Molecular phylogeny of Atractus (Serpentes, Dipsadidae), with emphasis on Ecuadorian species and the description of three new taxa

Abstract We present a molecular phylogeny of snake genus Atractus, with an improved taxon sampling that includes 30 of the 140 species currently recognized. The phylogenetic tree supports the existence of at least three new species in the Pacific lowlands and adjacent Andean slopes of the Ecuadorian Andes, which we describe here. A unique combination of molecular, meristic and color pattern characters support the validity of the new species. With the newly acquired data, we propose and define the Atractus iridescens species group, as well as redefine the Atractus roulei species group. The species Atractus iridescens is reported for the first time in Ecuador, whereas Atractus bocourti and Atractus medusa are removed from the herpetofauna of this country. We provide the first photographic vouchers of live specimens for Atractus multicinctus, Atractus paucidens and Atractus touzeti, along with photographs of 19 other Ecuadorian Atractus species. The current status of Atractus occidentalis and Atractus paucidens is maintained based on the discovery of new material referable to these species. With these changes, the species number reported in Ecuador increases to 27, a number that is likely to increase as material not examined in this work becomes available and included in systematic studies.


Introduction
With 140 species, Atractus is the most speciose snake genus in the world, with 33 new species described only during the last ten years (Uetz et al. 2016). Most of these new species have been described using a combination of meristic and morphometric characters (Passos et al. 2009a, Passos and Lynch 2010, Schargel et al. 2013, Salazar-Valenzuela et al. 2014). However, with the exception of the preliminary phylogeny presented in De Oliveira and Hernández-Ruz (2016), no studies have involved a phylogenetic approach to test species arrangements and boundaries.
One recent work by Passos et al. (2009a) evaluated the taxonomic status of Atractus species from the Pacific lowland of Colombia and Ecuador, using a combination of meristic, morphometric, color pattern, and hemipenial characters. These authors described three new species and provided a comprehensive review of all Atractus known to occur in the region. However, when referring to this work to compare previously unexamined material from Ecuador, it became clear to us that several Ecuadorian specimens of Pacific lowland Atractus could not be assigned to any taxa currently recognized to occur in the country. Some specimens identified as A. medusa (Passos et al. 2009a) matched the coloration of the first specimen reported in Ecuador by Cisneros-Heredia and Romero (2015), but they did not match the coloration of the holotype (Passos et al. 2009a). Other specimens were closer in coloration and lepidosis to A. iridescens (Peracca, 1860) from Colombia, and others resembled both A. microrhynchus (Cope, 1868) and A. occidentalis (Savage, 1955). To further complicate matters, the taxonomic validity of A. occidentalis and A. paucidens (Despax, 1910) was not recognized in Arteaga et al. (2013), owing to their close morphological resemblance to A. dunni (Savage, 1955) and A. modestus (Boulenger, 1894), respectively.
To resolve these pending issues and to shed light on potentially unclear species boundaries, we report on new material of Atractus from Ecuador, review current knowledge on the species occurring in the Pacific lowlands and adjacent Andean slopes, present a new molecular phylogeny, including most Ecuadorian species, and describe three new species of Atractus.

Ethics statement
This study was carried out in strict accordance with the guidelines for use of live amphibians and reptiles in field research compiled by the American Society of Ichthyologists and Herpetologists (ASIH), The Herpetologists' League (HL) and the Society for the Study of Amphibians and Reptiles (SSAR). All procedures with animals (see below) were approved by the Centro de Investigación de la Biodiversidad y Cambio Climático (BioCamb) of the Universidad Tecnológica Indoamérica. They also were reviewed by the Ministerio de Ambiente del Ecuador (MAE) and specifically approved as part of obtaining the following field permits for research and collection: MAE-DNB-CM-2015-0017, granted to Universidad Tecnológica Indoamérica; and permit N°012-IC-FAN-DPEO-MAE, granted to the Museo Ecuatoriano de Ciencias Naturales. Specimens were euthanized with 20% benzocaine, fixed in 10% formalin or 70% ethanol, and stored in 70% ethanol. Museum vouchers were deposited at the Museo de Zoología of the Universidad Tecnológica Indoamérica (MZUTI).

