Molecular phylogenetic analysis of a taxonomically unstable ranid from Sumatra, Indonesia, reveals a new genus with gastromyzophorous tadpoles and two new species

The presence of an adhesive abdominal sucker (gastromyzophory) allows tadpoles of certain species of anurans to live in fast-flowing streams. Gastromyzophorous tadpoles are rare among anurans, known only in certain American bufonids and Asian ranids. To date, Huia sumatrana, which inhabits cascading streams, has been the only Sumatran ranid known to possess gastromyzophorous tadpoles. In the absence of thorough sampling and molecular barcoding of adults and larvae, it has remained to be confirmed whether other Sumatran ranid species living in similar habitats, i.e., Chalcorana crassiovis, possesses this larval type. Moreover, the taxonomic status of this species has long been uncertain and its taxonomic position within the Ranidae, previously based exclusively on morphological characters, has remained unresolved. To study the diversity and relationships of these frogs and to establish the identity of newly collected gastromyzophorous tadpoles from Sumatra, we compared genetic sequences of C. crassiovis-like taxa from a wide range of sites on Sumatra. We conducted bayesian and maximum likelihood phylogenetic analyses on a concatenated dataset of mitochondrial (12S rRNA, 16S rRNA, and tRNAval) and nuclear (RAG1 and TYR) gene fragments. Our analyses recovered C. crassiovis to be related to Clinotarsus, Huia, and Meristogenys. The DNA barcodes of the gastromyzophorous tadpoles matched adults from the same sites. Herein, we provide a re-description of adult C. crassiovis and propose “C. kampeni” as a synonym of this species. The molecular evidence, morphological features, and distribution suggest the presence of two related new species. The two new species and C. crassiovis together represent a distinct phylogenetic clade possessing unique molecular and morphological synapomorphies, thus warranting a new genus.


Introduction
A fascinating aspect of Southeast Asian ranid frogs is that some of them possess tadpoles with large abdominal suckers. The presence of this adhesive structure has been referred to as gastromyzophory (Inger 1966). Altig and Johnston (1989) described gastromyzophorous tadpoles as an ecomorphological guild. These tadpoles are adapted to live in fast-flowing streams (McDiarmid andAltig 1999, Altig 2006). Their body profile is streamlined with an extended sloping snout. Their adhesive abdominal sucker allows them to cling to rocks even in the fast-flowing, turbulent water of cascades (Nodzenski and Inger 1990, Gan et al. 2015. The abdominal sucker occupies almost the entire ventral surface of the body immediately posterior to the oral disk; both act together to press the body to the substrate through suction. The sucker has raised thickened lateral and posterior rims that seal against the substrate; the oral disk itself is broadly expanded to almost full body width. On the ventral surface of this sucker, there are spots or bands of brown skin, i.e., keratinized epithelium, probably enhancing friction when the sucker engages with the rock surface (Inger 1985, Gan et al. 2015. The tadpoles are able to loosen the suction momentarily to drag themselves forward by action of their strongly developed jaws; algae and other organic rock overgrowth is scraped off by the jaws and keratodont rows of the oral disk while wandering over the rock surface (Inger 1966, AH pers. observ.). According to our field observations on Bornean Meristogenys tadpoles (AH unpubl.), this feeding mode restricts taxa with gastromyzophorous tadpoles to certain habitats and microhabitats: clear rocky streams with considerable water velocity and enough light reaching those rocks to form organic overgrowth for the tadpoles to graze on.
Chalcorana crassiovis (Boulenger, 1920) was originally described as Rana crassiovis Boulenger, 1920 based on two specimens (BMNH 1947.2.3.99 andBMNH 1947.2.4.1) collected from Kerinci, Sumatra, Indonesia at ~1219 m (4,000 ft.) a.s.l.. In the same publication describing C. crassiovis, Boulenger (1920) also re-described Rana pantherina van Kampen, 1910 as R. kampeni. The short original description of R. pantherina by van Kampen (1910) included a figure of one specimen. Boulenger (1920) based his description of C. kampeni on a specimen (ZMA unregistered number) collected at Bandar Baru, Batak Mts., Kabupaten (=Regency) Dili Serdang, Provinsi (=Province) Sumatera Utara at ± 900 m a.s.l.. Van Kampen (1923) later recorded another population of C. kampeni from Serepai and Sungai Kring in Kerinci. Van Tujil (1995) declared the holotype of C. kampeni as lost. Inger and Iskandar (2005) were the first to report on a large series of Chalcorana crassiovis from along the banks of Batang Tarusan, Provinsi Sumatera Barat and provide a re-description of C. crassiovis on the basis of these samples. The original description of C. kampeni was very similar to that of C. crassovis leading Inger and Iskandar (2005) to doubt the validity of C. kampeni and to conclude that it may be conspecific with C. crassiovis. Inger and Iskandar (2005) considered the larger tympanum of C. kampeni only a sexual dimorphism within C. crassiovis; judging by its small reported size (36.5 mm, van Kampen 1910), the C. kampeni type specimen was probably a male. Despite the conclusion of Inger and Iskandar (2005), the two taxa have not been synonymized and some authors have maintained the name C. kampeni and applied it to all known populations previously referred to as Rana kampeni (e.g., Frost et al. 2006, Che et al. 2007, Oliver et al. 2015. To date, no studies have included Chalcorana crassiovis (or C. kampeni) in a molecular phylogenetic context, and few have included Sumatran congeners , Pyron and Wiens 2011, Oliver et al. 2015, Chan and Brown 2017. Sound phylogenetic hypotheses based on robust sampling of the Chalcorana group remain to be proposed. This is significant given the ensuing debates over the relationships among the Asian Ranidae in recent decades. After its original description by Bou-lenger (1920) as Rana crassiovis, this species has been placed in various genera (i.e., Hydrophylax (Frost et al. 2006), Hylarana (Che et al. 2007), andChalcorana (Oliver et al. 2015)) on the basis of secondary taxonomic implications from analyses of other, putatively related taxa (Frost 2017). This past history of various placements clearly shows that C. crassovis needs to be analyzed in a larger phylogentic context amongst ranids. Phylogenetic analyses of new data have the potential to significantly contribute to the ongoing discussion and ultimately lead to more stable taxonomic amendments.
