Phylogeographic Pattern of the Striped Snakehead, Channa Striata in Sundaland: Ancient River Connectivity, Geographical and Anthropogenic Singnatures

A phylogeographic study of an economically important freshwater fish, the striped snakehead, Channa striata in Sundaland was carried out using data from mtDNA ND5 gene target to elucidate genetic patterning. Templates obtained from a total of 280 individuals representing 24 sampling sites revealed 27 putative haplotypes. Three distinct genetic lineages were apparent; 1)northwest Peninsular Malaysia, 2)southern Peninsular, east Peninsular, Sumatra and SW (western Sarawak) and 3) central west Peninsular and Malaysian Borneo (except SW). Genetic structuring between lineages showed a significant signature of natural geographical barriers that have been acting as effective dividers between these populations. However, genetic propinquity between the SW and southern Peninsular and east Peninsular Malaysia populations was taken as evidence of ancient river connectivity between these regions during the Pleistocene epoch. Alternatively, close genetic relationship between central west Peninsular Malaysia and Malaysian Borneo populations implied anthropogenic activities. Further, haplotype sharing between the east Peninsular Malaysia and Sumatra populations revealed extraordinary migration ability of C. striata (.500 km) through ancient connectivity. These results provide interesting insights into the historical and contemporary landscape arrangement in shaping genetic patterns of freshwater species in Sundaland. Copyright: ß 2012 Tan et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. The funder had no rule in study design, data collection and analysis, decision to publish, or preparation of the manuscript.


Introduction
Genetic patterning of ichthyofauna is greatly influenced by ecological processes, anthropogenic factors [1] and geological history [2]. Therefore, understanding palaeogeographical arrangement is as essential as evaluating its contemporary genetic differentiation. During the last glacial maximum, Sumatra Island, Malay Peninsula and Malaysian Borneo (comprising of Sarawak and Sabah), Figure 1a, were bridged by the exposed lowland known as the Sunda shelf. However, melting of the ice sheet during the late glacial period submerged and covered large portions of the Sunda shelf leaving disconnected islands [3] which remain up to the present day. Postglacial invasion of sea water formed the Malay Peninsula isolating it from the Sumatra on the west by the Straits of Malacca and from Borneo on the east by the South China Sea. As a consequence, the subsequent distribution and colonization of the freshwater ichthyofauna was greatly affected as the obligate freshwater taxa found the sea water an insurmountable barrier to dispersion due to the disjunction of the geographic areas [4]. Although the regional separation events are relatively recent (approximately 10,000 years ago), the various large and isolated patches of habitats allowed for independent evolution of surviving individuals along different paths from one another. Thus, investigations of contemporary spatial genetic structuring among these isolated groups could provide insights into the ecological processes and demographic causes of the phylogeographic structuring.
The striped snakehead, C. striata (Channidae) is native to, and found naturally throughout freshwater sources across many of Southeast Asian countries and is the most widely distributed species among the snakehead members [5,6]. Within its native range, it is economically very important, both in the culture and capture sectors [6]. Due to its high value as a food fish, it has been extensively introduced outside of its native range [7] for aquaculture purposes. However, in certain instances introduction has been unintentional and due to the carnivorous and aggressive behavior of this species, it may have serious impact on endemic species, if not well managed. In Malaysia, C. striata is often known as haruan or ruan (and several other local names). This species is now commonly found in many freshwater habitats. Its wide occurrence in natural waters coupled by its well-known nutraceutical and pharmaceutical properties, has made it one of the most popular protein food sources, especially among the rural communities [8,9].
The ability of C. striata to colonize natural and artificial reservoirs [10] correlated with its extensive distribution across wide range of natural environmental conditions, suggests that the genetic variation in the wild population could be high. However, how this variation may be geographically distributed, is currently unknown. Indeed, very limited genetic information is known to date for this species. A previous study on genetic structuring and differentiation of C. striata in Malaysia using RAPD markers [11] revealed highly significant genetic structuring between the eastern and western divide of Peninsular Malaysia, but contains no record for populations from Malaysian Borneo. The genetic separation in Peninsular Malaysia was effectively marked by the main mountain range that acts as a natural divider, and this finding is in accord with reports on investigations of other local freshwater biota [12][13][14]. Hence the current study is a timely attempt to document molecular data on this channid species in the wider Sundaland region.
Matrilineal markers are potent tools for assessing genetic relationships among individuals within species due to their rapid mutation rates and consequent high genetic variability [15,16]. Multiple sequences from maternal inherited loci can serve as an informative tool that may reveal demographic changes and genealogical history from millions of years ago by tracing their ancestry back hundreds of generations [17]. The same data could also serve as a guideline for future selection of broodstock by the aquaculture sector and population management [18,19]. To date, mitochondrial DNA genes have been widely utilised as genetic markers to study the relationships of present day biological populations and to reveal their historical lineages as well as seeking evidence for the existence of ancient biogeographic barriers [20][21][22]. However, they have several disadvantages in population and phylogeographic studies; they can potentially introgress between species [23,24], are prone to selective sweeps that lead to the loss of mitochondrial diversity within populations [24][25][26] and symbiont-driven changes in mtDNA variation over space specifically in arthropods [27]. Therefore, such investigations must be carefully designed, incorporating supporting evidence from nuclear markers in order to obtain a more robust indication.
The present study was focused on the Malaysian populations from both Peninsular Malaysia and Malaysian Borneo. Two populations from Sumatra, Indonesia were also included. These areas represent a major part of the Sundaland biogeographical region. The NADH dehydrogenase subunit 5 (ND5) gene was utilized as an analytical target. The specific objectives of this study were to characterize genetic diversity at each sampling locality and phylogeographic structuring across the region with respect to natural physical barriers. The initial hypothesis was that adjacent populations share common haplotypes, but even close populations may be significantly structured according to prominent geographical barriers.

