Subdoluseps, Freitas & Datta-Roy & Karanth & Grismer & Siler, 2019

Freitas, Elyse S., Datta-Roy, Aniruddha, Karanth, Praveen, Grismer, L. Lee & Siler, Cameron D., 2019, Multilocus phylogeny and a new classification for African, Asian and Indian supple and writhing skinks (Scincidae: Lygosominae), Zoological Journal of the Linnean Society 186, pp. 1067-1096 : 1080-1085

NON-MONOPHYLY OF LYGOSOMA S. L. AND PARAPHYLY OF LYGOSOMA S. S. AND MOCHLUS

A stable taxonomy reflects evolutionary relationships of species and clades and is of paramount importance for studies in biological science. Diverse fields, from ecology to development, rely on accurate species- and supra-specific-level identifications for their research ( Mayr, 1976; Felsenstein, 1985; Winston, 1999; Wheeler et al., 2004). Furthermore, taxonomy plays a critical role in biodiversity conservation and management, with agencies using recognized nomenclature for identification and classification of regional fauna, including rare and threatened species (e.g. CITES and IUCN; Kaiser et al., 2013; Groves et al., 2017; IUCN-SSC Species Conservation Planning Sub-Committee, 2017).

In supra-specific taxonomy, the genus category is included in the binomial name of a species, so although it is not based inherently on biological criteria, it is an important communication tool in the name of a species, depicting a close relationship between species in the same genus to the exclusion of other species ( Cain, 1956; Winston, 1999). Therefore, the genus reflects information about the evolutionary history of the species it composes. Inger (1958) proposed a definition of genera that uses ecological criteria to determine the species that are placed in a genus, with ‘mode of life’ (i.e. adaptive zone; Vences et al., 2013) as a major diagnostic character of the genus. However, currently this approach is problematic, especially for little-known clades, because it requires ecological knowledge of all species included in a genus and of closely related species excluded from that genus. Furthermore, congeners that live in sympatry may have undergone niche displacement (e.g. genus Brachymeles ; Huron & Siler, unpubl. data), making the adaptive zone difficult to define empirically ( Vences et al., 2013). Accordingly, the only current defining characteristic of a genus is that it represents a clade in a broader family-level clade.

Among scincid lizards, studies have shown that many taxonomic groupings are not supported as monophyletic, e.g. Amphiglossus ( Whiting, Sites & Bauer, 2004) ; Sphenomorphus ( Linkem et al., 2011) ; Anomalopus and Eulamprus ( Skinner et al., 2013) ; Trachylepis ( Karin et al., 2016) ; and Afroablepharus ( Medina et al., 2016) . These inconsistencies between historical nomenclature and the evolutionary relationships recovered through molecular datasets necessitate the revision of genus-level classifications for taxonomic stability and for discussions of evolutionary patterns and processes within and among clades ( Kaiser et al., 2013; Vences et al., 2013).

Our concatenated Bayesian Inference (BI) phylogenetic and coalescent-based species tree analyses reveal that Lygosoma s.l. is not monophyletic. Additionally, Lygosoma s.s. is paraphyletic, with respect to Mochlus and Lepidothyris , and the genus Mochlus is paraphyletic with respect to Lepidothyris ( Figs 1–3). These results are consistent across all analyses and are in line with the findings of previous studies: Datta-Roy et al. (2014) observed similar relationships between Lamprolepis and Lygosoma s.l., and Lygosoma s.s. and Mochlus in their study, albeit with low support at some of their deeper nodes. In our concatenated and coalescent-based analyses, Lygosoma s.s. Clade A, containing Lygosoma quadrupes , the type species of the genus, is supported as divergent from the other two major clades of Lygosoma s.s. ( Figs 1–3), again corroborating the results of Datta-Roy et al. (2014).

