Introduction

As a part of the Sri Lanka-Western Ghats biodiversity hotspot, Sri Lanka features an exceptional species richness and endemism in amphibians [1, 2]. At present, about 120 amphibian species are recognized from the island, of which 104 are endemic. Nineteen species are considered extinct, while 72 are threatened with extinction due to a plethora of human induced activities [3].

The family Rhacophoridae, spread across tropical to sub-tropical Asia and Africa, is a diverse group of frogs containing some 432 species, 6% of the world’s anuran fauna [4, 5]. Their taxonomy, evolution and biogeography have been studied in several recent analyses [2, 6–12]; much of this work has involved molecular analyses. Despite this, molecular data are available for only about 73% of the species (Genbank, last accessed January 2021). This highlights the fact that rhacophorid taxonomy and systematics could still benefit from species-level analyses [10]. Accuracy of species identification, together with a clearly resolved taxonomy, are critical to biological research, especially in evolution, ecology, conservation, and biogeography [13].

Rhacophorid tree frogs of the genus Pseudophilautus, characterized by terrestrial direct development [7], represent about 63% of amphibian species recognized from Sri Lanka. Such high diversity has been generated over the course of their complex evolutionary history, spanning ~30 MY [9, 11]. The key innovation of terrestrial direct development, together with factors such as climatic fluctuations, topography, orogeny, ecological opportunity and terrestrial connectivity with India mediated by sea-level fluctuations, have shaped this remarkable diversification [2, 7].

The genus Pseudophilautus currently contains 59 extant species in Sri Lanka [5]. Many of these were described relatively recently [14–21]. Some of these species are cryptic and hence difficult to identify from morphology alone, though validated in molecular analyses; others (e.g., [20, 21]) are based on morphological data alone. As of now, molecular data are available, at least for a single locus, for 49 of the 59 extant Sri Lankan Pseudophilautus (See S1 Table). However, population level data remains scarce for many of these species.

Though some species have been validated using molecular data, sometimes through an integrative approach, a detailed multi-gene phylogeny became available for them only recently [2]. Meanwhile, species delimitation analyses carried out recently [10] on a part of the previously available data, mostly based on singletons of the popular 16S frog barcoding gene [22, 23], suggested that a few closely related species of Pseudophilautus, e.g., P. hankeni with P. dilmah and P. schmarda; P. papillosus with P. reticulatus (and several species from other genera) do not reach the molecular thresholds of species delimitation in Rhacophoridae [10].

The initial descriptions of Pseudophilautus were often based on a few samples drawn from a few populations, which may not have captured the full breadth of the molecular and morphological variation of the species. With most species having been diagnosed from morphology alone, an objective assessment of species boundaries is needed to stabilize the species-level taxonomy of the group.

Pseudophilautus has been shown to evolve at a moderate and constant rate [2], a pattern now shown to hold true for all rhacophorids [10]. The slower and predominantly allopatric mode of speciation, the narrow-range endemism characteristic of many species, and the relatively small genetic distances that separate morphologically distinct species, seem to be characteristic of Rhacophoridae [2]. These factors too, need to be considered in evaluating the species boundaries of this diversification.

Molecular species delimitation analyses are now being increasingly used to inform taxonomic judgements [10, 24–29]. The goal of these analyses is to build a taxonomic scheme for a set of reliably-identified samples and infer a de novo delimitation of operational taxonomic units (OTUs) [27, 29, 30–33] in the context of the phylogenetic species concept (PSC). Currently, many molecular species delimitation methods exist, which are based on either molecular-distance or gene-tree approaches. Widely used methods include automatic barcode gap discovery (ABGD; [34]), the generalized mixed Yule-coalescent model (GMYC; [35, 36]), the Poisson tree processes model (PTP; [37, 38]), the Bayes factor delimitation [25, 39], the Bayesian coalescent method in the software Bayesian Phylogenetics and Phylogeography (BPP; [26]), and phylogeographic inference using approximate likelihoods [27]. However, these algorithm-based species delimitation approaches have their own limitations and been subjected to criticism [40]. Despite the uncertainty surrounding their results, these methods are still being used in delimiting species. They are used either as stand-alone methods or as integrative contributors to objective species delimitation [28, 29, 41].

Here we address the following questions in the context of Pseudophilautus. How many species meet the validity of molecular delimitation methods? Are there commonalities in distribution patterns of the sister taxa that fail delimitation thresholds? Which clades harbor species difficult to identify from morphology alone? What lesson can a study such as this contribute to traditional morphology-based taxonomy?