Ethics Statement
Live specimens were collected from local fishermen and wet markets. Sample locations were determined by interview and confirmed to have originated from a single source prior to collection. Clips were taken from the dorsal or caudal fin rays (approximately 0.2 cm x 2 cm) and preserved in 95% ethanol and stored at room temperature (,25uC) until use. The fish were then returned to the dealers or brought back to the Aquatic Research Centre at Universiti Sains Malaysia (USM), Penang for further research. This study has been approved by the USM Ethics Committee. All practical steps to ameliorate suffering by specimens were taken throughout this study.

Sampling Location and Collection
Random samples of individuals from a total of 22 wild C. striata populations were collected throughout its distribution in Peninsular Malaysia and Malaysian Borneo (Sarawak and Sabah) between 2007 and 2010. Populations were provisionally divided into five regions, northwest Peninsular, central west Peninsular, east Peninsular, southern Peninsular and Malaysian Borneo comprising of Sabah and Sarawak (Table 1)  Range were labeled as southern Peninsular. All populations sampled from the Malaysian Borneo states of Sarawak and Sabah were classified as Malaysian Borneo. Two populations from Sumatra, Indonesia were also included.

Mitochondrial DNA Extraction and Analysis
DNA templates were isolated using AquaGenomic TM DNA isolation kits (MultiTarget Pharmaceuticals, Salt Lake City, Utah 84116) following the manufacturer's protocol. Aliquots of purified DNA isolates were used as templates for PCR amplification of the complete ND5 gene. The primer pair L12321-Leu (59-GGTCTTAGGAACCCAAAACTCTTGCTG-CAA -39) and H13396-ND5 (59-CCTATTTTKCGGATGT-CYTG-39) [12] were used. The PCR mixture contained 50-100 ng of genomic DNA, 0.05 mM of each primer, 0.17 mM of dNTP, 1.46PCR buffer, 1 mM MgCl 2 and 1.67 U of Taq polymerase (all from iNtRON). The PCR was conducted in 30 ml total reaction volume in an MJ PTC-200 Thermal Cycler (MJ Research, Waltham, MA, USA). Amplification conditions were: initial incubation at 94C (2 min); 35 cycles at 94C (20 sec), 55C (20 sec), 72C (1 min 10 sec), final extension at 72C (5 min) and a final hold at 10C. The PCR products were visualized on a 1.7% agarose gel and stained with ethidium bromide to confirm successful amplification. The PCR product was purified using QIAGEN purification kits (QIAGEN Sciences, Maryland 20874, USA) according to the manufactur-er's instruction. At the final step, 30 ml product was eluted and 5 ml of the total elution was used to assess the quality of purified product on a 1.7% agarose gel. All purified products were sent for DNA sequencing (First BASE Laboratories Sdn Bhd, Selangor, Malaysia) and reading from both DNA strands. Multiple sequences were aligned and all unambiguous operational taxonomic units (OTUs) were compiled and edited using ClustalW implemented in MEGA 4.0 [28]. All haplotype sequences have been submitted to GenBank under accession numbers HQ384453-HQ384478 and HQ438583. DNA sequences were translated into protein to ensure accurate alignment and detection of numts, if present. The aligned sequences were then exported to Collapse 1.2 [29] to construct a haplotype datasheet. Haplotype distribution among all populations was summarized manually from DnaSP programme output [30]. Then, the intra-and inter-population variation patterns of these haplotypes were analyzed. The complete aligned dataset was analyzed for variable nucleotide sites, parsimony informative sites, number of haplotypes, synonymous and non-synonymous amino acid substitutions and nucleotide frequencies in MEGA 4.0. Haplotype/gene diversity (H d ) and nucleotide diversity (p) were calculated to describe DNA polymorphism at each sampling site using Arlequin 3.1 [31].