Considering the confusing and unstable taxonomic history of Chalcorana crassiovis and its relatives, it became clear that a thorough resampling and molecular analysis of cascade-dwelling frogs of Sumatra was necessary. Herein, we present our analyses of newly sampled material of C. crassiovis. The objectives of this study were: 1) to examine the phylogenetic relationships and taxonomic status of C. crassiovis and morphologically similar taxa based on new molecular data; 2) to evaluate the phylogenetic position and taxonomy of material topotypic with C. kampeni; 3) to assess material from extensive sampling along the longitudinal axis of Sumatra in an effort to elucidate the diversity and distribution of this group of frogs; 4) to assign samples of collected gastromyzophorous tadpoles to specific species based on molecular evidence.

Sampling strategy
We conducted rapid biological sampling (Ribeiro-Junior et al. 2008) at sites across Sumatra between 2013-2016. All specimens examined were collected during these sampling activities, and additional specimens were collected during 2008 and 2012 (Fig. 1). Rapid sampling entails visiting many sites but with limited time at each site in order to gather as much data as possible from as many sites as possible. This approach is cost effective and indispensable for sampling potentially cryptic species (Ribeiro-Junior et al. 2008). We collected frogs that were morphologically similar to Chalcorana crassiovis at their torrential stream habitats along with any gastromyzophorous tadpoles found in the same streams. The sampling included specimens from the reported type locality of the enigmatic taxon C. kampeni. Its type locality, when originally described as Rana pantherina van Kampen, 1910 was Bandar Baru, a village in the Kabupaten Deli Serdang, Provinsi Sumatera Utara. We collected stream frogs that are morphologically similar to C. crassiovis from the hillside streams of Bandar Baru and consider our materials (Appendix 1) topotypic to the original types of C. kampeni. The type locality of C. crassiovis is "Korinchi, Sumatra, 4,000 feet" (Boulenger 1920). Today, the modern spelling, "Kerinci" is applied to Mt. Kerinci as well as the Kabupaten Kerinci area; the original description does not provide hints as to where exactly the type specimens were collected from within that area. We visited Mt. Ker- Figure 1. Sampling localities of adult and larva of Chalcorana crassiovis specimens for this study. Black circles represent localities of specimens which were examined. White triangles represent localities of specimens which were examined and measured. Red stars represent localities of specimens which were examined, measured, and sequenced. Type locality of C. kampeni shown by number 1 (Bandar Baru), number 2 (Kerinci) for C. crassiovis. Provinces are shown by alphabet: A Aceh, B Sumatera Utara, C Riau, D Sumatera Barat, E Jambi, F Bengkulu, G Sumatera Selatan, H Lampung. Borders between provinces are represented by black lines. The map was prepared using GeoMapApp (Ryan et al. 2009). this study are deposited at one of these following museum: The Natural History Museum (BMNH), London, United Kingdom; the Museum Zoologicum Bogoriense (MZB), Bogor, Indonesia; the Zoologisches Museum Hamburg (ZMH), Hamburg, Germany; the Museum of the University of Texas Arlington (UTA), Arlington, USA; and the Museum of Vertebrate Zoology (MVZ), Berkeley, USA.
In order to uncover the true diversity of Chalcorana crassiovis, we acquired DNA sequences from tissue samples of adults (n = 20) from 19 localities across Sumatra. We selected the 20 specimens after a preliminary assessment of the qualitative morphological features of all specimens (n = 329) that were examined. Additionally, we included a subsample of four Sumatran gastromyzophorous tadpoles in the genetic analysis for identification. We inci but could not find frogs of the group in question, so our nearest samplings were approximately 10 km north and northeast of Mt. Kerinci and several localities were still within the Kabupaten Kerinci area (Appendix 1).

Laboratory protocols
We extracted DNA from tissue samples (liver, muscle) using Crystal DNA mini Kit (Biolab), PeqGOLD Tissue Kit (Peqlab), or Qiagen DNeasy Blood and Tissue Kit. We then amplified mitochondrial genes (12S rRNA, 16S rRNA, and tRNA val ) and nuclear genes (recombination-activating gene 1, RAG1, tyrosinase exon 1, TYR) for all frog samples. For tadpoles, we sequenced the 12S rRNA and 16S rRNA (which include tRNA val ) genes as barcode tool to associate them with adults. Primer information and PCR annealing temperatures applied for this study are provided in Table 1. We cleaned the PCR products using ExoSAP-IT TM and let a contractor (Macrogen, LGC, or Microsynth) sequenced the purified forward and reverse strands. We used GENEIOUS v 8.0 (Kearse et al. 2012, Biomatters Inc., www.geneious.com) to check sequence quality of both strands by comparison to their respective chromatograms, and to assemble and edit if necessary. Furthermore, we aligned sequences for each gene loci using MAFFT v7.017 (Katoh and Standley 2013, module implemented in GENEIOUS v 8.0) with default setting. We eliminated poorly aligned positions and divergent regions of an alignment of each DNA loci using GBLOCK 0.91b (Castresana 2000, Talavera andCastresana 2007) which included in the online software http://www.phylogeny.fr (Dereeper et al. 2008), with setting for a less stringent selection (allows smaller final block and allows gap positions within the final block).