Here, using all available validated molecular data for Pseudophilautus, we test the validity of all species for which molecular data are available, using multiple molecular delimitation methods. We show that several nominal species fail to reach delimitation thresholds. We also show the existence of several potential new species, underscoring that Pseudophilautus would benefit from further taxonomic assessment.

Materials and methods

Results

The phylogenetic analysis indicates that the internal nodes of the Pseudophilautus clades are, for the most part, well supported. Building on these relationships, six major clades can be seen in the phylogeny (Fig 1, clades A–F), concordant with the topology recovered by [2]. Topologies of the consensus trees of Bayesian and ML analyses are similar in their assembly of major clades (S1 Fig). Gene trees based on 16S rRNA, 12S rRNA and Rag-1 nDNA are congruent in most instances and clade-level relationships are well established for many taxa (S2 Fig). However, the gene trees have slight differences in their topologies as well; in the 16S tree Clade A is divided into two distantly related clusters and the clade involving P. mooreorum appears as an early-diverging lineage. In the nuclear Rag-1 tree, the P. fulvus and P. cuspis of clade C are more closely related to Clade F while Clade D is recovered as two different clusters (see S2 Fig). These differences were also reflected in the delimitation analyses carried out for different genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Delimitation summary results.

Molecular phylogenetic relationships of Pseudophilautus, based on Bayesian inference of the concatenated data set of the 16S rRNA + 12S rRNA + Rag1 (2189bp) loci. Asterisks (*) above nodes represent ≥ 95% Bayesian posterior probabilities. The scale bar represents number of changes per site. Recovered species delimitation based on overall significance of results for molecular species delimitation methods (ABGD, bPTP, mPTP and bGMYC) using 16S rRNA, 12S rRNA, Rag-1 nDNA loci and the concatenated data set are shown as black rectangles on the right. Six major clades of Pseudophilautus, A-F, are shaded in different colours on the phylogeny. Images of frogs indicate species suggestive of synonymyzation based on recovered species delimitations. Rectangles indicated by * in similar colours represent species pairs that were recovered as close relatives or recovered in different clades based on their relative position on the concatenated tree.

https://doi.org/10.1371/journal.pone.0258594.g001

Pairwise genetic distances among taxa are provided in the S3–S5 Tables. Table 1 show estimates of within clade uncorrected mean phylogenetic distances (p-dist). Overall genetic distances for Pseudophilautus are low given their slow accumulation of changes over a long period. Within-clade average distance for 16S rRNA is lowest in Clade C, while the highest average distance is recorded in Clade E.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Mean within clade uncorrected percent divergences (p-dist) over sequence pairs of mitochondrial 16S rRNA and nuclear Rag-1 gene fragments.

https://doi.org/10.1371/journal.pone.0258594.t001

In summary, out of 104 Pseudophilautus haplotypes, ABGD, bPTP, mPTP and bGMYC recovered a maximum of 60, 61, 67 and 57 potential species, respectively, for the 16S rRNA gene fragment (Fig 1). The ABGD analysis recovered a total of seven partitions with prior maximal intraspecific distances (p) ranging from 0.001 to 0.031 and number of delimited species ranging from 1 to 60. The recursive approach recovered a higher number of species compared to the initial run, except partitions one, three and four, which indicated a similar number of species. We selected the partitions that had the smallest maximal intraspecific distances below 0.002 for our interpretations to recover a maximum number of OTUs. Therefore, we use the conservative first partition for discussion, below.

At least three analyses merged the following nominal species: P. procax and P. abundus (p-dist = 0.21%); P. hallidayi and P. fergusonianus (p-dist = 0.44%); P. reticulatus and P. papillosus (p-dist = ~0.00%); P. pleurotaenia and P. hoipolloi (p-dist = 0.63%); P. hoffmani and P. asankai (p-dist = 1.04%); P. silvaticus and P. limbus(p-dist = 0.42%); P. dilmah and P. hankeni (p-dist = ~0.69%); P. fulvus and P. silus group (p-dist = 1.04%).

The analyses of 16S rRNA gene fragment recovered 8 distinct OTUs from clade A, together with three putative unnamed species (P. cf. frankenbergi WHT2729, P. cf. mooreorum WHT6306 and P. cf. poppiae WHT5051). Although the phylogenetic placement of P. bambaradeniyai was confirmed in clade A, we did not include it in our final analysis as the sequence identified for that species in Genbank was of doubtful provenance (M. Wickramasinghe, pers. comm.). Clade B contained the greatest number of previously identified taxa (29), of which 18 were recovered as seven clusters by all methods, while P. cf. popularis WHT6010, P. cf. popularis WHT6074 and P. cf. cavirostris WHT6381 were recovered as unidentified species. The highest degree of clustering (~85%) was observed in Clade C, while clade D suggests two potential synonymies (P. pleurotaenia—P. hoipolloi, and P. hoffmanni—P. asankai). About 45% clustering was evident in clade E, which recovered four unidentified species (P. cf. limbus WHT2540, Pseudophilautus sp. 4 WHT2669, P. cf. folicola1 WHT2525 and P. cf. folicola2 WHT2531), whereas clade F recovered P. cf. simba WHT3221, M5 P. cf. semiruber and M7 P. cf. semiruber as putative new species.