Sampling
Tissue samples from 39 individuals representing 22 species (including three new species described here) were obtained throughout Ecuador. The majority of individuals were located by space-constrained visual examination of ground-level substrates (Campbell and Christman 1982). The remaining individuals were detected by turning over logs, rocks and other surface objects. All specimens included in the genetic analyses were morphologically identified according to Savage (1955Savage ( , 1960, Cisneros-Heredia (2005), Passos et al. (2009a), Arteaga et al. (2013), Schargel et al. (2013) and Salazar-Valenzuela et al. (2014). We generated sequence data for samples marked with an asterisk under Appendix I, which includes museum vouchers at the Museo de Zoología de la Universidad Tecnológica Indoamérica (MZUTI), the División de Herpetología del Museo Ecuatoriano de Ciencias Naturales (DHMECN) and the Fundación Herpetológica Gustavo Orcés (FHGO).

Laboratory techniques
Genomic DNA was extracted from 96% ethanol-preserved tissue samples (liver, muscle tissue or scales) using a modified salt precipitation method based on the Puregene DNA purification kit (Gentra Systems). We amplified the 16S gene using the primers 16Sar-L and 16Sbr-H-R from Palumbi et al. (1991). Additionally, the Cytb gene was obtained with the primers L14910 and H16064 developed by Burbrink et al. (2000), whereas the gene coding for the subunit 4 of the NADH dehydrogenase was ampli-fied with the primers ND4 and Leu developed by Arévalo et al. (1994). PCR reactions contained 2 mM (Cytb and ND4) or 3 mM (16S) MgCl 2 , 200 µM dNTP mix, 0.2 µM (16S and Cytb) or 0.8 µM (ND4) of each primer and 1.25 U (16S and Cytb) or 0.625 U (ND4) Taq DNA Polymerase Recombinant (Thermo Fisher Scientific) in a 25 µL total volume. The nucleotide sequences of the primers and the PCR conditions applied to each primer pair are detailed in Appendix II. PCR products were cleaned with Exonuclase I and Alkaline Phosphatase (Illustra ExoProStar by GE Healthcare) before they were sent to Macrogen Inc (Korea) for sequencing. All PCR products were sequenced in both forward and reverse directions with the same primers that were used for amplification. The edited sequences were deposited in GenBank (Appendix I).

DNA sequence analyses
A total of 126 mtDNA sequences were used to build a mitochondrial phylogenetic tree of the genus Atractus. 69 were generated during this work and 57 (all available sequences for the sampled gene fragments) were downloaded from GenBank. A mitochondrial marker dataset, though less powerful to study higher-level phylogenetic relationships, was chosen because it is the most effective to successfully resolve specieslevel phylogenies (Patwardhan 2014). Recently published works looking to resolve intrageneric relationships within Neotropical dipsadines have done so using phylogenies that are largely based on mitochondrial data (Krysko et al. 2015, Pyron et al. 2016). Specifically, we use the gene Cytochrome-b because it is reported as the most powerful in recovering phylogenetic relationships among closely related taxa (Patwardhan 2014), which is the case for the species of Atractus studied here. The mitochondrial genes 16S and ND4 were used to be able to compare with Atractus sequences available in GenBank. Novel sequences were edited and assembled using the program Geneious ProTM 5.4.7 (Drummond et al. 2010) and aligned with those downloaded from Genbank (Appendix I) using MAFFT v.7 (Katoh and Standley 2013) under the default parameters in Geneious ProTM 5.4.7. Genes were combined into a single matrix with seven partitions, one per non-coding gene and three per protein coding gene corresponding to each codon position. The best partition strategies along with the best-fit models of evolution were obtained in PartitionFinder 1.1.1 (Lanfear et al. 2012) and jModeltest (Darriba et al. 2012) under the Bayesian information criterion. Phylogenetic relationships were assessed under a Bayesian approach in MrBayes 3.2.0 (Ronquist and Huelsenbeck 2013). Four independent analyses were performed to reduce the chance of converging on a local optimum. Each analysis consisted of 6.7 million generations and four Markov chains with default heating settings. GenBank accession numbers are listed in Appendix I. Trees were sampled every 1,000 generations, resulting in 5,000 saved trees per analysis after 25% of those were arbitrarily discarded as ''burn-in." Stationarity was confirmed by plotting the-ln L per generation in the program Tracer 1.2 (Rambaut and Drummond 2003). Genetic distances between A. esepe and its closest morphological relatives were calculated using the uncorrected distance matrix in PAUP 4.0 (Swofford 2002).