Phylogenetic analyses
We ran PARTITION FINDER v.1.1 (Lanfear et al. 2012) on our concatenated dataset using Bayesian Information Criterion (BIC) to find the best models by testing a variety of models and partitioning strategies for each loci. Four partitions were proposed by the analysis: 12S rRNA, 16S rRNA, and tRNA val : GTR+I+ G; RAG1 codon 1, RAG1 codon 2, and TYR codon 1: HKY+I; RAG1 codon 3: HKY+ G; TYR codon 3: K80+G. We then employed Maximum Likelihood (ML) and Bayesian Inference (BI) to infer phylogenetic trees. To explore partitions, we constructed trees using individual loci, concatenated sequences for mitochondrial loci only, concatenated sequences for nuclear loci only, and concatenated sequences for combined mitochondrial and nuclear loci; the later was used for optimal tree reconstruction (Kluge 1989(Kluge , 2004. ML tree search included 1000 bootstrap replicates in RAXML v. 8 (Stamatakis 2014) and was performed using the CIPRES Science Gateway V 3.3 (Miller et al. 2010, www.phylo.org/sub.sections/ portal), with default parameters. We also used the CIPRES Science Gateway to find optimal phylogenetic trees with MR. BAYES v 3.2.6 Ronquist 2001, Ronquist andHuelsenbeck 2003) in two independent runs, each with four chains, and running for 50 million generations with sampling every 1000 generations. Convergence was assessed by examining all parameters and the effective sample sizes in TRACER v.1.6 (Rambaut et al. 2014) after discarding the first 25% of samples as burn in. We viewed trees that resulted from .3 (http://tree.bio.ed.ac.uk/software/figtree/) and prepared the tree in Fig. 2 using CORELDRAW X6. Nodal support with Bootstrap values (BS) ≥ 70 for ML tree (Hillis and Bull 1993) and Posterior Probability value (PP) ≥ 0.95 for Bayesian analyses (Huelsenbeck and Ronquist 2001) are herein considered as strong support (Huelsenbeck and Ranala 2004, Mulcahy et al. 2011). We also calculated genetic p-distances using MEGA 7.0.25 (Kumar et al. 2016) from 16S ribosomal subunit.

Adult and tadpole morphology
We measured a total of 175 adult Chalcorana crassiovis group frogs (males = 133, females = 42). These represent a subsample of all specimens examined (n = 329, Appendix 1). Measurements were taken with digital calipers with 0.01 mm reading accuracy. The subsample of 175 specimens included the sequenced specimens (except for MVZ271526, tissue only). Measurements were taken by UA, following current standards for morphological mea-surements of frogs (e.g., Matsui et al. 2010, Shimada et al. 2011, Waser et al. 2016, and Watters et al. 2016. All acronyms and definitions of measured distances are explained in Table 2 and illustrated in Suppl. material 2A. We determined sex by the presence of nuptial pads and vocal sacs in males, and their absence and presence of eggs, respectively, in females. We analyzed sexes separately to control for bias resulting from sexual dimorphism. We collected tadpoles from rocks in fast flowing water using a fishnet and followed the procedures suggested in Haas and Das (2011). We preserved tail tissues of the photographed specimens in either 96% ethanol or RNAlater. We fixed and stored the remaining specimens from the series in neutral-buffered formalin solution (4%). We staged the tadpoles (n = 29) according to the table in Gosner (1960). The range of Grosner stages was 25-42, with the majority of specimens at Stage 25 (n = 12). We assigned the 25 tadpoles that were not sequenced to the respective clade of the genetically examined tadpoles based on their morphological similarity. Standard measurements for tadpoles (Altig 2007, Shimada et al. 2007, Haas and Das 2011, Oberhummer et al. 2014 were taken from digital images with a calibrated digital microscope VHX5000 KEYENCE Corporation, Japan (Table 3 and Suppl. material 2B) by UA. We slightly edited all images in this study using Adobe PHOTOSHOP CS6 (contrast adjustment, background, cleanup, cropping, sharpening). We prepared image plates with CORELDRAW X6.
We followed the morphological terminology of Duellman (2001) and Kok and Kalamandeen (2008). For webbing we used the formula in Guayasamin et al. (2006). We adopted the suggestions for glands cluster definitions from Shimada et al. (2015).  . Red branches represent relationship between Clinotarsus and Huia melasma. Tadpole sequences named with specimen number_Tad_locality (province). Adult sequences named with specimen number_locality (province). MZB.AMPH.29336 and ZMH.A14197 were collected from the type locality of C. kampeni and C. crassiovis, respectively.

Phylogenetic analyses and morphology
We inferred optimal phylogenetic trees from our concatenated dataset (3611 bps) comprising all gene markers (12S rRNA+16S rRNA+tRNA val +RAG1+TYR), of which 12.16% gaps and undetermined characters state.
The best log likelihood of ML tree was -25426.240268. The tree topologies recovered from ML and BI, respectively, were identical, except for the arrangement of Clinotarsus and Huia melasma (Fig. 2). Our BI tree ( Fig. 2 right) suggested Clinotarsus to be sister taxon of the Chalcorana crassiovis group and H. melasma to be the sister taxon of H. sumatrana+H. masonii. In the ML tree ( Fig. 2 left), however, Clinotarsus+H. melasma and the C. crassiovis group were sister taxa. Based on a dataset of two nuclear markers (RAG1+TYR) and lacking C. crassiovis, Stuart (2008) suggested Clino-tarsus+H. melasma to be the sister taxon of a clade comprising other Huia species from Sumatra, Java, and Borneo, and Meristogenys. In contrast, based on a larger dataset, Pyron and Wiens (2011) identified Clinotarsus as sister taxon to H. sumatrana+H. masonii, whereas H. melasma was sister taxon to all other species in a clade comprising Huia+Meristogenys+Clinotarsus. However, all of these scenarios for the arrangement of Clinotarsus and H. melasma within ranid phylogeny had low nodal support. Consequently, we prefer not to draw any phylogenetic conclusions or recommend taxonomic amendments concerning Clinotarsus or H. melasma.
