dam inside the reserve. Tropidurus pinima (Rodrigues, 1984) was collected from rock crevices and while basking on large rock outcrops. The third species corresponded to the morphotype sampled in 2009. Specimens were found using small rocks along trails that cut through the sandy caatingas and dry forests of Serra do Arame. After observing that adults shared the same unique combination of characters previously found in the juveniles, we confirmed the occurrence of an unnamed Tropidurus for that locality.

FIGURE 2.

Syntopic species of Tropidurus found at the Reserva Particular do Patrimônio Natural Fazenda Pé da Serra, Serra do Arame, Ibotirama, Bahia, Brazil, and their respective habitats: (A, B)T. sertanejo, n. sp. (MZUSP 104274, allotype); (C, D) T. hispidus (MZUSP 104276); (E, F) T. pinima (MZUSP 104271).

The new taxon is most similar to species of the T. torquatus species group, with a robust body that is not flattened dorsoventrally and the absence of a vertebral crest. It is easily diagnosable from other members of this group, however, by having a unique dorsal coloration pattern and by lacking mite pockets laterally on the neck. During the investigation of its phylogenetic relationships we were able to identify samples with similar molecular profile, which were originally collected by M.T.R. in the Municipality of Caetité, Bahia, in 1991. Morphological analysis of the specimens from Caetité confirmed the presence of the diagnostic characters and allowed the identification of this second population ∼150 km to the south of Ibotirama. Herein, we provide a detailed description of the species based on specimens from both localities and analyze its phylogenetic position, external morphology, morphometric profile, and distribution. We conclude with a discussion of the impacts of our discoveries to the phylogenetic reconstruction of tropidurine lizards.

FIGURE 3.

Map on left shows the distribution of the Brazilian biomes and highlights the State of Bahia, predominantly covered by the semiarid Caatinga. Map on right (altimetric profile) shows the distribution of Tropidurus sertanejo, n. sp.: northernmost dot indicates the type locality (RPPN Fazenda Pé da Serra, Serra do Arame, Ibotirama, Bahia: 12° 08′ 45.21 S, 43° 03′ 20.83 W) and southernmost dot indicates the only known additional locality of occurrence of the new species (Caetité, Bahia: 14° 04′ 17.82 S, 42° 29′ 48.33″ W).

Material and Methods

Samples: Adult and juvenile specimens of Tropidurus were collected at the Reserva Particular do Patrimônio Natural (RPPN) Fazenda Pé da Serra, Serra do Arame, Municipality of Ibotirama, State of Bahia, Brazil (12° 08′ 45.21 S, 43° 03′ 20.83 W, WGS84 system; fig. 3). Some of us visited the area in two nonconsecutive years, the first time between 22–23 July 2009 (A.L.G.C., H.R.S., and R.M.) and a second time in 28 July 2013 (A.L.G.C., P.L.V.P., and R.M.). The specimens collected during the first expedition were originally housed at the herpetological collection of the Universidade Federal Rural do Rio de Janeiro (RU) and recently donated to the Museu de Zoologia da Universidade de São Paulo (MZUSP). Those collected in 2013 were directly deposited at MZUSP. All individuals were collected under permits 39914-3 and 10689-1, granted by the Brazilian Ministério do Meio Ambiente (ICMBio-SISBIO).

Lizards were observed during the day along trails that cut through the caatingas of the Serra do Arame. In total, 13 specimens were collected with the aid of rubber bands or nooses. All specimens were euthanized with an overdose of 2% lidocaine, preserved with 10% unbuffered formalin, then transferred to 70% ethyl alcohol solution. Before fixation, we collected tissue samples (muscle) from the left thigh of four individuals and stored them in Eppendorf^© tubes containing absolute ethyl alcohol for subsequent molecular analyses. Additional ethanolpreserved tissue samples (muscle, liver, or tail tips) of other species selected for phylogenetic analyses were obtained from the Ambrose Monell Cryo Collection, American Museum of Natural History (AMCC-AMNH), and Miguel Trefaut Rodrigues Tissue Collection, Instituto de Biociências, Universidade de São Paulo (MTR-USP). Thirteen specimens originally collected by M.T.R. in the municipality of Caetité, Bahia, in 1991, and housed at MZUSP, were also examined for morphological descriptions and were incorporated in the type series. Tissue samples associated to three of them were included in our phylogenetic analyses. See appendix 1 for a list of tissue samples, voucher specimens, and collection localities.

We determined the sex of the specimens based on the observation of colored patches of glandular scales varying from yellow to black on the ventral side of thighs and precloacal flap in males. Females lack ventral colored patches. Adult males have wider heads and thinner bodies than females of the same body size. This is a common sexually dimorphic pattern observed in Tropidurus that facilitates sex identification of adults (Pinto et al., 2005; Ribeiro et al., 2012). Sex determination of juveniles is not as easy based on external morphology alone because young individuals lack the attributes listed above. To avoid dissecting specimens included in type series, we chose to forgo the sex determination of juveniles.

Taxonomic framework and morphological analysis: We performed morphological and morphometric comparisons based on the primary analysis of 373 Tropidurus representing all 15 valid species currently assigned to the T. torquatus group (Frost et al., 2001) plus 25 specimens of the new taxon. We also analyzed data compiled from the literature for scale counts, morphometric measurements, and coloration (see below). Because several nominal species represent unresolved cryptic species complexes, we opted for comparing exclusively specimens from type localities and/or closely related populations. Although the number of individuals considered for comparative analyses was reduced, this conservative approach allowed us to assess morphological differences among populations that unequivocally represent valid nominal taxa, excluding those of uncertain taxonomic status.

