A new species of Erythrolamprus from the oceanic island of Tobago (Squamata, Dipsadidae)

Abstract Tobago is a small island on the southeast edge of the Caribbean Plate with a continental flora and fauna. Using DNA sequences from Genbank, new sequences, and morphological data from the snakes Erythrolamprusepinephalus, E.melanotus, E.reginae, and E.zweifeli, the species status of specimens of a Tobago snake previously considered to be Erythrolamprusreginae was assessed. Erythrolampruszweifeli, long considered a subspecies of E.reginae, was found to be a northern Venezuela-Trinidad endemic and the sister to E.reginae. The trans-Andean species E.epinephalus is shown to be non-monophyletic while the Costa Rican lineage of E.epinephalus is weakly supported as the sister to the Tobago population. The Tobago Erythrolamprus is described as a distinct taxon based upon five specimens from four localities in lower montane rainforest. Much of the new species range includes the Main Ridge Forest Reserve of Tobago, the oldest protected forest in the Western Hemisphere. All known locations fall within a 400-ha area, and its total geographic distribution is likely to be less than 4,566 ha. The restricted distribution of this new snake makes it a likely candidate for threatened status. The new species also becomes another biogeographic link between northern Venezuela and Tobago.


Introduction
The Cordillera de Costa (CC) is a sky island archipelago that extends 925 km in an east-west orientation from western Venezuela, across the Northern Range of Trinidad to the island of Tobago. The CC is separated from the Andes by the Yaracuy River depression, and in the east, the CC is separated from the Guyana shield by the Llanos grasslands. The Gulf of Paria separates the Peninsula de Paria from Trinidad, and Trinidad is separated by 35 km of open water from Tobago. The CC formed between the late Cretaceous and the Miocene (Sisson et al. 2005). Rising and falling sea levels, marine incursions, changing climates, and tectonic events have continually remodeled the landscape isolating and reconnecting populations of organisms.
Tobago is at the eastern edge of the CC sky island complex and is slightly more than 300 km 2 , and its highest peak is about 576 m above sea level (ASL). The island has two physiographic regions: a flat coastal plain composed of a coral terrace in the southwest and the Main Ridge, a mass of metamorphic and igneous rocks, covered by dense tropical forest. The Main Ridge runs in a northeast-southwest direction.
Tobago's snake fauna contains 23 species, and eleven of these belong to the Dipsadidae clade. Molecular studies on the Western Hemisphere snake clade Dipsadidae (or Dipsadinae) (Zaher et al. 2009;Vidal et al. 2010;Grazziotin et al. 2012) suggest Erythrolamprus Boie 1826 is not monophyletic unless most of the snakes formerly placed in the genera Liophis Wagler, 1830, Leimadophis Fitzinger, 1843, and Umbrivaga Roze, 1964 are included. This action increased the number of Erythrolamprus species from six, mostly coral snake mimics, to 50 species (Uetz and Jacob 2018) with a variety of color patterns and habits. Thus, Erythrolamprus became one of the most species-rich genera of Neotropical snakes. This arrangement has not been accepted by everyone (Wallach et al. 2014). Here we consider the genus Erythrolamprus in the broadest sense, including the species traditionally allocated to Leimadophis, Liophis, and Umbrivaga, acknowledging that future taxonomic changes are likely.
There is no known synapomorphy for the genus Erythrolamprus (Myers 2011). That said, members of the genus are usually less than 1.6 m in total length; nine scales are normally present on the crown; the number of dorsal scale rows is 15-19 and in some species they are reduced once, in others, they may be reduced twice posteriorly; apical pits may be present or absent on some or all of the scales; ventral counts range from 129-212; subcaudal counts range from 38-106; the temporal formula is usually 1+2; the preocular is usually single; the postoculars are usually two; upper labials are usually eight; lower labials are usually ten, and two pair of enlarged chin shields are present.
Erythrolamprus ranges from Costa Rica southward to Argentina and occurs on both sides of the Andes as well as in the Lesser Antilles. Some taxa reach an elevation of 3,500 m ASL. Members of the genus have life styles that range from fossorial to terrestrial to semi-aquatic in habitats spanning rainforests, savannas, and páramo (Savage 2002).

