When the tail shakes the snake: phylogenetic affi nities and morphology of Atractus badius (Serpentes: Dipsadidae), reveals some current pitfalls on the snake’s genomic age

: Atractus badius has a long and controversial nomenclatural history due to both its antiquity and the brevity of its original description. This species was described based on two syntypes from Java. Later, a lectotype was designated and the distribution range restricted to the Guiana Shield. Although this species has been repeatedly recorded throughout Amazonia and the Andes, these records have been erroneously assigned to A. badius because of a considerable level of confusion in the literature. We found 13 additional specimens of this poorly known snake, expanding our knowledge on its morphological variability, phylogenetic relationships and distribution. In this paper, we report new localities and data on meristic, morphometric, coloration, scales micro-ornamentation, osteology and hemipenial morphology to A. badius . We also determine the phylogenetic position of A. badius as nested in a composite Guiana Shield clade. In addition, we discuss recent advances on the systematics of Atractus and comment on some ‘new arrangements’ with respect to previously recognized species groups and available morphological evidence. Finally, we highlight the importance of accurate voucher identifi cation before promoting taxonomic changes or implementing nomenclatural acts derived from new phylogenetic hypotheses, and notably propose the synonymy of the recently described A. pyroni with A. roulei .


INTRODUCTION
The cryptozoic snakes of the genus Atractus Wagler, 1828 are distributed widely in the New World, occurring from Panama to Argentina (Giraudo & Scrocchi 2000, Myers 2003).Atractus is the most species-rich snake genus in the world, encompassing 147 valid species to date (Passos et al. 2019a, b, Uetz et al. 2021).Despite a substantial number of taxa originally known from small series with restricted distributions, several species have been recently rediscovered in the wild or reported based on old misidentifi ed material (sometimes for decades) stored in collections (Köhler & Kieckbusch 2014, Passos et al. 2018a).In the last ten years, the taxonomy of Atractus has undergone an unprecedented flux of improvements based on detailed morphological studies (see Passos et al. 2017Passos et al. , 2018b)).As an expected result, the taxonomy of several species complexes has been enhanced regarding species delimitation and morphological variation (Passos et al. 2010, 2019a, Passos & Prudente 2012, Melo-Sampaio et al. 2019, 2021).However, other taxonomic assemblages within the genus still need further study to accurately characterize species boundaries and phylogenetic relationships.
In parallel, new phylogenetic hypotheses derived from molecular datasets have much improved our knowledge about the relationships among species and clades within Atractus (Zaher et al. 2009, Grazziotin et al. 2012, Pyron et al. 2013, Figueroa et al. 2016, Arteaga et al. 2017).However, these analyses are misleading because they are based both on several misidentified voucher specimens and/or chimeras created between distinct and non-closely related species (Passos et al. 2017; Table I).Such identification problems may be overcome by simple efforts toward the examination of the corresponding voucher specimens.Many recent reviews provide unambiguous diagnoses and detailed comparative data for taxa distributed along South American provinces (Passos et al. 2009a(Passos et al. , b, 2010(Passos et al. , 2013a)), with sequences for autochthonous species available in GenBank.As for recent voucher specimens (Table I), they should be accurately identified by direct examination or reference to relevant literature.
In this paper, we report new material unambiguously assigned to Atractus badius and provide new data on variation for distinct and putatively independent morphological systems (e.g., scales, bones and male genitalia).We also infer the phylogenetic position of A. badius based on previously available sequences and new ones obtained from examined voucher-based specimens ensuring accurate identifications.