Species comparisons and taxonomic placement followed the general framework proposed by Frost et al. (2001), in addition to our own phylogenetic results. Morphological comparisons among clades and species within clades followed Rodrigues (1987), Rodrigues et al. (1988), and Harvey and Gutberlet (1998) for the T. torquatus species group; Rodrigues (1984), Manzani and Abe (1990), and Passos et al. (2011) for the T. semitaeniatus (Spix, 1825) species group; Roze (1958) and Myers and Donnelly (2008) for T. bogerti Roze, 1958; and Alvarez et al. (1994) (accepting the species rank proposed by Frost et al. [1998] for T. spinulosus (Cope, 1862) and T. guarani Alvarez et al., 1994) and Harvey and Gutberlet (1998) for the T. spinulosus species group. All specimens used for comparisons are housed at the herpetological collections of the American Museum of Natural History, New York (AMNH); Museo de Historia Natural Alcide d'Orbigny, Cochabamba, Bolivia (MHNC); Museu de Zoologia da Universidade de São Paulo, São Paulo, Brazil, (MZUSP); Universidade Federal do Mato Grosso, Mato Grosso, Brazil (UFMT); and Universidade Federal do Rio Grade do Sul, Porto Alegre, Brazil (UFRGS). A list of catalog numbers and collection localities of examined specimens is available for download from the AMNH Digital Library Repository ( http://dx.doi.org/10.5531/sd.sp.16).

External morphology: We adopted the general terminology of Frost (1992) for description of morphological structures, but also adapted and extended protocols for scale counts and scale nomenclature following Harvey and Gutberlet (1998) and others, as noted. The rostral is conserved among Tropidurus, however, there is variation in the pattern of contact between rostral and adjacent scales, i.e., postrostrals, supralabials, lorilabials, loreals, and nasal. Postrostrals correspond to the series of scales bordering the rostral posteriorly and are divided or undivided. The canthal ridge is covered with one or two (occasionally three or four) enlarged canthals that contact anteriorly the nasal (or tiny scales surrounding the nasal) and grade posteriorly into superciliaries. Counts of superciliaries included exclusively the row of elongate-laminate overlapping scales beginning with the first scale to flank the supraoculars. On the side of the head, we distinguish loreals and lorilabials. The former corresponds to the scales on the side of the head between nasal and preocular, below the canthals. The latter consists of scales below loreals and subocular, and between these and supralabials. In contrast to the loreals lorilabials are not adherent to the underlying periosteum and are easily lifted by forceps or dissecting needles (Etheridge and Williams, 1991). Supra- and infralabials were counted as the sum of the enlarged scales that give shape to the superior and inferior lips, respectively. Reduced (sometimes nearly granular) scales that follow the enlarged labials posteriorly to reach the end of the rictus oris were considered separately.

Scales on the dorsal surface of the orbit are referred to as supraoculars. Two or three (occasionally four) series of supraoculars are found, one row of enlarged scales positioned internally and one or tow rows of smaller scales closer to the border of the orbit. Tropidurus have supraoculars bordered by one or two rows of small angulate circumorbitals separating the former from the median head shields. In between supraoculars and superciliaries one can find a row of short semilaminate scales. Temporals are located on the temporal region, in between the orbit and anterior border of external ear, below the parietals and occipitals in lateral view. We distinguish infra- and supratemporals (generally slightly enlarged in relation to infratemporals), separated by a line drawn horizontally at the level of the center of the pupil. Parietals are positioned behind supraoculars, separated by a considerably enlarged interparietal bearing a visible pineal eye. One or a few rows of irregular angulate scales referred to as occipitals separate interparietal from dorsals.

The general shape of the mental scale, positioned at the medial edge of the lower lip, is nearly invariable in Tropidurus, but the length of its posterior end in relation to adjacent scales was considered. Chinshields (postmentals sensu Etheridge 1968, 1970) consist of paired series of enlarged scales extending posteriorly from mental and contacting one more infralabials anterior to the first sublabial. Sublabials form a row of scales contacting the infralabials (Harvey and Gutberlet, 1998).

Species of the T. torquatus group as per Frost et al. (2001) lack a vertebral crest, resulting in vertebrals and paravertebrals indistinguishable from other dorsals. Therefore, scales covering the dorsal surface of the body were considered altogether as dorsals and counted from the posterior head scales (occipitals) in a straight line to the posterior edge of the hind limb where dorsals grade into enlarged caudals. Ventrals were counted midway from a line corresponding to the margin of the antegular fold to the anterior margin of hind legs. Following the same midline, we counted as cloacals the scales located in between the anterior margin of the hind legs and the anterior margin of the cloaca. Scales were also counted around midbody, halfway between the forelegs and hind legs. Supracarpals/tarsals and infracarpals/tarsals refer to scales on the dorsum and palm of hands and feet, respectively. Subdigital lamellae were counted from the digital articulation to, but not including, the ungual scale (noncarinate or lightly carinate scale just proximal to the claw) in both fingers and toes. Scales covering the dorsum of the fingers and toes are referred as supradigitals.

FIGURE 4.