The genus Erythrolamprus in the Cordillera de la Costa
Eighteen species of Erythrolamprus occur in northern Venezuela, of these, two are Pantepui species: E. trebbaui (Roze 1958a), E. ingeri (Roze 1958b). The remaining 16 species are associated with the CC either as montane species, lowland species, or species that are not restricted by elevation. Eight of the 16 species occur on the Guyana Shield and seven species of Erythrolamprus are recognized on Trinidad and Tobago: E. aesculapii (Linnaeus, 1758); E. bizona Jan 1863; E. cobellus (Linnaeus, 1758); E. melanotus (Shaw, 1802), E. ocellatus Peters, 1868; E. zweifeli (Roze, 1959); and E. reginae (Linnaeus, 1758) (Murphy et al. 2018). Both E. aesculapii and E. bizona are coral snake mimics, and each is known from a single specimen from Trinidad (but better known from elsewhere in their ranges).
Erythrolamprus ocellatus is a Tobago endemic, with a bright red dorsum and black ocelli, and is best considered an imperfect coral snake mimic, keeping in mind that there are no extant species of coral snakes on Tobago (Hodson and Lehtinen 2017). Erythrolamprus cobellus is a semi-aquatic, lowland species; while E. melanotus and E. zweifeli are forest species often associated with stream-edge habitats and mountains from sea level to at least 2,000 m. However, in Venezuela, E. zweifeli is usually associated with montane environments. Noting significant differences in coloration, as well as distinct ventral and subcaudal counts from E. reginae, Rivas et al. (2012) returned Erythrolamprus reginae zweifeli to species status. They noted E. zweifeli differs from E. reginae in having a salt-and-pepper dorsal pattern or a more uniform olive-green or olive-brown pattern. Wallach et al. (2014) concurred and recognized the elevation of zweifeli to species level. E. reginae has a dor sum with dense pale and dark paravertebral flecking. The two species also differ in subcaudal counts (69−88 in E. zweifeli as opposed to 55−78 in E. reginae) with the ranges overlapping, but different means. Following this arrangement, E. zweifeli occurs through out the Cordillera de Mérida and the CC in Venezuela, including Trinidad.

Natural history of Trinidad and Tobago´s Erythrolamprus
There are some ecological differences between the Trinidad and Tobago Erythrolamprus. The two poorly known coral snake mimics (E. aesculapii, E. bizona) are forest dwellers and snake predators (Campbell and Lamar 2004). Dietary differences between the two better known forest and forest-edge species are apparent. Erythrolamprus melanotus feeds on the microteiid lizards in the genus Bachia, the rain frog Pristimantis urichi, the puddle frog Engystomops pustulosus and the gecko Gonatodes vittatus, and unidenti-fied fish have been reported. Erythrolamprus zweifeli feeds on stream frogs of the genus Mannophryne, hylid frogs, Leptodactylus validus, salamanders, lizards of the genus Ameiva, and small birds (Michaud and Dixon 1989;Murphy 1997;Esqueda et al. 2009). While the diets overlap the presence of Bachia, Gonatodes, and Pristimantis urichi in the diet of E. melanotus suggest it is hunting in more terrestrial situations in forests or at forest edges. Mannophryne in the diet of E. zweifeli suggests it is hunting along forested stream-edges. It supports the fact that E. zweifeli was the most common snake encountered during a study in a canal system used for water collection from a mountain stream located in Naiguatá, Venezuela (Silva et al. 1985;Silva and Valdez 1989).
Erythrolamprus epinephalus (Cope, 1862) is widespread and polytypic, ranging from Costa Rica to Ecuador, Colombia, and Venezuela and has not been previously associated with Trinidad or Tobago. The examination of a single specimen (USNM 22069) from Tobago led Dixon (1983b) to conclude that it was Liophis (= Erythrolamprus) reginae with an atypical color pattern that resembled a Liophis (= Erythrolamprus) epinephalus population from eastern Colombia. Dixon's remark was the only mention of E. epinephalus associated with Trinidad and Tobago. He noted the most striking difference in the Tobago animal was a dorsolateral tan stripe not present in the Trinidadian E. zweifeli.
Here, we examine the genetic divergence and morphology of a Tobago snake, previously considered part of the E. reginae group, in an attempt to understand its phylogenetic relationship to other Erythrolamprus and the biogeography in northeastern South America.