Some differences between our concatenated and coalescent-based topologies are seen regarding the relationship between Lamprolepis smaragdina and Lygosoma s.l. In concatenated analyses, Lamprolepis smaragdina is recovered as part of Lygosoma s.l. with strong support ( Fig. 1), although its position in Lygosoma s.l. is unresolved, suggesting that Lygosoma s.l. is paraphyletic with respect to Lamprolepis smaragdina . In contrast, the relationship between Lamprolepis smaragdina and Lygosoma s.l. is resolved fully in our coalescent-based species tree analyses, which recovered Lamprolepis smaragdina as the sister taxon to Clade A with strong support ( Fig. 2). Although the finding of a paraphyletic Lygosoma s.l. with respect to Lamprolepis smaragdina is consistent with previous studies ( Honda et al., 2000, 2003; Datta-Roy et al., 2014), it is surprising nevertheless given the highly divergent life histories of the species in question: Lamprolepis smaragdina is a larger, more robust, bright-coloured, arboreal skink, whereas most of the species in the genus Lygosoma are small, inconspicuously coloured and semi-fossorial ( Greer, 1977; Das, 2010). In fact, Greer (1977) cited this ecological difference as evidence that the genera Lygosoma and Lamprolepis were not each other’s closest relatives. The differences between our concatenated and coalescent-based analyses may be attributed gene tree discordance ( Degnan & Rosenberg, 2009; Linkem et al., 2016). Given the presence of discordance between loci in our nuDNA dataset, concatenation of our sequences may have resulted in a misleading BI topology.

The relationships of Clades B, C and D are fully resolved in our concatenated analyses, but not in our coalescent-based analyses ( Figs. 1, 2). In our concatenated analyses, Clades C and D are supported highly as sister taxa and together are recovered as sister to Clade B. However, in coalescent-based analyses, the relationships between the three clades are not resolved, although they are still recovered as a clade distinct from Clade A with high support ( Fig. 2). We suspect that incomplete taxonomic sampling across the radiation and low sample sizes for some rare or secretive species contributed to this lack of resolution. To estimate the multispecies coalescent process for each gene, sequences from at least two individuals per lineage need to be included in the dataset ( Heled & Drummond, 2010), which suggests that increasing the taxonomic sampling per lineage will increase resolution of the species tree. Additionally, studies have shown that increased taxonomic sampling across the group being investigated improves species tree accuracy ( Hovmöller et al., 2013; Lambert et al., 2015). Unfortunately, these two issues could not be addressed fully at this time given the rarity or absence of tissues in collections for some focal taxa. However, as next-generation sequencing techniques are revolutionizing approaches to phylogenetic studies by providing datasets of thousands of loci at increasingly lower costs ( Ekblom & Galindo, 2011), these datasets are becoming more common in skink population and phylogenetic research ( Barley et al., 2015b; Brandley et al., 2015; Rittmeyer & Austin, 2015; Linkem et al., 2016; Bryson et al., 2017). These techniques have the power to resolve difficult intra- and interclade relationships (e.g. Crawford et al., 2012; McCormack et al., 2012; Streicher & Wiens, 2017) and may be a promising tool for resolving the relationships among Clades B, C and D.

CLADES ARE NOT DIFFERENTIATED BY MORPHOLOGY

Researchers have struggled to find diagnostic characters for Lygosoma s.l., which has resulted in challenges to understanding the systematics of the group ( Boulenger, 1887; Smith, 1937a; Mittleman, 1952). As a result, species relationships have been in flux for almost two centuries, with species sometimes placed together in a single genus ( Boulenger, 1887) or separated into multiple genera ( Smith, 1937a; Mittleman, 1952). In performing multivariate analyses, we investigated whether combinations of characters commonly used in delimitating species and genera could differentiate Lygosoma s.l. species and clades in morpho space. However, our principal components analysis (PCA) and discriminant analysis of principal components (DAPC) showed that species and clades were not separated in morpho space. This result underscores the historical difficulties of using morphology to classify Lygosoma s.l. skinks ( Fig. 4), illustrating how traditional morphological approaches have largely failed in diagnosing clades with Lygosoma s.l., because of the large amount of morphological overlap between species. Among the species examined, our PCA results show transitions in Lygosoma s.l. between robust and elongated body forms, with species overlapping along a morphological gradient ( Fig. 3). As a result, among the major clades, we find that none form distinct clusters in morpho space ( Fig. 4A), although it appears that Clade A contains the most elongated species, followed by Clade D and then by Clades B and C, with the highest amount of morphological overlap between Clades B, C and D. Given our phylogenetic results, which indicate that Clades B, C and D together form a clade to the exclusion of Clade A, our observations of these clades having the highest amount of morphological overlap makes sense.