Measurements used for morphometric analyses of Tropidurus. Abbreviations: AL, arm length; EOS, ear opening—snout distance; FAL, forearm length; FOL, foot length; HDL, hand length; HH, head height; HW, head width; SL, shank length; SVL, snout-vent length; THL, thigh length; TL, tail length.

Harvey and Gutberlet (1998) explained that the “flash” coloration on the ventral surface of the thighs and precloacal region of tropidurids results from pigment in the glandular scales. Rows of glandular scales on the ventral side of the thigh were counted along the short axis of the limb. Continuous rows of both fully and partially pigmented scales were counted. The same applied to rows of glandular precloacal scales, counted along the midline of the cloacal flap. We referred as caudal scales to those covering the dorsal, lateral, and ventral sides of the tail, from the posterior edge of hind limbs to the tip of the tail. Differences in shape and size of caudal scales are noted accordingly.

Morphometrics: Eleven morphometric measurements from the right side of 239 adult specimens (133 males and 106 females) were taken with aid of digital calipers (to the nearest 0.1 mm; fig. 4): SVL (snout-vent length), from the tip of snout to the anterior margin of the cloaca; HH (head height), from interparietal scale to gular region, measured with caliper positioned frontally; HW (head width), distance between temporal regions, measured below the level of the dorsal limit of ear opening; EOS (ear opening—snout distance) from tip of snout to the anterior margin of ear opening; AL (arm length), from insertion of the arm to tip of humerus; FAL (forearm length), from tip of proximal end of brachium to carpals; HDL (hand length) from carpals to tip of longest digit (fourth), including the claw; THL (thigh length), from insertion of the leg to distal end of thigh; SL (shank length), from proximal end crus to heel; FOL (foot length), from heel to tip of longest digit (fourth), including the claw; TL (tail length), from anterior margin of the cloaca to tip of the tail—taken exclusively from specimens with fully grown tails.

The mean, standard deviation, minimum, and maximum values were calculated for all morphometric variables of each species. To identify outliers, we conducted a visual inspection of the dispersion graphs constructed for each variable, plotted against SVL. The assumptions of normality and variance homoscedasticity were tested using Shapiro-Wilk and Bartlett tests (Sokal and Rohlf, 1995). Because some variables did not show normal distribution and homoscedastic variance, all morphometric measurements were log-transformed to meet these assumptions. Log-transformed variables were double-checked and confirmed for normality and homoscedasticity.

To estimate TL for individuals with broken, regrown, or missing tails, we adopted a three-step approach. First, we centered all variables by subtracting the group mean from each individual morphometric measurement. Second, we performed a multiple linear regression (MLR) between TL and all other mean-centered variables, employing the morphometric measurements of all individuals with fully grown tails. We finally employed the MLR function calculated to estimate missing TL values, and group means were then added back to individual values to obtain the final TL estimates.

We performed an exploratory analysis of morphometric variation employing a principal-component analysis (PCA; covariance matrix), and tested for morphometric differences among species using multivariate analysis of variance (MANOVA) and linear discriminant analysis (LDA). Because size accounted for a large portion of the variation summarized in the first principal component, we performed the Burnaby's size-correction procedure (Burnaby, 1966) by back-projecting the original log-transformed observations into a plane orthogonal to an isometric vector (Somers, 1986) prior to recalculation of MANOVA and LDA. To assess accuracy of species reclassification, we repeated LDA employing a leave-one-out cross-validation procedure and compared correct reclassification rates.

We used scores associated with PC1 as a proxy for size variation and tested for differences in size among species with analysis of variance (ANOVA) and a post hoc Tukey-Kramer test. All morphometric analyses were carried out for adult males and females separately using the packages “nortest” (Gross, 2012), “car” (Fox and Weisberg, 2011), DTK (Lau, 2013), MASS (Venables and Ripley, 2002), and Lattice (Sarkar, 2008) in “R” (version 3.0.2; R Core Team, 2013).

Meristics: Species description and taxonomic comparisons were based on a large number of meristic characters. Statistical analyses were performed on six scale counts for all species of the T. torquatus group: dorsal scales, gular scales, ventral scales, scales around midbody, tibial scales, and subdigital lamellae on fourth toe. The distribution of these scale counts among species was investigated to detect characters potentially useful to distinguish the new taxon from other species and/or identify major species groups based on meristic characters. Because we failed to identify any consistent correlation between SVL (as a proxy for age) and scales counts (results not shown), we pooled juveniles, subadults, and adults of each species for subsequent analyses. In total, we analyzed 337 specimens, including individuals of both sexes and all ages (120 males, 141 females, and 76 undetermined).

The assumptions of normality and homoscedasticity of scale-count distributions were tested using Shapiro-Wilk and Bartlett tests (Sokal and Rohlf, 1995). About 20% of the tests rejected the null hypothesis of normality, yet no consistent pattern was identified with respect to specific variables or taxa. In addition, the Barlett test rejected homoscedasticity of variances among species for all variables analyzed. Because basic assumptions of parametric methods were violated, we opted for a nonparametric multivariate analysis of variance to test for differences between species, sexes, and potential interaction between these factors (Anderson, 2001). The significance of the nonparametric MANOVA was based on 10,000 permutations. Because LDA is robust to violation of assumptions of normality and homoscedasticity (Krzanowski, 1977), we employed this multivariate technique to investigate differences in scale counts among species, avoiding multiple pairwise comparisons. We repeated LDA employing a leave-one-out cross-validation procedure and compared correct reclassification rates.

