A new species of groundsnake genus Atractus Wagler, 1828 (Serpentes, Dipsadidae) from the Peruvian Andes revealed by unequivocal morphological characters

Based on an exhaustive revision of external morphological characters we describe a new species of Atractus from the humid montane forest of the Andes of northern Peru, Cajamarca department, occurring at elevations of 1641 to 2161 m. This new species was misidentified as A. gigas in the literature and for more than a decade represented the southernmost record of the that species. In the absence of molecular data and limited by a small sample, we use some underreported characters in the genus Atractus such as the presence of apical pits. Thus, the combination of apical pits as well as other characters mentioned in the literature (i.e., head scute-llation and number of subcaudals) distinguishes the Peruvian population from A. gigas , and strongly supports the morphological separation of this taxon from the rest of its congeners.


Introduction
During last five years, the taxonomy of the genus Atractus has improved notably with molecular data leading to systematization of this fascinating group of cryptozoic snakes in the tree of life (Arteaga et al. 2017;Melo-Sampaio et al. 2019;Murphy et al. 2019;Jowers et al. 2021;Melo-Sampaio et al. 2021a;Arteaga et al. 2022).Those studies collectively led to an increase in diversity with seven new species "molecular-based" or "phylogenetically positioned" as well as generating relevant new data for poorly known ones.On the other hand, in the same timespan morphological data was also critical to shedding light on eight new species (Passos et al. 2016;Melo-Sampaio et al. 2019;Meneses-Pelayo and Passos 2019;Passos et al. 2019a, b;Melo-Sampaio et al. 2021a).Faced with this encouraging scenario of discoveries, Passos et al. (2022) argued that phylogenies are complements equally important when available, but data integration should be solid enough in the test of hypotheses and grounded by multiple lines of evidence.
Despite having a usually secretive habit that makes sighting by humans difficult, many species of Atractus are relatively large, reaching more than 600 millimeters (Silva-Haad 2004;Tolhurst et al. 2010).Among the largest species of the genus are A. albuquerquei Cunha & Nascimento, 1983, A. atlas Passos, Scanferla, Melo-Sampaio, Brito & Almendáriz, 2019, A. gigas Myers & Schargel, 2006, A. major Boulenger, 1894, A. obesus Marx, 1960, A. pachacamac Melo-Sampaio, Passos, Prudente, Venegas & Torres-Carvajal, 2021, A. serranus Amaral, 1930, A. touzeti Schargel, Lamar, Passos, Valencia, Cisneros-Hereda & Campbell, 2013, A. trihedrurus Amaral, 1926, and A. ukupacha Melo-Sampaio, Passos, Prudente, Venegas & Torres-Carvajal, 2021.Descriptions of "giant" Atractus in the Andes region are uncommon and primarily based on small samples (Marx 1960;Passos et al. 2019a).Atractus gigas was described based on a single specimen from Bosque Protector Río Guajalito, between Quito and Santo Domingo, Pichincha, Ecuador (Myers and Schargel 2006).Later, Tolhurst et al. (2010) found the first live specimen in primary cloud forest, Bosque Protector Santa Lucía, extending the distribution 48 km NE from the type locality.Passos et al. (2010), provided new information about phenotypic variation based on 11 specimens from Ecuador and Peru, highlighting that those Peruvian specimens have a smaller number of subcaudals than Ecuadorian specimens of Atractus gigas.Additionally, they have more supralabials, more infralabials, and four infralabials contacting chinshields.Despite these significant differences in characteristics between the Ecuadorian and Peruvian populations, Passos et al. (2010) argued that they represented the first occurrence of one Atractus species (A.gigas) on both sides of the Andes.Recently Arteaga et al. (2022) properly raised concerns about the status of Atractus gigas along the eastern versant of the Andes.They argue that scutellation and ontogenetic variation between Ecuadorian and Peruvian specimens are easily diagnosable and suggest re-identification as Atractus atlas or perhaps as a new species.Analysis of other species occurring along the Andes led us to exhaustively verify diagnostic morphological features hitherto unknown in some Atractus and allows us to describe the Peruvian species previously confused with A. gigas as a new species.

