The identity of the Sri Lankan Amblypharyngodon (Teleostei, Cyprinidae)

Abstract Morphological and molecular analyses of specimens representative of the geographic range of the cyprinid genus Amblypharyngodon in Sri Lanka suggest the presence of only a single species in the island, for which the name Amblypharyngodongrandisquamis Jordan & Starks, 1917, is available. Amblypharyngodongrandisquamis is a species endemic to Sri Lanka, distributed across the lowlands of both of the island’s main climatic zones. It is distinguished from all other species of Amblypharyngodon, including the three species recorded from peninsular India (A.mola, A.microlepis, and A.melettinus), by a suite of characters that includes a body depth of 26.9–31.2% of the standard length (SL), 42–56 scales in the lateral series (of which usually 8–16 are pored), 20–24 circumpeduncular scales, 14–17 scale rows between the origins of the dorsal and pelvic fins, a dorsal-fin height of 21.1–27.6% SL, 18–19 caudal vertebrae and an eye diameter of 22.7–30.5% of the head length. Amblypharyngodongrandisquamis differs from A.melettinus and A.mola by uncorrected pairwise genetic distances of more than 9% and 6%, respectively, for the mitochondrial cytochrome oxidase subunit 1 (COI) gene.


Introduction
The cyprinid genus Amblypharyngodon Bleeker, which has a range extending from the Indian subcontinent to Southeast Asia, is considered to contain just five valid species: A. atkinsonii (Blyth), A. chulabhornae Vidthayanon & Kottelat, A. melettinus (Valenciennes), A. microlepis (Bleeker), and A. mola (Hamilton) (Vidthayanon and Kottelat 1990). The genus is characterised by having pharyngeal teeth that are compressed, with the crown blunt and enlarged; the lateral line incomplete, with 42-79 small scales in the lateral series; seven branched dorsal-fin rays; a small maximum size (standard length, SL), ranging from approximately 40 to 150 mm; and the absence of an upper lip, barbels, and fleshy labial folds (Vidthayanon and Kottelat 1990).
The genus was first reported from Sri Lanka by Günther (1868: 202), who identified specimens from the island as Leuciscus melettina Valenciennes, in Cuvier andValenciennes 1844 (hereafter Valenciennes 1844), type locality Bombay [Mumbai], India. Jordan and Starks (1917) described a second species from Sri Lanka, A. grandisquamis (type locality "river at Colombo"), which they distinguished from A. melettinus as follows: "This species is closest to Amblypharyngodon melettinus, but is deeper and has much larger scales". Since then, the identity of the species of Amblypharyngodon inhabiting Sri Lanka has been confused, with some authors considering only A. melettinus to be present in the island (e.g., Deraniyagala 1952: 45, Senanayake 1980: 167, Pethiyagoda 1991, while others consider both A. melettinus and A. grandisquamis to be present (Pethiyagoda 2006, MOE 2012, De Silva et al. 2015. No specimenbased study has been conducted to resolve this confusion until now. Here we review Sri Lankan Amblypharyngodon using an integrative taxonomic approach (Dayrat 2005) that combines morphological and molecular data, and show that only a single species is present in the island, for which the name A. grandisquamis Jordan & Starks, 1917 is available.