Evolutionary Relationships Among Haplotypes
Gene trees were constructed using Neighbor-Joining (NJ) [32] and Bayesian phylogenetic tree building methods in MEGA 4.0 [28] and BEAST v1.7.1 [33], respectively. GenBank sequences of C. micropeltes (HQ 438584) and C. argus (10251173: [34]) were included as outgroups.  evolutionary distances were used with the NJ method and the confidence levels at each node assessed by 1000 bootstrap replications [36]. Bayesian inference of phylogeny was implemented in BEAST and data files were compiled using the BEAUTi routine in the BEAST package with the following parameters: General Time Reversible nucleotide substitution model incorporating Gamma site heterogeneity model (GTR+G), a relaxed molecular clock with uncorrelated lognormal distribution [37], randomly generated starting trees with tree prior coalescent-constant size [38]. GTR+G substitution model was selected for it is an independent, finite site and generalized time reversible model while gamma allows substitution rate variation among sites in the data [33]. Age calibration or time divergence analysis was not considered in this study as it is not our prior focus and thus only gene tree with posterior probability results were discussed. The analysis incorporated 10,000,000 generations with parameters logged every 1000 generations. This analysis was run three times and the log output files were combined using LogCombiner (in BEAST package). The software package Tracer v1.4 [39] was used to visualize the performance of the analysis by checking the Effective Sample Size (ESS) values. The final target tree was viewed in FigTree [40] after summarization from a sample of trees produced in BEAST by using TreeAnnotator (in BEAST package). To view the evolutionary relationships among haplotypes, a phylogenetic network of all haplotypes was constructed by median joining calculation in Network 4.6 [41].

Defining Groups of Populations, Genetic Differentiation and Gene Flow Estimates
A spatial analysis of molecular variance was conducted using SAMOVA v.1.0 [42] to identify genetically similar groups of populations and to evaluate the amount of genetic variation among the partitions. The optimal number of groups (k) was determined based on the highest value of variance among groups (F CT ), incorporating information on haplotype divergence and geographical proximity. Subsequently, a hierarchical analysis of molecular variance (AMOVA) was conducted to infer the relative contribution to variance among groups (F CT ), among populations within groups (F SC ) and within populations. Based on the SAMOVA population structure estimate (k), the population pairwise comparison statistic, F ST that calculates relative genetic differentiation between populations was determined in Arlequin 3.1 to evaluate the significance or otherwise of differences among populations and spatial population structuring. The analysis used Kimura 2-Parameter data and statistically significant pairwise comparisons were tested with 10,000 permutations. Significant probability values were adjusted by performing the False Discovery Rate Procedure (FDR) at a = 0.05 which controls the family wise error rate (FWER), a conservative type I error rate that originates from multiplicity [43]. Haplotype-based statistics (H ST ) and sequence-based statistics (N ST [44] and K ST *) were also employed using 1000 permutations [45] in DnaSP programme [30] as additional measures of genetic differentiation. Using the same programme, gene flow estimates (Nm) based on both haplotype-based and sequence-based statistics were derived as in [46] and [45], respectively. Genetic distances between populations were calculated based on Kimura 2-Parameter distance method as implemented in MEGA 4.0.