Morphological data
Our terminology for Atractus cephalic shields follows Savage (1960), diagnoses and descriptions generally follow Passos et al. (2009a), and ventral and subcaudal counts follow Dowling (1951). We examined comparative alcohol-preserved specimens from the herpetology collections at the MZUTI, DHMECN, Fundación Herpetológica Gustavo Orcés (FHGO), Museum d'Histoire Naturelle de la Ville de Genève (MHNG), Museo de Historia Natural de la Escuela Politécnica Nacional (EPN), Museo de Zoología de la Pontificia Universidad Católica del Ecuador (QCAZ), National Museum of Natural History (USNM), Muséum National d'Histoire Naturelle (MNHN) and Museo de Zoología de la Universidad San Francisco de Quito (ZSFQ). (Table 1). Morphological measurements were taken with measuring tapes to the nearest 1 mm. When providing the standard deviation, we use the ± symbol. Sex was determined by noting the presence or absence of hemipenes through a subcaudal incision at the base of the tail.

Nomenclatural acts
The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix "http://zoobank.org/". The LSID for this publication is: urn:lsid:zoobank.org:pub:7CBF7FB1-EFEA-4DC1-8F64-5BF862694AA0. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: PubMed Central, LOCKSS.

Molecular phylogeny
The overall topology and support ( Fig. 1) is similar to that of Pyron et al. (2015). We consider strong support to be posterior probability values >95%, following Felsenstein (2004). Overall, there is low support for many backbone nodes. Strong support was found for the clade colored in yellow under Fig. 1.
The resulting hypotheses of species relationships for our mitochondrial phylogenetic tree supports Savage's (1960) assumption suggesting independent evolution of the 15 dorsal scale row lineage within Atractus, since species with this number of dorsal scale rows, like A. elaps, A. roulei and A. duboisi, belong to different lineages. However, the tree does show that A. carrioni (Parker 1930), A. lehmanni (Boettger 1898), Table 1. Locality data for specimens examined in this study. Coordinates represent georeferencing attempts from gazetteers under standard guidelines, though some variation from the exact collecting locality will be present. Similarly, elevations are taken from Google Earth, and may not exactly match the elevations as originally reported. A. roulei (Despax, 1910) and A. pyroni sp. n., species with 15 scale rows, form a monophyletic group that includes two more species than was suggested by Passos et al. (2013) when naming the A. roulei species group (Fig. 1). Atractus gigas (Myers and Schargel, 2006), A. modestus, A. paucidens, A. savagei (Salazar-Valenzuela et al. 2014), A. typhon (Passos et al., 2009a) and A. zidoki (Gasc and Rodrigues, 1979) form a poorly supported clade that does not include A. microrhynchus and A. iridescens, as was suggested by Passos et al. (2009a) when naming the A. paucidens species group (Fig. 1). Six species, Atractus cerberus sp. n., A. dunni, A. esepe sp. n., A. iridescens A. microrhynchus, and A. occidentalis, form a strongly supported clade sister to the A. paucidens species group. Here, we name this lineage as the A. iridescens species group (Fig. 1).
Atractus occidentalis forms a strongly supported distinct lineage, sister to A. microrhynchus. Together, these two species are sister to A. dunni. Atractus typhon is shown to be the strongly supported sister lineage of A. gigas, as is the case for a relationship between A. roulei and A. pyroni sp. n.

