Phylogeny and diversity of neotropical monkey lizards (Iguanidae: Polychrus Cuvier, 1817)

Neotropical monkey lizards (Polychrus) are arboreal lizards with compressed bodies, partially fused eyelids and strikingly long, whip-like tails. The eight currently recognized species occur in the lowlands of South and Central America. Based on the largest taxon and character sampling to date, we analyze three mitochondrial and one nuclear gene using Bayesian methods to (1) infer the phylogeny of Polychrus under both concatenated-tree and species-tree methods; (2) identify lineages that could represent putative undescribed species; and (3) estimate divergence times. Our species tree places P. acutirostris as the sister taxon to all other species of Polychrus. While the phylogenetic position of P. gutturosus and P. peruvianus is poorly resolved, P. marmoratus and P. femoralis are strongly supported as sister to P. liogaster and P. jacquelinae, respectively. Recognition of P. auduboni and P. marmoratus sensu stricto as distinct species indicates that the populations of "P. marmoratus" from the Amazon and the Atlantic coast in Brazil represent separate species. Similarly, populations of P. femoralis from the Tumbes region might belong to a cryptic undescribed species. Relative divergence times and published age estimates suggest that the orogeny of the Andes did not play a significant role in the early evolution of Polychrus.


Introduction
Neotropical monkey lizards Polychrus Cuvier, 1817 are restricted to South America on both sides of the Andes, except for P. gutturosus Berthold, 1845, which ranges from the Pacific coast in Ecuador and Colombia into Central America as far north as Nicaragua. Two species, P. gutturosus and P. auduboni Murphy et al., 2017, have colonized islands off the coast of South America-Gorgona Island in Colombia and Trinidad and Tobago, respectively [1,2]. The eight recognized species [1,3] of monkey lizards are remarkable among New World lizards in that they resemble Old World chameleons in both morphology and behavior [3][4][5]; they are arboreal, slow-moving lizards with a laterally compressed body and cone-shaped eyes with a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Liolaemus lineomaculatus Liolaemus scapularis Phymaturus somuncurensis Pristidactylus scapulatus Pristidactylus torquatus Thus, in spite of its relatively low diversity (8 species), a molecular phylogeny of Polychrus based on a complete dataset of more than two mitochondrial genes and more than four species has not been published. Moreover, no attempts have been made to explore the genetic variation and diversity within Polychrus despite the wide distribution of most species (Fig 1). In this paper we analyze the phylogenetic relationships among all currently recognized species of monkey lizards based on broad geographic sampling. Using one nuclear and three mitochondrial genes, we (1) test the monophyly of Polychrus and its currently recognized species based on the largest taxon and character sampling to date; (2) identify lineages that could represent putative undescribed species; and (3) co-estimate divergence times and a species tree of Polychrus under a coalescent model.

Fieldwork and data sampling
A total of 35 specimens representing different species of Polychrus were collected during several field trips to different localities in Bolivia, Ecuador, Costa Rica, Panama, and Peru. After lethal anesthetization of voucher specimens with an intracoelomic injection of Nembutal or T611, tissue samples were taken from the thigh muscle and the specimens were stored in 70% ethanol and deposited in the collections of the Museo de Zoología de la Pontificia Universidad Católica (QCAZ), Quito, Ecuador; the Centro de Ornitología y Biodiversidad (CORBIDI), Lima, Peru; the Zoologisches Forschungsmuseum Alexander Koenig (ZFMK), Bonn, Germany; the Museum of Southwestern Biology (MSB; POE field numbers) at the University of New Mexico, Albuquerque, United States; and the Museum of Comparative Zoology at Harvard University, Cambridge, United States. Six additional tissue samples were taken from specimens previously housed at ZFMK (Table 1).
We used ArcGis to generate a distribution map of all species of Polychrus based on locality data from databases of the collections listed above, as well as data from the literature and Vert-Net (www.vertnet.org).
Voucher specimens and tissue samples were obtained following ethical and technical protocols [27]. Collecting and export permits were kindly provided by the Ministerio de Agricul
Genomic DNA was isolated from frozen muscle or liver tissues using a guanidinium isothiocyanate extraction protocol. Polymerase Chain Reaction (PCR) amplification of gene fragments was performed in a final volume of 25 μl reactions using 1X PCR Buffer (-Mg), 3 mM MgCl 2 , 0.2 mM dNTP mix, 0.2 μM of each primer, 0.1 U/μl of Platinum1 Taq DNA Polymerase (Invitrogen, Carlsbad, CA) and 1 μl of extracted DNA. Negative controls were run on all amplifications to check for contamination. Primers and PCR amplification protocols are presented in Table 2. Polymerase chain reaction products were analyzed on 1% agarose gels by horizontal electrophoresis (the target fragment size was estimated from molecular weight markers), using SYBR1 Safe (Invitrogen, Carlsbad, CA) staining, and analyzed with a Molecular Imager1 Gel Doc TM XR+ Imaging System (Bio Rad, Hercules, CA). Amplified products were treated with ExoSAP-IT (Affymetrix, Cleveland, OH) to remove remaining dNTPs and primers, and extraneous single-stranded DNA produced in the PCR. Double stranded sequencing of the PCR products were performed in both directions by Macrogen Inc. New sequences were deposited in GenBank (Table 1).