Results
A total of 280 individuals from 24 populations were successfully PCR amplified for mtDNA sequence variation in the ND 5 gene. A final truncated target length of 1017 bp was obtained after alignment and editing of ambiguous sequences. Sample size varied from five to 19 individuals per population with an average of 12. The final alignment of sequences revealed a total of 53 segregating sites (54 mutation sites and 22 parsimony informative sites) defining 27 putative haplotypes with 14 (51.9%) of them being private haplotypes (Table S1). The average nucleotide composition was 28.4% A, 27.6% T, 30.3% C and 13.7% G. Nucleotide substitution rate was approximately 1:3.9 transversion to transition, occurring with a ratio of 2.3:1:14.3 at codon position 1, 2 and 3 respectively. Out of a total of 339 amino acid sequences, nine amino acid mutations occurred as a result of nucleotide substitutions at the various codon positions (5 at the first codon position, 3 at the second and 1 at the third). This is typical of ND5 gene that is known to accumulate more informative protein variation at the first and second codon positions [47]. As an illustration, Hap25 from collection location LG had 18 polymorphic sites (see Table S1) which resulted in two amino acid substitutions.
Closer observation revealed that these amino acid substitutions were mainly regional specific, if not population specific. Thus, Hap01, Hap05, Hap19, Hap21 and Hap23 were specific to the six northwest Peninsular populations (TT, JN, KN, SP, TK and KR) ( Table S1 &

Genetic Diversity within Population
Overall haplotype diversity was 0.960.006. In general, genetic diversity was the highest in central west Peninsular. Here, TR was most polymorphic with 20 variable sites, three being singletons ( Table 2). This was followed by KJ (16), KK (14) and SG (12) polymorphic sites. Number of haplotypes in each population ranged from one to eight with a mean of 2.42 per population where TR possessed the highest number of haplotypes while nine populations showed a total absence of genetic variation. Hence, a wide range of within population genetic variability was observed (H d = 0.00 to 84.62% and p = 0.00 to 0.65%). In common with findings made using other variability measures, nucleotide diversity was highest in SG (0.65%), followed by KJ (0.58%), TR (0.57%), JN and KK (0.39%).

Evolutionary Relationships among Haplotypes
Gene trees inferred by NJ and Bayesian tree clustering methods recovered the same topology. All 27 haplotypes could be assigned to three major lineages marked by more or less robust bootstrap values (.70%) as shown in the NJ tree ( Figure 2). However, relatively low bootstrap values/posterior probabilities were recovered for Clade II and III (58 and 49%, respectively) while the internal node of Clade III showed bootstrap values higher than 80%. Clade I consists of haplotypes from northwest Peninsular Malaysia and is the sister taxon to Clade II that clustered haplotypes in populations from central west Peninsular and Malaysian Borneo (except SW). Clade III consists of haplotypes mainly from east Peninsular, southern Peninsular, Sumatra and SW. The genetically distant haplotype, Hap25, from the LG population forms the basal group in the NJ tree.
The haplotype network diagram ( Figure 3) illustrated that Hap01, Hap02 and Hap03 were the dominant haplotypes in Group 1. northwest Peninsular, Group 2. central west Peninsular and Malaysian Borneo (except SW) and Group 3. east Peninsular, southern Peninsular, Sumatra and SW, respectively. The ubiquitous haplotypes (Hap02, 05 and 06) were shared by all three groups while another three (Hap01, 08 and 16) were shared between any two of the three groups. Homoplasy, as indicated by multiple substitutions of nucleotide at a single site [48] was detected between Hap03 (east Peninsular) and Hap27 (Sumatra).
Interestingly, there were some slight discrepancies as observed in the assignation of Hap11 and 15 using gene tree and network diagram. In the gene tree, they were grouped closely to Clade III while they seemed to be closer to Clade II in the haplotype network diagram.