HISTORICAL RÉSUMÉ
Atractus badius has a long and controversial nomenclatural history due to both its antiquity and brevity of its original description.Friederich Boie recognized the species early in the XIX Century based on an unpublished manuscript entitled 'Erpétologie de Java'.Later, his brother H. Boie (1827) 1858) reported on two specimens of "R.badius" from Pará, remarking the white-banded pattern among them.Jan (1862) recognized other species outside the Guiana Shield as "varieties" of Rabdosoma badium (e.g., R. badium var.multicintum from Peru; see Passos et al. 2009b).Boulenger (1894) proposed the synonymy of Rabdosoma with Atractus, but still recognized Atractus badius as a widespread and polytypical taxon.After Boulenger, the name A. badius has been incorrectly applied to several Atractus species that occur both in Amazonia and the Andes.Virtually all authors prior to Hoogmoed (1980), associated this name to several specimens or species with 17 dorsal scale rows and a dorsal coloration with dark alternate bands or blotches on the back.This was likely also reinforced, in part, by the seminal work of Savage (1960), in which the definition of the Atractus badius species group may have influenced subsequent species' identification.This scenario with a generalized definition of A. badius remained until Hoogmoed (1980) rediscovered the syntypes of A. badius, A. flammigerus and A. schach.He established the distinctive nature of the latter two species and removed them from synonymy of A. badius.Hoogmoed (1980) redescribed A. badius, designated a lectotype, and transferred A. mitcheli and A. subcinctum to the synonymy of A. badius and further emphasized that, "it is very likely that this presumed species (= A. badius) does not have such an extensive distribution and that it is composed of several species.To be able to make conclusive statements on this subject a revision of all South American material referred to this species would be necessary."Even after Hoogmoed's comprehensive review, Atractus badius continued to be inadvertently associated with many species outside the Guiana Shield without any justification (Peters & Orejas-Miranda 1986, Pérez-Santos & Moreno 1988, Carrillo & Icochea 1995).More recently, Ávila-Pires et al. (2010) reported on the second vouchered specimen of A. badius from the Brazilian part of the Guiana Shield (see Hoogmoed 1980 for the additional Brazilian record to Amapá) to ESEC Grão Pará, state of Pará.All the specimens outside the Guiana Shield previously reported in the literature as A. badius (sensu Hoogmoed 1980) refer in fact to the A. schach or A. snethlageae species complexes which have undergone recent reviews (Melo-Sampaio et al. 2019, 2021).Finally, despite Savage's early definition of the Atractus badius species group, no other species was later associated to that group, presumably because of the lack of detailed descriptions and comparisons.

Material and techniques for morphological characters
Institutional acronyms are as listed in Sabaj (2020).All specimens of A. badius examined by us are listed in Appendix I. Terminology for cephalic shields follow Savage (1960), whereas ventral and subcaudal counts follow Dowling (1951).Measurements were taken with a dial caliper to the nearest 0.1 mm, except for snoutvent length (SVL) and tail length (TL), which were measured with a ruler to the nearest 1 mm.Sex was determined by presence/absence of hemipenes after a ventral incision at the base of the tail.We examined maxillae of all specimens under a stereoscope, through a narrow lateromedial incision between the supralabials and the maxillary arch.After removing tissues covering the maxillary bone, we counted teeth and empty sockets.Terminology for microornamentation descriptions follows Price (1982) and Price & Kelly (1989), with a few modifications.The superficial layer of the dorsal scales was sampled from the middorsal body region of nine individuals of Atractus badius.We removed scale layers (= Oberhäutchen) with forceps and stored them separately in 70% ethyl alcohol.The layers were affixed to metal plates with doublefaced carbon tape, then metallized using a Denton Vacuum Desk IV Metallizer.They were photographed using a JEOL JSM 6390LV Scanning Electron Microscope (SEM) under 500-10.000xmagnification and 10-20 kV at the Scanning Electron Microscopy laboratory of the Museu Nacional/UFRJ.As first stated by Price & Kelly (1989), different microdermatoglyphics might occur in basal and apical portions of scales.Thus, we describe both scale portions for A. badius (MNRJ 26712 and USNM 438).Terminology for description of skull osteology follows Cundall & Irish (2008).The head of one individual (MNRJ 26712) was scanned on a Skyscan 1173 invivo high-resolution µ-CT scan at the Nuclear Instrumentation laboratory COPPE/UFRJ.The specimen was scanned at 50 kV and 160 µA and rendered in three dimensions using CTVox for Windows 64 bits version 2.6.Terminology for hemipenial descriptions follows Zaher (1999) with a few minor adaptations based on Passos et al. (2013b).The method for preparation of preserved hemipenes was modified from Pesantes (1994) in replacing KOH with distilled water according to Passos et al. (2016).Prior to the inflation with petroleum jelly, the organs were placed in an ethyl alcohol (70%) solution with Alizarin red for 15-20 min in order to stain the ornamented calcareous structures according to adaptations from original procedures used by Uzzell (1973).