All statistical procedures were performed in “R” (version 3.0.2; R Core Team, 2013), as described above, and with aid of the package “vegan” (Oksanen et al., 2015).

Phylogenetic inference: To infer the phylogenetic relationships of the new species within the T. torquatus group, we constructed a matrix composed of four mitochondrial (12S, 16S, CO1, Cyt b) and six nuclear loci (BACH1, kif24, NTF3, PRLR, PTPN, SNCAIP). In addition to seven samples corresponding to the new taxon, we included as ingroup all 15 valid species currently assigned to the T. torquatus group as per Frost et al. (2001): T. catalanensis, T. chromatops, T. cocorobensis, T. erythrocephalus, T. etheridgei, T. hispidus, T. hygomi, T. imbituba, T. insulanus, T. itambere, T. montanus, T. mucujensis, T. oreadicus, T. psammonastes, and T. torquatus. We selected the stenocercine Stenocercus quinarius Nogueira and Rodrigues, 2006, to root the tree, and the tropidurines Microlophus quadrivittatus (Tschudi, 1845), Plica plica (Linnaeus, 1758), T. semitaeniatus (Spix, 1825), T. spinulosus (Cope, 1862), and Uranoscodon superciliosus (Linnaeus, 1758) as additional outgroups.

Laboratory procedures: We extracted and isolated DNA from frozen ethanol-preserved tissues (muscle, liver, or tail tips) using the Qiagen DNeasy kit following the manufacturer's guidelines. Polymerase chain reactions (PCR) for amplification of DNA fragments were carried out in 25 µl reactions using Illustra PuRe Taq Ready-To-Go PCR Beads (GE Healthcare Life Sciences). Primers and PRC parameters adopted for amplification and sequencing are listed in table 1. PCR products were cleaned and desalted in an AMPure (Agencourt Biosciences Corporation) reaction in a Beckman Coulter Biomek 2000 robot (“Becky”) or by hand. Cycle sequencing using BigDye Terminators, v. 3.0 (Applied Biosystems) was run in 8 µL reactions, adapting the protocol of Platt et al. (2007), and products were cleaned and desalted using cleanSEQ (Agencourt Biosciences Corporation) in “Becky.” Sequencing was performed in a Roche ABI 3730 XL automated sequencer. Samples were sequenced in both directions to check for sequencing errors and ambiguities. Sequence contigs were assembled and edited in Geneious 6.1.8 (Biomatters,  www.geneious.com). Genbank accession numbers are given in appendix 2.

Alignment, model selection, and optimality criteria: Alignments were conducted separately for each locus using the MAFFT (Katoh and Toh, 2008) plugin in Geneious 6.1.8 (Biomatters,  www.geneious.com); we employed the 200 PAM (k = 2) scoring matrix, gap open penalty 1.53, offset value 0.123 and used the “auto” function to select the best algorithm according to data size. Subsequently, we concatenated the alignments in Sequence Matrix 1.8 (Vaidya et al., 2011). All alignments produced in this study were made available for download from the AMNH Digital Library Repository ( http://dx.doi.org/10.5531/sd.sp.16).

To investigate the contribution/conflict of mitochondrial and nuclear loci determining/ affecting phylogenetic reconstruction, we performed partial and total evidence analyses based on different data sets: (1) mitochondrial loci concatenated, (2) nuclear loci concatenated, and (3) mitochondrial plus nuclear loci concatenated; hereafter referred as mitochondrial, nuclear, and total evidence data sets, respectively. Each dataset was phylogenetically analyzed under maximum parsimony (MP) and maximum likelihood (ML). Analyses were based on the same alignments and under the same treatment of gap characters (as missing data). Although we are well aware that gap characters can be informative in a phylogenetic context—hypothesized to represent natural length variation, i.e., insertion/deletion events (Giribet and Wheeler, 1999; Simmons and Ochoterena, 2000)—we made a practical decision not to treat gap characters as an additional state, to make the parsimony and maximum likelihood results more comparable to each other, at least with respect of use of evidence. The issue of comparability among optimality criteria and its implications are discussed thoroughly in Peloso et al. (2015).

Parsimony tree searches were carried out in TNT version 1.1 (Goloboff et al., 2000). Heuristic searches were based on a combination of the random addition of sequences algorithm (RAS) and subsequent rearrangement of branches through tree bisection-reconnection (TBR), parsimony ratchet (Nixon, 1999), tree fusing (Goloboff, 1999), and sectorial searches (Goloboff, 1999) under driven searches. For each search the best solution was reached 500 times before the search was stopped (command hits 500). In cases where more than one equally parsimonious MP tree was found, support values were plotted over a strict consensus tree summarizing competing hypotheses.

TABLE 1.

Primers and PCR profiles for DNA amplification. Conditions for denaturation, annealing, and extension steps for each cycle, followed by the number of cycles. All reactions included a 4 min initial denaturation at 94° C and a 6 min final extension at 72° C. Mitochondrial sequences encoding the mitochondrial genes 12S rDNA, 16S rDNA, COI and Cyt b, and nuclear genes BACH1, BNDF, kif24, MKL1, NTF3, PRLR, PTPN, RAG1, and SNCAIP, were employed for phylogenetic analyses.

