Unveiling an enigma from the Cerrado: taxonomic revision of two sympatric species of Apostolepis Cope, 1862 (Dipsadidae: Xenodontinae: Elapomorphini) from central Brazil

. Apostolepis albicollaris and A. cerradoensis are two Elapomorphini snake species, described within a short timespan, from the Cerrado of central Brazil. In their brief descriptions, these two species were diagnosed from congeners largely based on highly variable external morphological characters. Interestingly enough, A. cerradoensis has remained known based on a single specimen since its description. Here, we present a reanalysis of both type specimens, as well as a careful examination of a large series of specimens formerly assigned to these species, based on the comparison of morphological variation, geographic distribution, hemipenial morphology, phylogenetic relationships and an osteological description are also provided. We also discuss its conservation status, suggesting that the species is under threat and qualifies to be listed as Vulnerable (VU ab(iii)), considering its rarity, small geographic range, and persistent environmental threats. & Unveiling an enigma from the Cerrado: taxonomic revision of two sympatric species of Apostolepis Cope, 1862 (Dipsadidae: Xenodontinae: Elapomorphini) from central Brazil. European Journal of Taxonomy 817: 143–182.

Of these, Apostolepis is the most speciose genus, with 35 species that range from central Colombia and the Guyana Shield, southwards into northern Argentina and southern Brazil (Entiauspe-Neto et al. 2020a, 2020c, 2021b. A considerable amount of this diversity can be found in Brazil, which harbors approximately 30 species, of which most are associated with the South American Dry Diagonal and other open landscapes (Rodrigues 1992;Lema 2003;Ferrarezzi et al. 2005;Curcio et al. 2011;Guedes et al. 2014a;Entiauspe-Neto et al. 2021b).
The taxonomy of Apostolepis has been historically unstable and confusing; several species of Apostolepis are known from less than 10 specimens, being poorly represented in scientific collections and seldom found in the field, which has likely contributed to its instability Nogueira et al. 2012;Entiauspe-Neto et al. 2019). During the last decades, a considerable number of species of Apostolepis were described based on small series and subjective morphological diagnostic characters (e.g., Lema , 2003Lema , 2004aLema , 2004bLema & Renner 2011), which were later found to be synonymous with other congeners, as these diagnoses relied largely upon polymorphic characters (Ferrarezzi et al. 2005;Entiauspe-Neto et al. 2019, 2020a, 2020b, 2021a. Lema (1993), in his work on the morphological variation of Apostolepis dimidiata Jan, 1863, proposed the synonymization of A. barrioi Lema, 1978 (later revalidated by the same author in Cabral et al. (2017)), A. ventrimaculata Lema, 1978, andA. villaricae Lema, 1978 with the former species, while also reporting on the extremes of morphological variation for the analyzed specimens. Some of the specimens reported by Lema (1993) that bore white nuchal collars were then reanalyzed by  and briefly described as Apostolepis albicollaris , based solely on external morphology, with a holotype (MCP 8355) from "Parque Zoológico Distrito Federal", Brasília, Distrito Federal (mistakenly reported as the state of "Goiás" by the author), Brazil. Although only four specimens were part of the type series of A. albicollaris, its description reported 27 additional specimens, from Distrito Federal, Goiás, and Minas Gerais, all from the Cerrado of Central Brazil.
Shortly after, Lema (2003) described another species of Apostolepis from the Central Brazil, Apostolepis cerradoensis Lema, 2003, known from a single specimen (reported by the author as UHESM 21800, now MCP 15219) from "Usina Hidrelétrica Serra da Mesa", Minaçu, in the Cerrado of Goiás, Brazil. Oddly enough, this species was compared exclusively to A. dimidiata, and also only briefly described based on external morphology. Until now, A. cerradoensis remains known only from its holotype (Nogueira et al. 2019). Nogueira et al. (2012) provided a redescription for A. albicollaris, with an account of its morphological variation, hemipenial morphology, and a key to some species of Apostolepis and speciesgroups; it should be noted that these authors did not analyze the type specimen of A. cerradoensis, and that the species is not mentioned in their work.
In this study, we review the taxonomic status of A. albicollaris and A. cerradoensis, reanalyzing and comparing both type specimens. Based on newly generated data from the types and additional Apostolepis specimens from the Cerrado of Central Brazil, we consider these species as synonymous, A. cerradoensis beign a junior synonym of A. albicollaris. We also provide an account of its updated diagnosis, morphological variation, geographic distribution, hemipenial morphology, phylogenetic relationships and an osteological description. We also discuss its conservation status, suggesting that the species is under threat and qualifies to be listed as Vulnerable (VU ab(iii)), considering its rarity, small geographic range, and persistent environmental threats.

Material and methods
We examined a total of 715 specimens of Apostolepis (Appendix 1, acronyms according to Sabaj (2019)), of which 437 belong to 19 species that range in the South American Dry Diagonal. Scale counts follow Dowling (1951) and Peters (1964), while sex determination was done with a ventral incision in the base of the tail. The emended diagnosis is based on the nomenclature used by Entiauspe-Neto et al. (2020a). We measured head length (from the center of the rostral to the corner of the mouth) and head width (at the corner of the mouth) to the nearest 0.01 mm using a dial caliper; snout-vent length (SVL, from the center of the rostral to the posterior margin of the cloacal scale) and tail length (TL, from the posterior margin of the cloacal scale to the distal tip of the terminal scale) to the nearest 1 mm using a flexible ruler. We measured scales on the right side of the head and defined our measurement within the description when appropriate. Data are summarized in Supplementary Table 1. We used the verified point record database of A. albicollaris (Supplementary Table 2) to map its distribution. We calculate its Extent of Occurrence (EOO) and Area of Occupancy (AOO; based on 2 km cell width) by using the Geospatial Conservation Assessment Tool (GeoCAT; Bachman et al. 2011). To better address the conservation status of the species, we included layers of land use and land cover changes ( We used the environment R ver. 3.2.3 (R Core Team 2015) software for all statistical tests. We employed an Analysis of Variance (ANOVA) in order to assess group differentiation in meristic (ventrals, subcaudals) and morphometric (SVL, TL) characters for the male samples of A. albicollaris (n = 11) and A. cerradoensis (n = 1), while these two samples were compared for their pairwise mean values with Tukey's Range Test (Zar 1999). These comparative analyses could not be extended to females, as the single known specimen of A. cerradoensis is a male. We also evaluated the presence or absence of sexual dimorphism for the whole sample series with a T-test, after we evaluated the assumptions of univariate normality by using a Shapiro Wilk test, and homoscedasticity through Levene's test (Zar 1999). When describing sexual dimorphism, we report ranges followed in parentheses by mean ± 1 standard deviation and sample size. For statistics, we provide samples of each group as subscripts.