New taxa and systematic arrangements derived from the analyses
We seek here to only name or redelimit Atractus species groups that are supported in our molecular phylogeny and share features of their coloration pattern and lepidosis. The first such groups is the clade comprising A. cerberus sp. n., A. dunni, A. esepe sp. n., A. iridescens, A. microrhynchus and A. occidentalis. The other is the one comprising A. carrioni, A. lehmanni, A. pyroni sp. n. and A. roulei.
Distribution. Pacific lowlands and western Andean slopes in Ecuador and Colombia (Fig. 3).
Comment. Passos et al. (2009a) included Atractus echidna, A. iridescens and A. microrhynchus in the phenetic A. paucidens species group. Later, Passos et al. (2012) placed A. microrhynchus in the A. multicinctus group based on hemipenial characters. Unlike A. paucidens or A. multicinctus (Jan, 1865), however, the former three species have a brownish color pattern (Fig. 2) and also a lower number of ventral scales (Appendix III). These differences, together with the phylogenetic placement of A. iridescens and A. microrhynchus support the allocation of these species in the newly formed A. iridescens group.
Diagnosis. Atractus cerberus is placed in the genus Atractus as diagnosed by Savage (1960), based on phylogenetic evidence (Fig. 1). It is included in the A. iridescens group due to its brown dorsal ground color (Fig. 5) and its phylogenetic position (Fig. 1). The species is diagnosed based on the following combination of characters: (1) 17/17/17 smooth dorsals; (2) two postoculars; (3) loreal moderate; (4) temporals 1+2; (5) seven supralabials, third and fourth contacting orbit; (6) seven infralabials, first four contacting chinshields (7) seven maxillary teeth; (8) three gular scale rows; (9) Comparisons. Atractus cerberus is included in the A. iridescens species group and compared to other Pacific lowland congeners that have a brownish ground color (Fig. 2) (Fig. 5). From all others, it differs in having yellow ventral surfaces (as opposed to cream or dingy white) and having more than 150 ventrals in males. Finally, the dorsal pattern of A. cerberus is less clearly marked than in the majority of the known specimens of the species included in the A. iridescens group. Instead of having conspicuous spots, blotches or lines, A. cerberus has a series of feebly visible dashes made of pigment slightly darker than the surrounding ground color.
Color pattern. The dorsal ground color is brown with five feebly visible darkbrown to black longitudinal lines that are not continuous throughout the length of the body but broken into spots along some sections (Fig. 5). Between the dark longitudinal lines on each side of the body, there are fields of lighter pigment that on some sections of the body correspond to lines. The head is darker than the rest of the dorsal surfaces and is marked by a dark, irregular postocular stripe that reaches the corner of the mouth (Fig. 5). The top of the supralabials is tinged with black. The ventral surfaces are yellowish cream with scattered brownish speckling that becomes more concentrated towards the tail, which is almost completely brown. The iris is carmine and the pupil is black.
Description of holotype. Adult male, SVL 212 mm, tail length 23 mm (10.8% SVL); body diameter 6.5 mm; head length 7.9 mm (3.7% SVL); head width 4.8 mm (2.3% SVL); interocular distance 3.1 mm; head slightly distinct from body; snout-or-bit distance 2.8 mm; rostral 1.6 mm wide, about one time broader than high; internasals 1.0 mm wide; internasal suture sinistral relative to prefrontal suture; prefrontals 1.7 mm wide; frontal 2.3 mm wide, with a curvilinear triangle shape in dorsal view; parietals 2.1 mm wide, about twice as long as wide; nasal divided; loreal 1.5 mm long, about 2 times longer than high; eye diameter 1.4 mm; pupil round; supraoculars 1.4 mm wide; two postoculars; temporals 1+2, upper posterior temporal elongate, about four times longer than high, and three times as long as first temporal; seven supralabials, 3 rd -4 th contacting orbit; symphisial 1.0 mm wide, about twice as broad as long, separated from chin shields by first pair of infralabials; seven infralabials, 1 st -4 th contacting chin shields; anterior chin shields about three times as long as broad, posterior chin shields absent; three series of gular scales; dorsal scales 17/17/17 rows, smooth without apical pits; preventrals 3; ventrals 157; anal plate single; paired subcaudals 26. Natural history. The two known specimens of Atractus cerberus were found in an isolated patch of deciduous lowland forest surrounded by dry lowland shrubland. MZUTI 4330 was found active on leaf litter at 19h29, in 80% closed canopy secondary forest far from streams. The night was warm and there was drizzle the night before. MZUTI 5108 was found crossing a forest trail close to an open area at 10h00 during a sunny morning after a rainy night.