Materials and methods
Museum material examined (Appendix 1) included 105 specimens of five Erythrolamprus species. Snakes were examined for external morphological data; scale nomenclature follows Dixon (1983a, b). Museum acronyms are as follows:

AMNH
American Locality data was converted into coordinates using Google Earth. Measurements of the body and tail lengths were taken to the nearest millimeter; ventral scale count methods follow Dowling (1951). The anal plate and terminal scutes were not included in the number of ventrals or subcaudals. The dorsal scale row counts were made about ten ventrals behind the head, at mid-body, and about ten ventrals anterior to the vent. Values for paired head scales are given in left/right order. Scales were measured to the nearest 0.1 mm with the aid of a digital caliper and dissection microscope. Total length (TTL) and tail length (TL) measurements were taken to the nearest mm by carefully stretching the specimens along a ruler or placing a measuring tape along the length of the animal (Appendix 2). Statistical analyses were done with Excel-QI Macros (alpha = 0.05). Ventral and subcaudal counts were compared using ANOVA (Appendix 3).
Seaview v.4.2.11 (Gouy 2010) was used for preliminary alignments of sequences and were aligned thereafter in MAFFT (Katoh et al. 2002), and phylogenetic analyses were conducted using the concatenated mitochondrial and nuclear (12S+16S rDNA+cmos) alignment (with a length of 1332 bp) using a partitioned model of substitution by gene fragment. The most appropriate substitution model for each gene partition was determined by the Bayesian Information Criterion (BIC) in PartitionFinder v.2 (Lanfear 2012). The best-fitting models for the ribosomal and c-mos fragments were as follows: 12S rDNA + 16S rDNA (TRN+I+G), c-mos first and second codon positions (TrNef+I) and c-mos third codon position (HKY). Phylogenetic relationships between taxa were inferred using the Bayesian Inference (BI) optimality criterion under the best fitting substitution model for each gene partition. MrBayes Huelsenbeck et al. (2001) was used with default priors and Markov chain settings, and with random starting trees. Each run consisted of four chains of 30 million generations, sampled every 1,000 generations. Runs were evaluated for convergence and mixing by observing and comparing traces of each parameter in Tracer v.1.6 (http://beast.bio.ed.ac.uk/tracer) (Rambaut et al. 2014). We considered effective sampling size (ESS) values > 200 to be good indicators of parameter mixing. Phylogenetic relationships ( Figure 1) were also estimated using a Maximum Likelihood (ML) approach, as implemented in the software RAxML v7.0.4 (Silvestro and Michalak 2010), under the best partition scheme under the GTR model. All analyses were performed using the CIPRES platform (Miller et al. 2010). P-uncorrected distances were calculated in MEGA V7 (Kumar et al. 2016) under complete deletion of gaps and missing data.

