We assembled a genetic dataset that included sequences for the subfamily Cercosaurinae available on GenBank. Additionally, we newly sequenced 38 samples of nine genera (Anadia, Cercosaura, Euspondylus, Pholidobolus, Potamites, Proctoporus, the new genus described herein and two unnamed genera; Table 1, Suppl. material 1: Fig. S1) deposited in the SMF, NMP, IWU, JCM, and MUSM collections (for collection acronyms see below). All genera that are presently recognised to form the content of the subfamily (Sánchez-Pacheco et al. 2017b) were included in the dataset, represented as follows: Anadia (3 of 18 described species included), Andinosaura (9 of 11 described species included), Cercosaura (10 of 14 described species included), Echinosaura (5 of 7 described species included), Euspondylus (1 of 11 described species included), Gelanesaurus (2 of 2 described species included), Macropholidus (3 of 4 described and one yet undescribed species included), Neusticurus (2 of 5 described species included), Oreosaurus (4 of 6 described and one yet undescribed species included), Petracola (2 of 4 described species included), Pholidobolus (8 of 9 described and one yet undescribed species included), Placosoma (2 of 4 described species included), Potamites (6 of 7 described species included), Proctoporus (15 of 17 described and eight yet undescribed species included), Riama (14 of 15 described and three yet undescribed species included). We also included the unnamed clades identified by Torres-Carvajal et al. (2016) that could not be assigned to any described genus.

To avoid confusion, sequences of all loci were matched with the sample or museum code of the specimen to which they belonged as was used in the original reference (when available). As Cercosaurinae still contains non-monophyletic taxa (see below), we avoided combining sequences of more individuals into chimeric samples, even if they putatively belonged to the same species. As a result, each terminal in the tree represents an existing voucher specimen or tissue sample (Suppl. material 1: Table S1). For those species whose monophyly has been previously confirmed, we included only one (or a few) samples (e.g., Cercosaura ocellata; Sturaro et al. 2017).

Genomic DNA was extracted from ethanol-preserved tissue samples using a Geneaid kit. We PCR-amplified up to four loci, three from the mitochondrial DNA (mtDNA): 12S rRNA (12S), 16S rRNA (16S), cytochrome b (cytb), and the oocyte maturation factor MOS (cmos) from the nuclear DNA. Sanger sequencing of both the forward and reverse strands was carried out at Macrogen (Amsterdam, The Netherlands) using the same primers as for the PCRs. Details on the primers and amplification conditions are given in Table 2. Newly produced sequences were edited and contigs assembled in Geneious v.6 (Kearse et al. 2012). MAFFT v.7 (Katoh and Standley 2013) was used to align all loci individually with the ‘auto’ option selected for all. The Q-INS-I option that considers the secondary structure of RNA and that would therefore have been more suitable for the 12S and 16S datasets could not be used as the number of sequences in both datasets (332, respectively 343) exceeded the allowed limit. The alignments of cytb and cmos were translated into amino acids and no stop codons were detected, suggesting we did not amplify pseudogenes. We applied Gblocks (Castresana 2000) to the 12S and 16S alignments to trim regions that aligned ambiguously. We trimmed the tRNA-end of the ND4 and used only its coding part for the analyses. The final concatenated dataset for the phylogenetic analyses consisted of 2217 bp composed of the following loci with lengths given in parentheses: 12S (325 bp), 16S (454 bp), cytb (307 bp), ND4 (694 bp), and cmos (437 bp).

The dataset was partitioned by gene. Models of sequence evolution were assessed for each partition by Partition Finder v.1.1 (Lanfear et al. 2012) with the following settings: branch lengths linked, models available in BEAST evaluated, model selection based on BIC. The models identified as most suitable were as follows: GTR+I+Γ for the 12S, 16S, and ND4, SYM+I+Γ for cytb and HKY+I+Γ for cmos. As outgroups, we used 21 species representing the genera Rhachisaurus, Gymnophthalmus, Alopoglossus, Riolama, Ecpleopus, and Bachia that are known to be closely related to the cercosaurine genera but not being part of the subfamily (Pyron et al. 2013; Kok 2015). The outgroup species are also listed in Suppl. material 1: Table S1. In total the dataset consisted of 357 samples, of which 26 represented the outgroup taxa.