Defining Groups of Populations, Genetic Differentiation and Gene Flow Estimates
In the SAMOVA analysis, increasing the number of k clusters from 2,k,20 directly increased the values of variance among groups (F CT ) suggesting that the populations were highly structured. Therefore, k cluster grouping based on the highest F CT was not very useful to assess population relationships. Thus, based on the phylogenetic NJ analysis, k clusters value was set at 3 ( Figure 2, Table 3). Genetic variation was largely distributed among groups (F CT = 64.59%), followed by within populations (19.95%) and finally between populations within groups (F SC = 15.45%). The grouping obtained was precisely the same as that defined by the previous phylogenetic analysis.
Based on pairwise F ST values, 89.5% of the comparisons showed significant population differentiation (p,0.05) after FDR adjustment (table not shown). Out of 276 possible comparisons, three of the pairwise estimates were negative. Adjacent populations in the same group (as defined in SAMOVA-table 3) generally had closer relationships with each other (even though the population pairs were themselves fairly well differentiated compared with those between groups. Interestingly, however, pairwise F ST revealed that the KJ (central west Peninsular) population was closely related to the Malaysian Borneo population, SS, while the KP population (Sumatra) was not differentiated from SG (east Peninsular) and nor from two populations in southern Peninsular, LG and MS. These observations were in agreement with the gene trees which divided the populations we investigated into the three clades.
Genetic differentiation estimates for all populations based on haplotype-based statistics (H ST ) and nucleotide sequence-based statistics (N ST and K ST *) were high (0.67 to 0.72) and significant (p,0.001), consistent with the F ST statistic results. Gene flow estimates among populations were low (Nm = 0.25 and 0.20, haplotype-based and sequence-based, respectively).
Mean genetic distances between populations computed based on Kimura 2-parameter values ranged from 0.00%-1.20% (table not shown) with an overall mean distance of 0.62%. Comparisons of pairwise F ST between groups showed the highest differentiation between groups 1 and 3 (74.35%) followed by between groups 2 and 3 (58.86%) and the lowest between groups 1 and 2 (57.63%). This pattern was further supported by the divergence analyses. All pairwise comparisons were significantly structured in F ST at 95% confidence level.

Genetic Diversity
The SG, TR, KJ and LG samples were shown to have come from among the most highly variable populations ( Table 2) i.e. those that significantly contributed to the total genetic diversity in C. striata populations within the regions studied. Overall, relatively high haplotype diversities were observed in most C. striata populations (Table 2), a phenomenon often noted in freshwater fishes inhabiting non-glaciated regions (during the past glaciations era) or temperate regions [49,50]. In support of this, an earlier study of C. striata populations mainly from the Mekong and Chao Phraya rivers found them to exhibit relatively high population haplotype diversity (0.97) inferred from mtDNA cytochrome b gene sequence data [51]. Similar high genetic diversities have been observed in other freshwater fish populations; the cyprinid, Acrossocheilus paradoxus (H d = 1.00) inferred from multiple mtDNA segments [52], four Hawaiian freshwater fishes, Lentipes concolor, Stenogobius hawaiiensis, Sicyopterus stimpsoni and Awaous guamensis (0.47 to 0.98) inferred from both coding and non-coding regions of mtDNA [53]. In contrast, nine C. striata populations showed a total absence of genetic variation, which was not simply related to sample size. Several of the smaller populations (e.g. such as JN, N = 6, H d = 0.6) harbored higher genetic variation than some of the larger ones (N.10, e.g. TK, N = 15) which showed monomorphism or lower haplotypic variability (Table 2). Thus, the observed monomorphism may just be population specific. Being a commercially important fish could lead to a small effective population size as a consequence of one or, more likely, several plausible factors such as overexploitation [55], habitat fragmentation [56] or habitat loss due to environmental perturbation including human activities [57] resulting in genetic bottlenecks that may have led to inbreeding [54].