Molecular results
Runs showed high Effective Sample Size convergence (> 2300), indicating adequate sampling of the posterior distribution. The p-uncorrected distances between L. epinephalus from Costa Rica and E. sp. from Tobago were the highest of all terminal monophyletic clades (4.69%) indicating the high genetic divergence between both species (Appendix 5). The phylogenetic relationships of Erythrolamprus and the paraphyly of some species (E. typhlus, E. poecilogyrus, E. epinephalus, E. aesculapii) are similar to past published work (Hodson and Lehtinen 2017), suggesting the need for an in-depth systematic revision of the genus. Furthermore, the results show the paraphyly of Erythrolamprus reginae. Erythrolamprus melanotus from Trinidad and Tobago are monophyletic, and the Trinidad specimen shows no genetic differentiation from the most common Tobago haplotype. Erythrolamprus sp. from Tobago is the sister clade to E. epinephalus from Costa Rica. This clade, composed by the three species (E. melanotus + E. epinephalus+E. sp. Tobago), is strongly supported in the Bayesian analyses. The Trinidadian E. zweifeli form the sister clade to E. reginae from Guyana but are a distinct lineage.  fifth border the orbit, and the temporal formula is 1+2. Figure 3 compares the crowns and chins of four of these species (including E. zweifeli from three different populations). They all share nine plate-like scales on the crown in similar proportions and two pair of enlarged chin-shields. Figure 4 illustrates the distribution of the five species in northern South America, Trinidad, and Tobago.