Alignment, model selection, and phylogenetic analyses
Data were assembled and aligned in Geneious v9 [33] under default settings for the alignment program MAFFT [34]. Ribosomal (12S and 16S) gene regions with multiple gaps were realigned to minimize indels and optimize nucleotide identities among different individuals. ND2 and RAG1 sequences were translated into amino acids for confirmation of alignment. The best-fit nucleotide substitution models and partitioning scheme were chosen simultaneously using Par-titionFinder v1.1.1 [35] under the Bayesian Information Criterion (BIC). The "greedy" algorithm was used with branch lengths of alternative partitions "linked" to search for the best-fit scheme.
A Bayesian inference method was used to obtain the optimal tree topology of the combined, partitioned dataset using MrBayes v3.2.1 [36]. All parameters except topology and branch  lengths were unlinked between partitions, and rate variation (prset ratepr = variable) was invoked. Four independent runs, each with four MCMC chains, were run for 10 7 generations, sampling every 1,000 generations. Results were analyzed in Tracer v1.6 [37] to assess convergence and effective sample sizes (ESS) for all parameters. Additionally, we verified that the average standard deviation of split frequencies between chains and the potential scale reduction factor (PSRF) of all the estimated parameters approached values of 0.01 and 1, respectively. Of the 10,000 trees resulting per run, 25% were discarded as "burn-in". The resultant 30,000 trees were used to calculate posterior probabilities (PP) for each bipartition in a maximum clade credibility tree in TreeAnnotator v1.8.3 [38]. Phylogenetic trees were rooted with the acrodont iguanians Brookesia and Uromastyx [39]

Chronophylogenetic analysis
We estimated a Polychrus species tree from the mitochondrial and nuclear trees under a coalescent model-and simultaneously estimated relative divergence times-using the Starbeast method (Heled & Drummond, 2010) implemented in Beast 1.8.3. For this analysis we included only species of Polychrus (i.e., tree root was estimated by the clock model [41]). Models of nucleotide substitution and partition scheme were selected in PartitionFinder as explained above. The analyses were conducted under a model with uncorrelated substitution rates among branches and the rate for each branch independently drawn from an underlying lognormal distribution (Drummond et al., 2006). Because our sampling of the "marmoratus" species complex (i.e., including P. auduboni) was limited (Fig 1), we considered this complex as a single species for this analysis. Previous studies differing in gene data, taxon sampling and analytical methods have produced a wide range of age estimates and sister taxa for Polychrus (Table 3). Therefore, here we consider that reliable internal and nearby external calibrations are not yet available. To reflect the absence of calibration dates, default parameter priors were used except for the mean of branch rates parameter (ucld.mean), which was fixed to 1.0 resulting in time being measured in units that have been arbitrarily chosen so that 1 time unit corresponds to the mean time required for the accumulation of 1 substitution per site (Drummond et al., 2006; Drummond and Rambaut, 2007). Search parameters and tree construction were similar to the Bayesian analysis described above, with three runs and a 'Yule Process' species tree prior under the 'Piecewise linear & constant root' population size model. Results were analyzed in Tracer v1.6 [37] to assess convergence and effective sample sizes (ESS) for all parameters. All phylogenetic analyses were carried out in the CIPRES Science Gateway [42].

Species delimitation analysis
We identified clades or single branches within currently recognized species of Polychrus as putative species if (1) branches were much longer with respect to other branches within the clade corresponding to the currently recognized species (see below), and (2) their geographic distribution was disjunct with respect to other terminals within the currently recognized species. We evaluated diagnosability and monophyly of putative species using the Species Delimitation plugin [45] in Geneious 7.1.9 [33]. We calculated (1) the mean probability of correctly identifying an unknown member of the putative species using the criterion that it must fall within, but not sister to, the (putative) species clade in a tree (P ID(strict) ); (2) the probability that a putative species has the observed degree of distinctiveness due to random coalescent processes (P RD ); and (3) the probability of reciprocal monophyly under a random coalescent model (Rosenberg's PAB [46]). Because this method is applied to gene trees we chose 16S, the gene region for which we had the largest number of sequences (N = 55; Table 1) after incorporating recently published data [1], to compute an ultrametric (time) tree in Beast 1.8.3 (Yule speciation process; lognormal uncorrelated relaxed clock). We performed four independent runs for 10 7 generations each, sampling every 1,000 generations. Results were analyzed in Tracer v1.6 [37] to assess convergence and effective sample sizes (ESS) for all parameters. After a 10% "burn-in", trees were used to calculate posterior probabilities (PP) for each bipartition in a maximum clade credibility tree in TreeAnnotator v1.8.3 [38].