Our DAPC, which used the principal components from the PCA as descriptor variables, was conducted to compare within-clade variance to between-clade variance and revealed Clades B and C to have the highest amount of overlap and occupy more restricted areas of morpho space when compared with Clades A and D ( Fig. 4B). Interestingly, Clade A appears the most morphologically distinct clade with only two samples falling in the inertia ellipses of other clades and only a single individual from another clade (Clade D) recovered in its inertia ellipse ( Fig. 4B). However, this pattern may be driven by the large number of individuals from the Lygosoma quadrupes species complex in our morphological dataset, which have a highly derived body form in comparison to other species in Clade A and in Lygosoma s.l. ( Greer, 1977). It is likely that the inclusion of additional samples of other species in Clade A (e.g. L. corpulentum and L. isodactylum ) and from other clades would temper this pattern.

Four species are labelled incertae sedis in our PCA analysis because they were not represented in our phylogenetic analyses. Among these, Lygosoma koratense from Southeast Asia appears morphologically most similar to species in Clade B, and L. pembanum and L. tanae from Africa appear morphologically most similar to Clade C ( Figs 4, 5A). The remaining species, L. kinabatanganense , a large and robust species from Malaysia (Sabah, Borneo), does not fall within the morphological boundaries of any of the clades in our PCA. ( Figs 4, 5A). Interestingly, a previous phylogenetic study of Lygosoma s.l. suggested a close relationship between Lygosoma quadrupes and L. koratense ( Honda et al., 2000) , which was corroborated in subsequent studies using the same sequence data ( Ziegler et al., 2007; Wagner et al., 2009; Skinner et al., 2011 Pyron et al., 2013; Datta-Roy et al., 2014). Unfortunately, vouchered tissue samples of L. koratense were not available for this study. If the relationship of L. quadrupes and L. koratense holds true in future phylogenetic analyses, it would expand the extent of the occupied morpho space of Clade A and would have interesting implications for the evolution of body form in the clade.

The results of our PCA and DAPC analyses show that, like traditional morphological approaches, multivariate approaches have largely failed to differentiate clades in Lygosoma s.l. While there exists variation in body form among species in the group, this appears to change along a morphological gradient that only partially conforms to phylogeny ( Fig. 4A). However, there are two characters not included in our PCA and DAPC analyses that have been employed historically in Lygosoma s.l. systematics, which are worth discussing further because they may be of use to differentiating phylogenetic clades in Lygosoma s.l. These characters are the morphology of the secondary palate and the character state of the lower eyelid. Of these characters, the morphology of the secondary palate is the least controversial. Greer (1977) used this character to unite L. quadrupes with Riopa , and he described all species of Riopa recognized at the time (31 species) as having processes that project from the posteromedial edge of the palatine bones, which separate the two pterygoid bones. Interestingly, Greer (1977) noted two character states of the secondary palate in Lygosoma : an open state (pterygoids emarginated along their posterior edge) and a closed state (pterygoids not emarginated along their posterior edge), each of which corresponds consistently with clades in our phylogenetic analyses ( Figs. 1, 2). Greer (1977) listed all species found in our Clade B (with the exception of the recently described L. samajaya , which he did not examine) and our Clade C as having a closed palate, and he listed all species found in our Clade A (with the exception of L. corpulentum , which he did not examine) and our Clade D (with the exception of L. vosmaerii , which he did not examine and L. punctatum which was variable) as having an open palate. The palate of L. koratense was listed as closed, again morphologically linking this species more closely with Clade B than Clade A. Furthermore, Greer (1977) used the morphology of the secondary palate to diagnose the genus Lamprolepis from Lygosoma s.l. However, examination of written descriptions and drawings of the palate of Lamprolepis indicates that Lamprolepis smaragdina has posteromedial projecting processes separating the pterygoid bones ( Greer, 1970b: Fig. 1; 1977: Fig. 5), similar to, but not as pronounced as, the processes in Lygosoma s.l. Therefore, the morphology of the secondary palate is useful in diagnosing the larger Lygosoma s.l. group of clades and may also be a useful descriptor variable for clades within Lyogosoma s.l.