Molecular sampling, techniques, species identification and selection of sequences
We obtained tissue samples from 36 individuals representing 12 nominal species, comprising new sequences for the following species: Atractus badius, A. carrioni, A. elaps, A. favae, A. flammigerus, A. gigas, A. latifrons, A. major, A. riveroi, A. roulei, A. torquatus (Hoegg et al. 2004)].In addition, we included GenBank sequences for 42 terminals, limiting the sampling only to specimens checked for identification, except for two (DHMECN 10179 and KU 214837).However, in all cases we confirmed previous identifications by examination of digital photographs.We assigned specimen DHMECN 5105 tentatively as Atractus ecuadoriensis due to its proximity to the type-locality and because this species is still known only from its typeseries.Unfortunately, this specimen was not available for examination by us.Based on first hand examination, we re-identified several terminals from previously published datasets.
We excluded sequences for which we couldn't trace the voucher, provenance or both, and separated the previously chimeras found in Grazziotin et al. (2012), Pyron et al. (2015, 2016) and Arteaga et al. (2017).

Edition, alignment, evolution model and phylogenetic analyses
All amplified loci were obtained from single voucher specimens to avoid "chimeric terminals" (terminals including sequences taken from distinct individuals representing different species).Data were assembled and aligned in Mega 7.0 (Kumar et al. 2016) under default settings for the alignment program Clustal W (Thompson et al. 1994) and trimmed at both ends.
The best-fit nucleotide substitution models and partitioning scheme were determined simultaneously using PartitionFinder 2 (Lanfear et al. 2016) under the Bayesian Information Criterion (Sullivan & Joyce 2005).We built a concatenated matrix in SequenceMatrix (Vaidya et al. 2011).We employed the Bayesian inference (BI) method to obtain the optimal tree topology of the combined, partitioned dataset using MrBayes v3.2.1 (Ronquist et al. 2012).
All parameters except topology and branch lengths were unlinked between partitions.Four independent runs, each with four MCMC chains, were ran for 20 millions generations, sampled every 10,000 generations.We used Tracer v1.6 (available from http://beast.bio.ed.ac.uk/Tracer) to assess convergence and stationarity to ensure effective sample sizes (ESS) >200 of model parameters.We combined runs using LogCombiner 1.8 after discarding the first 10% of generations as burn-in and summarized them in a maximum clade credibility tree in TreeAnnotator v1.8.3 (Drummond et al. 2012).
We also performed Maximum likelihood (ML) analyses on the partitioned dataset using RAXML (Stamatakis 2014), under GTRCAT approximation.Support of nodes was assessed using the rapid-bootstrapping algorithm with 1000 nonparametric bootstraps.We edited and visualized the phylogenetic trees using FigTree v1.4.2 (available in http://tree.bio.ed.ac.uk/software/ figtree/).We consider the clades with posterior PAULO PASSOS et al.

Phylogenetic analyses
The results from both BI and ML analyses are overall very similar (Fig. 1).We recovered a Guiana Shield clade, maximally supported (Ahighlighted in green), containing A. torquatus sister to A. flammigerus + A. riveroi.This clade is sister to moderately supported clade B (highlighted in pink), which includes two sister clades, (Atractus duboisi, (A.resplendens, A ecuadorensis)) and (A.pachacamac, (A.dapsilis, (A.trefauti, A. schach))), each strongly supported.Color pattern after preservation (n=8): Dorsum of head uniformly black from snout to posterior margin of parietals (= cephaliccap); white band (2-3 scales long) covering the occipital and temporal regions between cephalic-cap and first body dyad; white temporal band usually with triangular descending shape and sometimes interrupted on level of posterior parietal suture by a connection between black cephalic-cap and first body dyad; head laterally black including supralabials, except for white spots (diagonally arranged) covering anterior portion of supralabials among its sutures; seventh supralabial usually part of white temporal band, except in melanic specimens, in which the white pigment is restricted to the anterior portions of supralabials and infralabials near the sutures of each scale; symphysial, first four to six infralabials and almost all portions PAULO PASSOS et al.

PHYLOGENETIC AFFINITIES AND TAXONOMY OF Atractus badius
An Acad Bras Cienc (2022) 94(1) e20191254 9 | 22 of chinshields dark brown to black, with some creamish white pigment; mid-posterior portion of chinshields, gulars and preventrals uniformly whitish cream; belly whitish cream with black spots or blotches concentrated on the lateral edges of ventral scales or on the middle of the body; first third of body with a few dark brown to black squared spots or small checkered bars; dark marks most concentrated on the terminal third of body, mainly after midbody region; usually squared spots arranged linearly on the midline resembling conspicuous longitudinal stripes; black spots often laterally concentrated on the posterior region of body; melanic specimens frequently present ventral marks restricted to lateral margins of ventral scales, giving an impression of ventral invasion of black pigment from dorsal dyads; ventral surface of tail with high concentration of brown or black irregular dots, with cream pigment on the margins of the subcaudals; sometimes median suture of subcaudals predominantly cream in melanic specimens; dorsum of body reddish brown with black dyads (three to four scales long) separated by cream band (1-2 scales long); interspaces between the dyads (4-6 scales long), with black pigment concentrated on the posterior portion of each scale; interspaces usually shorter than dyads; frequently, posterior dyads of body completely darkened with brown to black pigment, similar to background color.