Phylogeographic Structure
The mtDNA ND5 gene data defined the populations into three major lineages (Table 3, Figure 1) i.e. 1)northwest Peninsular, 2)central west Peninsular and Malaysian Borneo except SW, and 3)east Peninsular, southern Peninsular, Sumatra and SW. However, the less robust support for Clade II and Clade III could possibly be due to insufficient time for lineage sorting to complete. Hence, a better support for the inferences could be obtained by utilizing multiple genes with different modes of inheritance (58- Malaysia. Yet, when constructing the gene tree, both Hap11 and 15 seemed to be related closer to Clade III., i.e. southern Peninsular, east Peninsular, Sumatra and SW. This is not surprising as both haplotypes were found in KJ, a population located at the border between Clade II and III. The close genetic relationship could be explained as the result of gene flow between these two clades. Ancient river connectivity. Our study provides yet further evidence of historical events associated with the palaeo North Sunda River system during the Pleistocene era [3] as highlighted by the genetic distinctiveness of the SW population (west Sarawak) from other Malaysian Borneo populations in addition to its genetic proximity with several populations in southern Peninsular, east Peninsular Malaysia and Sumatra (Figure 1). A study of the river catfish, Hemibagrus nemurus within the Southeast Asian region has shown a genetically and morphologically characteristic form of H. nemurus is found in the Kapuas River (west Borneo) as distinct from other Sarawak populations (Sadong River at Serian and Rajang River at Kapit and Sibu) [61]. This finding was attributed to ancient isolation of these two regions during the Pleistocene Epoch. However, the genetic affinities of our SW C. striata population with the southern Peninsular and Sumatra populations suggest ancient drainage connectivity, possibly via the North Sunda River system. This ancient drainage was believed to have connected many rivers of Sumatra and Borneo, and thus having a major influence on freshwater fish dispersal between these two regions [62]. Based on the cytochrome b gene, haplotype sharing pattern noted between H. macrolepidota populations from southern Peninsular Malaysia with southern and western Sarawak populations has been taken to reflect recent geographic isolation between the two regions [63]. In another study, the sharing of haplotypes in Tor tambroides between Sarawak and Perak (central west Peninsular) populations, inferred from partial COI gene was hypothesized to be a consequence of a historical drainage connection during the last Pleistocene glaciation period [64]. Furthermore, genetic similarity found in samples of the freshwater species, Barbonymus schwanenfeldii from Peninsular Malaysia and Sarawak inferred from mitochondrial cytochrome b gene sequence data was in concordance to the separation of the Borneo Island from mainland Peninsular during the late Pleistocene [4]. In our study, when the partition statistics, k, was set at a value of 3, the SW population (west Sarawak) showed closest relationship to the populations belonging to east Peninsular, southern Peninsular and Sumatra though they are presently separated by a great expanses of sea namely the South China Sea and Straits of Malacca respectively. This mountain acted as an effective barrier to gene flow, which would have constrained short-range migrations between populations separated by this divide. The other significant structuring was observed between populations from the central west Peninsular and east Peninsular Malaysia (Clade III). The main mountain range (Titiwangsa Mountains) which was formed during the late Triassic (200mya) had acted as a natural divider to terrestrial as well as riverine biota since its formation. A similar pattern of separation has been reported by several researchers investigating freshwater fishes within this region; the freshwater cyprinid, Labiobarbus leptocheilus [65]; the marble goby, Oxyeleotris marmoratus [12], the river terrapin Batagur baska [13] and the climbing perch, Anabas testudineus [14]. Not unexpectedly, the three central west Peninsular Malaysia populations; TR, TP and KJ were significantly differentiated from all of the east Peninsular populations (KB, KT, BJ, SG, TL and KK) and of the southern Peninsular populations (LG, YP and MS) too. More interestingly, these same central west Peninsular populations were also closely related to the Malaysian Borneo (except SW); specifically SS, SB and KS (Group 2). However, while ancient connectivity was the likely reason for the close relationships between west Sarawak, east Peninsular, southern Peninsular and Sumatra populations as discussed in the previous section, evidence for such connectivity has never previously been reported between central west Peninsular and Malaysian Borneo (except SW) populations.