In contrast to the morphology of the secondary palate, the taxonomic utility of the lower eyelid character state has been more controversial. Mittleman (1952) proposed the state of the lower eyelid, which has been defined broadly as either scaly or with a transparent window, as a diagnostic character separating groups, and he relied on eyelid state to split Mochlus from Riopa . Subsequent authors have disagreed with the taxonomic value of this character ( Broadley, 1966; Greer, 1974, 1977), arguing that the character is highly variable within clades. Nevertheless, several recent skink taxonomic studies have mentioned the state of the lower eyelid as part of the combinations of diagnostic characters for some skink genera descriptions ( Euprepis and Eutropis [ Mausfeld & Schmitz, 2003]; Brachymeles [ Siler et al., 2011]; Heremites and Toenayar [ Karin et al., 2016]), although the presence of both states in the genus Scincella was noted by Linkem et al. (2011). In our study, the state of the lower eyelid does not appear consistent with our clades, with the exception of Clade A in which all our sampled members have a scaly lower eyelid. Instead, the lower eyelid character state appears highly variable between species and may also exhibit intraspecific variation. In Clade B, four of the five sampled species have a scaly lower eyelid; the exception being L. pruthi , which has a transparent disc on its lower eyelid ( Sharma, 1977). In Clade C, all sampled species have a scaly lower eyelid, with the possible exception of M. guineensis . In its original description, M. guineensis was recorded as having a lower eyelid with a transparent disc ( Peters, 1879), but the eyelid state was revised subsequently as scaly by Greer (1977). Additionally, four species from Africa that we lack genetic data for, but include provisionally in Clade C ( M. laeviceps ( Peters, 1874) , M. mabuiiformis ( Loveridge, 1935) , M. simonettai ( Lanza, 1979) and M. tanae ; see justification in our taxonomic revision section), also were described originally as having a lower eyelid with a transparent disc ( Peters, 1874; Loveridge, 1935; Lanza 1979). One of these lineages, M. laeviceps , later was reclassified as having a scaly lower eyelid ( Greer, 1977). In Clade D, all of our sampled species have a transparent disc on their lower eyelid, but there is a record of one specimen of L. albopunctatum from Sarbhog, Assam, India in the Indian Museum, Kolkata that possesses a lower eyelid with a transparent disc on its right side and a scaly lower eyelid on its left side ( Hora, 1927). Additionally, L. lineolatum was described originally as having a scaly lower eyelid ( Stoliczka, 1870), but Smith (1935) reclassified the species as having a transparent disc in its lower eyelid. Nevertheless, several Lygosoma sp. individuals from Myanmar appear to have a scaly lower eyelid (ESF, unpubl. data), suggesting that the lower eyelid state is variable in Clade D. Therefore, unlike the morphology of the secondary palate, the lower eyelid character state seems to be inconsistent across most clades of Lygosoma s.l. and not useful for clade-level diagnosis.

A REVISED CLASSIFICATION OF LYGOSOMA S. L.: OVERVIEW

Currently, Lygosoma s.l. comprises 49 nominal species: 31 species in the genus Lygosoma s.s., 15 species in the genus Mochlus and three species in the genus Lepidothyris . Of these 49 species, we were able to include 22 in our phylogenetic analyses, representing all three genera, for the most complete assessment of the radiation to date. The results of our phylogenetic analyses suggest that Lygosoma s.s. does not form a monophyletic group with respect to the other genera in Lygosoma s.l. ( Lepidothyris Cope, 1892 and Mochlus Günther, 1864a ) and the genus Lamprolepis Fitzinger, 1842 . Instead, Lygosoma s.s. is separated into three clades: one comprising elongate-bodied species from Southeast Asia, Indonesia and the Philippines (Clade A), one comprising the widespread species L. bowringii and other small, stouter-bodied species from India, Southeast Asia and Christmas Island (Clade B) and one comprising species from India and Southeast Asia (Clade D). Additionally, we do not recover Mochlus as monophyletic, with our results suggesting it is paraphyletic with respect to Lepidothyris . Given these results, we propose several taxonomic changes to this group ( Fig. 5). First, we redefine the genus Lygosoma to include only Clade A, comprising the type species Lygosoma quadrupes and other elongate-bodied taxa. Second, we resurrect the genus Riopa for Clade D, comprising the type species Riopa punctata and other species from India and Southeast Asia. Third, we synonymize the genus Lepidothyris with Mochlus . Last, we describe a new genus, Subdoluseps gen. nov. for Clade B, comprising the type species S. bowringii and additional species distributed across India, Southeast Asia and Christmas Island. We recognize that our taxonomic sampling is incomplete considering the large diversity of species that are recognized currently in Lygosoma s.l. and we, therefore, advocate for continued efforts to voucher and include additional species in future studies to better understand the diversity, distribution and boundaries of this unique radiation of Old World scincid lizards.