Metrics and meristics
Measurements and counts follow Sudasinghe et al. (2018a) except that the lateral-line scale count is given as the number of pored lateral-line scales + the scales between the last pored scale and the base of the hypural plate. In addition, the pelvic-anal distance was measured between the origins of the pelvic and anal fins, respectively. All counts and measurements were taken on the left side of specimens whenever possible. Body measurements and head length are given as proportions of standard length; and subunits of the head as proportions of head length. Values in parentheses following a count indicate the frequency of that count. Vertebral counts and osteological descriptions are based on cleared and stained specimens following the single-staining method of Taylor and Van Dyke (1985). Osteological descriptions follow Vidthayanon and Kottelat (1990) and Conway (2011).
A Principal Component Analysis (PCA) in a covariance matrix was carried out to identify variables that best discriminate among the species. In view of one of us (HS) having measured the recent material and another (RP) having measured the historical material (the types of Leuciscus melettinus and Amblypharyngodon grandisquamis), and in order to account for the deforming of specimens through long-term storage in preservative, the respective datasets were analysed separately. Prior to the PCA, all measurements were size-corrected by using the equation in material (the types of Leuciscus melettinus and Amblypharyngod order to account for the deforming of specimens through long-ter the respective datasets were analyzed separately. Prior to the PCA size -corrected by using the equation where is the size corrected measurement, is the measured (arithmetic) mean standard length for all individuals from all pop species, and is the standard length of the specimen, while is where M s is the size corrected measurement, M o is the measured character, L s is the overall (arithmetic) mean standard length for all individuals from all populations of all putative species, and L o is the standard length of the specimen, while b is calculated for each character from the observed data by using the allometric-growth equation (Elliott et al. 1995). The software PAST (Hammer et al. 2001) was used to carry out the PCA.

Molecular analysis
The protocols for DNA extraction, PCR amplification and PCR product purification for the mitochondrial cytochrome oxidase subunit 1 (COI) and 16s rRNA subunit (16s) follow Sudasinghe et al. (2018a), while those for the mitochondrial cytochrome b (cytb) gene follow Sudasinghe et al. (2018b). Sequences were checked and assembled in ChromasPro v1.34 (Technelysium Pty Ltd) and contig sequences of the two strands were constructed using MEGA v.7.0 (Kumar et al. 2016). The 16s, COI, and cytb contig datasets were prepared and aligned separately using ClustalW in MEGA v. 7.0 (Kumar et al. 2016) and verified manually. Protein-coding genes (COI and cytb) were translated and checked for internal stop codons or frameshift mutations. Details of the specimens used in the molecular analysis, together with additional GenBank sequences used, are given in Table 1. The uncorrected pairwise genetic distances for species of Amblypharyngodon for the three genes were calculated using MEGA. For each independent dataset, 16s (497 bp) and COI (654 bp), and a combined dataset of 16s+COI (1151 bp), a Bayesian phylogenetic inference was carried out using MrBayes v3.2 (Ronquist et al. 2012). A phylogenetic analysis for cytb was not carried out due to the paucity of comparative sequences in GenBank. For the Bayesian phylogenetic inference , the best-fitting nucleotide substitution model for each gene partition (16s and COI) was selected using jModelTest v.2. 1.6 (Guindon andGascuel 2003, Darriba et al. 2012) under the Bayesian information criterion (BIC). Four Metropolis coupled Markov chain Monte Carlo (MCMCMC) chains were run for 1 million generations for each analysis with a sample frequency of 100. The first 0.1% of generations was discarded as burn-in, after determining with Tracer v1.6 (Rambaut et al. 2014). The posterior probabilities (PP) of the clades (Huelsenbeck et al. 2001) were computed by using the frequency of the remaining clades in trees that were sampled every one hundred generations. Figtree v.1.4.3 (http://tree.bio.ed.ac.uk/software/figtree) was used to visualise the trees obtained from the Bayesian analyses.
Haplotype network reconstruction for the 16s, COI, and cytb genes of the various populations of Amblypharyngodon was inferred by TCS network (Clement et al. 2002) in PopArt (Leigh and Bryant 2015). DNAsp v.6 (Rozas et al. 2017) was used to compute the nucleotide diversity (π), haplotype diversity and neutrality tests: Tajima's D (Tajima 1989) and Fu and Li's F (Fu and Li 1993).