Phylogenetic analyses were conducted by means of maximum likelihood (ML) and Bayesian inference (BI). The ML analysis was conducted using RAxML-HPC2 v.8.2.9 (Stamatakis 2014) with a heuristic search that included 100 random addition replicates and 1000 thorough bootstrap pseudoreplications. We applied the GTR+CAT model to all partitions as the CAT model has been shown to be a faster and computationally less demanding alternative to the Γ model (Stamatakis 2006; Stamatakis et al. 2008). We skipped the +I parameter because the 25 default rate categories of the CAT model account for potentially invariant sites (Stamatakis 2006).

The Bayesian analyses were conducted using MrBayes v.3.2 (Ronquist et al. 2012) and BEAST 2.4.5 (Bouckaert et al. 2014). MrBayes settings were the following: GTR as the preferred substitution model for the mtDNA genes and HKY for cmos, invariable proportions of among-site rate variation and a gamma-distributed rate parameter applied, ploidy set to haploid for the mtDNA genes, four parallel runs ran each with four chains, number of generations set to 10⁸ with a 10⁵ sampling frequency, 10% of trees discarded as burnin. Stationarity was confirmed by the value of average standard deviations of the split frequencies being lower than 0.01. Convergence of the four runs was confirmed by the values of PSRF (potential scale reduction factor) reaching 1.00. Estimated parameter values were inspected and a 50% majority-rule consensus tree was generated in MrBayes.

The second Bayesian analysis was run in BEAST. In order to avoid over-parameterisation caused by the large size of the dataset, we applied the HKY model for all partitions instead of the GTR as preferred by PartitionFinder. The Γ parameter was selected to have four categories and shape estimated. We applied the Yule process tree prior with uniformly distributed birth rate (lower: 0, upper: 1000) and an independent relaxed uncorrelated lognormal clock prior for each partition. Ambiguities in the cmos alignment coded by the IUPAC ambiguity codes were accounted for. Clock rates were set to have lognormal distributions with the mean = 1 and st. dev. = 1.25 for the mtDNA genes and mean = 0 and st. dev. = 1.0 for the cmos relative to the first partition of the concatenated alignment, which was the 12S. Standard deviation of the clock parameter (among-lineage rate heterogeneity) was for all partitions estimated with an exponential distribution with the mean = 1. Four independent runs were made, each for 2.5×10⁸ MCMC generations and parameters logged every 10⁵ generation. 10% of sampled trees were discarded from each analysis as burnin. Stationarity, convergence of the runs, and effective sample sizes (ESS) of all parameters were inspected in Tracer v.1.5 (Rambaut and Drummond 2007). Post-burnin posterior trees were combined using LogCombiner v.2.4.5 and the maximum clade credibility (MCC) tree with mean node heights was identified and posterior probabilities calculated using TreeAnnotator v.1.7.5.

All phylogenetic analyses were run through the CIPRES Science Gateway (Miller et al. 2010). Tree nodes were considered strongly supported when they received ML bootstrap support ≥ 70% and posterior probability (pp) values inferred by the two BI analyses ≥ 0.95.

Genetic distances between the clades of cercosaurines were calculated for all genetic markers analysed except cytb using MEGA 6 (Tamura et al. 2013) with the pairwise deletion option selected.