We used PartitionFinder (version 1.1.1; Lanfear et al., 2012) to identify the optimal partition schemes for our data, and the best-fit nucleotide substitution model for each partition. We ran PartitionFinder allowing all available models of molecular evolution to be compared, and used the “greedy search” algorithm and linked branch lengths in calculations of likelihood scores. Bayesian information criterion (BIC) was adopted for selecting among alternative partitioning strategies. ML tree searches were performed in Garli (version 2.1) using the molecularevolution.org web platform (Zwickl, 2006; Bazinet et al., 2014). Starting tree topologies were generated using the stepwise-addition algorithm and the number of attachment points evaluated for each taxon to be added (attachmentspertaxon) was set to 57 (= two times the number of taxa + 1), meaning that all attachment points were evaluated for each taxon. We performed an adaptive best-tree search using a minimum of 10 replicates and determined the necessary number of search replicates to perform by calculating the number of replicates needed to recover the best topology with 0.95 probability. The relative support for each clade was assessed through 1000 nonparametric bootstrap replicates (Felsenstein, 2004). We summarized bootstrap results by plotting support values over the best tree using the Python package SumTrees of the DendroPy phylogenetic computing library (Sukumaran and Holder, 2010).

SPECIES ACCOUNTS

FIGURE 5.

Holotype of Tropidurus sertanejo, n. sp. (MZUSP 104273): (A) head in dorsal view; (B) head in ventral view showing intense pigmentation toward gular region; (C) head in lateral view; (D) dorsal body; (E) ventral body showing the typical dark flash marks on the underside of the thighs and cloacal flap of adult males.

FIGURE 6.

Allotype of Tropidurus sertanejo, n. sp. (MZUSP 104274): (A) head in dorsal view; (B) head in ventral view; (C) head in lateral view; (D) dorsal body illustrating the spotted pattern typical of the new species; (E) ventral body showing the unpigmented underside of the thighs and cloacal flap characteristic of females.

FIGURE 7.

(A) Dorsal, (B) ventral, and (C) lateral views of the head of the holotype (MZUSP 104273), illustrating in detail the scutellation of Tropidurus sertanejo, n. sp.

FIGURE 8.

Lateral view of the neck of (A) Tropidurus psammonastes (AMNH 138852) and (B) T. sertanejo, n. sp. (holotype MZUSP 104273), illustrating the lateral mite pockets associated with types I and J, respectively. See table 2 for details on mite pocket patterns in Tropidurus.

TABLE 2.

Expanded classification of mite pockets of the Tropidurus torquatus species group. Types A–H are treated in detail by Rodrigues (1987). We classify the mite pockets of T. psammonastes (originally described by Rodrigues et al., 1988) as Type I, and describe for the first time Type J as exclusive to T. sertanejo, n. sp. Internal surfaces of antegular-oblique neck, axillary, and inguinal mite pockets are coated with diminutive granular scales, unless noted.

FIGURE 9.

Boxplot of PC1 generated by principal component analysis of log-transformed morphometric variables as proxy for body size variation among Tropidurus species. Tropidurus sertanejo, n. sp., highlighted in dark gray.

TABLE 3.

Mean ± standard deviation, minimum and maximum values of morphometric measurements (in millimeters) of adult males of the Tropidurus torquatus species group. SVL, snout-vent length; HH, head height; HW, head width; EOS, ear opening—snout distance; AL, arm length; FAL, forearm length; HDL, hand length; THL, thigh length; SL, shank length; FOL, foot length; TL, tail length. The number of individuals measured followed by the number of individuals with fully grown tails. See Morphometrics in Material and Methods for details on statistical treatment of individuals with broken or missing tails.

Continued.

TABLE 4.

Mean ± standard deviation, minimum and maximum values of morphometric measurements (in mm) of adult females of the Tropidurus torquatus species group. SVL, snout-vent length; BH, body height; HH, head height; HW, head width; EOS, ear opening—snout distance; AL, arm length; FAL, forearm length; HDL, hand length; THL, thigh length; SL, shank length; FOE, foot length; TL, tail length. The number of individuals measured followed by the number of individuals with fully grown tails. See Morphometrics in Material and Methods for details on statistical treatment of individuals with broken, regrown or missing tails.

Continued.

FIGURE 10.

Scatterplot of PC1 and PC2 generated by the principal component analyses and LD1 and LD2 generated by the size-free discriminant analyses performed on the log-transformed morphometric variables of Tropidurus. Tropidurus sertanejo, n. sp., highlighted in dark gray. See table 5 for corresponding summary statistics.

The number of subdigitial lamellae on fourth toe had the strongest contribution on both LD1 and LD2, yet not allowing complete separation of any species in both axes. By plotting LD1 against LD2 we were able to identify two species groups that differed in number of scales. The first group included the psammophilous species T. cocorobensis, T. hygomi, and T. psammonastes, characterized by a higher number of subdigital lamellae in relation to the number of dorsals and ventrals. The second included all other species with a relatively lower number of subdigital lamellae. We observed that species with proportionally larger feet had higher numbers of subdigital lamellae on the fourth toe (Pearson correlation: males: r = 0.634, p = 0.008; females: r = 0.509, p = 0.044). Positive associations between body size and other meristic variables were similarly expected, but we failed to find significant correlations between SVL and number of dorsal, ventral, gular, and midbody scales. These results are in agreement with our previous observations that SVL and scales counts lack consistent correlation within species (results not shown), suggesting that the number of scales is determined from birth and shows no significant ontogenetic, static, or interspecific allometric trends. Although meristic variables are commonly used to distinguish lizard species, our results showed that we cannot rely on scale counts to separate T. sertanejo, n. sp., from most species of the T. torquatus group.