In general, the bPTP, mPTP and bGMYC models yielded similar results for the 16S rRNA gene fragment, with a few deviations. While mPTP differentiated P. cf. schmarda from P. hankeni—P. dilmah, other models recovered P. schmarda as conspecific with P. hankeni—P. dilmah. Additionaly, mPTP recovered P. fergusonianus and P. hallidayi; P. cf. microtympanum WHT6305 and P. microtympanum; P. asankai and P. hoffmani; P. rus and Pseudophilautus sp. WHT2797 as distinct OTUs, while all other models recovered them as single OTUs. The bGMYC model identified P. cf poppiae WHT5051 and P. poppiae; P mittermieri and P. decoris; P regius and Pseudophilautus sp WHT2669 as a single OTU, while other models recovered them as distinct. All analyses recovered the previously unnamed P. cf. mooreorum WHT6306, P. cf. frankenbergi WHT2729, P. cf. popularis WHT6010, P. cf. popularis WHT6074, P. cf. cavirostris WHT6381, P. cf. silus1 WHT2489; P. cf. silus3 WHT6070, P. cf. singu WHT2658, P. cf. limbus WHT2540, P. cf. folicola1 WHT2525, P. cf. folicola2 WHT2531, M5 P. cf. semiruber, M7 P. cf. semiruber and P. cf. simba WHT3221 as valid OTUs, which may be new species.

However, the delimitation analyses carried out on other loci (12S rRNA and Rag-1 nDNA) as well as the concatenated dataset shows that there are a few incongruencies for some of the species. Compared to 16S rRNA results, Rag-1 nDNA clustered P. reticulatus, P. pappilosus and P. silus together; synonymization of P. limbus and P. silvaticus was not supported. It was also unable to recognize P. cf. frankenbergi WHT2729 as a potential new species. However, lack of Rag-1 nDNA molecular data for many other taxa was a limitation in interpreting most of the results based on 16S rRNA data. The results based on the concatenated data as well as the data comprised only of singletons (S3 Fig) were largely congruent with the delimitations derived from 16S rRNA data.

The geophylogeny for the Sri Lankan Pseudophilautus diversification revealed complex patterns of geographical structuring (Fig 2). With narrow exceptions, members of the genus are mainly restricted to four distinct geographic regions, namely Central Hills, Knuckles Hills, Rakwana Hills and the Lowland wet zone [2]. However, the six clades identified in the phylogeny occur across these geographic regions. A few clades, however, are altitudinally stratified. Clade A, for example, was restricted to the uplands (>700m asl) while clade E includes mostly lowland (<700m asl) species. In some of the clades, sister species were distributed between adjacent mountain ranges or inhabited different parts of large mountainous regions, though local communities were composed of species from disparate clades.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Geophylogeny of Sri Lankan Pseudophilautus: The phylogeny of Sri Lankan Pseudophilautus laid across the geographic distribution of the species considered in the analysis.

Type localities are indicated with circles (see S6 Table for all known locations of Pseudophilautus). Different colors represent major clades of Pseudophilautus and the three major mountain ranges are highlighted on the overview map of Sri Lanka. A pattern where the six clades identified in the phylogeny, sharing representatives among different geographical regions, is evident. In some of the groups, sister species are distributed across adjacent mountain ranges or from different parts of large montane regions and local species communities are composed of species from disparate clades.

https://doi.org/10.1371/journal.pone.0258594.g002

Discussion

We tested the species boundaries of Sri Lankan Pseudophilautus using single-locus molecular species delimitation methods, which have found wide utility [30, 31, 57, 58]. Methods of molecular species delimitation differ from one another in a number of respects, each with its own limitations [13]. Among the most widely used methods, the generalized mixed Yule-coalescent (GMYC) and Poisson tree processes (PTP) are tree-based methods designed for the analysis of single-locus data but are often applied to concatenations of multilocus data. In contrast, ABGD is a method based on genetic distances computed from a single locus [34]. Generally, due to differences between these methods and their associated issues [40], testing with multiple methods as well as multiple genes and assessing congruence amongst them has been a widely used approach in species delimitation [51]. Nevertheless, the results of any of the methods tested should be compared and interpreted cautiously, integrating also the phenetic and ecological aspects of the concerned species [59].