The format of the descriptions and terminology of the morphological characters follow mostly Oftedal (1974), Chávez et al. (2017), and Sánchez-Pacheco et al. (2017b). Specimens were fixed in 96% and stored in 70% ethanol. Sex and maturity of specimens were identified through dissection of gonads. Specimens with SVL ≤ 30.2 mm were considered juveniles. The following metric characters were taken using a digital caliper and dissecting microscope (to the nearest 0.1 mm):

SVL snout-vent length – distance from the snout tip to cloaca;

HL head length – distance from the snout tip to the angle of jaw;

HW head width – greatest width of the head;

HD head depth – greatest depth of the head;

TL tail length – distance from cloaca to the tail tip, if original;

E–N eye-snout distance – straight distance from the snout tip to anterior corner of eye;

FLL forelimb length – from axilla to tip of distal claw;

HLL hindlimb length – from groin to tip of distal claw;

AGD axilla-groin distance – distance between limbs;

hemipenis length distance from hemipenial base to distal margin of hemipenial lobes.

Meristic and qualitative pholidotic characters were counted and evaluated as follows: number of supralabials from the rostral to the mouth corner, last labial defined by its considerably larger size compared with the posteriorly adjacent shields; dorsal scales by the number of transverse rows of dorsal scales from the third row behind the interparietal to the level of the rear edge of the hindlimb; ventral scales, the number of transverse rows of ventral scales (from collar to the anterior row of anal scales); lateral scales, the number of longitudinal rows of considerably smaller lateral scales lying between larger dorsal and ventral scales at midbody; scales around midbody; preanal plates are the number of large plates in the posterior row of anal scales; number of lamellae under Finger IV including the number of single and divided lamellae (left/right, lamellae divided into segments counted as one individual lamella); number of lamellae under Toe IV refers to the number of single and divided lamellae (left/right, lamellae divided into segments counted as one individual lamella); number of preanal pores (left/right).

Description of colouration in life was based on field notes and photographs. Collection acronyms are: MUSM Museo de Historia Natural Universidad Nacional Mayor de San Marcos, Lima, Peru; NMP6V National Museum Prague, Prague, Czech Republic; SMF Senckenberg Forschungsinstitut und Naturmuseum, Frankfurt, Germany. Field number codes are: IWU Illinois Wesleyan University; JCM Juan Carlos Cusi collection. Threat status was evaluated using the IUCN criteria (2016). High-resolution versions of photographs presented in this article and additional pictures of the type specimens have been uploaded to MorphoBank (project number: 3136; http://www.morphobank.org) where they are available for download.

All drawings were made by the senior author using a stereomicroscope and a camera lucida. Maps were made with QGIS (Quantum GIS Development Team 2014).

Figure 1. 

Map showing localities of samples newly sequenced for this study. Locality numbers correspond to those in Table 1. Localities of the new genus described here, Selvasaura gen. n., are marked with triangles; red triangle indicates the type locality of its type species, S. brava sp. n.; green triangle locality of paratypes MUSM 32718 and NMP6V 75655; yellow triangles localities published by Torres-Carvajal et al. (2016): A Provincia de Napo, Wildsumaco Wildlife Sanctuary, Ecuador B Provincia de Zamora Chinchipe, El Pangui, Ecuador C region San Martin, Provincia Mariscal Cáceres, Laurel, Peru. White circles denote major cities.

All three analyses performed here resulted in topologies concordant with previous studies (Torres-Carvajal et al. 2016, Sánchez-Pacheco et al. 2017b). The subfamily Cercosaurinae was monophyletic, although the pp support from the BEAST analysis did not exceed the 0.95 threshold (ML bootstrap: 100; MrBayes pp: 1; BEAST pp: 0.92; nodal support values in the same order hereafter; Suppl. material 1: Figs S1–S3). Most of the cercosaurine genera were strongly supported in all the analyses: Anadia (100, 1, 1); Andinosaura (86, 1, 1); Cercosaura (78, 1, 1); Euspondylus (98, 1, 1); Gelanesaurus (100, 1, 1); Macropholidus (92, 1, 1); Neusticurus (98, 1, 1); Petracola (93, 1, 1); Pholidobolus (63, 0.97, 0.96); Placosoma (100, 1, 1); Potamites (93, 1, 1); Riama (100, 1, 1). Two undescribed clades recovered by Torres-Carvajal et al. (2016) and referred to as Unnamed clade 2 and 3 were also strongly supported here: Unnamed clade 2 (100, 1, 1), Unnamed clade 3 (96, 1, 1), the latter being part of the ramification of the new genus described herein. Additionally, we identified yet another clade that may deserve recognition as an independent genus and that we term tentatively Unnamed clade 4 (support 91, 1, 1) and whose phylogenetic affinities to the other genera remained unclear due to low support (Figs 2, 3). Unnamed clade 4 consisted of two samples collected by EL and collaborators in cloud forests in the surroundings of Alfamayo (Region of Cuzco, Peru) and in a montane forest close to the National Park Yanachaga-Chemillén (Region Pasco, Peru).

