Molecular phylogeny and phylogeography of ricefishes (Teleostei: Adrianichthyidae: Oryzias) in Sri Lanka

Abstract Ricefishes of the genus Oryzias occur commonly in the fresh and brackish waters in coastal lowlands ranging from India across Southeast Asia and on to Japan. Among the three species of Oryzias recorded from peninsular India, two widespread species, O. carnaticus and O. dancena, have previously been reported from Sri Lanka based on museum specimens derived from a few scattered localities. However, members of the genus are widespread in the coastal lowlands of Sri Lanka, a continental island separated from India by the shallow Palk Strait. Although recent molecular phylogenies of Adrianichthyidae represent near‐complete taxon representation, they lack samples from Sri Lanka. Here, based on sampling at 13 locations representative of the entire geographic and climatic regions of the island's coastal lowlands, we investigate for the first time the molecular phylogenetic relationships and phylogeography of Sri Lankan Oryzias based on one nuclear and two mitochondrial markers. Sri Lankan Oryzias comprise two distinct non‐sister lineages within the javanicus species group. One of these is represented by samples exclusively from the northern parts of the island; it is recognized as O. dancena. This lineage is recovered as the sister group to the remaining species in the javanicus group. The second lineage represents a species that is widespread across the island's coastal lowlands. It is recovered as the sister group of O. javanicus and is identified as O. cf. carnaticus. Ancestral‐range estimates suggest two independent colonizations of Indian subcontinent and Sri Lanka by widespread ancestral species of Oryzias during two discrete temporal windows: late Miocene and Plio‐Pleistocene. No phylogeographic structure is apparent in Sri Lankan Oryzias, suggesting that there are no strong barriers to gene flow and dispersal along the coastal floodplains, as is the case also for other generalist freshwater fishes in the island.