Our results further corroborate previous studies (Stuart 2008, Pyron and Wiens 2011) in that the genus Huia is paraphyletic in its current composition. Yet, our phylogenetic trees were different from these previous studies concerning other genera. For example, our trees suggest Odorrana to be more closely related to Amolops (PP = 0.98, BS = 62, Fig. 2) than to Chalcorana+Hylarana+Hydrophylax. Stuart (2008) and Pyron and Wiens (2011) presented evindence that Odorrana was as closely related to some Rana or Lithobathes, embedded in a more inclusive assemblage (including, among others, Chalcorana, Hylarana, and Hydrophylax, in current generic assignment). To corroborate that was beyond the scope of our analysis and, thus, we did not include samples of Rana and Lithobates.
Within the clade of the crassiovis-group (Fig. 2), unexpected genetic diversity was revealed along the Sumatran transect. Our phylogenetic tree showed three distinct, well supported clades within our samples that previously would have been all be assigned to Chalcorana crassiovis, i.e., Clade A, Clade B, and Clade C (PP = 1, BS = 100). These three clades showed high genetic divergence among each other (Clade A-B: 6.61-8.53%, Clade A-C: 7.46-9.59%, and Clade B-C: 7.74-8.74%, respectively, Suppl. materials 3). Clade A comprises frogs from northern part of Provinsi Aceh to the southern part of Provinsi Lampung, including samples from the type localities of C. crassiovis (ZMH.A14197) and of C. kampeni (MZB. AMPH.29336), respectively. We found no evidence, that specimens from the type locality of C. kampeni were significantly divergent genetically from the remaining lineages in Clade A (uncorrected p-distance = 2.56%). Clade B encompass samples from Aceh, Sumatera Utara, and Bengkulu provinces, whereas Clade C consists of samples from the northern part of Provinsi Aceh. Apart from clearly being genetically distinct, we also found morphological features distinguishing both Clades B and Clade C, respectively, from Clade A. The morphology of our specimens in Clade A, however, fit well the description of C. crassiovis (sensu Inger and Iskandar 2005 assuming synonymy with C. kampeni). In the expanded morphological dataset, both quantitative data (morphometric values and body ratio values) and qualitative data (e.g., skin texture and coloration, iris coloration, pattern of rear of thigh, see Fig. 3) clearly clustered the C. crassiovis specimens and their respective geographic division into Clades A-C. Morphological analyses are detailed in the taxonomic section below.
Frogs in Clade A share a similar elevational range (425-1545 m a.s.l.) and a similar habitat type (primary forest or good secondary forest) with Clade C (314-1000 m a.s.l.). Clade A also overlaps in elevational range with Clade B (1190-2033 m a.s.l.). In Aceh, we observed specimens of Clade A and Clade B at the same stream (1190 m a.s.l.), as well as frogs of Clade A and Clade C in another stream (1000 m a.s.l.). These observations suggest independent evolution occurring with the syntopic species. Two genetic samples (MZB.AMPH.29200 and MVZ271526) from Cagar Alam (=Nature Reserve) Rimbo Panti, Kecamatan (=District) Panti, Kabupaten (=Regency) Pasaman, Provinsi (=Province) Sumatera Barat, were separated by 4.05-4.90% uncorrected p-distance from their nearest relatives (Suppl. materials 3) and were sister to all other samples in Clade A (Fig. 2). Although this could be indicative of a separately evolving lineage, we could not find unambiguous morphological evidence that could separate these two with certainty from that of the remaining samples in Clade A. Some morphological features in the Rimbo Panti specimens, such as rear of thigh pattern and webbing formula ( Fig. 4) overlap with other populations in Clade A. Rimbo Panti specimens (n males = 9, n females = 3) are bigger in size (SVL males = 46.45-48.87 mm, females = 78.00-83.99 mm) com-pared to the remaining samples of this clade (SVL males = 30.30-41.75 mm, females = 40.98-77.73 mm). However, the specimens of Rimbo Panti were collected at 450 m a.s.l. whereas the smallest body size of the remaining specimens of Clade A were from 1355 m a.s.l. at Gunung Kunyit, Kabupaten Kerinci, Provinsi Jambi (SVL males = 30.03-32.81 mm). The rear of thigh of Rimbo Panti specimens is typically mottled, light on dark background (Fig. 4d). The mottling pattern varied among specimens and some specimens are similar in pattern to the specimens from other regions in Clade A. The majority of specimens in Clade A were fully webbed, except for one free phalanx on Toe IV. Six specimens from Rimbo Panti were fully webbed, and six (all males) had webbing only reaching the base of the disc of Toe IV but deeply incised). This webbing pattern is also present in other specimens in Clade A. At present we conservatively consider these differences as interspecific variation, despite the genetic distance.
Three of the four tadpoles sequenced belonged to Clade A and one tadpole belonged to Clade C. Morphological characters such as shape of the jaw sheath and number of keratodont rows showed distinct separation Clades A and C (see below) and were in accordance with the genetically justified assignment.