Figure 2. 

Phylogenetic tree showing relationships between cercosaurine genera or, in cases when genera were not recovered as monophyletic, their major lineages. The tree is a strict consensus tree based on the results of three analytical approaches undertaken: ML, MrBayes, BEAST. The 24 lineages shown were supported in all three phylogenetic analyses. Relationships between genera are shown as dichotomies only for nodes that were strongly supported in all three analyses; otherwise, nodes were collapsed into polytomies to emphasise how little we can tell about the phylogeny of the subfamily Cercosaurinae with the data currently available. Outgroups are not depicted. For a tree that shows variability within genera see Fig. 3, for full trees see Fig. S1–S3. The pie charts on the right show i) species richness of the genera indicated by circle size with the proportion of species included in the analyses highlighted in red (left column), and ii) number of samples (log scale) available for each genus indicated by circle size with the proportion of material newly sequenced in this study in red (right column).

Figure 3 

. Maximum clade credibility tree for 107 species (both described and candidate) of the subfamily Cercosaurinae from the BEAST analysis. The dataset for the analysis contained 357 samples with most species being represented by multiple samples, but for visual purposes only one sample was retained for each species in this tree. Nodal support is shown in the ML/MrBayes/BEAST order; supported nodes are marked with asterisks, unsupported with dashes. Monophyletic groups at the genus level are highlighted by grey rectangles. Vertical grey bars connect species that supposedly belong to one genus, but whose monophyly was not supported in any of the phylogenetic analyses: the genera Proctoporus, Echinosaura, and Oreosaurus. Outgroups are not depicted. For a full BEAST tree see Fig. S3.

Monophyly was not supported for three of the described genera (Fig. 3). First, one species of the genus Echinosaura, E. sulcarostrum, did not cluster with the remaining four species of the genus included in the dataset and which formed a clade (100, 1, 1; as previously found by Torres-Carvajal et al. 2016). It was instead topologically closest to Euspondylus, although support of this sister relationship was low in two of the analyses (35, 0.96, 0.92). Second, monophyly of the recently resurrected genus Oreosaurus (Sánchez-Pacheco et al. 2017b) was also questionable. Although the topology of the BEAST tree shows all Oreosaurus species to form one group, monophyly of this group was not supported in any of the analyses and the phylogenetic position of O. serranus was unstable across the analyses. The other four Oreosaurus species that were included in the dataset formed a clade (80, 1, 1). Third, in concordance with Torres-Carvajal et al. (2016), but contrary to Goicoechea et al. (2012) and Sánchez-Pacheco et.al. (2017b), monophyly of Proctoporus was found to be questionable as the genus was supported only in the MrBayes analysis while it received no support in the ML and BEAST analyses (13, 0.95, 0.51). The Proctoporus species formed five well supported groups: i) a clade of P. pachyurus, P. oreades, P. spinalis, P. rahmi, P. chasqui, P. sucullucu, and three yet undescribed species (support 88, 1, 1; termed Proctoporus pachyurus group); ii) a clade comprising a single species, P. xestus (98, 1, 1); iii) a clade of P. carabaya, P. kiziriani, P. iridescens, P. lacertus (98, 1, 1; termed Proctoporus lacertus group); iv) a clade of P. bolivianus and one undescribed species (93, 1, 1; termed Proctoporus bolivianus group); v) a clade of P. unsaacae, P. guentheri, P. laudahnae, and four undescribed species (96, 1, 1; termed Proctoporus guentheri group). Mutual relationships between these five Proctoporus groups as well as their relationships to the other cercosaurine genera remained unresolved.