| INTRODUC TI ON
Members of the family Adrianichthyidae, commonly called ricefishes, are small fishes inhabiting fresh and brackish waters throughout the lowlands of Southeast Asia, East Asia, and the Indian subcontinent (Parenti, 2008;Yamahira et al., 2021). The family contains two genera: Adrianichthys, whose four species are confined to Lake Poso in Sulawesi, and Oryzias, represented by 34 species (Fricke et al., 2021;Yamahira et al., 2021). More than half the known adrianichthyid species are native to Sulawesi (Mokodongan & Yamahira, 2015). Peninsular India, by comparison, has been considered to harbor only three species: Oryzias carnaticus (Jerdon), O. dancena (Hamilton), and O. setnai (Kulkarni) (Parenti, 2008). Among these, the former two are considered widespread: they are reported from lowland coastal habitats of eastern India, Sri Lanka, and Bangladesh (Parenti, 2008;Yamahira et al., 2021). In addition, the distribution of O. carnaticus and O. dancena extends to the Andaman Islands and Southeast Asia, respectively (Parenti, 2008;Roberts, 1998;Yamahira et al., 2021).
Oryzias setnai, meanwhile, is confined to the lowlands of the west coast of Peninsular India, in rivers and estuaries draining into the Arabian Sea. The phylogenetic position of O. setnai, originally assigned to the monotypic genus Horaichthys, has been ambiguous.
The morphology-based phylogeny of Parenti (2008) recovered it as the sister group of the diminutive O. uwai Roberts from Myanmar, while recent molecular phylogenies recover it as the sister group to all other Adrianichthyidae, but with a long branch (Britz et al., 2022;Yamahira et al., 2021). The latter view is supported by O. setnai possessing a unique apomorphy in having the third to fifth anal-fin rays fused into a prominent gonopodium that is more than half the length of the body. It is the only adrianichthyid species to have such a structure.
Several combinations of specific names have been variably applied to the ricefishes of Sri Lanka in the past (see Pethiyagoda, 1991).
However, the taxonomic revision of the Adrianichthyidae of Parenti (2008), based on museum specimens, recognized two species (O. carnaticus and O. dancena) from the island. The Sri Lankan series examined by Parenti (2008), however, was derived from only a few scattered localities, though the genus Oryzias is ubiquitous in the island's coastal lowlands (Pethiyagoda, 1991;. Further, although the recent molecular phylogenies of Adrianichthyidae cited above include near-complete taxon sampling, they lack samples from Sri Lanka. Recent molecular phylogenetic and phylogeographic studies of freshwater fishes of Sri Lanka have revealed interesting biogeographic patterns as well as insights into their evolutionary history . However, these studies were based principally on Cypriniformes, which are usually confined to freshwater habitats. Pethiyagoda and Sudasinghe (2021) showed that although Sri Lanka was terrestrially connected to India by the erstwhile Palk Isthmus (now submerged by the Palk Strait) for much of the Plio-Pleistocene and until as recently as around 10,000 years ago, biotic exchange with India was mediated largely by the climate of the Isthmus. Except during brief pluvial periods, it appears to have been too arid to facilitate the dispersal of freshwater organisms between the mainland and Sri Lanka.
Oryzias, however, are not entirely confined to freshwaters; they occur also in brackish and estuarine environments . In this context, a phylogeographical comparison of a widespread, saline-tolerant species such as Oryzias would enhance our understanding of the biogeography of freshwater fishes in the island.
Given the lack of obvious barriers to dispersal within the island, we hypothesize a weak phylogeographic structure in the Sri Lankan ricefishes, as observed also in some of Sri Lanka's more widespread generalist cyprinid species. Further, the morphology-based phylogeny of Parenti (2008) (Parenti, 2008).
To test these hypotheses, we sampled Oryzias from across the geographic and climatic regions of the coastal lowlands of Sri Lanka and investigate, for the first time, the molecular phylogenetic relationships, phylogeography and the ancestral-range reconstruction of the Sri Lankan species based on a dataset derived from a combination of both mitochondrial and nuclear markers.  Figure 1). Specimens were tentatively identified based on the descriptions given in Jerdon (1849: 331) and Parenti (2008). The deep-bodied specimens with no yellow-orange dorsal and ventral margins on the caudal fin were tentatively identified as O. dancena, while the shallow-bodied specimens with yelloworange dorsal and ventral margins on the caudal fin were tentatively identified as O. cf. carnaticus (see Section 4).  Phylogenetic inference for the 73-taxa dataset was carried out based on a maximum-likelihood (ML) framework using RAxML-NG (Kozlov et al., 2019). The optimal nucleotide substitution model for the dataset was determined using ModelTest-NG v0.1.7 (Darriba et al., 2020), providing each codon position of each gene as the starting subset, with model selection based on the Akaike information criterion (AIC). Statistical support for the nodes in the ML tree was determined by non-parametric bootstrapping for 1000 replicates in RAxML-NG.

| Phylogenetic analysis
The haplotype networks for cytb, nd2, and rag1 for the populations of Oryzias in the island were constructed through a median-joining network (Bandelt et al., 1999) in PopArt (Leigh & Bryant, 2015). The third codon position of the protein coding mitochondrial genes was included in reconstructing the haplotype networks.

| Divergence-time estimation
The divergence-timing analysis was carried out in BEAST 2

| Ancestral-range reconstruction
The distribution of ancestral lineages of the species of Oryzias present in Sri Lanka was reconstructed using the dispersal-extinctioncladogenesis (DEC) model of BioGeoBears (Matzke, 2013;Ree & Smith, 2008), as implemented in RASP 4.2 (Yu et al., 2020). The DEC analysis was run on the MCC tree obtained from the BEAST analysis.
We did not impose any constraints on our model, and the maximum number of areas at ancestral ranges were specified as the maximum The optimal model was assessed using scores derived from the Akaike information criterion (AIC).