Distribution. Known only from the type locality, Pacoche, in the Ecuadorian province of Manabí at 280-324 m (Fig. 3). This locality is 3 km airline distance from the shoreline.
Etymology. The specific epithet "cerberus" is derived from the name of the Greek monster Kérberos. In Greek mythology, Kérberos is a monstrous multi-headed dog that guards the gates of the underworld, preventing the dead from leaving. Here, we use this word in allusion to the type locality, at the gates of the newly formed "Refin-ería del Pacífico", a massive industrial oil-processing plant that can easily be likened to the underworld.
Conservation status. Although Atractus cerberus belongs to a poorly studied genus of snakes and is known only from two specimens collected recently in a single locality, we consider this species to be Critically Endagered following B1a,b(iii) IUCN criteria because: i) its extent of occurrence is estimated to be less than 50 km 2 (i.e. total area of continous semideciduous forest in the Refugio de Vida Silvestre Pacoche); ii) it has not been detected in any other locality in the province of Manabí despite numerous surveys (Almendáriz and Carr 2007, Cisneros-Heredia 2004, MECN et al. 2013; and iii) and its habitat is severely fragmented, isolated from other such habitats and declining in extent and quality due to deforestation. Paratopotype. MZUTI 3759, adult female collected by Jaime Culebras. Diagnosis. Atractus esepe is placed in the genus Atractus as diagnosed by Savage (1960), based on phylogenetic evidence (Fig. 1). It is included in the A. iridescens group due to its brown dorsal ground color and its phylogenetic position (Figs 1, 6). The species is diagnosed based on the following combination of characters: (1) 17/17/17 smooth dorsals; (2) two postoculars; (3) loreal long; (4) temporals 1+2; (5) seven supralabials, third and fourth contacting orbit; (6) seven infralabials, first four contacting chinshields (7) seven maxillary teeth; (8) 2-3 gular scale rows; (9) 2-3 preventrals; (10) 149 ventrals in the male holotype, 156 in the female paratype; (11) 41 subcaudals in the male holotype, 30 in the female paratype; (12) dorsal ground color brown with a pattern of complete (MZUTI 3759) or broken (MZUTI 3758) (Fig. 6a) dark lines running parallel along each side of the body and separated from each other by a cream line, but rendering the appearance of a row of dorso-lateral blotches in the broken pattern (MZUTI 3758); (13) venter cream faintly speckled with brownish pigment (Fig.  6b); (14) 232-241 mm SVL; (15) 34-53 mm TL.
Comparisons. Atractus esepe is included in the A. iridescens species group and compared to other Pacific lowland congeners who have a brownish ground color (Figs 2, 5): A. boulengerii, A. cerberus, A. dunni, A. echidna, A. iridescens, A. medusa, A. microrhynchus, and A. occidentalis. From these, A. microrhynchus and A. occidentalis have striped pattern and cream ventral surfaces similar to that of A. esepe, but they occur parapatrically (Fig. 3) and can be distinguished from A. esepe by a genetic divergence of 5.3-5.7% in a 506 bp long fragment of the mitochondrial Cytb gene and by having a greater number of subcaudal scales in males (Table 2). Furthermore, adult specimens of A. microrhynchus have light brown dorsal surfaces instead of dark brown, and their pattern can be better described as a series of blotches rather than broken longitudinal lines. Specimens of both A. esepe and A. occidentalis have a pattern of longitudinal lines, but A. esepe has a greater number of ventral plus caudal scales than A. occidentalis (more than 180 in A. esepe) ( Table 2).