Higher-level relationships between the cercosaurine genera were difficult to infer for the generally low node support at this phylogenetic depth. Only a few clades could be identified that were common to the three different phylogenetic analyses undertaken (Fig. 2). In all analyses, Neusticurus was sister to Placosoma (100, 1, 1) and the clade of these two was sister to all the remaining genera of Cercosaurinae (76, 0.99, 1). Of the remaining genera, only the sister pair Pholidobolus/Macropholidus was recovered in all analyses with strong support (100, 1, 1) and a large clade comprising the five Proctoporus groups, Potamites, Cercosaura, Petracola, Pholidobolus, Macropholidus, Anadia, Euspondylus, “Echinosaura” sulcarostrum, both Oreosaurus lineages, Unnamed clades 2 and 4, and the new genus described herein (100, 1, 1). Otherwise, no other genera clustered into clades that would be supported by all three phylogenetic approaches.

Phylogenetic affinities of the new genus described herein to the other cercosaurine genera remained unresolved (Fig. 3). Although it was reconstructed as a sister lineage to the genus Potamites in all analyses, the topology was not supported in any of them (47, 0.55, 0.89). Within the genus, all analyses unambiguously identified four well genetically differentiated lineages (Fig. 4), albeit their mutual relationships remained unresolved due to the lack of nodal support. The first was a cluster of five samples from the PPPF, Peru that represent the new species described below, the second of one sample from El Pangui, Zamora-Chinchipe Province, Ecuador (voucher QCAZ 12891; sample code Cerc_s3_5), the third of three samples from Laurel, Mariscal Cáceres Province, Peru (vouchers CORBIDI 15117–19; sample codes Cerc_s3_1–3), and the fourth of one sample from Wildsumaco Wildlife Sanctuary, Napo Province, Ecuador (voucher QCAZ 12798; sample code Cerc_s3_4) (Fig. 1). The three latter lineages were published by Torres-Carvajal et al. (2016), who also found the weekly supported sister relationship between Potamites and the new genus.

Figure 4. 

A Maximum clade credibility tree of Cercosaurinae based on the BEAST analysis with the position of Selvasaura gen. n. and Potamites highlighted in red B A close-up of the red part of the tree in the left showing the phylogenetic relationships between and within Selvasaura gen. n. and Potamites. Nodal support is shown in the ML/MrBayes/BEAST order; supported nodes are marked with asterisks, unsupported with dashes. Note that the basal node in the inset is not supported and that the sister relationship of the two genera may not be real.

The examined morphological characters were used for comparisons with other genera and for the formal descriptions of the new genus and species provided below.

Recent studies that examined the phylogeny and systematics of the Cercosaurinae on the basis of thorough sampling of taxa (Goicoechea et al. 2012; Torres-Carvajal et al. 2016; Sánchez-Pacheco et al. 2017b) detected previously unknown evolutionary lineages being present in the cercosaurine tree. While most of them could be assigned to currently existing genera (and some have been taxonomically revised since (Goicoechea et al. 2013; Sánchez-Pacheco et al. 2017a)), there were clades whose high levels of genetic divergence and morphological disparity indicated towards the existence of yet unknown clades at the level of genera (Torres-Carvajal et al. 2016). Similarly, our phylogenetic analyses also identified several previously undetected evolutionary lineages.

By formally describing the genus Selvasaura we extend the list of currently recognised genera of Cercosaurinae to 16 (Sánchez-Pacheco et al. 2017b). Apart from the formally named genera, there are two clades within cercosaurines that merit genus-level distinction and that are termed here in accordance with previous studies, Unnamed clades 2 and 4. The phylogenetic position of neither of them could be inferred with certainty. Results of both studies in which Unnamed clade 2 was included, this work and Torres-Carvajal et al. (2016), vary in its placement depending on the method of phylogenetic inference. Although all analyses tend to show it close to Proctoporus (at least to some groups), lack of support hampers any definitive conclusions regarding its evolutionary origins. We herein provide additional material of ten voucher specimens for Unnamed clade 2 from the PPPF (Table 1), which extends the range of the clade further north compared to previously published localities (Mantaro Valley and Colcabamba-Quintao District, Peru; Torres-Carvajal et al. 2016).