Morphology, species boundaries and presentation rationale
Terminology for cephalic shields follows Savage (1960) and Peters (1964), whereas ventral and subcaudal counts follow Dowling (1951).Body proportions were measured from snout-vent length (SVL) and tail length (TL), with a flexible ruler to the nearest 1 mm.We used a Mitutoyo digital calliper to the nearest 0.1 mm in measurement of head scalation.Measurements and descriptions of paired cephalic scales are strictly based on the left side of the head.Dentition was examined in situ under an Olympus stereomicroscope through a narrow lateromedial incision between the supralabials and the maxillary arch.After removing tissues covering the maxillary bone, we counted the number of teeth or empty sockets.Colouration was described following a standardized catalogue (Köhler 2012).

Taxonomy
Atractus paulus sp.nov.4).In addition, other diagnostic features between A. paulus and species with specimens recorded above 600 mm SVL were provided in table 1 of Passos et al. (2010).However, inasmuch as size has evolved independently in various clades (Arteaga et al. 2017;Melo-Sampaio et al. 2019a, 2021a;Arteaga et al. 2022).None of those giant species shares the presence of two apical pits on dorsal scales, so we restrict the comparisons to two species possessing two apical pits on dorsal scales and occurring at high elevations in montane forest of the eastern Andes: Atractus duboisi (Boulenger, 1880) and Atractus orcesi Savage, 1955 (Figs 5, 6).
We did not directly examine the holotype of Atractus ecuadorensis Savage, 1955 and Atractus resplendens Werner, 1901, so it is not known whether they have apical pits, likewise, it does not allow us to conclude that these species constitute a natural group with the previous ones (see Fig. 7).Although Arteaga et al. (2022) noticed that a black stripe on a yellow belly is a characteristic shared by A. duboisi, A. discovery, and A. orcesi, they mistakenly assumed that it is absent in A. resplendens.This is clearly stated to be present in the description given by Werner (1901).Description of the holotype.An adult female, SVL 830 mm, tail length 83 mm (9.1% of total length); head distinct from body; head length 22.6 mm (2.7% SVL); head width 19.8 mm (87.6% head length); rostral-orbit distance 9.3 mm; nostril-orbit distance 6.8 mm; interorbital distance 10.9 mm; head rounded in lateral view; snout sub-acuminate in dorsal view, truncate in lateral view; canthus rostralis not conspicuous; rostral subtriangular in frontal view, 4.6 mm wide, 3.0 mm high, slightly visible in dorsal view; internasal 2.5 mm long, 2.30 mm wide; internasal suture sinistral with respect to prefrontal suture; prefrontal 6.1 mm long, 4.9 mm wide; supraocular subtrapezoidal, 4.5 mm long, 3.4 mm wide at broadest point; frontal pyramidal, 7.0 mm long, 5.9 mm wide; parietal 10.8 mm long, 7.1 mm wide; nasal entirely divided, nostril well-spaced into both pre-and postnasal; prenasal 2.4 mm high, 1.5 mm long; postnasal 2.6 mm high, 2.5 mm long; loreal 5.0 mm long, 1.9 mm high; second, third and fourth supralabials contacting loreal on left side; third to fifth supralabials contacting loreal on right side; eye diameter 2.9 mm; pupil rounded; two postoculars similar in height, being upper longer than lower; upper postocular 2.2 mm long, 2.7 mm high; lower postocular 1.4 mm long, 1.6 mm high; temporal formula 1+2; first temporal 5.3 mm long, 4.3 mm high; upper posterior temporal, 5.6 mm long, 4.3 mm wide; supralabials nine, fifth and sixth contacting eye on right side, supralabials eight, fourth and fifth contacting eye on left side; first supralabial narrower (1.4 mm wide) than second (1.7 mm wide) and similar in height; third supralabial trapezoidal, similar in height and wider (2.1 mm) than second; sixth (right) and seventh supralabial