TABLE 5.

Summary of the principal component analyses and size free linear discriminant analyses performed with 11 morphometric measurements of male and female Tropidurus. PC, component loadings; LD, discriminant coefficients, EVL, eigenvalues; SD, standard deviations; % Variance, percentage of explained variances.

FIGURE 11.

Boxplot showing variation in scale counts among Tropidurus species (ordinated by mean): (Tcat) T. catalanensis, (Tchr) T. chromatops, (Tcoc) T. cocorobensis, (Tery) T. erythrocephalus, (Teth) T. etheridgei, (This) T. hispidus, (Thyg) T. hygomi, (Timb) T. imbituba, (Tins) T. insulanus, (Tita) T. itambere, (Tmon) T. montanus, (Tmuc) T. mucujensis, (Tore) T. oreadicus, (Tpsa) T. psammonastes, (Tser) T. sertanejo, n. sp. (highlighted in dark gray.), (Ttor) T. torquatus.

TABLE 6.

Mean ± standard deviation, minimum and maximum values of scale counts of males (M), females (F), and juveniles, subadults and adults whose sex could not be determined based on external morphology plus males and females (All) of the Tropidurus torquatus species group.

Continued.

TABLE 7.

Summary of the results of the nonparametric MANOVA performed on six meristic variables of Tropidurus.

PHYLOGENETICS

Alignment And Partitioning: Alignments of four mitochondrial and six nuclear loci included 19–28 species (69%– 100%), with average taxon coverage of 25.9 terminals (92.5%). Number of aligned sites varied from 477–1211 per locus, summing up to 7098 characters. Of those, 4977 were conserved among all species, 2078 were variable, and 1159 were parsimony informative; singletons totaled 917 sites (table 9). The proportion of parsimony-informative sites over the total number of aligned sites varied from 19.78%–32.32% to 4.93%–21.10% among mitochondrial and nuclear loci, respectively. Numbers of parsimony-informative characters were up to 6.5 times higher in mitochondrial loci when compared with nuclear fragments.

FIGURE 12.

Scatterplot of LD1 and LD2 generated by the discriminant analysis performed on the scale counts of Tropidurus. See table 8 for corresponding summary statistics. Tropidurus sertanejo, n. sp., highlighted in dark gray.

BIC analyses favored five distinct partitions of the mitochondrial and nuclear data sets, and 10 partitions of the total evidence data set. The number of models of nucleotide evolution selected varied from three to seven depending on the data set. In total, the models summed up 80, 90, and 118 parameters employed in subsequent ML analyses of the nuclear, mitochondrial, and total evidence data sets. Parameters relative to partition schemes and models of nucleotide evolution selected are shown in table 10.

Phylogenetic Results: MP and ML phylogenetic analyses consistently placed Uranoscodon superciliosus as sister taxon of all tropidurines, and more internally the trans-Andean Microlophus quadrivattatus was recovered as the sister taxon of the remaining cis-Andean tropidurine species. Plica plica was placed as closely related to Tropidurus spinulosus in all MP and ML trees, rendering a paraphyletic Tropidurus (figs. 13–15).

Tropidurus semitaeniatus was highly supported as sister taxon of a monophyletic T. torquatus species group. Within this group, T. hygomi formed the sister clade of all other species in nuclear and total evidence trees. Mitochondrial analyses placed this species either as closely related to T. itambere + T psammonastes (ML tree) or as sister of all species of the T. torquatus group but T. itambere + T. psammonastes (MP tree).

FIGURE 13.

Maximum parsimony (upper) and maximum likelihood trees (lower) of Tropidurus based on four mitochondrial (12S, 16S, CO1, Cyt b) loci. Numbers associated to nodes refer to bootstrap values.

TABLE 8.

Summary of the LDA performed on six meristic variables of Tropidurus. LD, discriminant coefficients; EVL, eigenvalues; SD, standard deviations; % Variance, percentage of explained variances.

Mitochondrial, nuclear, and total evidence trees conflicted with respect to the relationships of several taxa, including T. sertanejo, n. sp. The first incongruence concerns the placement of T. hispidus, T. insulanus, and T. oreadicus. Although these species have always formed a well-supported monophyletic group, T. hispidus + T. insulanus were recovered in a more derived position (in relation to T. oreadicus) in MP mitochondrial and total evidence analyses, whereas T. oreadicus + T. insulanus occupied that same position (in relation to T. hispidus) in ML mitochondrial and total evidence topologies. Nuclear trees either placed the three species in a polytomy (MP tree) or supported T. insulanus as the sister taxon of T. hispidus + T. oreadicus (ML tree) (figs. 13–15).

The second incongruence in topological arrangement involves the placement of T. montanus as sister species of T. mucujensis + T. erythrocephalus in all mitochondrial and total evidence trees, but recovered as closely related to (T. torquatus, (T. chromatops + T. etheridgei), (T. catalanensis + T. imbituba)) in MP and ML nuclear trees (figs. 13–15). Labile phylogenetic placement was also observed for T. cocorobensis. This species was recovered as closely related to T. chromatops + T. etheridgei in the MP mitochondrial tree ((T. torquatus + T. imbituba) (T. catalanensis (T. chromatops + T. etheridgei))) in ML mitochondrial and total evidence trees, as sister species of T. sertanejo, n. sp., in MP and ML nuclear trees, and, finally, as the sister of the clade (T. catalanensis ((T. chromatops + T. etheridgei) (T. imbituba + T. torquatus))) in the MP total evidence tree (figs. 13–15).