This study is the first to identify a clade that is termed here Unnamed clade 4. As in the case of Unnamed clade 2, the phylogenetic affinities of this clade remain obscure as a result of the low support of deeper level relationships within cercosaurines. Two of the analyses (ML, MrBayes) placed it as sister to Selvasaura gen. n. plus Potamites, while BEAST placed it as a sister lineage to Unnamed clade 2 (Suppl. material 1: Figs S1–S3). However, none of the topologies was supported. The two specimens that form Unnamed clade 4 represented arboreal lizards caught in cloud forests of southern and central Peru. For the moment, no other data such as morphology, ecology, and natural history are available for the clade as they are being collected and will be published with the formal description of the genus (work in progress).

Besides the newly discovered genera, there are currently 19 unnamed lineages at the level of species in the Cercosaurinae (Fig. 3). They are either morphologically disparate from existing taxa, represent unique genetic lineages or have been delimited as candidate species (this study; Doan et al. 2005; Goicoechea et al. 2012; Torres-Carvajal et al. 2016; Sánchez-Pacheco et al. 2017b) and their systematics and taxonomy should be revised. The presence of so many unidentified species points at the disturbing fact of how little we know about the real diversity of cercosaurine lizards and, perhaps, Neotropical biota in general.

Some previous studies have already pointed out problems with certain genera not being monophyletic when samples of more species of that genus were included in a phylogenetic analysis. Although many of these issues have been resolved, some still persist or were identified in our study and are to be addressed.

One of the recent examples is paraphyly of the genus Echinosaura. As Torres-Carvajal et al. (2016) revealed and we confirmed herein, E. sulcarostrum does not cluster with the other species of the genus. The phylogenetic position of this species varies across different phylogenetic analyses. Torres-Carvajal et al. (2016) reconstructed it in their BEAST analysis as a strongly-supported sister lineage to a large clade containing Anadia, Euspondylus (a name resurrected by Chávez et al. [2017] for the Unnamed clade 1 of Torres-Carvajal et al. [2016]), Macropholidus, Pholidobolus, Petracola, Cercosaura, Selvasaura (termed Unnamed clade 3 in their paper), Potamites, Proctoporus, and Unnamed clade 2; their ML analysis recovered it as sister to Proctoporus xestus. Subsequently, the genetic analysis of Sánchez-Pacheco et al. (2017b) recovered it as sister to a clade of Oreosaurus, Potamites, Petracola, Cercosaura, and Proctoporus, although with a limited taxon sampling. Finally, our analyses yielded it as sister to Euspondylus, although the topology was supported only in the MrBayes analysis. Regardless, none of the analyses found it close to the other Echinosaura species including E. horrida, the type species of the genus. As a result, its taxonomy, as well as proper phylogenetic placement, remain to be revised.