Color pattern in life (n=6): Cephaliccap uniformly black; temporal and occipital regions white except in melanistic specimens; supralabials and infralabials black with white pigment restricted to anterior regions of scales near sutures; gular region creamish white with black spots or dots; dorsum of body red, covered by large black dyads separated by narrow white rings; belly cream scattered with squared dark brown to black blotches, dots and spots; ventral surface of tail mostly dark brown to black with a few cream spots (Fig. 2).
Microdermatoglyphics (n=2): Basal portion of dorsal scales with lamellate and imbricate cells caudally oriented; cell borders 5-6 µm distant from each other, with small and slightly triangulated denticulations, higher than wide and rarely exceeding 0.5 µm; microornamentation on the cell surface composed by small and inconspicuous pores.On the apical portion, cells are lamellate and overlapping; cell borders with long and narrow spinulated denticulations (2-3 µm high) that seem to be embedded in the adjacent cell and are also caudally oriented (Fig. 3).
Skull osteology (n=1): Cranium with colubroid structure; skull height, in lateral view, slightly increases toward end of parietal and decreases towards exoccipitals.Premaxilla: slightly triangular anteriorly with concave lateral edges expanding dorsally; ascendant process of premaxilla expands posteriorly, contacting vertical lamina of nasal and anterior-medial edge of nasals; nasal process absent; laterally, transverse process expands towards maxilla; ventrally, premaxilla expands posteriorly, forming short vomerian process, which contacts anterior-ventral tip of septomaxilla.Septomaxillae: located between nasals and vomers, forming floor of nasal cavity; its anteriordorsal edge fits posterior portion of ascendant process of premaxilla; septomaxillae lateraldorsally projected, comprising short conchal process, approximately trapezoidal without contact with nasals or prefrontals; septomaxillae posteriorly attached to frontals, forming prokinetic joint, and dorsal contact with vertical lamina of nasals.Vomers: premaxillary process slightly expanded anteriorly, delimiting anterior edge of olfactory capsule; premaxillary processes contact each other and septomaxillae dorsally; caudal processes of vomer expanded posterior-medially, constituting exochoanal fenestra.Nasals: trapezoidal, located between premaxilla and frontals, covering nasal cavity dorsally; nasal septum situated vertically between nasals, forming inner wall of nasal cavity; dorsal edges convex; anterior-medial edges contact ascendant process of premaxilla; nasals do not compose prokinetic joint; small process originated on posterior-lateral edge contact prefrontal; vertical lamina contact ventrally middorsal surface of septomaxilla.Frontals: trapezoidal, in dorsal view, located anterior-medially, contributing to dorsal roof and covering ventral and lateral edges of neurocranium; laterally, it forms dorsal edge and inner wall of eye socket; frontals contacting prefrontals anterior-laterally, parietal posteriorly, and postorbital posterior-laterally; frontals ventrally expanded anteriorly into septomaxillary process, contacting septomaxilla at prokinetic joint; orbital lamina descending and converging medially, comprising narrow subolfactory process, resting on medial keel of parabasisphenoid.Parietal: pentagonal, in dorsal view, with concave anterior edge attached to frontals; small anterior-lateral projections articulating with postorbital bone and, exceeding its lateral limits; posteriorly, parietal sutured to supraoccipital, posterior-laterally to prootics and ventral-medially to parabasisphenoid; dorsal surface of parietal smooth with two parietal crests, converging posteriorly reaching anterior edge of supraoccipital.Supraoccipital: fused, pentagonal, located on posterior-dorsal region of skull; supraoccipital contacting parietal anteriorly, prootics anterior-laterally, and exoccipitals posteriorly; anterior edge of supraoccipital concave; lateral margins do not contact supratemporal; dorsally, supraoccipital with oblique crest, which makes its posterior portion ventrally positioned.Exoccipitals: trapezoidal, in dorsal view, and located on posterior edge of skull, comprising dorsal edge of foramen magnum; each exoccipital contacting basioccipital ventrally and prootic laterally; two foramina lateral-ventrally located on each exoccipital, posteriors to foramen ovale.