The four delimitation methods we used for 16S rRNA recovered 57–67 OTUs. These include 14 previously unrecognized OTUs that were unambiguously recovered by all four methods; these may represent undescribed species. Among extant, previously recognized Sri Lankan Pseudophilautus (59 species), only 49 were represented by genetic data in the present analysis. Our delimitation criteria applied to the 16S rRNA amphibian barcoding gene suggested that 16 of these 49 nominal species in fact represent eight OTUs. Analyses based on other gene fragments, concatenated data set as well as the data set comprised only of singletons showed low levels of incongruences with the results obtained from 16S data thus strengthening the synonymy of above 16 species.

In general, the rate of evolution of mitochondrial genes is faster than of nuclear genes [60, 61] and this pattern can be seen when 16S rRNA (mtDNA) and 12S rRNA divergences are compared to that of Rag-1 nDNA (S3 and S4 Tables). As expected, the delimitation boundaries for Rag-1 nDNA are stricter than for mtDNA-based analyses, recovering only 42 species from 60 OTUs tested, too conservative to validate, given their morphological diversity.

Subject to verification based also on phenotypic criteria, we therefore propose the following putative synonymies: Pseudophilautus hallidayi [15] is likely a synonym of P. fergusonianus [62]. Pseudophilautus papillosus [14] is likely a synonym of P. reticulatus [63]. Pseudophilautus hoipolloi [14] is likely a synonym of P. pelurotaenia [64]. Pseudophilautus dilmah [65] is likely a synonym of P. hankeni [18]. For the following pairs of putative simultaneous synonyms, we act as first revisers in allocating precedence as follows under Art. 24.2 of the International Code of Zoological Nomenclature (1999): Pseudophilautus abundus [14] is likely a synonym of P. procax [14] (precedence to P. procax). Pseudophilautus silvaticus [14] is likely a synonym of P. limbus [14] (precedence to P. limbus). Pseudophilautus silus [14] is likely a synonym of P. fulvus [14] (precedence to P. fulvus). Pseudophilautus hoffmani [15] is likely a synonym of P. asankai [14] (precedence to P. asankai).

Therefore, based on the present sampling, the overall species richness of Sri Lankan Pseudophilautus would be 65 (33 species unaffected by delimitation + 8 revised species + 14 putative new species + 10 nominal species lacking genetic data, presumed valid). However, the 14 putative new species remain to be assessed phenetically in the context of the general lineage concept of species, where independent evolutionary lineages are recognized using multiple criteria [66]. The criteria that have been used hitherto to test species hypothesis for Sri Lankan Pseudophilautus are morphology, mtDNA, geographic distribution and vocalization.

Further, the validity of the remaining ten species lacking genetic data (see S1 Table) too, need to be tested using molecular approaches, as sequences for these species from their topotypes become available. These are P. nemus [14], P. bambaradeniyai, P. dayawansai, P. jagathgunawardanai, P. karunarathnai, P. newtonjayawardanei, P. puranappu, P. samarakoon, P. sirilwijesundarai [20] and P. conniffae [21].

Although our analyses recovered P. bambaradeniyai and P. cf. frankenbergi as a single entity, the original description of P. bambaradeniyayi shows it to be distinguished from P. frankenbergi by a relatively smaller body size, a dorsally convex head, a laterally truncated snout, a rounded canthus rostralis, a convex interorbital space, and the absence of fringe on fingers [20]. We therefore consider P. bambaradeniyayi to be a valid species and consider the Genbank sequence assigned to it (Accession number KP272047) a misidentification. In view of this confusion, we excluded P. bambaradeniyai from the final analysis due to unavailability of molecular data from a specimen from its type locality.

The geophylogeny highlights the geographic distribution of species pairs which are shown to be discrete OTUs by the delimitation analyses. Contiguity for these species has been established through subsequent field sampling since their original description, where new populations have come to light. Their geophylogeny shows a pattern of sister species distributed across adjacent mountain ranges, or from different parts of large montane regions, highlighting the importance of mountains as a mechanism of isolation and allopatric speciation. Hence, local communities are composed of species from disparate clades that, in most cases, have been assembled through hybridization [67, 68] or migration rather than in situ speciation [2]. Further, lineages could be incompletely isolated for millions of years after their formation. Hence, much evolution of eventual reproductive isolation can occur while nascent species are in gene-flow contact, in sympatry or parapatry, long after divergence begins [67].