Color pattern. The dorsal ground color is dark brown with either six longitudinal black lines separated by lighter areas or a pattern of dark brown longitudinally arranged spots that correspond to the longitudinal lines. On each side, the line or series of dark spots along the 2 nd and 3 rd dorsal scale row is feebly visible, but the other lines or spots are conspicuous. The dorsal surface of the head is dark brown and there is a clearly marked dark postocular stripe running from behind the eye to the edge of the mouth (Fig. 6). The ventral surfaces are dingy white, finely speckled with brown pigment that becomes more concentrated towards the tail. The iris is carmine and the pupil is black.
Description of holotype. Adult male, SVL 232 mm, tail length 53 mm (22.8% SVL); body diameter 7.0 mm; head length 7.9 mm (3.4% SVL); head width 4.8 mm (2.2% SVL); interocular distance 3.4 mm; head slightly distinct from body; snout-orbit distance 3.3 mm; rostral 1.8 mm wide, about one time broader than high; internasals 0.9 mm wide; internasal suture sinistral relative to prefrontal suture; prefrontals 1.9 mm wide; frontal 2.2 mm wide, with a curvilinear triangle shape in dorsal view; parietals 2.1 mm wide, about twice as long as wide; nasal divided; loreal 2.5 mm long, about 3 times longer than high; eye diameter 1.5 mm; pupil round; supraoculars 1.2 mm wide; two postoculars; temporals 1+2, upper posterior temporal elongate, about four times longer than high, and three times as long as first temporal; seven supralabials, 3 rd -4 th contacting orbit; symphisial 0.8 mm wide, separated from chin shields by first pair of infralabials; seven infralabials, 1 st -4 th contacting chin shields; anterior chin shields about three times as long as broad, posterior chin shields absent; three series of gular scales; dorsal scales 17/17/17 rows, smooth without apical pits; preventrals 3; ventrals 149; anal plate single; paired subcaudals 41. Natural history. The two known specimens of Atractus esepe were found actively foraging among soil and roots in secondary evergreen lowland forest at least 400 m from the nearest natural body of water. They were found by night at 20h00 after a warm, sunny day.
Distribution. Known only from the type locality, Caimito, in the Ecuadorian province of Esmeraldas at 102 m (Fig. 3). This locality is 1.3 km airline distance from the shoreline.
Etymology. The specific epithet esepe is derived from the Spanish pronunciation of "sp.", which is the abbreviation for the Latin word species. Here, we use this word in allusion to how the majority of Ecuadorian researchers refer to Atractus specimens found in the field.
Conservation status. We consider Atractus esepe to be Data Deficient following IUCN criteria because it is known only from its type locality but its occurrence in the biogeographic Choco suggests that it might as well be present in other localities. The Chocoan forests of Caimito do not appear to be isolated from other similar habitat by geographical or ecological barriers. Therefore, we consider there is inadequate information to make a direct, or indirect, assessment of its extinction risk based on its scarce distribution data.
Comparisons. Atractus pyroni is compared to members of the A. roulei species group: A. carrioni, A. lehmanni, and A. roulei (Fig. 2). From A. carrioni, it differs by having a loreal. From A. lehmanni and A. roulei, it differs in size and color pattern. Atractus pyroni is 443 mm in SVL; whereas A. lehmanni is 262-321 in SVL, and A. roulei is 230-396. Both A. lehmanni and A. roulei have uniform dorsal ground color, whereas A. pyroni has a distinct dorsal bicolored pattern (Fig. 7). Finally, in life, A. pyroni is darker than the remaining members of the A. roulei species group and has a ventral pattern that, instead of having fine speckles, has conspicuous scattered blotches of a contrasting color.