taller (left) and seventh (right) and eighth longer (left) than remaining supralabials; symphysial subtriangular, 3.1 mm wide, 1.3 mm long; first pair of infralabials preventing contact symphysial-chinshields; infralabials eight (right) and seven (left), first four contacting chinshields; chinshields 8.9 mm long; gular scale rows four; preventrals four; ventrals 167; subcaudals 26/26 respectively from left to right side; dorsal scale rows 17/17/17, with apical pits at level of cloaca; midbody diameter 27.3 mm (3.3% of SVL); caudal spine 3.2 mm long, shorter than last fused subcaudal scale (2.8 mm).Maxillary bone arched upward anteriorly in lateral view, ventral portion curved on anterior and nearly flattened on median to posterior portion; maxillary with eight teeth; teeth angular in cross section, robust at base, narrower at apices, curved posteriorly; lateral process of maxilla well developed (Figs 2, 3).Paratype ZFMK 89147 agrees in most of characters of holotype, differing only in colouration and scale counts (166) in venter.
Colour in life and preservative.Passos et al. ( 2010) described the colouration as follows: dorsum and background of head mostly dark brown.We observed that dorsum and head are Raw umber (colour 23) with gradation to cinnamon in contact internasals-prefrontals (colour 21) becoming cinnamon rufous (colour 31) on snout region; supralabials cinnamon with small invasion of cinnamon rufous; mental and gular regions mostly dark brown, with few cream (colour 12) dots; venter light to dark brown, with few cream to greyish brown dispersed dots (colour 284); underside of tail dark brown.Ontogenetic variation is pronounceable (see fig. 2 of Passos et al. 2010), where juvenile has head and dorsal ground colour of body dark greyish brown to sepia (colour 286), with few dispersed  Distribution and natural history.The new species is known from two close localities Alto Ihuamaca and El Chaupe, at elevations of 1641-2161 m in the northern portion of the Cordillera Occidental of the Andes, San Ignacio province, Cajamarca department, Peru (Fig. 8).Both localities are in agricultural areas with coffee plantations and pastures for cattle ranching with scattered small patches of montane forest and secondary vegetation.According to Passos et al. (2010), the adult female holotype was collected by day crossing a trail at the forest edge in an open area and the juvenile paratype inactive under a log near a coffee plantation.The holotype con-tained 12 eggs in the oviduct, measuring 30.4-36.3 mm (X = 33.7 mm) length and 14.5-16.3mm (X = 15.5 mm) width (Passos et al. 2010).
Etymology.The species epithet "paulus" is a Latin word being patronym for our friend Paulo Gustavo Homem Passos.Dr. Paulo Passos has described more than 34 Atractus species, approximately one-fifth of the astonishing diversity of this complex genus.The Latin word "Paulus" also means "small" and thus, we also refer to the type series composed only by the holotype and paratype.
Remarks.The montane forest below 2000 m in the San Ignacio province has almost disappeared owing to agricultural activities.Only scattered patches of montane forest and secondary vegetation can be observed in the landscape.The National Sanctuary Tabaconas Namballe is between approximately 5 and 9 km in a straight line from the localities of Atractus paulus preserve forests from 2000 m.Even other protected private areas in San Ignacio possess most of their natural surfaces above 2000 m.
However, due to scarcity of knowledge about the distribution of this new species, especially its altitudinal range, we could not objectively propose a conservation category based on the IUCN criteria, but due to the levels of habitat destruction in the area efforts should be allocated for the conservation status of this species to be rapidly reassessed.