Another possible case of paraphyly is the genus Oreosaurus that was recently resurrected by Sánchez-Pacheco et al. (2017b). While the authors found the genus monophyletic, our analyses yielded no support for the clade as a whole with O. serranus placed separately from the other species. It was found sister to the remaining species only in our BEAST analysis, but support for the basal node was very low (pp = 0.3) indicating that 70% of the posterior trees actually had a different topology. Such a discrepancy between our and Sánchez-Pacheco et al. (2017b) results may stem from the difference in the analytical approaches undertaken. All our phylogenetic analyses were model-based, i.e., we assumed the sequences to evolve under evolutionary models that take into account the variation of substitution rates among sites in the alignments and the possibility of recurrent mutations at one site. On the contrary, Sánchez-Pacheco et al. (2017b) performed a maximum parsimony analysis that reconstructs the phylogeny based on the smallest number of evolutionary events necessary to explain the sequence data. Both methods may under some circumstances result in different topologies, especially when analysing relatively distant taxa where long branch attraction can occur (Felsenstein 1978; Alfaro et al. 2003). This may be the case here considering that we deal with a group whose origin has been estimated to have taken place in the early Tertiary (Zheng and Wiens 2016). Regarding the morphology, Sánchez-Pacheco et al. (2017b) found the genus Oreosaurus being clearly different from Andinosaura and Riama by lacking a narrow band of differentiated granular lateral scales. On the other hand, Sánchez-Pacheco et al. (2017a) show that Oreosaurus serranus can be distinguished morphologically from all other Oreosaurus species by having only one pair of genial scales. Therefore, taking into account the above-mentioned discrepancies in the results of molecular analyses, and the morphological distinctiveness of O. serranus from other Oreosaurus species, the character of nucleotide divergence between O. serranus and the other Oreosaurus species should be examined in detail in order to trace the inconsistency in the phylogenetic reconstructions.

At the species level, recent taxonomic advances made possible by the tremendous effort of many authors are making considerable progress in stabilising the taxonomy of cercosaurines (e.g., Goicoechea et al. 2013; Torres-Carvajal et al. 2014; Venegas et al. 2016; Sánchez-Pacheco et al. 2017a; among many others), yet non-monophyletic species are still present in the phylogeny. For example, the seven samples of Potamites ecpleopus used in this study form two groups, one is distributed in eastern Ecuador and the other in northeastern and southern Peru. Paraphyly of this species was already noted by Torres-Carvajal et al. (2016). Because the type locality of the species lies approximately in the centroid of the crescent delineated by the sampled localities, it is impossible to assign the species name to either of the groups with certainty until a comprehensive revision with specimens from the type locality is undertaken. Another example are two Cercosaura species, C. parkeri and C. schreibersii, which are paraphyletic with respect to each other as also noted by Sturaro et al. (2017). Cercosaura parkeri was originally described as a subspecies of C. schreibersii (Ruibal 1952) and elevated to species status by Tedesco and Cei (1999), but a further taxonomic revision of the species complex is apparently needed to resolve the remaining issues.

Yet another genus in which between-species relationships have proven difficult to infer is Proctoporus (Doan 2003; Doan et al. 2005; Torres-Carvajal et al. 2016), and this study supports this notion (but see Goicoechea et al. 2012). These semi-fossorial lizards inhabit primarily montane forests between 1000 m and 4000 m of altitude from central Peru to central Bolivia (Uzzell 1970; Goicoechea et al. 2012). Although certain species groups have traditionally been identified, mutual relationships between them and to other cercosaurine genera remain poorly resolved. Given the amount of cryptic species present within the genus (8 undescribed or candidate species; Fig. 3) indicating that future taxonomic revisions are to be expected, we herein propose the following terminology of the species groups in order to facilitate addressing this issue in future studies. The species groups are: (1) Proctoporus pachyurus group that contains P. chasqui, P. oreades, P. pachyurus, P. rahmi, P. spinalis, P. sucullucu, and three yet undescribed species (labelled P. sp., P. sp. 1, P. sp. 5); (2) Proctoporus lacertus group that consists of four recently described or resurrected species (P. carabaya, P. iridescens, P. kiziriani, P. lacertus), which were formerly considered part P. bolivianus (Goicoechea et al. 2012; 2013); (3) Proctoporus bolivianus group of two species, P. bolivianus and a confirmed candidate species (labelled P. Ca1 following Goicoechea et al. [2012, 2013]); (4) Proctoporus guentheri group, which contains the highest proportion of undescribed species (four, labelled P. sp. 2, P. sp. 3, P. sp. 4, P. Ca2 following Goicoechea et al. [2012, 2013]) besides three described species (P. guentheri, P. laudahnae, P. unsaacae); and (5) a clade of a single species, P. xestus. Most recent phylogenetic reconstructions of the genus were based on identical sampling of loci (three mitochondrial [12S, 16S, ND4] and one nuclear marker [cmos]) and our study has added sequences of one additional mtDNA marker (cytb) for only two species. The congruence of results obtained across studies and showing little support for the basal nodes is thus not surprising. We believe that getting a better resolution of the relationships between the Proctoporus groups is a matter of better sampling of loci and that more nuclear markers sequenced would shed more light on this subject.