Molecular analysis
The HKY+G model was chosen as the best-fit nucleotide-substitution model under BIC in the jModelTest for the COI dataset. In the Bayesian phylogenetic inference for the COI dataset, A. mola was recovered as the sister species of A. grandisquamis, though weakly supported (PP = 66, Figure 1A). The GenBank sequence JX260817 Amblypharyngodon sp nests within the rest of the sequences of A. mola.
The SYM+G model was chosen as the best-fit nucleotide-substitution model under BIC in the jModelTest for the Bayesian phylogenetic inference of the 16s dataset. Here, Amblypharyngodon melettinus was recovered as the sister species of A. grandisquamis (PP = 85, Figure 1C). The GenBank sequences FJ751266 as A. chakaiensis, FJ751264-65 as A. melettinus 'muriyadensis' nest with the other sequences of A. melettinus (FJ751267-68) available in GenBank.
In the combined 16s+COI (1151 bp) Bayesian phylogenetic inference, A. mola was recovered as the sister species of A. grandisquamis (PP = 89) while A. melettinus was recovered as the sister species of [A. grandisquamis + A. mola] (PP = 100).
The uncorrected pairwise genetic distances obtained for the 16s, COI, and cytb genes for the species of Amblypharyngodon is given in Table 2. Amblypharyngodon grandisquamis differs genetically from A. melettinus, A. mola, and A. chulabhornae by uncorrected pairwise distances of more than 4% for the 16s gene fragment, while the intraspecific divergence within A. grandisquamis for the same gene is only 0.0-0.4%. For COI, A. grandisquamis differs from A. melettinus, A. mola, and A. chulabhornae by uncorrected pairwise genetic distances of more than 9%, 6%, and 10%, respectively, while the intraspecific genetic divergence within A. grandisquamis for the same gene is only 0.0-1.9%. With respect to cytb, A. grandisquamis differs from A. mola and A. chulabhornae by uncorrected pairwise genetic distances of more than 7% and 14%, respectively, while the intraspecific genetic divergence within A. grandisquamis for the same gene is 0.0-4.2% (there are no cytb sequences available for A. melettinus). DZ 3292 from Remuna, Kalu basin (wet zone) differs from the other Sri Lankan samples by 3.2-4.2%, while WHT 101 from Nochiyagama, Kala Oya basin (dry zone) differs from them by 1.8-2.5%.
In the COI haplotype network, the dry zone and wet zone samples of A. grandisquamis form a shared a haplotype (H1) ( Figure 1B). A single unique haplotype (H3) occurs in the wet zone, while three unique haplotypes are recorded from the dry zone (H2, H4, H5) ( Figure 1B). Populations of A. grandisquamis (ten sequences) included 12 segregating sites and five parsimony-informative sites. The nucleotide diversity and haplotype diversity for A. grandisquamis were 0.0073 and 0.800, respectively. Tajima's D test and Fu and Li's F* test statistic were both negative (-0.16285, -0.65621) but not significant (p > 0.05, p > 0.02).
Only three haplotypes occur in the 16s haplotype of A. grandisquamis ( Figure 1D). These include a haplotype (H1) shared between populations of the dry and wet zones, together with a unique dry zone (H3) and unique wet zone (H2) haplotype. Populations of A. grandisquamis (14 sequences) included two segregating sites and a single parsimony-informative site. The nucleotide diversity and haplotype diversity for A. grandisquamis were 0.00139 and 0.582, respectively. Tajima's D test was positive (0.17874) and Fu and Li's F* test statistic was negative (-0.32441), but neither was significant (p > 0.05, p > 0.02).