Morphological results
Comparisons and summaries of the meristic characters for taxa under consideration are given in Table 1. Ventral counts for all Erythrolamprus taxa under consideration have ranges that overlap, although they have different means, some of which are significantly different. The ranges for the subcaudal counts are similar. The Tobago E. pseudoreginae sp. n. can be separated from E. melanotus but not the other taxa. The results of single factor ANOVAs are presented in Appendix 2. Some support the separation of E. zweifeli from E. reginae, E. zweifeli from the Tobago E. pseudoreginae sp. n., and E. reginae from the Tobago E. pseudoreginae sp. n. Substantial genetic differences (0.047) (Appendix 5) and relatively minor morphological differences (different means for ventral counts, distinctive coloration, the absence of apical pits on dorsal scales) and its geographic isolation support the description of the Tobago population as a new species. Figure 5 compares the color morphs of Erythrolamprus zweifeli found in Trinidad and Venezuela with Erythrolamprus reginae from Guyana. Figure 6 illustrates E. pseudoreginae sp. n.  Diagnosis. Ventrals 143-154; subcaudals 76-79; second pair of chin shields longest; some anterior dorsal scales have an apical pit; lateral stripe on scale rows 3-4-5, dark stripe (row 3) and a pale stripe (rows 4-5) on posterior body and tail, the black stripe continues to the forebody as a series of black spots on scale row three; and the ventral surface has scattered flecks of pigment toward mid-body. Otherwise, the belly is uniform cream with fine speckling in preserved material, and red in life, tail uniform cream in preservative, red in life.
In life the crown is dark moss green with black spots, the upper labials are cream, with a dark stripe on the upper edge that runs from nasal to orbit, and widens posteriorly onto the temporals. Dorsal spots on scale rows 2-3 about two ventrals apart, start above the 12 th ventral, and coalesce into a stripe at about the 96 th ventral and extend posteriorly to the tip of the tail; lateral stripe mostly on scale row three on body, goes onto scale row one on tail. About one-third down the body, about ventral 40, scale rows 1−4 blue-gray, row five is brown, row six and above blue-gray; except for the mossy green on the anteriormost dorsal surface for about 40 ventrals. Ventral surface mostly uniform yellow to orange with light mottling starting about the 50 th ventral; tail has a mid-line zigzag stripe.
Variation: The smallest specimen measured 347 mm SVL with a 129 mm tail; the largest specimen 420 mm SVL with a 119 mm damaged tail. Dorsal scale rows 17-17-15. Ventrals range from 143-154 (n = 5, X = 147.5, SD = 3.35); subcaudals 76-79 (n = 2, X = 77.5, SD = 1.5). Upper labials eight or nine, 2-3 contact loreal, 4-5 border the orbit (one specimen has 5-6 bordering the orbit on one side), the tallest can be seventh, (or eight if nine labials are present); the sixth labial is the largest in the area. Loreal is quadrangular to pentagonal. Lower labials 9-10; first four or five contact the anterior chin shields. Longest pair of chin shields is the second. Eye diameter is greater than eye-nostril distance. The dark posterior lateral stripe is usually on scale rows 2-3-4, but one specimen has it on scale rows 2-3 only.
Color in life. The following is based on the holotype ( Figure 6) and a color photograph in Brown (2013). Crown and face olive brown, upper labials white, a short black subocular stripe extends from the nasal scale under the eye and posteriorly to the last labial. Immediately behind the head, the interstitial skin is yellow; dorsum brown with an indistinct vertebral stripe and scales partially edged with black pigment most obvious on posterior two-thirds of the body. First three scale rows are blue-gray and separated from brown dorsum by a row of black spots.
Color in alcohol. Head, body, and tail dark blue to brown with a black stripe on the posterior lateral body that becomes a series of dark spots extending anteriorly on the body. The belly is a uniform cream with fine speckling of pigment.
Comparisons. Erythrolamprus pseudoreginae sp. n. differs from E. zweifeli in the presence of apical pits on some dorsal scales, an almost uniform yellow to red venter, and a dark stripe on the posterior body on scale rows 3-4 bordered above by a pale stripe on scale row five. The new species lacks the well-defined postocular stripe that runs from the postocular scales across the temporals to a point just above the rictus in most E. zweifeli. In E. zweifeli the postocular stripe may also have a pale dorsal border.
Erythrolamprus pseudoreginae sp. n. differs from all populations of E. epinephalus in having more than 75 subcaudal scales, except for some Venezuelan and Colombian populations. The E. epinephalus populations with more than 75 subcaudals have a dorsal or ventral pattern that includes transverse bars, black checks, or a pattern with irregular black spots on the outer edges of the ventral scales that may extend onto the first row of dorsal scales (Dixon 1983a, Escalona 2017. The new species differs from Erythrolamprus reginae in having a uniform venter (E. reginae) has yellow to orange venter with black checks, and a dark stripe on the last fourth of the body on scale rows 3-4 which is not bordered by a pale stripe. Erythrolamprus pseudoreginae sp. n. has uniform yellow to red ventral surface and a very distinctive, pale posterior lateral stripe on row five above the black stripe on rows 3-4 that extends anteriorly as a row of dark spots. Erythrolamprus reginae has fewer ventrals and a lower mean ventral count than E. pseudoreginae sp. n.
The pattern will readily distinguish it from the two coral snake mimics (Erythrolamprus aesculapii and E. bizona) which are on Trinidad but not Tobago. The endemic Tobago Red Snake, E. ocellatus, has a bright red dorsum with black ocelli. The semi-aquatic Erythrolamprus cobellus has a uniform dark green or black dorsum and is known from Trinidad but not Tobago. The absence of a black stripe five scale rows wide on the vertebral line separates it from Shaw's Black Back Snake, E. melanotus, a species known from both islands.
Distribution. It occurs in northeastern Tobago and appears to be restricted to the forested ravines along the crest of the Main Ridge (Fig. 4). Tobago's Main Ridge is about 16 km long and covered with lower montane rain forest on schist soil above 224 m ASL. The ridge crest reaches elevations of 487-576 m ASL and forms steep terrain with deep gullies and fast-moving streams. The area receives about 318 cm of rainfall per year, and no month receives less than 10 cm (Beard, 1944). Tobago's Main Ridge Forest Reserve is the oldest protected forest in the Western Hemisphere (since April 1776) and encompasses 3958 hectares. At this writing, five specimens of Erythrolamprus pseudoreginae sp. n. are known, all of which came from the northeast end of the Main Ridge. The locality and elevation data available suggest it occurs within an area of about 400 ha at elevations between 430-500 m ASL. Three types of rainforest occur on Tobago: lowland rainforest covers 4,844 ha, lower montane rainforest covers 4,566 ha, and xerophytic rainforest covers 937 ha (Helmer et al. 2012). All of the localities for E. pseudoreginae sp. n. fall within the lower montane rainforest, suggesting its maximum area of occupancy may be 4,566 ha, if it is restricted to that forest type.
Natural History. Erythrolamprus pseudoreginae sp. n. is diurnal, all of the specimens with time of collection data were found in the morning or afternoon. Nothing is known about the diet and reproduction of this snake. Its close relatives have been reported to eat anurans, and it likely preys upon small ground-dwelling frogs. A The distribution of the species of Erythrolamprus under discussion in northern Venezuela and Trinidad and Tobago B More detailed view of the distribution on Trinidad and Tobago C Tobago with the known localities for E. pseudoreginae sp. n. Note that two of the markers closely overlap. Key: black stars = E. zweifeli from Cordillera de Costa in Venezuela and the island of Trinidad; green circles = E. epinephalus from the Cordillera de Mérida, Venezuela. Note that these markers denote the closest population to Tobago based on Roze (1966). Specimens examined came from several different locations. Purple stars = E. reginae from the Guianas including Orinoco Delta in Venezuela; red stars = E. melanotus from Venezuela, Trinidad, and Tobago; blue star = Erythrolamprus pseudoreginae sp. n. on Tobago.