Adult Sumaterana gen. n. can be distinguished from Huia, Meristogenys, and Amolops by: lacking posttympanic fold (present in Huia, Meristogenys and Amolops; Yang1991; UA unpubl. data); the disc of Finger III wider or almost equal to that of Toe IV (subequal in Huia, less or equal to in Meristogenys, wider in Amolops; Yang 1991); Finger I length shorter or subequal to that of Finger II (Finger I ≥ Finger II in Huia, Finger I > Finger II in Meristogenys, Finger I ≤ Finger II in Amolops; Yang 1991); lacking an outer metatarsal tubercle (present in Huia except for H. cavitympanum, present in Meristogenys except for M. kinabaluensis; Yang 1991); tibia length relative to SVL 58.08-78.39% (> 70% in Huia and in Meristogenys; Yang 1991); furthermore, Sumaterana gen. n. differs from Huia by having a translucent but non-transparent tympanum; tympanum not encased by dark Π-shaped marking (Manthey and Denzer 2014); and dorsolateral folds less distinct or absent. Sumaterana gen. n. differs from Amolops by having diamond-shaped finger and toe tips (rounded in Amolops) and relatively smaller fingers and toe discs.
Etymology. Sumaterana is a compound generic epithet created from the Indonesian proper noun Sumatera, the Indonesian name for the island of Sumatra, and rana, the feminin Latin word for frog. Sumatera itself is named after the kingdom of Samudra Pasai, which was located along the coast of Aceh, Sumatra from the 13th to the 16th centuries CE. Samudra is a sanskrit word that means gathering of the seas, a place where the Andaman, Java, and South China seas meet the Indian Ocean. Rana, was also the very first generic name to be assigned to a member of the S. crassiovis group, endemic to the island of Sumatra.
Phylogenetic definition and content. Sumaterana gen. n. is a node-based genus that consists of three known species: Sumaterana crassiovis comb. n. (Fig. 2 Clade A, Fig. 5a), S. montana sp. n. (Fig. 2 Clade B, Fig. 5c), and S. dabulescens sp. n. (Fig. 2 Clade C, Fig. 5b), and their most recent common ancestor. Chalcorana kampeni is considered a junior synonym of S. crassiovis comb. n. based on Inger and Iskandar (2005) and the new molecular evidence. The monophyletic clade of Sumaterana gen. n. is restricted to the island of Sumatra, Indonesia. Our phylogenetic analyses and morphological examination supports these taxonomic recognitions (uncorrected p-distances in Suppl. materials 3).
Distribution and habitat. Species of Sumaterana gen. n. inhabit riparian habitats in primary or secondary forest in Sumatra, Indonesia. Inhabited streams are typically fast flowing, 5 m wide or less, dominated by big rocks (diameter > 1 m). The known elevational range is from 314-2033 m a.s.l.. Adult frogs of these genus usually perched on rocks or vegetation at the stream. Tadpoles of these frogs can be found in groups attached to the top or sides of rocks in fast moving water.  Che et al., 2007. Chalcorana kampeni Fei et al., 2010;Oliver et al., 2015. Chalcorana crassiovis Fei et al., 2010Oliver et al., 2015. Syntypes. Two adult females (BMNH1947.2.3.99 and BMNH1947.2.4.1- Fig. 7), Kerinci, Sumatra, Indonesia, 4000 feet (~1219 m a.s.l.), coll. Robinson-Kloss Expedition on the Batrachians. Based on the lack of morphological distinguishing characters (Inger and Iskandar 2005) and low genetic divergence (2.56%, Suppl. materials 3) of topotypic specimens (this study), we consider C. kampeni a junior synonym of S. crassiovis comb. n..
Coloration. Dorsal skin background green in life, with dark blotches around tubercles, lighter areas on the dorsum forming irregular network pattern; dark line connects the eye and the snout; the upper and lower lips with dark blotches on a light background; iris golden yellow, reddish anteriorly and posteriorly, with a dark netting pattern; tympanum pale brown, encircled by a dark line; flanks lighter than dorsum, lighter ventrad and with dark spots; venter whitish, throat and chest with or without dark marking; distinct cross-bars on dorsal limbs; the rear of thigh with dark vertical bars (usually a continuation from dorsal surface and separated by narrow lighter areas) or mottling (dark marking on lighter background); ventral legs are dusted with brown pigment; webbing color brown. In preservative, dorsal background light brown; flanks becoming gray; iris changed to gray.

Variation.
(1) number of tubercles on dorsum and flanks: few to dense; (2) size of tubercles on dorsum: small and round to larger and elongated; (3) dorsolateral fold absent, but row of few small tubercles form incomplete dorsolateral series, dorsal to the posterior of trunk (not in continuation of tympanic fold); (4) dorsal coloration: dark blotches on green background vary from few and isolated, to dense, and forming irregular green background network between the dark blotches; (5) flank color yellowish-green to green (as dorsum), lighter ventrad, with distinct spots; (6) upper and lower lips: whitish to greenish, with dark markings, small distinct bars to wide and connected, lip from one phalanx free of webbing to webbing reaching intercalary tubercle of Toe IV (Fig. 4e-h). See Fig. 8 for images of Sumaterana crassiovis comb. n. from different localities and for morphometric variation Tables 4-5.
Sexual dimorphism. Males significantly smaller than females. Tympanum diameter 45.27-71.68% ED in males, 33.33-48.51% ED in females. Male with distinct undi-markings absent or very thin in few individuals; (7) ventral dark markings: from none (ventral side whitish) to dark on throat and reaching venter, pale to dark; (8) rear of thigh with dark bars, complete or broken, or occasionally dark mottling on whitish/grayish background (Fig. 4a-d); (9) iris: golden to pale yellow, from faint and thin to dense and dark netting; (10) number of cross bars: 3-4 on lower arm (from elbow to wrist), 4-7 on thigh; (11) Toe IV:  vided nuptial pads, covering base of the first finger to subarticular tubercle in dorsal and medial surface, paired subgular vocal sacs, humeral glands absent.