Statistical analysis
The morphometric PCA of the syntypes of A. melettinus (MNHN 3812) and the paratypes of A. grandisquamis (SU 22868) clearly separate the two species into two distinct clusters, with body depth and caudal-peduncle length explaining most of the variation (Figure 2A, Table 3). Similarly, the PCA of the recent material separate A. grandisquamis, A. melettinus, and A. mola in morphological space ( Figure 2B, Table 4), with only a slight overlap between A. grandisquamis and A. mola. The identity of the two specimens identified as A. cf. mola (BNHS 560-61) is doubtful (see Discussion). Most of the variation on PC1 is explained by pre-anal length and body depth, while the variation on PC3 is explained mostly by the dorsal-fin height.    Dorsal fin with two unbranched and seven branched rays, its origin just posterior to vertical through pelvic-fin origin, its distal margin straight. Anal fin with three unbranched and five branched rays, its origin slightly posterior to vertical through origin of dorsal fin, its distal margin slightly concave. Pectoral fin with a single unbranched and 11 (1), 12 (10), 13 (2), or 14 (1) branched rays, its origin anterior to posteriormost point of opercular opening, not reaching pelvic-fin origin when adpressed. Pectoral-fin axillary lobe rudimentary. Pelvic fin with one unbranched and 7 (3) or 8 (13) branched rays, its origin slightly closer to anal-fin origin than to origin of pectoral fin, its tip not reaching anal-fin origin when adpressed. Pelvic 'axillary' scale present. Caudal fin with 9 + 8 (14) branched rays, forked, lobes rounded distally, upper and lower lobes subequal.     Vertebrae 15 + 18 = 33 (1), 15 + 19 = 34 (4). Pharyngeal teeth 5 + 2 + 1 (4), 5 + 3 + 1 (1) (Figure 5A-C). Fifth ceratobranchial with 2-4 minute foramina in addition to a large foramen at base of lateral arm; no foramen at base of medial arm ( Figure 5A-C).
Coloration. In 70% alcohol ( Figure 4B), head and body silvery brown, darker dorsally, becoming lighter laterally, off-white ventrally. Head darker than body. Duskybrown stripe 1-1½ scales wide on side of body, from immediately behind operculum, extending to caudal fin base, broader at middle, scales above it with prominent melanophores throughout, scales below it with scattered melanophores on margins, disappearing ventrally. Fins hyaline. Dorsal and caudal fins with scattered melanophores, more prominent on caudal fin.
In life ( Figure 4A), head and body silvery grey to iridescent gold, lighter laterally. Scattered melanophores on side of body. A faint yellowish stripe extending from behind operculum to caudal fin base. Caudal fin yellowish, other fins mostly hyaline.
Habitat, distribution, and natural history. Amblypharyngodon grandisquamis occurs in lotic habitats such as rivers and canals as well as in lentic habitats such as reservoirs and marshes. The species is recorded primarily from the lowland floodplain of Sri Lanka, in both the dry and the wet zones of the island (annual precipitation less than, and greater than, 2,000 mm, respectively), though much more frequently encountered in the dry zone ( Figure 6). The highest elevation from which we recorded A. grandisquamis was at ca 460 m a.s.l., in the Samanala reservoir on the Walawe River basin. It is a slow-swimming fish, usually encountered in large groups close to the surface. With the onset of the rains, adults are observed in rice paddies, probably migrating there to spawn. The relative abundance of A. grandisquamis, at least in the wet zone, appears to be seasonal, with more adults usually observed during the rainy season.

Discussion
While Menon (1999) did not consider A. mola to occur in the peninsula of India, most other recent authors (e.g., Talwar andJhingran 1991, Jayaram 2010), following Day (1878), recognise three species of Amblypharyngodon from the peninsula: A. melettinus, A. microlepis and A. mola. We therefore discuss the identity of each of these in the context of A. grandisquamis.