We describe the skull morphology of the A. albicollaris (MCP 8355) and A. cerradoensis (MCP 15219) holotypes, based on high-resolution micro-CT scans, performed with a Bruker SkyScan 1173 at the MCP. The scans were recorded over a 180° rotation with a frame averaging of 2, an X-ray beam with 60 kV source voltage, 133 μA current, and an exposure time of 800 ms per projection, without the use of a filter. For the scan of A. albicollaris a rotation step of 0.2° degrees was used, resulting in 1200 projections, an isotropic voxel size of 7.08 μm, and a total scan duration of 49:17 min. For A. cerradoensis a rotation step of 0.23° degrees was used, resulting in 1043 projections, an isotropic voxel size of 8.51 μm, and a total scan duration of 55:47 min. We reconstructed the CT-datasets using N-Recon software ver. 1.7.1.6 (Bruker MicroCT) and rendered the images in three dimensions through the aid of Amira visualization software (FEI, Thermo Fisher Scientific). Segmentation to separate and color the bones was also performed using Amira. For the description of the skulls, we use the osteological terminology of Bullock & Tanner (1966) and Cundall & Irish (2008), and the description of the skulls follows Entiauspe-Neto et al. (2020a, 2021b. We compared the skulls of A. albicollaris and A. cerradoensis to descriptions of the skulls of A. ambiniger, A. assimilis, A. cearensis Gomes, 1915, andA. sanctaeritae (Ferrarezzi et al. 2005;Entiauspe-Neto et al. 2020a, 2021a, 2021b. In order to evaluate genetic distances among species of Apostolepis, we selected available sequences at GenBank (NCBI), of small unit ribosomal RNA (12S) gene sequences (JQ598793-JQ598797, GQ457781, GQ457782) and oocyte maturation factor (CMOS) gene fragment sequences (JQ598965-JQ598969, GQ457843, GQ457844). Each gene fragment was aligned separately with the MUSCLE Alignment plugin (Edgar 2004), in Geneious Pro (Drummond et al. 2010) software, with default settings ( Supplementary Files 1-4). For the protein-coding gene CMOS, we translated the alignment into proteins to check for sequence integrity and stop-codons. Uncorrected genetic distances (p-distances) were calculated using MEGA 11 (Tamura et al. 2011), for each gene separately, using a d parameter (Transitions + Transversions), while assuming uniform rates among sites and homogeneous pattern among lineages. The p-distance calculation was made based on the proportional (p) differences among nucleotide sites in which two compared sequences differ, as inferred through the division of nucleotide differences by the total number of nucleotides (Tamura et al. 2011).
In order to evaluate the phylogenetic relationships among Apostolepis taxa and verify phylogenetic information contained in the available sequences, in addition to the aforementioned gene fragments, we pooled available Caenophidian sequences of mtDNA [12S, large unit ribosomal RNA (16S), cytochrome b (CYTB), NADH dehydrogenase subunit 2 (ND2), and NADH dehydrogenase subunit 4 (ND4)] and nuDNA [brain-derived neurotrophic factor (BDNF), CMOS, neurotrophin-3 (NT3), and recombination activating gene 1 (RAG1)] gene fragments into a combined dataset (Supplementary Table 3). Each gene fragment was aligned separately with the MUSCLE Alignment plugin (Edgar 2004) in Geneious Pro ver. 7.1 (Drummond et al. 2010) software, with default settings, and then incorporated into a concatenated alignment (Supplementary File 5). In order to evaluate topological congruence, we employed inferences of Maximum Likelihood (ML) and Bayesian Inference (BI). In order to avoid topological effects of model-induced redundancy or overparameterization biases associated with the selection of substitution models, we employed separately two distinct models, namely the Hasegawa, Kishino, and Yano 1985 (HKY85, hereafter) model (Hasegawa et al. 1985), which only accounts for the rates of transitions and transversions and unequal base frequencies (πA ≠ πG ≠ πC ≠ πT), and the General Time-Reversible model (Tavaré 1986;GTR, hereafter), which accounts for six substitution rate parameters and four equilibrium base frequency parameters, each applied with BI and ML analyses, respectively. A phylogenetic inference of ML was employed with RAxML ver. 7.2.8 (Stamatakis 2014), using the GTR+Γ model, and a Rapid Bootstrapping with search for best-scoring ML tree (-f a), using 100 bootstrap replicates (Supplementary File 6). A phylogenetic inference of BI was also employed with MrBayes ver. 3.2.6 (Huelsenbeck & Ronquist 2001), using an HKY85 substitution model and Gamma rate variation parameter, with 1 100 000 Markov-Chain Monte Carlo chains, with a subsampling frequence of 200, a Burn-In Length of 100 000 and four heated chains; priors were set to unconstrained branch lengths, with a compound Dirichlet prior on branch lengths of (1, 01, 1, 1) and an exponential shape parameter (10). Resulting topologies of the Bayesian inference analysis were then analyzed for their consensus tree on TreeAnnotator, with a consensus support threshold arbitrarily defined ( ≥ 80 %) and pruned to contain Echinantherini Zaher, Grazziotin, Cadle, Murphy, Moura-Leite & Bonatto, 2009, Philodryadini Cope, 1886, Tachymenini Bailey, 1967, and Elapomorphini outgroups only (Supplementary File 7). The ML topology was then used to infer branch lengths and patristic distances (absolute time and mutation rate, to which patristic distances represent the sum of branch lengths used to link two species terminal nodes in a tree) as reinforcement proxies of genetic distance (Supplementary Table 4), following Montingelli et al. (2020) and Entiauspe-Neto et al. (2021c), inferred with the package Ape for R (R Core Team 2015; Paradis & Schliep 2019). For patristic distances, paraphyletic terminals were treated separately.