Common name. We propose Kerinci Cascade Frogs as the common English name (to replace the old spelling in "Korinchi Frog", Iskandar and Mumpuni 2004) and Katak Jeram Kerinci as the Indonesian name.
Distribution and ecological remarks. This species is widespread on the island of Sumatra, ranging from the northern part of Provinsi Aceh to Kabupaten Pasawaran, the southern part of Provinsi Lampung (Fig. 9). Elevational range 425-1545 m a.s.l.. This species is abundant along rocky streams (usually 1-5 m wide) in primary or good secondary forest. The inhabited streams are typically rocky with boulders (usually diameter > 1 m) and with rock formations along the stream, water current velocity 0.2-1.1 m/s (Fig. 10). Males of this species commonly can be observed perching on rocks or vegetation at the stream banks. Females were rarely observed near the streams. It seems that they approach the streams only during breeding activities. Tadpoles were often found in groups, on rocks in the stream, overflown with water in cascading sections.
In life (Fig. 11a-c), dorsum light brown, orangeish anteriorly and posteriorly to eyes; trunk darker than head; tail muscle light brown with fine-orange stippling; lower flanks region whitish; lateral tail vein very obvious, including dorsal branching along myosepta; upper and lower fins mostly transparent without iridophores; iris black, with dense gold to orange iridescent stippling; abdomen whitish laterally and densely stippled with fine-orange iridophores medially; abdominal sucker mostly transparent with white iridocytes in the center. In preservation, upper side gray with dark stippling; dense-dark stippling laterally; iris black; lens gray; ventral side uniformly transparent with some grey pigments in the anterior region of snout and lateral parts.
Holotype coloration. In life, dorsum and upper head brown with scattered light spots; dark dorsolateral line from eye to groin; flanks brown lighting up ventrad, with yellowish color in the posterior region, and many round dark spots; venter yellowish, dark markings on throat up to half of abdomen; golden brown color in at the upper quarter sector of iris, the remaining parts of iris with dense red stippling on black background; a series of dark spots encircled base of upper eye lid; dark brown line from eye to nostril (along canthus rostralis) towards snout tip, not connected to counterpart at tip of snout; dark brown area between eye and tympanum; tympanum pale brown with darker spot in the center; upper lip background brown, lighter posteriorly, with dark brown spots; lower lip brown with few light spots; arm with four dark cross-bars, from elbow to wrist; dorsal face of thigh and tibia brown, each with 6 dark bars; yellow spots on groin; rear of thigh mottled, whitish and yellow spots on brown background; ventral skin of thigh dusted brown on cream background, denser on both lateral side of posterior region; webbing color brown. Color in preservative similar to life coloration; dorsum brown and markings remain the same; yellowish color on flanks and venter changed into white; iris color became gray.
Sexual dimorphism. Males smaller than females. Tympanum diameter 52.31-92.89% ED in males and 41.54-60.47% ED in females. Adult males with single, undivided nuptial pad covering base of the first finger to subarticular tubercle on dorsal and medial surface. Paired subgular vocal sacs visible, humeral glands absent.
Etymology. The specific epithet is the Latin adjective montana in allusion to the distribution of this species at high elevations of the Bukit Barisan mountain range of Sumatra.
Common name. We propose Mountain Cascade Frogs as common English name and Katak Jeram Gunung in Bahasa Indonesia.
Distribution and natural history. Only known from high elevations of northern (Provinsi Aceh and Provinsi Sumatera Utara) and mid (Provinsi Bengkulu) Sumatra (Fig. 9). Known elevation was from 1190-2033 m a.s.l.. The holotype was perching on moss on a root of a dead tree, about 120 cm above a small creek (50 cm wide), ~50 m from Camp 4.5 of Gunung Baru, Desa Ulu Seblat, Taman Nasional Kerinci-Seblat, Kabupaten Lebong, Provinsi Bengkulu (~2000 m a.s.l.). The paratype ZMH.A14194 was observed sitting on the branch, about 300 m away, at the same creek where the holotype was collected, 200 cm above the ground. Accompanying fauna included species of Rhacophorus and Philautus. Paratypes from the vicinity of Tele, Kecamatan Samosir, Kabupaten Toba-Samosir, Provinsi Sumatera Utara were collected along the stream in the rainforest with patches of coffee plantation. The two specimens of Sumaterana montana sp. n. from the stream at Marpunge, Taman Nasional Gunung Leuser, Kabupaten Gayo Lues, Provinsi Aceh were found within low vegetation in the middle of the stream, S. crassiovis were abundant syntopically. Specimens from Gunung Sibuatan, Kabupaten Karo, Provinsi Sumatera Utara were found on the stream bank about 1-4 m away from water.
Holotype coloration. In life, dorsum and flanks generally gray; scattered tubercles on the dorsum and the upper part of flanks usually embedded in dark color; lighter area of the dorsum form an irregular network; golden color with dark spot between eye and nostril; upper lip grayish-white with dark spots (right: 4; left: 4); lower lip whitish with dark spots (right: 3; left: 2); iris silver-gray with dark netting, golden orange in the upper part; tympanum gray with light spot in the center; venter, chest, and throat fully whitish; forearm with four distinct dark cross-bars; hind limbs with thick dark cross-bars dorsally (thigh: 5; tibia: 5); rear of thigh with dark mottling on light gray background; legs light brownish ventrally; webbing brown. Dorsal coloration turned from gray with dark spots into uniformly dark brown in preserved specimens; flanks remained gray, lighter ventrad; iris color changed to uniform gray; no color change in the dark markings or pattern.

Variation.