Phylogeny and divergence times
Monophyly of Polychrus is strongly supported (PP = 1) by the concatenated gene tree (CGT), which includes representatives of most major iguanid lineages (Fig 2). This tree is similar in topology to the species tree (SPT; Fig 3) in that it strongly supports (PP = 1) a sister taxon relationship between P. marmoratus and P. liogaster and between P. femoralis Werner, 1910 and P. jacquelinae. Nonetheless, CGT and SPT have two major differences. First, according to the CGT, P. gutturosus is sister to all other species of Polychrus, which are clustered in a weakly supported (PP = 0.62) clade, where P. acutirostris Spix, 1825 is sister to a clade (PP = 0.75) composed of two subclades-(P. marmoratus, P. liogaster) with PP = 1, and ((P. peruvianus, (P. femoralis, P. jacquelinae)) with PP = 0.50. In contrast, the SPT has P. acutirostris as sister to a strongly supported clade (PP = 1) containing all other species of Polychrus. In this clade, P. gutturosus is sister to a clade (PP = 0.43) composed of two subclades-(P. peruvianus, (P. marmoratus, P. liogaster)) with PP = 0.40, and (P. femoralis, P. jacquelinae).
The SPT also shows that the split between the Pacific-western Andean species Polychrus femoralis and Amazonian P. jacquelinae occurred later than the split among Amazonian P. peruvianus, P. marmoratus and P. liogaster. In addition, the two splits corresponding to the three putative species currently recognized as P. femoralis (see below) are more recent (Fig 3). In contrast to the CGT, where the "femoralis putative species" from the Tumbes region (i.e., extreme northwestern Peru and southern Ecuador) is sister (PP = 1) to the other two putative species, in the SPT the "species" from northern Peru is sister to the clade (PP = 0.62) formed by the two "species" from the Tumbes region and western Ecuador.

Phylogeny of Polychrus and divergence times
As expected by the relatively low number of characters and loci included in this study, the phylogenetic relationships among major lineages of iguanid lizards are poorly resolved (Fig 2);

Phylogeny of monkey lizards
Polychrus is weakly supported (PP = 0.24) as sister to the clade (Iguana, Basiliscus). Whether Polychrus is sister to Anolis remains controversial (see Introduction). Nonetheless, in agreement with previous hypotheses [9,10,18], here we show that Polychrus is monophyletic based on phylogenetic analyses of the largest taxonomic and geographic sampling of Polychrus to date, including all species and samples from throughout the range of the clade. Despite our sampling effort, the relationships among species of Polychrus were only partially resolved. Frost et al. (2001) inferred P. gutturosus as sister to all other species of Polychrus recognized at the time (i.e., excluding P. auduboni, P. jacquelinae, P. liogaster, and P. peruvianus). Although our CGT weakly supports this relationship, our SPT strongly supports a different scenario where P. acutirostris is sister to all other species of Polychrus (Fig 3). In both CGT and SPT, the relationships among remaining species remain unclear except for the sister taxon relationship of both (P. marmoratus, P. liogaster) and (P. jacquelinae, P. femoralis). Moreover, no pair of sister species is strongly supported (i.e., all PP values 0.71) by the 16S gene tree (Fig 4).
The age of Polychrus has been estimated by several authors using different methods, as well as different taxon and character sampling strategies. These estimates are incongruent, ranging between~32 and~125 million years ( Table 3). The limited taxon sampling of Polychrus (N = 1-4 species) and the lack of fossil calibrations within Polychrus in these studies evoke little confidence in any of these estimates and suggest that preference among them is arbitrary. In the absence of reliable calibration points, or reliable divergence time estimates, only arbitrary calibrations (e.g., ucld.mean fixed to 1.0) resulting in relative age estimates should be adopted. These estimates, however, still contain useful information on the relative timing of events (e.g., [47]). Based on the chronophylogenetic species tree analysis, here we conclude that the split between two species from west of the Andes occurred earlier than the split between two eastern Andean species, and that lineage divergence within P. femoralis is more recent. We refrain from drawing more time-related conclusions because they would be based on observations of poorly supported relationships (Figs 2 and 3). Taxon numbers are the same as those presented in the phylogenetic tree in Fig 4. The number of specimens per species (N); average pairwise tree distance among members of a putative species (D intra ); average pairwise tree distance between members of one putative species and members of the closest second putative species (D inter ); D intra /D inter ratio; the mean (95% confidence interval) probability of correctly identifying an unknown member of the putative species using the criterion that it must fall within, but not sister to, the species clade in a tree P ID(strict) ; the probability that a clade has the observed degree of distinctiveness due to random coalescent processes (P RD ); and the probability of reciprocal monophyly under a random coalescent model (Rosenberg's P AB ) are presented. https://doi.org/10.1371/journal.pone.0178139.t004 Phylogeny of monkey lizards  Table 4.