Results
Upon reanalysis of both type specimens, it became evident that the type of A. cerradoensis has a different voucher number than the one reported in its description; although reported as UHESM 21800 in Lema (2003: 125), it is here corrected to MCP 15219. The previous acronym seems to be a field number that actually refers to "Usina Hidrelétrica de Serra da Mesa", which refers to its type locality, rather than a scientific collection. According to Lema (2003: 126), A. albicollaris could be distinguished from A. cerradoensis (characters in parentheses) based on its black ventral coloration (white), fully black tail (white), and wider black lateral stripes (smaller). These coloration characters were previously shown to be highly variable in other Elapomorphini species (e.g., Entiauspe-Neto et al. 2019, 2020a, 2021a, 2021b. Another congener from the Cerrado, Apostolepis dimidiata Jan, 1862, has a remarkably wide variation in ventral and lateral stripe coloration, which ranges from uniformly cream venter and small lateral black stripes, to a fully black venter and dorsolateral area (Entiauspe-Neto et al. 2019, 2021a. Furthermore, no other specimens aside from the type are known for A. cerradoensis. During our analyses of specimens formerly referred to as A. albicollaris or A. cerradoensis, we observed a near total overlap of morphological characters that are reportedly diagnostic for congeners (ventrals, subcaudals, supralabials, infralabials, SVL, and TL), as seen in Table 1. For our ANOVA analyses, we encountered no statistical differences for SVL (F 11,1 = 0.184, p < 0.7, n = 12), as well as TL (F 11,1 = 0.115, p > 0.7, n = 12), and subcaudal (F 11,1 = 0.331, p < 0.5, n = 12) values. However, we did encounter differences among the ventrals (F 11,1 = 5.924, p < 0.03, n = 12) of A. albicollaris and A. cerradoensis. Tukey's Range Test also recovered no differences for SVL (p < 0.6), TL (p < 0.7), of subcaudal (p < 0.5) values, encountering differences in ventral (p < 0.05) values. These differences in ventral scales can likely be attributed to sample size limitations, as there were only 11 males of A. albicollaris examined, and a single known specimen of A. cerradoensis.
Both species are also mentioned as sympatric (Lema 2003: 126), with several specimens of A. albicollaris being collected in Minaçu (type locality of A. cerradoensis), which further blurs any geographic distinction between both taxa. Considering the lack of diagnostic characters and large overlap of morphology, we conclude that both taxa are conspecific, considering A. cerradoensis as a junior synonym of A. albicollaris. We provide an updated account on the variation for this species below, as well as an updated emended diagnosis.

Description
Head slightly distinct from body, twice as long as wide, triangular in dorsal view, arched in lateral view (Fig. 3); cervical constriction absent; snout conical, tapered in dorsal view, projected in lateral view; nasal  undivided, longer than wider, in contact with preocular (rarely separated), rostral, first two supralabials and prefrontals; prefrontals paired, longer than wider, in contact with rostral, nasal, supraocular, and frontal; supraocular longer than wider; frontal hexagonal, longer than wider; internasals absent (rarely present); loreal absent (rarely present); preocular present, longer than high; eye small; interorbital distance as long as parietal distance; postocular single, as long as high; temporals absent; occipital present, in contact with parietal and sixth supralabial; supralabials six (rarely four, five, seven), first in contact with nasal and rostral, 2-3 in contact with orbit, 3-4 in contact with postocular, 5-6 in contact with parietal; infralabials six or seven (rarely eight), first pair in contact with symphisal and anterior chinshields, second, third, and fourth pairs in contact with anterior chinshields, third, fourth, fifth, sixth, and seventh pairs in contact with posterior chinshields; dorsal rows 15/15/15; short tail, terminal spine rounded.

Coloration in life
Live specimens have dorsal portion of head black, with snout blotch absent or thin and divided, white or cream colored, covering rostral and prefrontals; white nuchal collar present, 3-4 scales wide; black nape collar present, covering 1.5-3 scales, rarely absent; lateral head portion black, white supralabial blotch elongated, covering three to five scales, infralabials white; ventral portion of head white, with a complete gular black collar; background dorsal coloration red or orange, vertebral stripe vestigial or absent, two lateral black stripes ( Fig. 4A-F); ventral coloration uniformly black, bicolor, or white; caudal blotch 5-9 scales wide on dorsum, incomplete (rarely complete) on venter, terminal spine black and white.

Coloration in preservative
Orange and red colorations turn gray or white in preservative (Fig. 1).

Comparisons
Apostolepis albicollaris might be misidentified as another congener from the Cerrado, Apostolepis dimidiata ( Fig. 4G-H). Both species share an uniformly red dorsal background coloration, with two lateral black stripes, a white or black venter, and have overlapping scale counts of supralabials, infralabials, and subcaudals. However, A. albicollaris can be distinguished from A. dimidiata (characters in parentheses) by having nuchal collars (absent), a small-sized semicapitate hemipenis (medium-sized capitate hemipenis), and lower ventral scale counts (220-244 total range, 226-240 in males, 220-244 in females). It should also be noted that A. dimidiata is not sympatric with A. albicollaris in most of its range, with the former species occurring in the Cerrado of southeastern and central-western Brazil, in the states of São Paulo, Paraná, Minas Gerais, and Mato Grosso do Sul, and in Paraguay, while the latter is restricted to Goiás, Distrito Federal, and Minas Gerais, in central Brazil; both species can be found in sympatry in a small stretch of Cerrado in western Minas Gerais, in the locality of Uberlândia.