Basioccipital: pentagonal and located on posterior-ventral portion of braincase; basioccipital contributing to posterior portion of braincase floor and median portion of occipital condyle; basioccipital contacts parabasisphenoid complex anteriorly, prootics and exoccipitals laterally, and atlas posteriorly, where it composes v e n t ra l e d g e o f f o ra m e n m a g n u m .Parabasisphenoid: triangular and located midventrally, contacting frontals anteriordorsally, parietal and prootic laterally, and basioccipital posteriorly; parabasisphenoid with two lateral-ventrally foramina.Prootics: irregular, contacting parietal dorsally and anteriorlaterally, parabasisphenoid and basioccipital ventrally, supraoccipital posterior-dorsally, exoccipital posterior-laterally; anterior portion of supratemporal lies on its dorsal surface; each prootic pierced by two foramina at ventrolateral surface; posterior region of prootics with small enlargement, comprising foramen ovale.Columela auris: small and slender, inserted on foramen ovale, composed by posterior-lateral portion of prootics and anterior-lateral portion of exocciptals; columela auris crosses foramen ovale towards process located on medial portion of quadrate.Prefrontals: irregular, contacting frontal on its latera-posterior region and ventrally maxilla and palatine; lateral descending lamina of prefrontal enlarged dorsally; ventrally, prefrontal with enlarged process supported by palatine; anterior lamina concave with medial process towards prokinetic joint pierced by lacrimal foramen, which crosses prefrontal ventrally.Postorbitals: small, slender with "C" shape, delimiting orbital cavity posteriorly; dorsal portion in contact with anterior-lateral process of parietal and, on its anterior tip, contacting slightly posterior-lateral portion of frontal.Maxillae: curved, limiting lateral edge of anterior portion of skull; maxilla, in lateral view, extends from anterior tip of nasals to posterior  (2022) 94(1) e20191254 12 | 22 portion of frontals; posterior tip of maxilla attached to anterior tip of ectopterygoid; two medial processes: palatine on medial portion contact dorsally prefrontal; and ectopterygoid on posterior portion contacting ectopterygoid; lateral surface of maxillary concave and pierced by single foramen; maxillary teeth 6-7 posteriorly curved and decreasing in size.Palatines: slightly shorter than maxillaries, in anterior-dorsal contact with prefrontal; posterior tip slightly forked and attached ventrally to anterior-dorsal tip of pterygoid; maxillary process of palatine located laterally, do not contact palatine process of maxilla, and supporting ventral portion of prefrontal dorsally; palatine with longitudinal row of five to six teeth; choanal process absent.Pterygoids: located posterior to palatine and its posterior end exceeding posterior limits of skull; peterygoids with longitudinal row of 12 teeth, smaller and reducing in size posteriorly; ectopterygoid fits to dorsal surface of pterygoid approximately on the level of gap between fifth and sixth teeth, extending third teeth.Ectopterygoids: elongated, connecting maxilla to pterygoid; anterior portion bifurcated and attached to ectopterygoid process of maxilla, exceeding maxillary row of teeth and reaching sixth to seventh teeth; ventrally attached to d o rsal-m e d i al p o r t i o n o f p te r y go i d .Supratemporals: narrow and elongated bones placed lateral-posteriorly to braincase, ventrally attached to posterior portion of prootics and posteriorly reaching posterior portion of exoccipitals; posterior-dorsal contact with quadrate.Quadrates: elongated and vertically positioned bones, located on posterior portion of skull and laterally angulated; proximal portion enlarged, medially articulated with posteriordorsal portion of supratemporal; distal end slightly enlarged, articulating with glenoid cavity of retroarticular process of mandible, forming quadrate-articular articulation; lateral lamina twined, facing to posterior portion of skull and with short medial process.Compound bone: elongated bone comprised by surangular, articular crests and retroarticular