Amblypharyngodon melettinus
The species we here refer to Amblypharyngodon grandisquamis was first reported from Sri Lanka by Günther (1868: 202) and Day (1878: 555), as Amblypharyngodon melettinus (Valenciennes 1844). Although Günther (1868) gave the distribution of A. melettinus as "Bombay; Coast of Malabar; Ceylon [= Sri Lanka]", it is clear from the listing of material at the end of his description that he had not seen specimens from Bombay. For his part, Day (1878) had the distribution of this species as "(Bombay, according to Cuv. and Val.) Malabar coast, and southern India, from the Neilgherries to Madras", making it clear that he too, had not seen specimens from Bombay. As such, the conception of A. melettinus by these authors was based essentially on the original description of Valenciennes (1844) and specimens from southern peninsular India and Sri Lanka: neither Günther (1868) nor Day (1878) had opportunity to examine the type specimens in Paris.
The syntypes of A. melettinus are in quite poor condition. Jean-Jacques Dussumier probably collected these on his visit to Bombay 1827-30 (see Bauchot et al. 1990: 46). We have been unable to determine whether, when Valenciennes (1844) gave the type locality as Bombay, he meant the city of Bombay (now Mumbai) or the erstwhile Bombay Presidency. The latter, in the mid-19 th century, encompassed a vast territory inland of British India's Arabian-sea coastline, extending from southern Karnataka across the Sindh Province of present-day Pakistan and on to Yemen (Collen 1909). The result of this uncertainty has been that subsequent authors (e.g., Talwar and Jhingran 1991, Menon 1999, Jayaram 2010 appear to have followed Day's (1878) conception of A. melettinus as the species of Amblypharyngodon inhabiting the south-western region of the Indian peninsula, characterised by the lateral line being incomplete, with 15-20 pored scales, and with 50-57 scales in the lateral series and 4 scale-rows between the lateral-line row and the pelvic-fin origin. These counts are broadly consistent with the original description of A. melettinus, "more than fifty rows of scales along the flanks…" (Valenciennes 1844: 305) and those syntypes in which scales can reliably be counted. The latter possess 16-21 pored lateral-line scales and 48-55 scales in the lateral series, plus a further 2-3 on the caudal-fin base. While these counts overlap partly with those of A. grandisquamis (8-16, rarely up to 19 pored lateral-line scales, 42-56 scales in lateral series on body), the eight paratypes of A. grandisquamis (SU 22868) are distinguished from the five syntypes of A. melettinus (MNHN 3812) by their greater body depth (29.3-33.6% SL, vs. 23.2-28.8% SL), bearing in mind that evisceration and age have resulted in the latter specimens becoming somewhat compressed and hence appearing deeper-bodied (see Figure 7A). Nevertheless, the body depth in recently collected A. grandisquamis (N = 23) is 26.9-31.2% SL, while that in recently collected A. melettinus (N = 14) is 22.9-26.3, and the two species are easily distinguished by the deeper body of the former.
Our search for recent specimens of A. melettinus from the vicinity of Mumbai led to the examination of two specimens of Amblypharyngodon, BNHS 560-61, 44.5 and 56.8 mm SL, from Tilase on the Vaitarna River, Maharashtra, ca. 50 km from Mum-bai. These have 66 and 67 scales in the lateral series, 25 and 27 scale rows between the dorsal-and pelvic-fin origins, and 31 and 35 circumpeduncular scales. Although these meristics are suggestive of A. mola, these specimens group with A. melettinus in morphological space ( Figure 2B). We provisionally assign them to A. mola noting that Wahab et al. (2003) observed that A. mola is popularly used in aquaculture in India along with other species of fish and prawns, which may explain its presence in the Vaitarna. Wagh and Ghate (2002) too, recorded A. mola from Mula and Mutha rivers of Maharashtra.

Amblypharyngodon microlepis
In differentiating A. microlepis from A. melettinus, Bleeker (1853: 141) relied primarily on the erroneous assumption that the lateral line was complete in the latter: "It has such a great resemblance to Leuciscus melettina…", observed Bleeker (1853), "that it might be kept to the same [species]. In Leuciscus melettina, however, the lateral line goes up to the caudal fin, the scales are somewhat fewer and, at least according to the picture, the pectoral fins appear a little longer than the head." Bleeker was evidently misled by the illustration of A. melettinus that accompanied the original description (Valenciennes 1844: pl. 501, reproduced here as Figure 7C), which seems to show a complete lateral line. In the syntypes of A. melettinus, as noted also above, the lateral line extends only to the first 16-21 scales (Valenciennes did not provide a scale count in the original description); see also Figure 7A, which shows a 65.7 mm syntype (MNHN 3812), in which the lateral line consists of approximately 20 pored scales, terminating just posterior to the pelvic-fin origin.
Bleeker's (1853) original description of A. microlepis contains some information of diagnostic value, including that the lateral line terminates just anterior to the tip of the pectoral fin, the lateral-line series consisting of 60 scales; that there were 20 scale-rows in transverse line on the body; and that the dorsal-fin origin lay between the pelvic-fin origin and anal-fin origin. Vidthayanon and Kottelat (1990), however, examined four specimens from Bengal (RMNH 7043) identified as A. microlepis by Bleeker, reporting that these contained 58-65 scales in the lateral-line series, the first 8-12 of which were perforated; and had a body depth of 24.3-26.3% SL. The specimen AMS B7593, from Orissa, India, identified as A. microlepis by Day, has 55 scales in the lateral-line series. These data enable A. microlepis to be unambiguously distinguished from A. grandisquamis, which has 42-56 scales in lateral series and a body depth of 26.9-31.2% SL.