Morphological investigations suggest that Clade E is a taxonomically challenging group. This clade comprises of small-bodied species, mainly from the lowland rainforests, which are difficult to distinguish. Further, there are no clear geographic barriers that seem to have isolated its constituent species. Their populations may, however, have been isolated historically but coalesced by migration or changes in the extent of suitable habitat [69]. While this clade contains some species that are difficult to distinguish by morphology alone, it also contains a higher number of species yet to be described.

A common pattern of sister species that are merged (synonymized) seem to be populations leading to a pattern of parapatry, where one of the species is widely distributed with the other being a montane isolate. This is seen between P. abundus (widespread), P. procax (montane isolate around Rakwana Hills: 1060m asl); and P. reticulatus (widespread) and P. papillosus (montane isolate around Handapan Ella: 1270 m asl). Another pattern is limited geneflow between what appeared initially to be montane isolates (P. hoffmanni and P. asankai), but later, populations were discovered in valleys connecting these mountains. The distribution of these species fits a pattern of parapatric speciation through limited gene flow.

In clade B, the grouping of the clade [P. hankeni, P. cf. schmarda and P. dilmah] suggests a possible pattern of parapatric speciation through restricted gene flow. Here, P. cf. schmarda and P. dilmah are distributed at similar elevations in different parts of the central mountains (Fig 2), while P. hankeni is confined to the adjacent Knuckles mountain range.

Further, the delimitation failure of P. hallidayi and P. fergusonianus in clade B is plausible given the marked overlap in their geographic distributions. The previously known distribution of P. hallidayi ranged from Namunukula to the mid-elevations of the Central Hills (Gadaladeniya and Namunukula), whereas P. fergusonianus had a wide range spanning Monaragala, Ginigathhena, Laggala, Pallegama, Deniyaya, Pitadeniya, Gannoruwa to Peradeniya, including both the lowlands as well as mid-elevations of the central and Knuckles hills.

Pseudophilautus cf. silus 2 in clade C seem to be species found at the edge of the distribution range of P. silus in the Central Hills; this lends confidence to our treating these populations as belonging to a single species. The recovery of P. fulvus and P. silus in clade C, as a single OTU, can be understood in the context of limited gene flow occurring among the populations of the two adjacent mountain regions, Knuckles and Central Hills, respectively, which are separated by the deep valley carved by the Mahaweli River.

Our delimitation analysis relies primarily on the 16s rRNA gene fragment, which is the most widely used barcoding marker in frogs [22]. However, future genetic sampling will help us better understand these frogs. The present framework can be used to develop and test hypotheses on distribution and speciation in Pseudophilautus.

Further, it is useful to have a standardized molecular benchmark for delimiting species for this group, in order to objectively validate species. Among the synonymized species the highest uncorrected p-distance was recovered among P. hoffmani and P. asankai (p-dist = 1.04%) for the 16S rRNA amphibian barcoding gene. This is perhaps close to the 16S threshold for pairwise species differentiation. Therefore, we propose an uncorrected 16S pairwise distance of 1.5% as a barcoding benchmark for species boundaries in new-species descriptions for Pseudophilautus, and also perhaps for rhacophorids in general.

Ideally, however, we suggest testing species boundaries using combinations of delimitation methods and criteria in the context of the PSC. We recommend that future new-species diagnoses and descriptions be based on multiple criteria, including both molecular and morphological approaches, in the context of the general lineage concept [66]. Clearly, at least some of the new synonymies we report here derive from original descriptions based on small sample size, highly localized sampling or polymorphic species. Integrative approaches may help avoid the creation of unnecessary names in future. Testing species boundaries using the general lineage concept will also help recognize young species which are separated by modest molecular divergences but which are nevertheless independent evolutionary entities.

Molecular species delimitation is a useful tool to refine the assessment of species hypotheses, which here we test in the case of Sri Lankan Pseudophilautus. For almost all species assessed using the IUCN Red List Criteria, it is criterion B (restricted distribution of a species), where Extent of Occurrence (the area contained within the shortest continuous imaginary boundary that can be drawn to encompass all the known, inferred or projected sites of occurrence, excluding cases of vagrancy) and Area of Occupancy (the area within its Extent of Occurrence which is occupied by a taxon, excluding cases of vagrancy) plays a major role in the assignment of threat categories, which lead to prioritization for conservation. Hence, a clearer understanding of operational taxonomic units provides a framework for targeted conservation action.

Supporting information

Acknowledgments

We thank the Department of Wildlife Conservation and the Forest Department of Sri Lanka for the permission to carry out rhacophorid based research over several decades.