In all analyses, individuals of T. sertanejo, n. sp., clustered together forming a highly supported monophyletic group, however, the phylogenetic affinities of the species itself were conflicting. The new species was recovered in MP and ML nuclear trees as the sister taxon of T. cocorobensis, but placed as sister of the clade formed by (T. cocorobensis (T. catalanensis ((T. chromatops + T. etheridgei) (T. torquatus + T. imbituba)))) and (T. cocorobensis ((T. torquatus + T. imbituba) ((T. catalanensis (T. chromatops + T. etheridgei))))) in MP and ML total evidence analysis, respectively; and by ((T. montanus (T. erythrocephalus + T. mucujensis)) ((T. catalanensis (T. torquatus + T. imbituba) (T. cocorobensis (T. chromatops + T. etheridgei))))) and (T. cocorobensis ((T. torquatus + T imbituba) (T. catalanensis (T chromatops + T. etheridgei)))) in MP and ML mitochondrial analyses.

TABLE 9.

Taxon coverage, number of aligned, conserved, variable, parsimony informative sites and singletons present in the alignment of mitochondrial and nuclear loci.

Total evidence and topological conflict: Inasmuch as separate analyses were useful for understanding differences and inferring the sources of topological conflict (Nixon and Carpenter, 1996; Peloso et al., 2015; but see Grant and Kluge, 2003, and Fitzhugh, 2006, for a critique on this kind of data exploration), we favored the phylogenetic hypotheses based on the analyses of our concatenated data set following the principle of character congruence or total evidence (Carnap, 1950, who credited much earlier workers; subsequently championed by others including Kluge, 1989; Rieppel, 2005, and Fitzhugh, 2006). The ML total evidence tree was preferred as the best hypothesis representing the phylogenetic relationships of tropidurines (fig. 15).⁸ This decision was based on the fact that ML analyses incorporate models of nucleotide evolution that account for heterogeneous rates of nucleotide substitution within and among loci, and also alleviate the detrimental effects of multiple substitutions affecting phylogenetic reconstruction and branch-length estimation (Felsenstein, 2004; Gadagkar and Kumar, 2005; but see Padial et al., 2014, and Peloso et al., 2015, for alternative views).

The observation that trees generated by independent analyses of nuclear and mitochondrial sequence data conflicted with our total evidence hypotheses caught our attention. In a modelbased framework, the topological incongruities noted in the previous section of this study might result from at least one of the following factors: misspecification of models of nucleotide evolution (Lemmon and Moriarty, 2004), inadequate determination of partition schemes (Brown and Lemmon, 2007; Lanfear et al., 2012), or conflicting phylogenetic signal in mitochondrial versus nuclear loci (Leaché and McGuire, 2006; Leaché, 2010; Leaché et al. 2013, 2014). However, since topological disagreements were detected among phylogenetic trees resulting from ML and MP reconstructions, and MP analyses do not require distinct evolutionary models or partition schemes to be defined prior to conduction of phylogenetic reconstruction, these factors were considered unlikely to explain all the variation in the observed results, favoring the interpretation of conflicting phylogenetic signal. Furthermore, phylogenetic analyses of the genus Microlophus performed by Benavides et al. (2007, 2009) reveled that conflicting phylogenetic signals promoted by mtDNA introgression might be more frequent among tropidurines than previously thought. mtDNA introgression and other processes that prevent complete fixation of alleles (e.g., incomplete lineage sorting, rapid speciation events) should not be ruled out as potential factors generating conflict between mitochondrial and nuclear genes, and ultimately influencing the phylogenetic reconstruction of Tropidurus.

FIGURE 14.

Maximum parsimony (upper) and maximum likelihood trees (lower) of Tropidurus based on four nuclear (BACH1, kif24, NTF3, PRLR, PTPN, SNCAIP) loci. Numbers associated to nodes refer to bootstrap values.

FIGURE 15.

Maximum parsimony (upper) and maximum likelihood trees (lower) of Tropidurus on four mitochondrial (12S, 16S, CO1, Cyt b) and six nuclear loci (BACH1, kif24, NTF3, PRLR, PTPN, SNCAIP). Numbers associated to nodes refer to bootstrap values.

TABLE 10.

Data partitions and respective models of nucleotide evolution selected for ML analyses of mitochondrial and nuclear loci.

Comparison with previous phylogenetic hypotheses: Inasmuch as our taxon sampling was originally designed with the sole objective of inferring the phylogenetic placement of T. sertanejo, n. sp., within the T. torquatus group, the results produced revealed a paraphyletic Tropidurus. Prior analyses that recovered Tropidurus as nonmonophyletic were affected by significant amounts of homoplasy resulting from morphological convergence among some tropidurine genera (Frost, 1992; Harvey and Gutberlet, 2000). The combined analysis of morphological characters and mtDNA sequence data performed by Frost et al. (2001) was the first to successfully support a restricted, but monophyletic arrangement of Tropidurus. This phylogenetic hypothesis remained unrevised until Pyron et al. (2013) identified a sister-taxon relationship between T. bogerti and Uracentron; although a monophyletic clade including T. bogerti, Uracentron, Plica, and Strobilurus had been previously found by the mtDNA-only analysis of Frost et al. (2001). However, Pyron et al.'s results suffered from poor character sampling (12S was the only locus shared by all tropidurine taxa in their analysis). Thus, it remains unclear whether their hypothesis accurately reflects the tropidurine relationships or is a byproduct of insufficient data. Our total evidence analysis benefited from complete or nearly complete taxon coverage for most mitochondrial and nuclear loci, surpassing all previous studies. Herein, for the first time, an analysis with high nodal support recovered a paraphyletic Tropidurus, resulting from a close relationship between T. spinulosus and Plica, which formed the sister clade of the remaining Tropidurus (fig. 15).