Three other similar congeners occur in sympatry with A. albicollaris: Apostolepis assimilis (Reinhardt, 1861), Apostolepis flavotorquata (Duméril, Bibrón & Duméril, 1854), and Apostolepis sanctaeritae Werner, 1924. Apostolepis assimilis can be diagnosed from A. albicollaris based on its dorsal pattern with two black stripes (uniformly red), a divided snout blotch (entire), elongated supralabial blotch, up to five scales wide (smaller, up to three scales wide). Apostolepis flavotorquata can be diagnosed from A. albicollaris based on its distinct dorsal pattern (uniformly red or with three black stripes), ventral coloration in life (yellow), and hemipenial morphology (organ long and simple, spinules restricted to proximal area, enlarged spines absent, asulcate surface calyculate) (Lema & Renner 2005). Apostolepis sanctaeritae can readily be diagnosed from A. albicollaris based on its single nuchal collar (additional white nuchal collar), smaller black nuchal collar, covering 1.5-3 scales (4-7 scales), elongated white supralabial blotch (small, tear-shaped), and a divided, white snout blotch in life (single, orange). An extralimital species, Apostolepis quirogai , bears striking similarity to A. albicollaris ( Fig. 4I-J). Both species share a uniformly red dorsal background coloration, with two lateral black stripes, wide white and black nuchal collars, and black ventral coloration. However, A. albicollaris can be distinguished from this species based on its lower ventral scale counts (269 in the single male, 276 in the single female), a divided snout blotch (entire), and a larger, elongated, supralabial blotch (smaller, covering up to three scales). Furthermore, A. quirogai occurs in Misiones, in northwestern Argentina, and northern Rio Grande do Sul, southern Brazil, in Alto Paraná Atlantic Rainforests, a very distinct environment from the Cerrado of Central Brazil in which A. albicollaris occurs.
From Apostolepis multicincta A. dorbignyi (Schlegel, 1837), two extralimital species from the Argentine and Bolivian Chaco open areas, A. albicollaris can be distinguished based on its dorsal pattern with two lateral black stripes (uniformly red coloration in life), divided snout blotch (entire), and bicolored tail tip (uniformly white).

Hemipenis
Hemipenis slightly bilobed, semicapitate, semicalyculate (Fig. 5). Lobes slightly differentiated, present on distal portion of an incomplete capitular ring; lobes rounded, present on distal portion next to incomplete capitulum; lobes and incomplete capitulum covered by papillate calyces on sulcate and lateral sides; incomplete capitulum inconspicuous on both asulcate and sulcate sides, slightly smaller than hemipenial body, and positioned above bifurcation of sulcus spermaticus. Bifurcation of sulcus spermaticus in middle of organ. Capitular groove slightly distinct on sulcate side. Sulcus spermaticus branching centrolineally, with wide invagination on apex and incomplete capitulum; margins of sulcus spermaticus thin and inconspicuous, bordered by expanded lips. Hemipenial body moderately long for Apostolepis, subcylindrical; on sulcate surface, hemipenial body covered by small to moderate-sized hooked spines; on asulcate surface, hemipenial body with large basal naked pocket, extended to medial and apical regions, covered by enlarged spines, with inconspicuous lamellae on its asulcate side apex; 1-5 rows of enlarged spines present on lateral portion of sulcate side and medial and apical region of asulcate side.

Geographic distribution and natural history
Known from the states of Goiás, Distrito Federal, and Minas Gerais, central Brazil, in the Cerrado biome (Fig. 7A). The locality reported by Nogueira et al. (2019) in Uruaçu, at Poções municipality, Bahia state, is considered here as erroneous. The specimen on which this record is based (IBSP 26712) is reanalyzed herein, and redetermined as being from Cana Brava, on the outskirts of Minaçu municipality, Goiás. Apostolepis albicollaris occurs in open areas, with dense shrub vegetation. Nogueira et al. (2012) report predation on legless lizards, twilight activity, and reproduction data.

Conservation status
It is known from 37 records representing 17 localities (unique georeferenced coordinates) (Fig. 7A). Its extent of occurrence was calculated to be 62 434 517 km² and its area of occupancy was calculated to be 64 000 km² (B2). The species occurs in less than 10 locations (a). It is under strong pressure by the continued decline in habitat quality by extensive and ongoing habitat loss and fragmentation (b(iii)) due to expansion of agriculture, pasturing, mining, and fire frequency (Fig. 7BC). Additionally, only a small part of the distribution of the species is located in just one Protected Area (Parque Nacional de Brasília, Fig. 7D). We suggest the species is under threat and qualifies to be listed as Vulnerable (VU ab(iii)).

Molecular analyses and systematics
Interspecific uncorrected p-distances range from 1 to 7% for 12S (Table 2) and 0 to 1% for CMOS (Table 3). Apostolepis albicollaris has a remarkably small genetic difference regarding its putative sister-species, Apostolepis dimidiata, having p-distances of 1% for 12S and 0% for CMOS. Further studies with larger samples and more molecular markers are needed in order to evaluate the presence or absence of gene flux and monophyletic reciprocity among species. Notably, the small distances recovered for the CMOS gene Fig. 6. Apostolepis albicollaris , sexual dimorphism in morphometric (SVL, TL) and meristic (ventrals, subcaudals) characters. Outliers are indicated as circles, A. cerradoensis Lema, 2003 holotype indicated as red star.
indicate that this single-copy, intronless, slow evolving nuclear marker may contain little phylogenetic information for this group, although proven useful for higher level relationships in squamates (Saint et al. 1998). Patristic distances among A. albicollaris and A. dimidiata were recovered as 0.055, and for other species of Apostolepis ranged from 0.053 to 0.090 (for A. flavotorquata and A. sanctaeritae, respectively). The mean interspecific patristic distance for Apostolepis is recovered as 0.061, and for the other Elapomorphini genus, Phalotris, as 0.116. The results can be seen in Supplementary Table 4. The phylogenetic trees inferred under maximum likelihood and Bayesian inference recovered as monophyletic the sampled Elapomorphini terminals (Apostolepis, Elapomorphus, Phalotris), with a    final ML optimized likelihood of -InL = -12757.360116. All Elapomorphini terminals are recovered as a sister group to Echinantherini genera. Both trees largely agree with each other regarding internal nodes within our clade of interest (Elapomorphini), and are depicted with a maximum likelihood topology and node values for both bootstrap (maximum likelihood) and consensus support (Bayesian inference) (Fig. 8). A clade composed of A. albicollaris and A. dimidiata is recovered as strongly supported, sister to A. flavotorquata, another red congener from the Cerrado. The Apostolepis assimilis species group (sensu Ferrarezzi 1993;Ferrarezzi et al. 2005), composed of A. assimilis, A. cearensis, and A. sanctaeritae, is also recovered as monophyletic. Regarding the intrageneric and intergeneric relationships of Apostolepis terminals, our topology agrees with the one presented by Zaher et al. (2019).