process (2/3 mandibular length); compound bone is positioned at posterior end of mandible, connected to skull through glenoid cavity on its posterior end; anterior portion projected as tapering process, which inserts into dentary and extends until level between sixth and seventh teeth of dentary; lateral lamina of compound bone concave, and its midventral portion contacts angular; mandibular fossa relatively deep, with anterior foramen located on 1/3 of bone, composed by surangular crest laterally and prearticular crest medially; prearticular crest slightly higher than surangular; anteriorlateral portion with foramen on each bone; retroarticular process short and rounded, medially curved.Dentaries: located at anterior portion of mandible, each curving medially on its anterior end towards opposite dentary; longitudinal row of eight teeth posteriorly curved and reducing gradually in size; lateralposteriorly inserting to compound bone on its forked portion; anterior end of compound bone fits into these processes, reaching level of sixth tooth; mentonian foramen located laterally at level of fifth tooth; midposterior end of dentary contacts anterior tip of angular and anterior half of splenial; Meckel's groove opening at fourth tooth level and extending posteriorly through whole dentary.Splenials: narrow and short bones, located on medial surface of mandible, with mesoventral orientation with respect to articulation between dentary and angular; anterior portion fits into dentary through tapering process, reaching gap between fourth and fifth tooth, and its posterior end attached to angular; posterior region of splenial pierced by mylohyoid foramen.Angulars: triangular and located on medial surface of mandible, with tapering projection directed posteriorly and anterior-dorsal process extending above splenial, reaching posteriordorsal tip of dentary dorsally; dorsal process of angular originates at level of seventh tooth; anterior-ventral portion of angular pierced by mylohyoid foramen (Fig. 4).
Hemipenial morphology (n=3): Organ in situ (retracted) extends at level of 8-10 th subcaudal and bifurcates at level of 5-6 th subcaudal.Fully everted and maximally expanded hemipenis bilobed, almost bicapitated and bicalyculated (sensu Zaher 1999); lobular region equivalent (based on previously semi-everted organs), or slightly wider than hemipenial body; lobes attenuated and centrifugally oriented with a distinct "V" orientation, only fully developed on the previously inverted organs (Fig. 5b, e); lobes symmetrical or asymmetrical (Fig. 5a, d) with right lobe slightly longer than left one; lobes covered with spinulate calyces; spinules replaced by regular papillae toward apices of lobes; basal region of each lobe with a welldefined capitular groove on asulcate and lateral sides; capitular groove indistinct on the sulcate side; capitulum with regular rows of papillae on lobes and intrasulcar region; intrasulcar region on the sulcate side of organ covered with irregularly distributed small size spines; hemipenial body elliptical (Fig. 5a, d) or triangular (Fig. 5b, e) and slightly longer than capitulum; hemipenial body ornamented with hocked spines ranging from moderate to large size; larger spines distributed along both sides of organ, except for the most basal portion of hemipenis; hooked spines sometimes with laterally expanded basis on both sides of organ; sulcus spermaticus bifurcating basally and extending for about 30% of organ length; sulcus spermaticus bifurcation occurring much earlier from capitular groove on the proximal region of hemipenial body with (Fig. 5e-f) or without (Fig. 5d) spines; each branch of sulcus spermaticus centrolinearly oriented, running to tip of lobes; sulcus spermaticus margins relatively thick at level of division and laterally expanded on capitular region; sulcus spermaticus not bordered by spinules; most proximal region of hemipenial body poorly ornamented, covered with a few spines (Fig. 5e), spinules (Fig. 5d) or longitudinal plicae (Fig. 5f ); naked pocket shallow, extending laterally (right side) from most basal region to length of hemipenial body.