Amblypharyngodon mola etc.
Amblypharyngodon grandisquamis is easily distinguished from A. mola by possessing fewer scales in the lateral series: 42-56, vs. 69-73 in A. mola from the putatively topotypical West Bengal specimens examined by us, 65-79 in the Orissa specimens  fig. 3), laterally inverted. examined by Vidthayanon and Kottelat (1990), and 65-75 reported by Day (1878: 555) from across the species' range in India. Additionally, A. grandisquamis differs from A. mola by having a shorter dorsal fin, smaller eye, fewer circumpeduncular scales and fewer lateral scale rows (see Diagnosis).
There remain five additional nominal species of Amblypharyngodon reported from the Indian peninsula: A. chakaiensis . jerdoni, which Day (1878: 555) himself relegated to the synonymy of A. melettinus, none of these is represented by type material or accompanied by a description enabling differential diagnosis with congeners. Elucidation of their identity must necessarily await a revision based on fresh collections in India. We note, however, that Day (1878), who was last to review the Indian species of Amblypharyngodon, considered A. indicus and A. jerdoni to be synonyms of A. melettinus; and A. chitul and A. macrocephalus synonyms of A. mola. Amblypharyngodon chakaiensis, distinguished from A. melettinus by Babu and Nair (1978) in having 16 pectoral-fin rays and six scale-rows between the lateral line and pelvic fin base, has been treated as valid by some subsequent authors (e.g., Talwar andJhingran 1991, Jayaram 2010) and as a synonym of A. melettinus by others (e.g., Menon 1999). No type material of A. chakaiensis is known, and an attempt in 1996 by one of us (RP) to recollect topotypes from the Aakulam 'lake' at Chacka, a suburb of Thiruvananthapuram, was abandoned owing to the high degree of aquatic pollution. In any event, A. grandisquamis can be unambiguously diagnosed from A. chakaiensis by having only 11-14 (14) branched pectoral-fin rays (vs. 16 in A. chakaiensis), although this by itself is hardly a character by which cyprinid species can be validated.

Genetic analysis
While our genetic analysis clearly separates A. grandisquamis from its peninsular-Indian congeners A. melettinus and A. mola, it does not fully resolve the sister-group relationships of A. grandisquamis. Our COI and the COI+16s combined analyses recover A. melettinus as the sister group of [A. grandisquamis + A. mola], whereas our 16s phylogram recovers A. mola as the sister group of [A. grandisquamis + A. melettinus].
The uncorrected pairwise genetic distances for COI between A. grandisquamis and A. melettinus, A. mola and A. chulabhornae, of 9%, 6% and 10%, respectively, are consistent with the distances observed between independently validated species (Ward 2009), lending confidence to our consideration of A. grandisquamis as a valid species.
The non-significant results of the neutrality tests suggest that there has been no recent range expansion or population bottleneck in A. grandisquamis in Sri Lanka. melettinus, had been frequently recognised in the Sri Lankan literature without, however, a specimen-based study. Here, by incorporating both morphological and molecular data, we conclude that only a single species of Amblypharyngodon (A. grandisquamis) occurs in the island, which is, morphologically as well as genetically distinct from the species of Amblypharyngodon in the Indian peninsula.