Color pattern. The dorsal ground color is blackish with a dark vertebral (mid-dorsal) scale row flanked by a dark yellow scale row on either side (the 7 th dorsal scale row), irregularly adjoined by one to few additional yellow scales on the 6 th dorsal scale row, rendering an appearance of an irregularly edged mid-dorsal striped pattern (Fig. 7). The dorsal and lateral surfaces of the head are dark grayish brown and the labials are dark mustard yellow. All ventral surfaces are glossy grayish black except for the throat and some scattered blotches, which are dark mustard yellow.
Description of holotype. Adult female, SVL 443 mm, tail length 34 mm (7.7% SVL); body diameter 11.6 mm; head length 14.4 mm (3.3% SVL); head width 9.8 mm (2.2% SVL); interocular distance 5.1 mm; head slightly distinct from body; snout-orbit distance 5.7 mm; rostral 2.8 mm wide, about two times broader than high; internasals 1.5 mm wide; internasal suture sinistral relative to prefrontal suture; prefrontals 2.8 mm wide; frontal 3.5 mm wide, with a curvilinear triangle shape in dorsal view; parietals 4.0 mm wide, about twice as long as wide; nasal divided; loreal 3.7 mm long, about 3 times longer than high; eye diameter 1.8 mm; pupil round; supraoculars 2.1 mm wide; one postocular; temporals 1+2, upper posterior temporal elongate, about five times longer than high, and twice as long as first temporal; six supralabials, 3 rd -4 th contacting orbit; symphisial 2.4 mm wide, separated from chin shields by first pair of infralabials; five infralabials, 1 st -4 th contacting chin shields; anterior chin shields about three times as long as broad, posterior chin shields absent; three series of gular scales; dorsal scales 15/15/15 rows, smooth without apical pits; preventrals 2; ventrals 143; anal plate single; paired subcaudals 16. Natural history. The only known specimen of Atractus pyroni was found dead on a dirt road surrounded by silvopastures and remnants of native montane cloudforest.
Distribution. Known only from the type locality, between Balzapamba and Bilován, in the Ecuadorian province of Bolívar at 2026 m (Fig. 7).
Etymology. Named after R. Alexander Pyron, one of the most prolific contemporary herpetologists, in recognition of his invaluable contribution to systematics and evolution of the world's reptiles.
Conservation status. We consider Atractus pyroni to be to be Data Deficient following IUCN because there is inadequate information to make a direct, or indirect, assessment of its extinction risk based on its scarce distribution data.

Discussion
Species relationships and taxonomy in the colubrid snake genus Atractus are still far from being resolved, and many infrageneric groups are either non-monophyletic, or poorly supported and weakly placed, which may reflect inadequate sampling of taxa (only 30 out of 140 species are included) or characters (only 1 locus is used). No monophyly was found for the groups defined by Savage (1960), which, until further phylogenetic evidence is accumulated or unambiguous diagnostic characters are defined, should not be used.
From the five members of the A. paucidens species groups of Passos et al. (2009a) that were sampled in our phylogeny, only A. paucidens, A. savagei, and A. typhon cluster together. Atractus microrhynchus and A. iridescens belong to another lineage, which is here named the A. iridescens species group. This group includes the aforementioned two species plus A. cerberus, A. dunni, A. echidna, A. esepe, and A. occidentalis. From the species included in this group, we expand the known distribution of all their members (Fig. 3). However, we do not include the specimens ANSP 18114 nor ANSP 26316, from the vicinity of Huigra and identified as A. occidentalis by Savage (1960), because their description disagrees with the observed morphological variation reported for A. occidentalis in this work. Upon a visit to Huigra, a dry valley dominated by xeric vegetation and rocky outcrops, it became clear to us that it is unlikely for a species like A. occidentalis, which is found in evergreen lower-montane forests (Arteaga et al. 2013), to occur in an isolated dry habitat type ca. 250 km airline distance south of the type locality.