DISCUSSION
Arteaga et al. ( 2017) recently provided a molecular phylogeny of the genus Atractus (with emphasis on the Ecuadorian species) based on three mitochondrial genes and including approximately 20% of the generic diversity.That study (with unprecedented taxonomic sampling for the genus) improved much of our knowledge about some specific relationships and supported Atractus badius as unrelated to the remaining Guiana Shield species.We herein confirm A. badius as an independent lineage related to Guiana Shield taxa (Fig. 1), which is corroborated by its distinctive hemipenial morphology (Fig. 5).Savage (1960) reviewed the Ecuadorian species of Atractus and proposed three species groups to accommodate the diversity known at that time.The Atractus badius group was defined by possession of the 'differentiate' condition (= capitulated organs) in the hemipenial morphology, but the nominal species of the group is not autochthonous to Ecuador (Savage 1960, Hoogmoed 1980).Hoogmoed (1980) provided a brief hemipenial description of A. badius based on its dissected and completely retracted organ, providing a limited number of characters for comparison with other congeners.We recovered A. badius in a highly supported clade with A. favae and A. trilineatus (Fig. 1).Unfortunately, we only had access to females of A. favae and its hemipenial morphology remains only known from its completely retracted condition (see Hoogmoed 1980).In addition, we obtained an unstable position of Atractus favae in both molecular analyses.Considering some hemipenial similarities (Fig. 5), A. badius shares a nearly bicapitate and bicalyculate organ with A. paisa (Passos et al. 2009a).However, other endemic species from the Cauca Valley in the Colombian Andes display a similar condition of the hemipenial ornamentation and general lobular structure (Passos 2008).We anticipate the possibility that some of them may be recovered in this clade once more species from the Andes of Colombia are incorporated in a  f=frontal, p=parietal, po=postocular, cb=compound bone, pro=prootic, so=supraoccipital, eo=exoccipital, rap=retro-articular process, sm=septomaxilla, st=supratemporal, q=quadrate, ca=columela, pt=pterygoid, ep=ectopterygoid, d=dentary, pal=palatine, v=vomer, vp=vomerian  broader phylogenetic study.Such hemipenial morphology does not occur in any other congener (Passos 2008).Consequently, we herein restrict the A. badius group to the nominal species waiting for new phenotypic or molecular evidences to render natural groups, even because two of the three Savage's species groups were recovered in the same clade (Fig. 1).
The diagnosis of Atractus pyroni was presented in its original description (Arteaga et al. 2017:102 p., Table 3).Arteaga and colleagues began by stating that the holotype of Atractus pyroni is an adult male (Arteaga et al. 2017:111 p.).However, in their figure 7 (Arteaga et al. 2017:112 p.) the caption presents the specimen as an adult female.The authors provided some additional morphological comparative data for the Atractus roulei species group (A. carrioni, 'A. lehmanni', A. pyroni and A. roulei).However, the meristic, pholidosis, and color pattern characters of A. pyroni completely overlap with those from A. roulei (Table III).Arteaga and colleagues distinguished A. pyroni from A. roulei by 'a distinct bicolored pattern'.Nonetheless, such supposedly different pattern of A. pyroni from the typical uniformly or reticulated colored A. roulei (see its color variation on the left margin of the Fig. 1) is because the specimen has been collected and preserved with partially retained ecdysis (Fig. 7 Arteaga et al. 2017:112 p., Fig. 7 from the present study).Consequently, the first author easily removed the outermost layer of some scales from the holotype of A. pyroni (= oberhäutchen, Fig. 7d), revealing uniform coloration typical from A. roulei.Furthermore, all meristic data from their specimen of 'A.lehmanni' also overlap with those from A. roulei.One of us (PP) also examined nearly all the specimens used by Arteaga et al. (2017), for this clade: a female holotype of Atractus pyroni (MZUTI 5107), a male specimen of 'Atractus lehmanni' (DHMECN 7644) and a female specimen of Atractus roulei (DHMECN 7645).We found no phenotypic differences among them, since all specimens possess dorsal scale rows in 15/15/15, supralabials six (3 rd and 4 th contacting the orbit), infralabials six (three or four contacting chinshields), postocular single, temporals 1+2, maxillary teeth 10 (not eight as reported by Arteaga et al. 2017) or 11, preventrals 0-2, ventrals 142-145 in females, subcaudals 21 in female, and 28 in male, and similar dorsal and ventral coloration (Table III).Boettger (1898), in the original description of Atractus lehmanni, diagnosed the species by having the following combination of characters: dorsal scales rows in 17/17/17; supralabials seven (four or five infralabials in contact with chinshields); postoculars two; incomplete reddish-white nuchal collar, dark-brown to black dorsum with reddish white tip of scales; venter black with reddish white square blotches; ventrals in females 148-153, 142-144 in males; subcaudals 20-21 in females, 25-27 in males.Later, Savage (1960) described the color pattern and the inverted condition for the hemipenis of A. lehmanni based on one male syntype (currently a paralectotype; see Mertens 1967).There is no reason to assume that specimens of A. lehmanni or A. roulei present such a disparate variation in many phenotypic