Discussion
The description of Atractus paulus raises questions and concerns about the lack of attention to detail in the use of some traditional morphological characteristics in snake taxonomy, such as the presence of apical pits, within the genus Atractus.The interest in making taxonomy increasingly integrative with the inclusion of molecular data combined with the need to publish data faster has minimized the use of informative and well-known character systems such as hemipenial morphology (Passos et al. 2012).Consequently, diagnostic comparisons are becoming increasingly basic and inadequate (see Arteaga et al. 2017Arteaga et al. , 2022)).Passos et al. (2009) removed Atractus duboisi and A. orcesi from the synonymy of Atractus occiptoalbus (Jan, 1862) as proposed by Savage (1960).Passos et al. (2009: 391-392) give the following diagnosis for Atractus duboisi: 15/15/15 dorsal scale rows without apical pits, supra-anal tubercles and keels; usually eight supralabials, with 4-5 th contacting orbit; usually seven infralabials; seven maxillary teeth; usually four series of gular rows; subcaudals 23-26 in males.However, after careful analysis of the specimens of A. duboisi, we found that all specimens possess apical pits (often more visible near the cloaca in females, sometimes distributed throughout the body, except in the first two dorsal scales and vertebral row), usually seven supralabials in males, eight maxillary teeth and 30-35 subcaudals in males similar to reported in Boulenger (1894).Furthermore, Passos et al. (2009) in the revalidation of Atractus orcesi Savage, 1955, did not mention presence of apical pits in this species.After examining specimens, we found that apical pits are also present (Figs 5,6).The ornamentation of the scales (with apical pits, keels and tubercles), which was an innovative character in the positioning of species in the genus Atractus in the past (Gasc and Rodrigues 1979;Hoogmoed 1980), was neglected for a long time, but finally received attention again with the publication of Passos et al. (2013) on A. caxiuana;Fraga et al. (2017) on Atractus riveroi; and Passos et al. (2022) on Atractus badius, and was also very important for this study.Arteaga et al. (2017), provided the first molecular phylogeny with mitochondrial genes including A. duboisi, A. ecuadorensis Savage, 1955 andA. resplendens Werner, 1901.Although A. ecuadorensis and A. resplendens are rare species very little has been discussed about their morphological variation.Arteaga et al. (2022) expanded their phylogeny to include A. orcesi and and a new species, A. discovery, and retrieved both species nested in a clade sister to A. duboisi + A. resplendens, respectively.
Based on previous phylogenetic results and the presence of apical pits, herein we name the Atractus duboisi group, including the following species: A. discovery, A. duboisi, A. orcesi, and A. paulus.Atractus discovery is tentatively assigned to this group based on molecular positioning retrieved in Arteaga et al. (2022).We refrain to include A. ecuadorensis and A. resplendens because we are unable to check the vouchers (DHMECN 5105 and MZUTI 3996 respectively) utilized by Arteaga et al. (2017).Complementarily, a recent study by Passos et al. (2022) shows that many sequences presented in Arteaga's work included misidentifications, but do not affect or influence the decisions made here.Furthermore, Arteaga et al. (2022) criticized Passos' work but themselves concede that some reidentifications were warranted by recognizing synonymy of one species that they described.On the other hand, two juvenile specimens QCAZ 5183 (Fig. 6C) and QCAZ 12666 described as paratypes of Atractus zgap have apical pits and therefore verification of the condition in the holotype (ZSFQ 4946) is crucial to understanding interspecific relationships of this group if they are not conspecific.Additionally, the status of A. discovery needs to be assessed through exhaustive comparison with A. resplendens, comparing the hemipenial morphology of both species.Despite that both species were recovered as sibling species and males are available in museum collections, preparation of hemipenes was not done.
Recent taxonomic revisions have improved knowledge about Atractus species described based on holotypes or small type series (Passos et al. 2013;Schargel et al. 2013;Köhler and Kieckbusch 2014;Arteaga et al. 2017Arteaga et al. , 2022;;Melo-Sampaio et al. 2019;Passos et al. 2019a, b;Melo-Sampaio et al. 2021a).As long as there is no systematic collection in the Andes, the rate of discovery of new species of reptiles will always be a product of serendipity so a better understanding of snake relationships in a vicarious context like the Andes is still in its infancy (see Melo-Sampaio et al. 2021a, b).However, the main forces that impede taxonomic progress in Atractus are: the large number of synonyms, the inaccurate description of species, the unjustified revalidation of synonymized species, the low representativeness in museums and the sample gaps established by the difficulty of access to the collection sites, financing or even for geopolitical and bureaucratic issues that do not allow the joint effort of researchers.

Figure 2 .
Figure 2. Dorsal pattern of juveniles for comparison: A A. gigas MEPN 8706 and B A. paulus.

Figure 3 .
Figure 3.Comparison of holotypes head in lateral, ventral and dorsal views.A-C Atractus gigas; D-F Atractus paulus.Note the contact of infralabials with chinshield (3 in A. gigas versus 4 in A. paulus).

Figure 4 .
Figure 4. Comparison of holotypes head in dorsal, ventral and lateral views.G-I Atractus atlas and J-L Atractus touzeti.