Premaxilla single, thick, robust, and edentulous, 1.5 (MCP 15219) to 1.6 (MCP 8355) times as broad as high; posterodorsally oriented ascending process has two broad lateral wings and a triangular dorsal apex, contacting anterior end of nasals; transverse processes are posterolaterally oriented, define widest part of premaxilla and are partly visible (MCP 8355) or not (MCP 15219) in dorsal view, long, approaching but not contacting maxillae; posteriorly oriented vomerine processes are robust thumb-like and almost parallel (MCP 15219) or slightly laterally diverging (MCP 8355), with a blunt tip, dorsally contacting anteroventral part of septomaxilla, distinctly separated from anterior end of vomers; ventral surface of premaxilla pierced by a slit-like foramen in anterior region and may have further perforations or notches (MCP 15219). Nasals paired, more or less rectangular in dorsal view, 1.8 (MCP 15219) to 2.2 (MCP 8355) times as long as broad, convex, in medial contact along a straight suture, with front edges forming a broad V-shaped anterior region, together framing ascending process of premaxilla; lateral edges slightly curved downwards; posteriorly each nasal shares a loose, slightly oblique suture with corresponding frontal, although right nasal of MCP 15219 has a deep, almost circular indentation posteromedially,
Frontals paired, about semicircular in dorsal view, about 1.8 (MCP 15219) to 2 (MCP 8355) times as long as wide, slightly convex, in medial contact with a straight suture; only a short part of lateral margin of each frontal, at region of greatest width of frontal, participates in formation of dorsal margin of orbit; anterior edge slightly oblique, forming a loose suture with posterior edge of nasal; anterolateral edge forming an oblique suture with prefrontal; posterior edge curved and framed by anterolateral processes of parietal; anteroventrally frontal contacts nasal and directly lateral to this contact zone has a bulge whose anterior surface contacts posterior end of posteromedial process of septomaxilla; laterally each frontal has a ventromedially oriented concave lamina, that contacts its counterpart medially along anterior third of its length; anteromedially, frontal exhibits a vertical lamina fused ventrally with lateral lamina to form a short tubular structure in anterior region of each frontal, with vertical laminae of both frontals in firm medial contact; posterior to tubular structure, lateral lamina of both frontals clearly separated, with their medial margins approximately parallel to each other; on ventral side of contact zone of vertical lamina of frontal a facet into which anterior part of dorsal process of parabasisphenoid penetrates, behind it dorsal process of parasphenoid protrudes into gap between both lateral laminae and contacts them with its lateral edges.
Parietal single, elongate, 1.4 (MCP 15219) to 1.8 (MCP 8355) times as long as wide, roughly rectangular in dorsal view except for anterolateral processes; dorsal surface slightly convex, except for a slight depression longitudinally along its midline; anterolateral processes moderately long and robust, framing Postorbitals absent.
Supraoccipital single, pentagonal, with straight anterior border and triangular posterior border, about twice as wide as long, slightly concave, and with anterior region slightly elevated; anteriorly it contacts parietal along a straight transverse suture, anterolaterally prootics with an slightly (MCP 15219) to strongly (MCP 8355) oblique suture, and posteriorly exoccipitals with an oblique suture; MCP 15219 has a short elongated notch anterodorsally right and left of midline, respectively, distinctly separated from supratemporals; internally, lateral borders of supraoccipital extend downwards to contribute to dorsomedial walls of otic capsules.
Prootics paired, ovaloid in lateral view, 1.2 (MCP 15219) to 1.45 (MCP 8355) times as high as long; each prootic contacts parietal anteriorly, supraoccipital dorsally, exoccipital posteriorly, parabasisphenoid complex anteroventrally, basioccipital posteroventrally, and anterior part of supratemporal on its dorsal surface; at suture with parietal, prootic forms posterior border of foramen for maxillary branch of trigeminal nerve and at suture with exoccipital it forms anterior margin of fenestra ovalis; slightly posterior and slightly below central part of prootic in lateral view a posterolaterally oriented foramen for mandibular branch of trigeminal nerve; two further foramina on lateral surface of prootic situated in between and ventral to both trigeminal nerve foramina; in dorsal view, prootic bears a depression in posterior region in which rests anterior tip of supratemporal; internally, prootic contributes to formation of anteroventral, anteromedial and anterolateral wall of otic capsule; few small foramina pierce medial laminae of each exoccipital.
Exoccipitals paired, irregularly shaped, and in medial contact along a straight suture; each has a moderately pronounced, dorsolateral ridge; anterodorsally, each exoccipital contacts supraoccipital with a straight, oblique suture, anterolaterally prootic, ventrally basioccipital and dorsolaterally, lateral to dorsal ridge, supratemporal; fenestra ovalis situated at suture between prootic and exoccipital, and exoccipital forms posterior margin of fenestra; posterior to fenestra ovalis halfway to foramen magnum a recess, bearing several foramina; internally, exoccipital contributes to formation of posteroventral, posteromedial and posterolateral wall of otic capsule; exoccipital most posteriorly projecting bone of skull roof and forms dorsal, lateral, and lateroventral border of foramen magnum, with exoccipital components of occipital condyle approaching but remaining slightly separated from each other, posterior process of basioccipital not excluded from participation in foramen magnum.
Parasphenoid and basisphenoid single and fused to form elongate parabasisphenoid, about 1.9 (MCP 15219) to 2.1 (MCP 8355) times as long as wide, and occupying most of skull floor; parabasisphenoid concave in parasphenoid region and convex in basisphenoid region; basisphenoid portion roughly rounded, parasphenoid rostrum lanceolate with a pointed anterior tip; anterior half of parasphenoid rostrum bears a dorsally oriented process, which protrudes into gap between vertical laminae of frontals, and contacts them with its lateral edges; anterior tip of parasphenoid rostrum does not surpass anterior border of frontals in ventral view; anteriormost part of dorsal process of parasphenoid rostrum contacts ventral surface of posteromedial region of nasals; parasphenoid rostrum distinctly separated from choanal process of palatine, from posterior ending of vomers, and from septomaxillae; dorsolaterally parabasisphenoid contacts frontals and parietal, and prootics posterior to it, and reaches greatest width at sutures between parietal and prootics; foramen located only on right side (both specimens) in posterolateral part of basisphenoid portion, at suture with prootic; in dorsal view (internally), parabasisphenoid pierced by some foramina in lateral region of bone.