traits.This fact could be checked through consultation of public repository as GBIF, in which photos of the lectotype of Atractus lehmanni are available or perhaps directly examining the available types.For that reason, the simplistic explanation based on available evidence suggests a misidentification of the Arteaga's  pyroni as a valid species we have made A. roulei paraphyletic (Fig. 1).The alternative taxonomic posture implies recognizing three or four species The inaccurate identifications of Atractus species in molecular phylogenies are more common than previously thought (above 30% of sequences or specimens available in GenBank are misidentified or doubtfully identified with respect to the study that produced such sequences; Table I).As pointed out by Tautz et al. (2003), DNA-based systematics must be firmly anchored within the knowledge, concepts, techniques and infrastructure of traditional taxonomy.More importantly, 'DNA-taxonomy' will have to be based mainly on sequences from newly collected individuals identified by experienced taxonomists (Tautz et al. 2003:71 p.).We argue that such caveats are less harmful for the macro-evolutionary or macro-ecological studies in which the focus is not direct towards promoting taxonomic changes or nomenclatural acts.Still, there have been systematic-based studies in which authors abstained to make shifts for taxa or clades with complex and/or unresolved taxonomy (see Pinna et al. 2018) and, sometimes, less representativeness of species as terminals (Grazziotin et al. 2012, Zaher et al. 2014).By contrast, we believe that there are three main hierarchical levels of implications to this question: (i) as DNA databases grow exponentially (e.g., GenBank, BOLD), the use of misidentified sequences and even chimeras, strictly with species-level taxonomic proposal represents a real peril; (ii) phylogenies built upon sequences without associated vouchers prevents from the possibility of falsifying first identification hypotheses for the new sequences added to a public repository; (iii) deliberate non-availability of data associated with uploaded sequences at GenBank, such as detailed location of the voucher collection have been omitted (e.g., Hamdan et al. 2017).Although the issues highlighted above are increasing fast with the reduction of the cost and new sequencing techniques, they affect the taxonomy of some groups disproportionately, notably for some snakes.As a rule, historically the snakes' sampling efforts, and availability of geographically balanced samples of taxa with secretive lifestyle both are elusive.Besides, some cryptozoic snakes, as Atractus, differ from other vertebrate groups in presenting a few locally exhaustively sampled regions, but with many subsampled portions along species distribution range.To complicate this matter, many Atractus species are apparently rare with supposedly restricted endemism, only because of the lack of knowledge (de Fraga et al. 2017), making the geographic range of variation of local samples sometimes unpredictable.Therefore, we claim to authors improving the accuracy on species identifications, integrating the data with exhaustive comparisons, prior to split new names and traditional taxa concepts, considering the long-standing (and functional) taxonomic system (Pinna et al. 2018).

Figure
Figure 2. General view in life of the body of Atractus badius, showing banded coralcolor pattern (a-c, e), specimen entirely melanic (d), and ventral view (f).Photos by Bernard Dupont (c, d, e) and Fausto Starace (a, b, f).

Figure 6 .
Figure 6.Geographical distribution of Atractus badius.Yellow circle represents the typelocality, white circles literature records and red circles the new records provided herein.
published such study describing Brachyorrhos badius together with B. flammigerus, B. schach and B. torquatum (see Hoogmoed 1980 for more details on its nomenclatural history).Wagler (1828) erected the genus Atractus to accommodate a new species A. trilineatus.Later, Wagler (1830) proposed the synonymy of Brachyorrhos Kuhl with Atractus, recognizing B. badius under the name Atractus badius.Duméril et al. (1854) proposed the genus Rabdosoma and used a new combination, R. badium.The same authors included B. flammigerus and B. schach as junior synonyms of R. badium, but recognized R. torquatum as a distinct species (see Passos & Prudente 2012 for additional details).Günther ( and A. trilineatus.

Table II .
Specimens of Atractus sequenced in this work with Genbank accession numbers.

Table III .
Meristic and morphometric variation for the holotypes of Atractus pyroni and A. roulei, and the syntypes of A. lehmanni, as well as additional samples of A. roulei.The range of each continuous variable is from our own sample, except for A. lehmanni in which we are based on Boettger (1898) complemented with high-resolution photographs from the type-series.Abbreviations correspond to: SVL (snout-vent length), TL (tail length) and "-" (unknown data).The numbers in parentheses represent the total number of individuals available in collections for each sex.We placed together for both sexes the variables lacking secondary dimorphism.The holotypes of A. pyroni and A. roulei are marked in bold.order to restore the reciprocal monophyly of A. pyroni and A. roulei.However, as highlighted above both are not diagnosable on the basis of available morphological evidence.Therefore, given the data gathered in the course of study, we propose herein the formal synonymy of A. pyroni with A. roulei. in