| Molecular phylogeny
The ML phylogeny of the concatenated dataset of Oryzias recovered a topology similar to that of Yamahira et al. (2021). Oryzias setnai was recovered as the sister group to the remaining Adrianichthyidae, supported by a high bootstrap (>95%) value and a long branch ( Figure 2). The monophyly of the three main species groups within ricefishes, the "latipes," "celebensis," and "javanicus" clades were well supported, with high bootstrap (>95%) values ( Figure 2).
The latipes group, which comprises species from East Asia, the Indochina + Sundaland and the Philippines, was recovered as the sister group to the celebensis + javanicus group with high node support (bootstrap > 95%). The celebensis group, comprised of species confined to the island of Sulawesi, and the javanicus group, comprised of species from India, Sri Lanka, and Southeast Asia, were recovered as sister groups to each other with high node support (bootstrap > 95%).
The Sri Lankan Oryzias represent two distinct lineages within the javanicus group, which do not show a sister-group relationship   (1) DZ5178 Eluwankulama, Kala (7) Oryzias sinensis

| Ancestral-range reconstruction
The reconstruction of ancestral ranges of ricefishes was evaluated under the DEC model, with BioGeoBEARS in RASP, under two different scenarios (Figure 3). Among the two scenarios evaluated, the best model with the lowest AIC score was obtained for analysis 2 (LnL -31.56, AIC 67.51) rather than analysis 1 (LnL -40.26, AIC 84.91).
The ancestral ranges of each scenario, together with the vicariance and dispersal events, are shown in Figure 3 and Table 2.
Based on analysis 2, the most probable distribution range of the common ancestor of Adrianichthyidae was a widespread species distributed in South and Southeast Asia (Figure 3). The ancestral range of the common ancestor of the latipes, celebensis, and the javanicus groups was estimated to be widely distributed in Southeast Asia, Wallacea and New Guinea (Figure 3). The ancestral range of the latipes, celebensis, and the javanicus groups was estimated to be Southeast Asia, Wallacea and New Guinea, and Southeast Asia, respectively. For both Sri Lankan species, the ancestral range of their common ancestor was estimated to be South and Southeast Asia   Parenti (2008), in her taxonomic revision of the Adrianichthyidae, recognized two species, O. carnaticus and O. dancena, among museum specimens collected from Sri Lanka. These two appear to be morphologically similar except that O. dancena is a markedly deepbodied species, with a body depth of 24%-34% of standard length (SL), while O. carnaticus has a lesser body depth of 21%-28% of SL (Parenti, 2008). She also observed the morphological similarity between O. carnaticus and O. javanicus, which too has a body depth of 24%-30% SL; these two taxa were recovered as having a sistergroup relationship in her morphology-based phylogeny. Oryzias carnaticus can be distinguished from O. javanicus by the former having the anterior margin of the ethmoid cartilage irregular and indented anteromedially, as opposed to straight in the latter (Parenti, 2008). F I G U R E 2 Molecular phylogenetic relationships of adrianichthyid taxa based on Maximum Likelihood inference of the concatenated cytb + nd2 + rag1 (2793 bp) dataset for 73 taxa. Asterisks (*) below nodes represent ≥95% ML non-parametric bootstrap values. Scale bar represents the number of changes per site. Newly generated Sri Lankan sequences are in gray. Numbers in parentheses represent the sampling localities listed in Table 1

F I G U R E 3
Bayesian time-calibrated tree, based on the cytb substitution rate, for the concatenated dataset of cytb + nd2 + rag1 (2793 bp) for 40 taxa. Bars on the nodes indicate the 95% HPD for divergence-time estimates. Pies at each node represent the ancestral-range reconstructions of adrianichthyids, using the DEC model. Numbers below nodes refer to the node identifiers in    (Parenti, 2008;Roberts, 1998). Whether O. carnaticus occurs naturally in the Andaman Islands has been a subject of doubt (Parenti, 2008).