Basioccipital single, about hexagonal, almost as wide as long, convex; contacts parabasisphenoid complex anteriorly with straight transverse suture, prootics anterolaterally with oblique sutures, and exoccipitals posterolaterally with oblique sutures; posterodorsally, basioccipital partly overlain by exoccipital components of occipital condyle; posterior end of basioccipital (part of occipital condyle), rectangular-shaped in ventral view and quadrant-shaped in posterior view; widest part of basioccipital at (MCP 15219) or slightly posterior (MCP 8355) to sutures with prootics and exoccipitals; ventral surface of basioccipital slightly elevated in anterior region; small foramen present medially in posterior region, at about where basioccipital part of occipital condyle begins.
Maxillae paired, elongate, account for about one-third of length of skull, about 4.5 (MCP8355) to 4.8 (MCP 15219) times as long as high, each extending from level of (MCP 8355) or shortly behind (MCP 15219) lateral process of premaxilla to posterior region of orbit, forming most of lower margin of orbit laterally; in lateral view, highest point at about mid-length of bone; anteriorly, maxillae slightly arched towards premaxilla; ventral surface of maxilla bears four tooth loci, with curved, and rear-facing anterior teeth approximately similar in size, followed posteriorly, after a small interspace, by a pair of larger and grooved fangs, situated below eye; posterior end of maxilla bears a tooth-like spur at lateral margin of ventral surface; maxilla contacts prefrontal on its dorsal surface at its highest point, anterior regions of lateral and medial processes of ectopterygoid posteriorly, and its short, knob-like palatine process contacts maxillary process of palatine (MCP 8355), or remains slightly separated from it (MCP 15219); a small, knob-like ectopterygoid process visible, but situated slightly anterior to contact zone with anterior tip of medial process of ectopterygoid.
Ectopterygoids paired, wishbone-shaped, deeply forked anteriorly, with a medial process 1.4 (MCP 8355) to 1.9 (MCP 15219) times as long as lateral process; in dorsal view, ectopterygoids either almost entirely visible (MCP 15219) or almost only lateral processes visible and rest of ectopterygoids covered by parietal (MCP 8355); anterior processes directed anterolaterally and frame posterior end of maxilla, forming a maxilo-ectopterygoid fenestra; ventral surface of flattened posterior process has facet, where it firmly contacts dorsal surface of anterolateral portion of pterygoid.
Pterygoids paired, either edentulous (MCP 15219) or with one small tooth in anterior region (MCP 8355), flattened, elongate and slender, 7.5 (MCP 15219) to 8.9 (MCP 8355) times as long as broad, corresponding to slightly less than half length of skull; not in contact with other skull bones except for firm contact with ectopterygoid; in dorsal view, pterygoids only partly (MCP 15219) visible or not visible at all (MCP 8355), completely covered by roofing skull bones; anteromedial tip of pterygoid approaching but not contacting palatine, either thumb-like (MCP 15219) or not pronounced (MCP 8355); in ventral view, lateral border of pterygoid slightly curved posterolaterally; medial borders of pterygoids nearly parallel to each other in anterior third, gradually diverging and tapering posterolaterally in posterior two-thirds, resulting in greatest distance between both pterygoids at their posterior tips; posterior end of pterygoid approaches ventromedial process of quadrate medially without touching it.
Palatines paired elongate and slender, about 6.2 (MCP 15219) to 6.3 (MCP 8355) times as long as wide, when not considering choanal and maxillary processes, and correspond to 25% (MCP 15219) to 27% (MCP 8355) of length of skull; medial edge slightly curved; ventral surface has three (MCP 8355), four (MCP 15219, left), or five (MCP 15219, right) tooth loci; teeth subequal, curved, and rear-facing; maxillary process of palatine contacts medioventral region of prefrontal on dorsal surface of its maxillary process, and palatine process of maxilla (MCP 8355), or remains slightly separated from it (MCP 15219); anterior portion of palatine, anterior to tooth line almost triangular and contacts ventrolateral part of vomer dorsally; dorsomedially, a long, thin choanal process rises and curves downwards in a semicircle, approaching but remaining separated from its counterpart medially; short, broad maxillary process situated on lateral surface of palatine at level of first tooth, directed anterolaterally; posterior part of palatine behind tooth line bifurcated, with equally long processes (MCP 15219) or slightly shorter ventrolateral process and slightly longer dorsomedial process (MCP 8355), both tapering towards posterior end.
Supratemporals laminar, elongate, more than five times as long as wide, slightly curved, and slightly oblique; in dorsal view, each supratemporal overlaps and firmly contacts more than posterior half of prootic and anterolateral part of exoccipital over well-defined facets; anteriorly approaching, but not reaching to parietal; posterolateral region approaches dorsomedial aspect of quadrate, without touching it; posterior end protrudes posteriorly slightly beyond dorsal region of quadrate, but ends before posterior end of exoccipital.
Quadrates flattened and broad dorsally, tapering dorsoventrally in lateral view, but gradually increasing in width in rear view; oriented obliquely, from anterodorsally to posteroventrally; posterodorsal part approaches posterolateral region of supratemporal medially; medial part has a short process, which corresponds to contact region with columella auris, but columella auris hardly visible; ventral part bifurcated, with medial branch broader than lateral branch and both together spanning glenoid cavity of retroarticular process of mandible; quadrate does not exceed posterior limit of skull roof.