The above problems with non-monophyletic genera raise an important nomenclatural issue regarding the application of generic names. Generic names apply to clades that contain the type species of the genus. In cases when genera are formed by more unrelated evolutionary lineages (e.g., Echinosaura) inferring the phylogenetic position of the type species is the only way to determine which of the lineages will bear the genus name; the other has to be renamed. In cercosaurines, most type species have been sequenced and placed in the phylogenetic context of the subfamily (for type species see Uetz et al. 2018), and this study provides an important addition to it.

For the first time, we sequenced the type species of the genus Anadia (A. ocellata). The sample clusters with other congeneric species in the dataset and thus fixes the name Anadia to this clade. Most species of Anadia have not been sequenced yet (Fig. 2), and it cannot be ruled out that once they are included in phylogenetic analyses, more cases of paraphyly will be detected. Similar situations have occurred in Anadia before when A. mcdiarmidi was found not to cluster with other species of the genus (Torres-Carvajal et al. 2016) and was subsequently reassigned to the genus Oreosaurus (Sánchez-Pacheco et al. 2017b).

Currently, the only cercosaurine genera with type species missing from the phylogenetic trees presented here are Euspondylus (type species E. maculatus) and Oreosaurus (type species O. luctuosus). Obtaining DNA sequences of the latter is particularly desired, as including its samples in phylogenetic analyses should help resolving the issue with the potential paraphyly in Oreosaurus (see above).

Montane forests (región yunga or selva alta) are found in the eastern Andes roughly between 800 and 3500 m a.s.l. (Perú, Ministerio del Ambiente 2015) and are known for their high biodiversity and an increasing endemism with increasing elevation (Young and León 2000). Yet, montane forests are among the least studied and least understood ecosystems (Ledo et al. 2012). In a recent vegetation map of Peru, 12 different types of montane forests were recognized within the región yunga, covering in total 9.58% of the national territory (Perú, Ministerio del Ambiente 2015). Its dense vegetation and steep slopes make the herpetofauna of montane forests relatively difficult to survey, and the canopy herpetofauna is probably the least known. However, exciting discoveries often happen by accident (e.g., arboreal species found in bromeliads on a fallen tree, Duellman et al. 2004).

All specimens of Selvasaura brava sp. n. were found in secondary forests, hiding in the roofs of simple wooden shacks where specimens could be easily seen and caught, whereas not a single specimen was found in primary forests. Our discovery of the new cercosaurine clade of arboreal lizards (Unnamed clade 4) together with a recent description of a new arboreal Euspondylus from central Peru (Chávez et al. 2017) indicate that arboreal species of cercosaurines may be much more diverse than previously thought and further research will be necessary to fully understand their diversity and ecology.

Some members of several cercosaurine genera (Anadia, Euspondylus, Selvasaura, Unnamed clade 4) are adapted to life in the above-ground vegetation (Oftedal 1974; Chávez et al. 2017; this study) and certain species of some other genera also show tendency to arboreality (e.g., Cercosaura; Vitt et al. 2003). Repeated convergent adaptation to arboreality in Neotropical lizards has been well documented and studied in anoles (Losos 1990; 1992; Kolbe et al. 2011). If such is the case for Cercosaurinae can only be answered when we have a better-resolved phylogeny. Furthermore, no such studies have been conducted to compare different arboreal lifestyles and arboreal locomotion (see Fischer et al. (2010) for definitions) in Cercosaurinae, nor have different ecomorphs, and their adaptations to their arboreal niches been described. We note the small size of Selvasaura brava (SVL = 42.1–45.9 mm, n = 4) and the relatively short front and hind limbs, yet detailed observations of their locomotory behaviour in nature are missing.