| Divergence-timing and ancestral ranges of Sri Lankan Oryzias
The divergence-timing analysis of Yamahira et al. (2021)   (95% HPD: 60-88 Ma). Britz et al. (2022), however, advocate caution with regard to some of the fossil calibrations used in Yamahira et al. (2021). The divergence-timing estimates in the present study were made primarily to understand the sequence of divergence of the Sri Lankan lineages from their most recent common ancestor.
Our divergence-timing estimates, using a cytb substitution rate for Adrianichthyidae, are substantially younger than those estimated by Yamahira et al. (2021). Mokodongan and Yamahira (2015) too estimated younger ages comparable to ours for the terminal nodes within the celebensis species group of the Sulawesi adrianichthyids, using only a cytb substitution rate. Based on our divergence-timing and ancestral-range estimation, two widely distributed ancestral ricefishes had colonized the Indian subcontinent and Sri Lanka twice, in two different temporal windows: one during the late Miocene and the other during the Plio-Pleistocene (Figure 3).

Lankan adrianichthyids
Previous studies exploring comparative phylogeographic patterns and genetic structure in Sri Lankan freshwater fishes focused primarily on Cypriniformes, which are obligatorily confined to freshwater habitats Sudasinghe et al., 2020aSudasinghe et al., , 2020bSudasinghe, Herath, et al., 2018;Sudasinghe, Pethiyagoda, Ranasinghe, et al., 2020;Sudasinghe, Ranasinghe, et al., 2021). In contrast, Sri Lankan adrianichthyids offer us, for the first time, an opportunity to explore the phylogeography of a widespread, saline-tolerant species. As hypothesized, we did not find any strong phylogeographic structure in the two species of Sri Lankan adrianichthyids. The pattern observed here is similar to that observed also in widespread generalist cyprinids such as in Dawkinsia filamentosa (Valenciennes), Devario malabaricus (Jerdon), Rasbora dandia (Valenciennes) and the snakehead, Channa kelaartii (Günther) in Sri Lanka (Sudasinghe, Pethiyagoda, Ranasinghe, et al., 2020;Sudasinghe et al., 2020b;. As in those species, it appears that gene flow in the island's adrianichthyids occurs freely along the lowland coastal floodplain, across which there are no physical barriers to dispersal. However, within the widespread O. cf. carnaticus, we observe two well-supported subclades. One of these, subclade 1, is a widespread lineage represented by samples from throughout the island's coastal lowlands, while subclade 2 is confined to the northern and western coastal lowlands. In some sampled localities in the northern and western coastal lowlands, representatives of both subclades occur in syntopy. The syntopic occurrence of genetically distinct mitochondrial lineages has been observed also in the Sri Lankan cyprinids Garra ceylonensis Bleeker and Pethia nigrofasciata (Günther); it may suggest that each of these populations represents historically separate evolutionary lineages Sudasinghe, Ranasinghe, et al., 2021). In contrast to O. cf. carnaticus, our samples of O. dancena derived only from the northern coastal regions of the island despite Parenti (2008) having recorded the latter species from several localities in the east and west coast as well.
It is interesting to note that at some localities in the northern coast, we recorded both species in syntopy. Pethiyagoda and Sudasinghe (2021) noted that "owing to aridity, the Palk Isthmus appears to have served more as a filter of-than as a conduit for-biotic dispersal as the Plio-Pleistocene advanced." A dearth of samples from southern India precluded us from assessing whether this holds true also for saline-tolerant fishes such as Oryzias.
Our results suggest that the systematics of Oryzias in Sri Lanka is more complex than was previously thought. Finer sampling throughout the coastal lowlands of the island accompanied by taxonomic revision based on morphological and genetic analyses will help construct a more complete picture of the identity and distribution of the adrianichthyids of Sri Lanka and South Asia. Supun Chandana for assistance in the field. We are grateful for the constructive commentary received from two anonymous reviewers and the subject editor.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analyzed during this study are included in this published article and available in the NCBI database (https:// www.ncbi.nlm.nih.gov/). The newly generated cytb, nd2, and rag1 sequences in this study are deposited in GenBank under accession numbers ON528953-ON528971, ON528990-ON529003, and ON528972-ON528989, respectively.