(1) dorsum generally with round tubercles, lighter spots vary from few to dense; (2) number of darkround tubercles on dorsum and flanks: few to many tubercles; (3) size of dark wound tubercles on dorsum and flanks: small to big tubercle; (4) life coloration of dorsum background: lighter grey or slightly grayish-green to dark gray; (5) iris upper sector: light yellow to orange; (6) dark netting of iris: loose to dense; (7) throat, chest, and venter with or without marking, ranging from none to marking reaching venter; (6) marking on upper and lower lip: variable in size; (9) number of cross bars on limbs: 2-4 (arm between wrist and elbow), 4-7 (thigh); (10) thickness of cross bars on limbs: variable; (11) composition of dark color on lighter background of mottling pattern on rear of thigh: dense to less dense dark pattern on lighter background. Metrics in Tables 4-5.
Sexual dimorphism. Males smaller than females. Tympanum diameter 38.54-72.94% ED in males and 28.18-45.70% ED in females. Adult males with divided nuptial pads and vocal sacs, covering dorso-medial face of proximal Finger I to level of subarticular tubercle, humeral gland absent.
Etymology. The species epithet dabulescens is an artificial construct of "dabul", "gray" in Gayo language, combined with the Latin ending "-escense", here in the sense of "tending to be", in allusion to the gray appearance of this species. The Gayo are a local tribe in the Aceh region of Sumatra and after which the Gayo highlands have been named.  (Fig. 9). Known elevation for this species was 314-1000 m a.s.l.. The holotype was caught 100 cm above water level on a rock wall at the stream slope. The paratypes were perching on vegetation above the stream (15-200 cm above water) or on rocks in the stream or at the stream bank. The other specimens were collected from rocks or vegetation either in stream or approx. 30-100 cm away from the water (e.g., Fig. 10c). Tadpoles were collected between 23:00-24:00 from rocks (diameter ~1 m) in a fast flowing stream (4 m wide), local protected forest, Kecamatan Mane, Kabupaten Pidie.
In life, dorsal coloration of body and tail densely mottled with brown and golden blotches on a grayish background with dense fine dark stippling; lower flanks with a conspicuous wedge-shaped white area; tail muscle dark with dense-dark stippling overlain by yellowish-golden to orange mottling; lateral tail vein visible in first third of tail, including dorsal branching along myosepta; upper and lower fin mostly transparent, stippled with melanophores, especially towards the fin margin; yellowish-golden stippling also present in the upper and lower fin; iris background color black, with dense golden to orange iridophore stippling; abdomen whitish laterally and densely stippled with golden iridophores medially; golden iridophores stippling also present in the anterior region of the snout and oral disc; abdominal sucker mostly transparent except for the central spot with golden iridocytes and scattered pigment along the rim. In preservative: color of dorsal region became gray with dense darker dots and dark brown mottling; darker region were obvious on the upper flanks and between eyes and naris; iris all black; lens grayish-white; ventrally uniformly transparent with dark pigments in the anterior region of snout, oral disc, and lateral.
Body proportions between Stage 25, Stage 28, and Stage 37 were variable, e.g., BW/BH in Stage  Fig. 11d-f). Color patterns were also variable among the specimens. For example, in life, MZB.AMPH.29411 (Stage 28, Fig. 11d-f) had less mottling on upper side of body and tail than MZB. AMPH.29413 (Stage 37), more extensive golden color in the iris, smaller orange blotches in the tail region, very few golden spots in both upper or lower fin region, golden iridopores and pigments were less in the ventral region. In preservative, MZB.AMPH.29411 is lighter than MZB.AMPH.29413.

Conclusive summary
The taxonomic status of the taxon previously known as Chalcorana crassiovis has been problematic for a long time. The case was confounded by the description of a morphologically similar species (C. kampeni), the loss of the C. kampeni type specimen, insufficient sampling, and a lack of evidence beyond morphology (viz., molecular data). After the original description by Boulenger (1920), only Inger and Iskandar (2015) collected substantial numbers of specimens from that taxonomic group. The exclusively morphological evidence in their validated the existence and provided a re-description of C. crassiovis, while questioning the existence of C. kampeni. Some authors still continue treating C. crassiovis and C. kampeni as distinct species, by implication of other evidence. Our study is the first to conduct molecular analyses for these doubtful taxa in a phylogenetic context. Our phylogenetic hypotheses strongly support C. crassiovis аs a distinct lineage, and a diverse, monophyletic group (Fig. 2) that is not closely related to other species previously assigned to the genus Chalcorana. Our comprehensive sampling along the Sumatran transect yielded specimens with astonishing genetic diversity and morphological differences among the clades comprising the crassiovis-group (Fig. 2). We recommend all taxa in Clades A-C to be moved to the proposed new genus Sumaterana gen. n.. The new genus currently comprises three known species: S. crassiovis comb. n., S. montana sp. n. and S. dabulescens sp. n.. We consider them to represent valid species (viz., independently evolving lineages) as indicated by genetic, morphological, and ecological differences in comparison to other related species (see above). Furthermore, our molecular data reveal the presence of gastromyzophorus larvae in the species belonging to Sumaterana gen. n..