Dentaries elongate and slender, making up about 43% (MCP 15219) to 47% (MCP 8355) of length of mandible, slightly curved anteromedially; dorsal surface bears six tooth loci; teeth subequal, curved and rear-facing; lateral face slightly convex with a mental foramen located at about level of third tooth, in anterior one-third of dentary; at (MCP 8355) or behind (MCP 15219) level of last tooth, dentary branches into a shorter dorsal process, which overlays anterior part of compound bone, and longer lanceolate ventral process; in medial view, gap between dorsal and ventral processes mainly filled by splenial; in ventral view, posteriormost part of ventral process contacts anteroventral part of angular; ventral process runs with its dorsal surface parallel and close along anterior part of medioventral region of compound bone without touching it; in medial view, dorsal process slightly bifurcated in its posteriormost region, distinctly behind last tooth, with a longer dorsolateral branch and a very short ventromedial branch.
Splenials elongate, triangular, tapered anteriorly, about 3.7 (MCP 8355) to 4 (MCP 15219) times as long as high, smallest bone of mandible, making up less than one-fifth of mandibular length; anterior mylohyoid foramen centrally positioned in posterior part of bone anterior to level of its greatest height; posterior edge of splenial firmly contacts anterior region of angular.
Angulars elongate, triangular, tapered posteriorly, almost 4.3 (MCP 8355) to 4.6 (MCP 15219) times as long as high, second smallest bone of mandible, making up a little less than one-third of mandibular length; each angular contacts splenial anteriorly, compound bone laterally and dorsally, posteriormost tip of medial process of dentary, and posteromedial surface of ventral process of dentary; posterior mylohyoid foramen centrally positioned in anterior third of bone.
Compound bones elongate and slender, about 7.3 (MCP 15219) to 8 (MCP 8355) times as long as high, largest bone of mandible, making up more than two-thirds of length of mandible; prearticular crest slightly higher than surangular crest and thus marginally visible in lateral view, the latter not visible in medial view; in lateral view, compound bone tapers anteriorly, loosely fitting between dorsal and ventral processes of dentary; an anterodorsally oriented foramen present in anterior region on lateral surface of compound bone; retroarticular process moderately long, slightly medially directed, and slightly extends (MCP 8355) or not (MCP 15219) beyond posterior end of exoccipital.

Osteological comparisons
The skull of A. albicollaris differs from that of A. assimilis (Entiauspe-Neto et al. 2021a, characters in parentheses) in having robust thumb-like posterior processes (sharply pointed); weakly to moderately pronounced parietal ridges (well-pronounced); nasal and prefrontal in contact (not in contact); parabasisphenoid and basioccipital bones not fused (fused); anterior end of supratemporal overlaps more than posterior half of prootic and posterior end does not reach posterior end of exoccipital and is not most posteriorly protruding bone of skull roof (anterior end overlaying only posterior part of prootic and posterior end protrudes posteriorly beyond posterior end of exoccipital and is thus most posteriorly protruding bone of skull roof); 0-1 pterygoid teeth (4); parabasisphenoid concave in parasphenoid region (straight); dentary with 6 tooth loci (7-8).
The skull of A. albicollaris resembles that of A. ambiniger (Entiauspe-Neto et al. 2021b, characters in parentheses) but differs from this species in having vomers medially in contact (not contacting each other medially); anterior border of parietal is strongly concave curved (anterior border is almost straight); exoccipital components of occipital condyle approaching but remaining slightly separated from each other, so that posterior process of basioccipital is not excluded from participation in foramen magnum (exoccipital components of occipital condyle are in close contact, excluding posterior process of basioccipital from participation in foramen magnum); prearticular crest is slightly higher than surangular crest of compound bone (prearticular and surangular crests are about similar in height).
It differs from that of A. cearensis (Ferrarezzi et al. 2005, characters in parentheses) by having 0-1 pterygoid teeth (2); prearticular crest is slightly higher than surangular crest of compound bone (same height as surangular crest); U-shaped fronto-parietal suture in dorsal view without parietal indentation between frontals (W-shaped suture, with an antero-median parietal indentation); dorsal laminae of nasal and frontals in contact (not in contact); dentary with 6 tooth loci (7-8).
The skull of A. albicollaris differs from that of A. sanctaeritae (Entiauspe-Neto et al. 2020a, characters in parentheses) by having dorsal laminae of nasal and frontal in contact (not in contact); nasal and prefrontal in contact (not in contact); 0-1 pterygoid teeth (4-5); weakly to moderately pronounced parietal ridges, both ridges slightly approach each other posteriorly, but remain clearly separated from each other (well-pronounced, both ridges merge in posterior fifth of parietal); dentary with 6 tooth loci (8-9).

Discussion
The number of species of Apostolepis is herein reduced to 34. As was the case for A. cerradoensis, several species of Apostolepis remain known based on a series smaller than five individuals (see Entiauspe-Neto et al. 2019). Although  and Nogueira et al. (2012) examined relatively large samples of A. albicollaris (28 and 14 specimens, respectively), these authors did not account for thorough interspecific comparisons of their evaluated morphological characters, that would reveal it as synonymous with A. cerradoensis. It was reported in both studies that A. albicollaris might present a white venter, which was considered as a diagnostic character from A. cerradoensis (Lema 2003). Other species of Apostolepis have also been reported as being highly polymorphic regarding their ventral coloration, such as A. dimidiata, which like A. albicollaris may also have gradually melanistic ventral coloration (Lema 1993;Entiauspe-Neto et al. 2019).
Throughout its range in the Cerrado of Central Brazil, A. albicollaris is sympatric or occurs close to at least seven congeners (Table 1). From these, it bears great similarity to A. dimidiata, sharing an overlapping dorsal pattern (Fig. 12), coloration, and supralabial, infralabial, and subcaudal counts. These species can be distinguished based on its ventral counts and nuchal collar presence (present in A. albicollaris, absent in A. dimidiata), and are also not sympatric in most of their ranges, although a small contact zone can be seen in the Cerrado of western Minas Gerais, southeastern Brazil (Nogueira et al. 2019). It should also be noted that A. albicollaris bears a small-sized semicapitate hemipenis, while A. dimidiata bears a medium-sized capitate hemipenis, with enlarged spines in three rows, with spinulate apices; the small sample precludes us from defining these subtle characters as diagnostic or not, although for the time being, we will treat them as such. It is noteworthy that A. albicollaris, A. ambiniger, A. dimidiata, A. goiasensis, and A. kikoi share similarly small to moderately sized hemipenes, covered by spinules on apical and medial surfaces, with papillate calyces on sulcate and lateral sides, and lamellae on their asulcate sides, which may bear minor interspecific differences, such as the incomplete capitulum on A. albicollaris, the well-developed enlarged lateral spines in A. ambiniger, A. goiasensis and A. kikoi, and the conspicuous lamellae on the asulcate side on A. dimidiata.
