Date Published: May 10, 2019
Publisher: Public Library of Science
Author(s): Hussam Zaher, Robert W. Murphy, Juan Camilo Arredondo, Roberta Graboski, Paulo Roberto Machado-Filho, Kristin Mahlow, Giovanna G. Montingelli, Ana Bottallo Quadros, Nikolai L. Orlov, Mark Wilkinson, Ya-Ping Zhang, Felipe G. Grazziotin, Ulrich Joger.
Caenophidian snakes include the file snake genus Acrochordus and advanced colubroidean snakes that radiated mainly during the Neogene. Although caenophidian snakes are a well-supported clade, their inferred affinities, based either on molecular or morphological data, remain poorly known or controversial. Here, we provide an expanded molecular phylogenetic analysis of Caenophidia and use three non-parametric measures of support–Shimodaira-Hasegawa-Like test (SHL), Felsentein (FBP) and transfer (TBE) bootstrap measures–to evaluate the robustness of each clade in the molecular tree. That very different alternative support values are common suggests that results based on only one support value should be viewed with caution. Using a scheme to combine support values, we find 20.9% of the 1265 clades comprising the inferred caenophidian tree are unambiguously supported by both SHL and FBP values, while almost 37% are unsupported or ambiguously supported, revealing the substantial extent of phylogenetic problems within Caenophidia. Combined FBP/TBE support values show similar results, while SHL/TBE result in slightly higher combined values. We consider key morphological attributes of colubroidean cranial, vertebral and hemipenial anatomy and provide additional morphological evidence supporting the clades Colubroides, Colubriformes, and Endoglyptodonta. We review and revise the relevant caenophidian fossil record and provide a time-calibrated tree derived from our molecular data to discuss the main cladogenetic events that resulted in present-day patterns of caenophidian diversification. Our results suggest that all extant families of Colubroidea and Elapoidea composing the present-day endoglyptodont fauna originated rapidly within the early Oligocene–between approximately 33 and 28 Mya–following the major terrestrial faunal turnover known as the “Grande Coupure” and associated with the overall climate shift at the Eocene-Oligocene boundary. Our results further suggest that the caenophidian radiation originated within the Caenozoic, with the divergence between Colubroides and Acrochordidae occurring in the early Eocene, at ~ 56 Mya.
Determining the phylogenetic affinities within snakes was viewed by many herpetologists in the past as an insurmountable challenge. Underwood  expressed his profound frustration with a simple sentence: “I have found snake systematics to be a hard test to intellectual honesty”. Although the phylogenetic affinities of snakes were indeed difficult to determine on morphological grounds, monophyly of some higher-level taxa represent a long-standing consensus. This is the case for the clade Caenophidia, a group of advanced alethinophidian snakes recognized formally by Hoffstetter  to accommodate the families Colubridae, Dipsadidae, Hydrophiidae, Elapidae, and Viperidae. Hoffstetter’s Caenophidia was characterized by the absence of a coronoid bone and included the colubrid subfamily Acrochordinae, already known to share several additional derived morphological traits with the remaining caenophidian families [3,4]. The same group of “advanced alethinophidian snakes” was also recognized by Romer , who preferred to accommodate them in a newly erected superfamily Colubroidea, equating the latter with Hoffstetter’s concept of Caenophidia. “Acrochordoids” and “colubroids” were only later recognized as two distinct superfamilies within Caenophidia after Groombridge [6,7] argued convincingly that acrochordids were the sister-group of the remaining caenophidians based on a number of synapomorphies derived from the vomeronasal capsule, musculature, hyoid and costal cartilages [8,9,10].
We compare our molecular data and phylogenetic results with three recently published large-scale molecular studies: Pyron et al. , Zheng and Wiens , and Figueroa et al. ). Pyron et al.’s  and Zheng and Wiens’  included representatives of all recognised squamatan families, whereas Figueroa et al.  focused on snake lineages. Caenophidian coverage in Pyron et al.  and Zheng and Wiens  were identical (1062 species) and included sequences from up to 12 and up to 52 genes, respectively. Figueroa et al.  combined up to 10 genes for 1358 species of caenophidian snakes (excluding multiple individuals, unidentified, and misidentified species). Our study combines sequences from up to 15 genes for 1263 species (see S3 Table for number of genes and accession numbers; see also S4 and S5 Tables for more details on taxon sampling).
The considerable recent efforts to use molecular sequence data to resolve deep interrelationships of caenophidians [24,26,27,28,132–137] have differed substantially in the inversely related sampling of genes and taxa. Most data sets have either many taxa and few genes (MTFG) mostly from Sanger sequencing (e.g., 10 loci for 1343 species) , or few taxa and many genes (FTMG) built using next generation sequencing (e.g., 333 loci for 31 species) , and many discrepancies in results are likely due to this important difference. FTMG analyses tend to provide strong support for higher-level relationships that may be misleading because of inadequate sampling of key taxa. MTFG studies provide much more information about low-level relationships but tend to fail to provide well-supported resolution of deeper divergences due to the limited number of loci and short branches. An important exception is the hybrid approach of Zhang and Wiens  which added many (up to 52) extra genes for a subset (3,8%) of the terminals included in Pyron et al.  MTFG analyses.
The traditional meaning of the superfamily Colubroidea [18,26,26,28] no longer accommodates our growing knowledge of the phylogenetic affinities [18,19,23,25,26,27,28] and morphological disversity and disparities in the group [8,11,23,25,29,30]. Despite some pleas against any change on the traditional usage of the names “Colubroidea” and “Colubridae” , the new morphological evidence provided here reinforces the need for a series of taxonomic changes to accommodate new phylogenetic and morphological knowledge. Xenodermids, pareids, and xylophiids represent ancient caenophidian lineages that are phylogenetically and morphologically distinct from the endoglyptodont radiation. All three lineages lack the dental specializations that gave rise to an advanced venom delivery system characteristic of endoglyptodonts, thus breaking a universally accepted definition of colubroids as representing the truly venomous snake radiation. Xenodermids further lack some vertebral and cranial features that are commonly used to determine colubroid fossil remains and share with acrochordoids cranial and vertebral specializations that are virtually absent in the remaining caenophidian lineages.