Biogeography of Polychrus
Given that our inferred phylogenies did not fully resolve the relationships among all species of Polychrus with high support, we refrained from carrying out phylogeny-based biogeographic analyses, such as ancestral area reconstruction. Nonetheless, our results provide a few insights into the biogeography of monkey lizards. First, the strongly supported position of P. acutirostris in the SPT (Fig 3) suggests that Polychrus has its origins in South America rather than Central America, because this species is presently widespread along the South American diagonal belt of open formations that goes from Argentina and Bolivia to northeastern Brazil, encompassing the Chaco, Cerrado, and Caatinga biomes [48]. Second, our hypotheses (Figs 2 and 3) do not support a basal split between species presently occurring west (P. femoralis, P. gutturosus) and east (all other species) of the Andes, suggesting that the orogeny of the Andes did not play a major role in the early evolution of Polychrus. Even though this is in agreement with most age estimates of Polychrus (Table 3), we believe that this biogeographic scenario should be tested more rigorously.

Diversity of Polychrus
Monophyly and diagnosability according to DNA sequence data are commonly used lines of evidence in species delimitation. As species properties, however, they are neither infallible nor essential (i.e., their absence does not constitute evidence contradicting a hypothesis of lineage separation) [49]. In this paper we explored species limits within currently recognized species of Polychrus by calculating monophyly and diagnosability statistics on a 16S gene tree (Table 4). A growing body of evidence suggests that the diversity of vertebrates from tropical South America is underestimated as widely distributed species usually represent species complexes, in which cryptic or poorly studied species await discovery [50][51][52][53]. Among species of Polychrus, P. acutirostris and P. marmoratus have large geographical ranges (Fig 1), which makes them suitable for species delimitation analyses. Cryptic diversity within P. marmoratus was recently reported by Murphy et al. (2017) on the basis of morphology and a phylogeny of two mitochondrial genes. They recognized populations from Trinidad, Tobago and northern Venezuela as a separate species, P. auduboni. Following Hoogmoed [54], Murphy et al. (2017) also restricted the name P. marmoratus to the populations in Guyana and Suriname (and possibly French Guyana and northern Brazil), and suggested that two additional species might occur in southeastern Brazil (see also [5]). However, the phylogenetic position of populations from the Amazon region (Brazil and Peru) were not clearly resolved [1]. Here we present phylogenies with better resolution (Figs 2-4), which along with the results of the species delimitation analyses (Table 4), support recognition of P. auduboni and restriction of P. marmoratus to Guyana and Suriname. These taxonomic changes leave populations of "P. marmoratus" from the Amazon region in need of a different specific name. Populations of "P. marmoratus" from southeastern Brazil might also represent different species, for which the names P. virescens Schniz, 1822 and P. neovidanus Wagler 1833 are available [1]. However, here we refrain from proposing additional taxonomic changes because we believe that both denser molecular and geographical sampling, as well as detailed morphological analyses are necessary to elucidate more objectively the taxonomic status of other populations traditionally assigned to P. marmoratus, as well as P. acutirostris.
Among species with more restricted distribution ranges, neither the P ID(strict) or Rosenberg's P AB statistics supported recognition of any of the three subclades within P. gutturosus (Panama, Costa Rica, and Ecuador; Fig 4) as separate species. In contrast, the same statistics suggest that populations of P. femoralis from the Tumbes region might belong to a cryptic undescribed species. If additional lines of evidence support this hypothesis (C. Koch, O. Torres-Carvajal and P.J. Venegas, unpubl. data), the name P. femoralis should be restricted to populations from western Ecuador based on type locality (Guayaquil, Ecuador). In this case, the disjunct set of populations from the Pacific slopes of the Andes in northern Peru (Piura and Lambayeque departments, Fig 4) could be either conspecific with P. femoralis, or represent a distinct undescribed species. We have eschewed describing new taxa in this paper, as our aim was to provide a general framework for future studies. Additional lines of evidence will lead to a better informed species delimitation process for Neotropical monkey lizards.