We also re-delimit the A. roulei species group of Passos et al. (2013) to include A. carrioni, A. lehmanni, A. roulei and A. pyroni. We expand the known distribution of A. roulei (Fig. 4), but do not include specimen AMNH 17492 from San José de Chimbo (Savage 1960) in the map because this specimen might actually be A. pyroni given the morphological similarities between the two species and the geographical proximity to the type locality of A. pyroni. Reports of A. lehmanni from Colombia (Passos et al. 2009b) are likely misidentifications since A. lehmanni has not been registered in Ecuador outside the type locality.
To further clarify the landscape of Atractus taxonomy in Ecuador, we analyze the presence of A. medusa, A. melas, A. typhon, A. badius, and A. bocourti in the country. Cisneros-Heredia and Romero (2015) presented the first country record of A. medusa in Ecuador (specimen DFCH-USFQ 191.101109 at Universidad San Francisco de Quito), based on similarities in scalation and coloration between that specimen and the holotype of A. medusa, form Gorgona island, Colombia. Certainly, the characters of scalation of the Ecuadorian specimen fit the diagnosis of A. medusa. However, they fit just as well the diagnosis of A. iridescens provided by Passos et al. (2009a), with the difference that the dorsal pattern of the Ecuadorian specimen resembles more the A. iridescens specimen, ICN 10902, pictured in Passos et al. (2009a). The dark brown ground color (as opposed to light cream), the light bordered brown blotches (as opposed to solid black blotches), and the absence of a black nape band are all characteristics shared by DFCH-USFQ 191.101109 and the other nine specimens of A. iridescens presented in Appendix III, with ICN 10902 of Passos et al. (2009a). Therefore, we consider that DFCH-USFQ 191.101109 actually represents the first country record of A. iridescens for Ecuador. Based on this new information and re-examination of museum material, we report on 9 additional specimens ( Table 1) that expand the current known distribution of this species. Cisneros-Heredia and Romero (2015) suggest that a photographic record of Atractus cf. melas from the Bilsa Biological Station, province of Esmeraldas, northwestern Ecuador (Ortega-Andrade et al. 2010) corresponds to A. multicinctus. The specimen differs from other material assigned to A. multicinctus in having whitish rings as opposed to red rings throughout the body (Fig. 2). Although photographic vouchers of A. typhon have been presented in MECN et al. (2013), we report on the first museum vouchers of the species in Ecuador (Table 1).
Finally, although Hoogmoed (1980) restricted the type locality of A. badius and pointed out that the upper Amazon basin specimens were misidentifications, the species has remained in Ecuadorian faunal lists (Torres-Carvajal et al. 2016), even after Schargel et al. (2013) made compelling cases to exclude this species from the upper Amazon Basin. Other snake, A. bocourti was included in the herpetofauna of Ecuador by Pérez-Santos and Moreno (1991) without pointing out to any museum voucher. These authors stated that although they have no information about the distribution of the species in Ecuador, its distribution in Colombia would suggest that it also occurs in Ecuador. Since there is no evidence that neither A. badius nor A. bocourti occur in Ecuador, we remove them from this country's herpetofauna.
Our analysis of new Atractus material supports the evolutionary phylogenetic distinctiveness of at least 22 of the total taxa currently recognized to occur in Ecuador. To include the remaining taxa in future phylogenetic analyses will certainly help resolve species relationships and taxonomic arrangements of cis-Andean Ecuadorian Atractus, since the five species that were not included in the phylogeny occur in the Amazonian slopes of the Andes. However, besides including more taxa in future phylogenetic analyses, we feel that a more adequate sampling of molecular markers is needed to overcome the difficulties that mitochondrial-based phylogenies have to capture higherlever evolutionary relationships. Certainly, future studies can benefit from a phylogeny based on both a nuclear and a mitochondrial dataset.
We hope that the novel genetic and morphological data provided herein will promote future researchers to examine species boundaries in Atractus, as additional work clearly is waiting.

Author contributions
Conceived and designed the work: AA. Performed the analyses: AA NP. Gathered morphological data: KB JHV DFCH CRP JLVF AA. Analyzed the data: AA KM DFCH JMG. Contributed reagents/materials/analysis tools: JMG NP. Wrote the paper: AA KM JHV DFCH NP CRP JLVF JMG.