Samples from the type localities of Sumaterana crassiovis comb. n. and "Chalcorana kampeni" were nested in Clade A in the phylogenetic analysis ( Fig. 2) with small amounts of genetic divergence (uncorrected p-distance 2.56%, Suppl. materials 3). Furthermore, Inger and Iskandar's (2005) morphological description of C. crassiovis and Boulenger's (1920) original description matched our Clade A samples well, except for small differences. For example, according to Inger and Iskandar (2005) the tubercles on the dorsum were large and rounded, but in our samples some tubercles were also elongated and variable in size. Inger and Iskandar (2005) noted Finger I equal or slightly longer than Finger II, but in our samples Finger I was consistently shorter than Finger II. This may partially be attributed to different methods of finger length comparison. Inger and Iskandar (2005) reported skin flaps on the outer phalanges of the second and third fingers. We observed flaps present on the outer phalanges of all fingers, although not all of them are movable. This character is difficult to express unambiguously in verbal form (i.e., some may consider them fringes rather than flaps) and graded character states can occur on different fingers. Thus, we do not believe our observations contradict Inger and Iskandar (2005). Based on low genetic divergence within Clade A, morphological homogeneity among samples corroborating the original description of the type (Boulenger 1920) and the re-description by Inger and Iskandar (2005), and the inclusion of topotypic specimens of both previously named taxa, we recommend "C. kampeni" be considered a junior synonym of S. crassiovis comb. n..
In this study we included four known species of Huia (H. cavitympanum-type species, Borneo; H. sumatrana, Sumatra; H. masonii, Java; and H. melasma, the mainland Asia). Nevertheless, we were unable to solve the phylogenetic problem of Huia, which has previously been considered paraphyletic (Stuart 2008, Pyron andWiens 2011). Our study perpetuates this conundrum as the type species of Huia (H. cavitympanum) was shown to be the sister taxon to Bornean Meristogenys in our analyses, rather than monophyletic with the other Huia species in our dataset. One possibility would have been to subsume all members of the assemblage (Sumaterana gen. n., Meristogenys, Huia, and Clinotarsus) under one name (Clinotarsus, the oldest available name). We did not choose this option in order to ensure taxonomic stability and because valuable biological information associated with the current monophyletic groups would be dissolved in one genus, such as island endemism (Sumaterana gen. n./Meristogenys), differences in adult and tadpole morphology and tadpole peculiarities (species of today's Clinotarsus with non-gastromyzophorous tadpoles). Because of the low support in parts of his tree, Stuart (2008) refrained from taxonomic amendments concerning Huia, and so do we. Much more effort needs to be invested to solve the perplexing phylogenetic uncertainties concerning Huia.
Another interesting subject arises from the optimized phylogenies in our analyses (Fig. 2) with respect to the evolution of larval gastromyzophory in Southeast Asian ranids. Previously all Asian ranid taxa with gastromyzophorous tadpoles were grouped under the genus Amolops (Inger 1966). Yang (1991) split the group into Amolops, Huia, and Meristogenys based on adult and tadpole morphological characters. Molecular systematic studies, however, suggested that the assemblage of Amolops, Huia, and Meristogenys was para-or polyphyletic (Frost et al. 2006, Stuart 2008, Pyron and Wiens 2011. Our phylogenetic analyses indicate that gastromyzophorous tadpoles have likely evolved independently, once in the most recent common ancestor of the group Huia+Sumaterana gen. n.+Meristogenys and again in the ancestor of Amolops. Tadpoles from both clades are perplexingly similar morphologically (Noble 1929, Gan et al. 2015), yet molecular evidence implies separate origins. Interestingly, although Clinotarsus does not possess gastromyzophorous tadpoles, this genus is nested within Huia+Sumaterana gen. n.+Meristogenys (Stuart 2008, Pyron andWiens 2011, this study). Therefore, it could be hypothesized that larval gastromyzophory might have been lost secondarily in Clinotarsus. Further studies are needed to test and understand the evolution of this larval type in these frogs.
A third case of ranids with gastromyzophorous tadpoles has been reported in Rana sauteri (Boulenger, 1909). Its tadpoles are clearly more morphologically (Kuramoto et al. 1984) and biogeographically (Taiwan) distant to Amolops, Huia, and Meristogenys. Gan et al. (2015) summarized that in R. sauteri the edge of the abdominal sucker was not as sharply defined as in Amolops, Huia, and Meristogenys (sucker is completely free and rim raised), particularly at the posterior. Moreover, the sucker seems to work differently in R. sauteri: the musculus diaphragmatopraecordialis is absent in R. sauteri, but well developed in Amolops, Huia, and Meristogenys (Gan et al. 2015, Kuramoto et al. 1984. Finally, other body features of R. sauteri (relatively narrow oral sucker and extensive dorsal tail fin) underline the morphological differences between this and to other Asian gastromyzophorous tadpoles, implying possible separate origins and different adaptive scenarios.
We are fully aware that phylogenetic and taxonomic problems persist in our studied taxa. These need to be addressed in the future. Broad thorough geographic sampling of adult and larval forms is a prerequisite to solve phylogenetic quandaries with any amphibian taxa, especially in the species rich tropical realm. Moreover, integrating independent sources of evidence (e.g. DNA, morphology, distribution) is an optimal strategy to accurately and convincingly validate the taxonomic position of doubtful amphibian taxa from hyperdiverse hotspots (Dayrat 2005, Padial et al. 2009, Padial et al. 2010. Distantly related frog species that converged onto similar morphotypes (i.e., ecomorphs) are common in tropical biodiversity hotspots (Bossuyt and Milinkovitch 2002) and can confound taxonomic decisions; examples are documented in Stuart (2008).
Our results are also further evidence that the taxonomic diversity of Sumatran frogs is still significantly underestimated (Iskandar and Colijn 2000, Stuart et al. 2006), despite the recent increase of am-phibian species described from the island (e.g. Teynie et al. 2010, Matsui et al. 2012, Hamidy and Kurniati 2015, Smart et al. 2017, Wostl et al. 2017). This also holds true for other herpetofauna, such as reptiles (Orlov and Ryabov 2002, David and Das 2003, Das 2005, Harvey et al. 2015, Wostl et al. 2016. Large scale and strategic sampling efforts are of the utmost priority in order to reveal the true faunal diversity and distribution patterns on this incredibly biodiverse island.