We corroborate T. semitaeniatus as the sister species of the T. torquatus species group, as previously suggested by Frost et al. (2001). However, better progress in understanding the phylogeny of Tropidurus came with the identification of T. hygomi as sister clade to the remaining species of the T. torquatus group. Tropidurus chromatops had been previously placed at the base of this group by Frost et al. (2001), but those authors made clear that this result was suggestive—observing the weak nodal support resulting from the absence of molecular data for T. chromatops. In our total evidence tree, T. chromatops was recovered in a highly derived position, in close association with T. etheridgei. This arrangement is not surprising if we take into account the overall morphological similarity they exhibit. However, the position of this species pair is clearly unstable, and the same can be noted about the relationships of T. catalanensis, T. cocorobensis, T. imbituba, and T. torquatus.

Our larger, multilocus data set supported greater resolution in the weakly supported clade associated with stem 11 of Frost et al.'s (2001: fig. 6) total evidence tree. We found greater support for a clade including T. hispidus, T. insulanus, and T. oreadicus, and also the sister taxon relationship between T. montanus and the species pair formed by T. erythrocephalus and T. mucujensis. However, the sister-group relationships between T. torquatus and T. hispidus, and between T. insulanus and T. oreadicus, hypothesized by Frost et al. (2001), are contradicted by our results. We argue for the first time that T. torquatus is actually more closely related to the recently described T. imbituba, with both placed in a larger clade that includes (T. cocorobensis ((T. torquatus + T. imbituba) (T. catalanensis (T. chromatops + T. etheridgei)))), although part of the relationships in this clade remain poorly supported.

TABLE 11.

Uncorrected genetic distance among pairs of tropidurine species calculated for 12S sequences.

Continued.

Phylogenetic placement of T. sertanejo, n. sp.: Most mitochondrial and total evidence trees showed that samples of T. sertanejo, n. sp., from Ibotirama and Caetité, Bahia, formed reciprocally monophyletic groups, but our nuclear trees did not recover the same result. Because mitochondrial loci sort at a faster rate than nuclear loci (Avise, 1994; Funk and Omland, 2003; Zink and Barrowclough, 2008), these results are not surprising and probably imply maintenance of ancestral polymorphism.

We were successful in firmly placing T. sertanejo, n. sp., within the T. torquatus species group. However, the phylogenetic relationships of the new species in relation to other representatives of this group remain unclear. Our trees resulting from the analyses of the mitochondrial, nuclear, and total evidence data sets conflicted considerably, and support values of nodes linking T. sertanejo, n. sp., to other taxa were as low as 58% in our MP mitochondrial tree. In addition to the previously discussed disagreement between mitochondrial and nuclear loci affecting the phylogenetic reconstruction of the T. torquatus species group, the extensive topological conflict was also consistent with the observation of a relatively longer branch supporting the clade formed by samples of T. sertanejo, n. sp. The exclusion of multiple representatives of T. sertanejo, n. sp., from our analyses in an attempt to eliminate that long branch resulted in only slightly increased nodal support in most analyses (see appendices 9–12).

Our preferred tree (fig. 15; ML total evidence tree) placed T. sertanejo, n. sp., as the sister lineage of the clade (T. cocorobensis ((T. imbituba + T. torquatus) (T. catalanensis (T. chromatops, T. etheridgei)))) with poor nodal support. Uncorrected pairwise genetic distances calculated for 12S indicated that T. sertanejo, n. sp., is more similar genetically to T. montanus and T. mucujensis (table 11). Indeed, our morphological analyses also showed that the latter species share both axillary and inguinal mite pockets with T. sertanejo, n. sp., which might indicate their close relationship. Nevertheless, we failed to determine with confidence the closest relative of T. sertanejo, n. sp., because of the conflict between molecular data and also with respect to morphological characters. The putative association between the new species and other Caatinga endemics distributed along the Espinhaço mountain range should be carefully investigated in future analyses including additional data.

Classification of tropidurines and future steps: Although the phylogenetic hypothesis presented herein represents the best-supported phylogenetic tree currently available for tropidurines, and recovers a surprisingly new arrangement of several species and genera in this group, we refrain from proposing a new taxonomic classification. This decision is founded on the fact that a solid phylogenetic classification of tropidurines would require broader taxonomic sampling than currently represented in our analyses. Nomenclatural changes resulting from a taxonomically incomplete phylogenetic assessment of the group might result in taxonomic instability and is, thus, unjustified. Nevertheless, our study achieved the major goals of confirming the placement of T. sertanejo, n. sp., within the T. torquatus species group and testing the monophyly of the group itself. Since Frost (1992), the relationships within the T. torquatus group have been proven extremely hard to solve either employing morphological (Frost, 1992; Harvey and Gutberlet, 2000), molecular (Pyron et al., 2013; this study) or combined evidence (Frost et al., 2001). Future studies might benefit from the inclusion of additional molecular loci, additional morphological characters, and greater coverage of terminals.