Conservation.
Given the restricted distribution of this snake as well as the fact that most, if not all, of its distribution lies within the oldest protected forests in the Western Hemisphere it may be assumed that it is well protected. However, as the climate changes the microclimate found in the lowland montane rainforest may be expected to change and potentially make the local environment inhospitable for this species and the other endemic taxa found here.
Etymology. The epithet pseudoreginae was chosen because prior investigators considered this snake to be Liophis reginae. We suggest Tobago Stream Snake as the common English name for this snake.

Discussion
Erythrolamprus pseudoreginae becomes the fifty-first species in the genus, and the eleventh member of the Tobago herpetofauna closely associated with the Main Ridge. The list of Main Ridge species includes the frogs Mannophryne olmonae, Pristimantis charlottevillensis, P. turpinorum, Hyalinobatrachium orientale; the lizards Bachia cf. flavescens, Gonatodes ocellatus, Anolis cf. tigrinus; and the snakes Atractus fuliginosus, Erythrolamprus ocellatus, and Leptophis haileyi.
Most of the Main Ridge endemic species seem to have their closest living relatives in the Costal Ranges of Venezuela as opposed to the more proximal island of Trinidad or the Guiana Shield. The Coastal Range endemic Mannophryne riveroi is the sister to M. olmonae (Manzanilla et al. 2009, Lehtinen et al. 2011). Tobago's Pristimantis charlottevillensis appears to be most closely related to P. terraebolivaris and members of the Pristimantis conspicillatus group (Hedges et al. 2008). Jowers et al. (2015) proposed a Pliocene land bridge connection between Tobago and Venezuela to explain the presence of Hyalinobatrachium orientale on Tobago and northeast Venezuela. An alternative explanation is that Tobago was accreted to Venezuela on its movement to its current position.
With this study, only 21 of the 51 named Erythrolamprus species have been included in molecular studies; thus, the tree contains only 41% of the known species in the genus. Therefore, its topography is likely to change with additional taxa from more locations. Erythrolamprus reginae and E. epinephalus are polytypic and given their distributions and morphological variation they represent a considerable challenge to resolving the lineages found within these taxa. Some of the color patterns have evolved multiple times in the different lineages and when combined with the conserved morphology, separating these taxa by morphology becomes a conundrum. It seems likely that some of the currently recognized subspecies will be found more closely related to lineages other than the one they are currently assigned.
The phylogenetic analyses suggest part of E. reginae is the sister to E. zweifeli. The results show for the first time the Trinidadian E. melanotus, has no genetic divergence from the most common haplotype from Tobago. This lack of differentiation suggests some recent or ongoing gene flow between islands. The position of E. ocellatus from Tobago suggests that it forms a highly divergent lineage to the remaining Tobago´s

Appendix 1
Locality data for specimens examined in this study. Coordinates represent georeferencing from Google Earth, variation from the exact collecting locality is expected.