Both Apostolepis albicollaris and A. dimidiata have been recovered as sister-species in our phylogenetic inferences, and are shown here to have a particularly small degree of molecular difference (p-distances of 1% for 12S and 0% for CMOS). Small genetic distances among both species are also supported by the use of patristic distances. We believe that the use of finer and larger molecular markers, such as ultraconserved elements (UCEs) or Restriction site-associated DNA sequencing (RAD-seq), should allow for the evaluation of gene flux, divergence time estimation among lineages, and testing of reciprocal monophyly among these species. Our results are also in contrast with the position of Nogueira et al. (2012), who proposed the Apostolepis dimidiata species group, containing A. ambiniger, A. albicollaris, A. breviceps, A. dimidiata, A. goiasensis, A. intermedia, A. lineata, A. phillipsi, A. vittata, and Parapostolepis polylepis (Amaral, 1922), and claimed that "no putative synapomorphy was found to support a sister-group relationship between these species" (for A. albicollaris and A. dimidiata). It should also be noted that no evidence was presented to support the aforementioned claim, and the relationships between A. albicollaris and A. dimidiata were only briefly mentioned in the abstract section. In general terms, the A. dimidiata group (sensu Ferrarezzi 1993) is diagnosed by having a rostral scale prominently enlarged (projected), a rounded terminal shield, naris located in the anterior part of nasal, anterior supralabials bent into the oral border, head narrowing towards snout (triangular), light snout blotches confined anteriorly, obscured to indistinct on prefrontal, and the white supralabial blotch elongated continuously throughout labial margin (Nogueira et al. 2012). However, these species also have noteworthy differences in other major diagnostic characters, such as the small-sized and rounded hemipenes for A. intermedia, and the elongated hemipenes, without prominent lamellae on its asulcate side, added to 17 dorsal rows without reduction for P. polylepis. Furthermore, the aforementioned diagnostic characters of the A. dimidiata group also integrally apply to other species, such as A. christineae, which bears great resemblance and shares several diagnostic characters with A. vittata and was seemingly overlooked in the revision proposed by Nogueira et al. (2012).
Another species that bears a striking resemblance to Apostolepis albicollaris is A. quirogai, that also has a red dorsum with two lateral black stripes, a black head and a white nuchal collar ( Fig. 13; Entiauspe-Neto et al. 2021b). However, the latter species occurs in Alto Paraná Atlantic Forests of Misiones, in northwestern Argentina, and northern Rio Grande do Sul, southern Brazil, far from the Cerrado of central Brazil (in which A. albicollaris occurs), has a wide circular snout blotch, and also bears a considerably higher count of ventral scales (269 in the single male, 276 in the single female) . Further molecular and morphological evidence is needed for A. quirogai in order to test its phylogenetic relationship regarding A. albicollaris, and clarify whether both species are close to each other, or if their similarities are rather a byproduct of convergent evolution. The small sample size of molecular data available for Apostolepis precludes us from making general statements on the systematic relationships of the group. Although Apostolepis is recovered as monophyletic by this study and other recent phylogenies (e.g., Zaher et al. 2019;Moraes et al. 2021), only seven terminals have limited gene fragments available (up to four) for comparison, out of the 34 valid species, and key taxa remain unavailable for comparison. For instance, the monotypic genera Coronelaps Lema & Hofstadler-Deiques, 2010 and Parapostolepis Amaral, 1930 have not yet been tested under a molecular systematic framework, and their relationships with their close counterparts (Elapomorphus and Apostolepis, respectively) remain hypothesized largely upon morphological characters interpreted as putatively autapomorphic (Lema & Hofstadler-Deiques 2010).
Nevertheless, the available molecular data provides support for the monophyly of the Apostolepis assimilis species group in its current sense (sensu Ferrarezzi 1993;Ferrarezzi et al. 2005), being composed of three red colored species (A. assimilis, A. cearensis, A. sanctaeritae) that inhabit the South American Dry Diagonal and share a nasal scale separated from preocular, fifth supralabial scale only in contact with parietal, temporals 0+1, terminal scale pointed in a sharp edge (although we have encountered specimens of A. assimilis with both pointed and rounded terminal scales), large white internasal blotch covering snout and reaching frontal shield, moderate to extremely broad and sharply evident white and black nucho-cervical collars, dorsal pattern uniformly red, white ventral surface (although some specimens may have a red venter), except gular region which is dark to mostly black pigmented, terminal shield black, undifferentiated from tail tip coloration, Duvernoy's gland moderately developed, harderian gland well developed, reaching temporal region, and adductor jaw muscles moderately developed, reaching dorsal surface of braincase (Ferrarezzi 1993;Ferrarezzi et al. 2005). However, these characters might not be shared exclusively by these species; A. dorbignyi and A. multicincta both have a nasal scale separated from preocular, fifth supralabial in contact with parietal, terminal scale pointed, large white internasal blotch, wide white and black nuchal collars, and a uniformly red dorsal pattern, with major differences from the A. assimilis group relying on number of infralabials contacting anterior chinshields (1-3 in A. dorbignyi and A. multicincta), temporal scales (absent in A. dorbignyi and A. multicincta), and terminal scale coloration (white in A. dorbignyi and A. multicincta). Both, A. dorbignyi and A. multicincta, also occur in the South American Dry Diagonal, in the Chaco of Argentina and Bolivia Burgos-Gallardo et al. 2017). The aforementioned morphological differences between A. dorbignyi and A. multicincta on one hand and species of the A. assimilis group on the other are subtle, and as most diagnostic characters of coloration used for Apostolepis, likely to be prone to individual variation and polymorphism, as more comprehensive series of specimens become available. We argue that taxonomic approaches based strictly on external morphology are likely to overestimate diversity in fossorial and poorly sampled species, and that further studies and data collection, seeking to integrate internal morphology and molecular studies, are needed in order to increase taxonomic stability and the reliability of diagnoses for Elapomorphini species.