Human-mediated translocation involving central west Peninsular and Malaysian Borneo (excluding SW)
populations. Lack of genetic differentiation between two adjacent populations is not unexpected but when two distantly located populations are found to be homogenous, then historical ecology and demographic explanations may be important factors to consider when trying to understand the lack of genetic patterning. As discussed above, based on similar investigations of other freshwater species in this area several of the unexpected findings of this study may be attributed to ancient connectivity. The native range of C. striata as reported by the United State Geological Survey (USGS, 2011) shows that this species is not indigenous to most of the eastern part of Sarawak and the state of Sabah. Thus, in this case data for populations SS, SB and KS strongly suggest human mediated translocation as the reason for the presence of these populations in these regions. However, there has been one report documenting C. striata as native to Sabah [62] and which now needs to be further investigated. Nevertheless, the special capabilities of the species to breathe air and stay alive during shipping [7], in addition to its attractive economic properties, has made it a popular species for introduction to other non-native areas. Shipping of live C. striata from Borneo to Singapore took place back in the 1950s as well as by unintentional translocation due to human activities in the past [66]. Unfortunately, there is no record of the fish being transported from Peninsular Malaysia to Sabah and Sarawak.However, as yet, there is no wider support for the alternative ancient river hypothesis to explain connectivity between the central west Peninsular and Malaysian Borneo particularly Sabah i.e. evidence for any such historical link is presently absent [3,4]. There is some circumstantial support; the reduced genetic diversity observed at these sites compared with its genetically close relatives in central west Peninsular populations is concordant with a founder effect. The colonization of newly created habitats by a low effective population number of founders is concomitant with a high rate Table 3. SAMOVA analysis on C. striata populations inferred from mtDNA ND5 gene.
of inbreeding and hence later characterized by a low genetic variation in the introduced population [67], exactly as is observed in the Sabah populations. Furthermore, the haplotypes found in populations of SB and KS (both from Sabah of Malaysian Borneo) were also a subset of the haplotypes found in the KJ population. In short, human translocation whether intentionally or otherwise seemed to be the most plausible explanation for the presence of related haplotypes separated by formidable marine divide.

High
genetic structuring among C. striata populations. A high level of genetic structuring was apparent among the populations of C. striata in this study as revealed by haplotype-based (H ST ) and sequence-based (N ST and K ST *) statistics. The population pairwise F ST values are also consistent with the low degree of gene flow estimates between populations (Nm = 0.25 and 0.20 respectively for haplotype-based and sequence-based). High genetic structuring particularly of nonmigratory freshwater fishes have been well documented [4,68,69]. This is due to restrictive physical barriers separating populations between pairs of adjacent populations, though C. striata is wellknown to be capable of short distance migration over land [10]. Therefore, when this occurs adjoining populations may effectively become panmictic, sharing or exchanging alleles as observed in this study among populations within regions.
The ancient Siam River connectivity. Another interesting finding in this study was the private unique haplotype, Hap27, from the KP population, Sumatra. Despite geographically distant separation, KP was grouped with the SG population (east Peninsular) and LG and MS (southern Peninsular); see Figure 2 and due to similarities between the sequences of Hap 27 and Hap 03. This is another probable indication of connectivity between Sumatra and east Peninsular Malaysia during the last glacial period when the sea level was lower than the present day. The existence of an ancient Siam River system has been hypothesized as ''a large river system that included Sumatra's Kampar River that ran through Straits of Singapore… likely joined branches from the Gulf of Thailand… and must have included major contributions from Endau River, Pahang River, Terengganu River and Kelantan River of the east coast of Peninsular Malaysia'' [3]. When the interconnecting river system between the Sumatra and Peninsular Malaysia was submerged by rising sea level, lateral escapes and recolonization inwards into east Peninsular Malaysia occurred as revealed by the present study.
In support of this account, another population from Sumatra, CS was also found to be clustered in the same group, further illustrating the close genetic relationship between populations from these two regions. Similar close genetic relationships were also observed between populations of the climbing perch, Anabas testudineus from Aceh (Sumatra) and Terengganu (east Peninsular Malaysia) [14]. The genetic proximity of these two regions was suggested as being due to a common origin, most probably through natural migration via the palaeo river system which traversed across the two regions during the most recent Pleistocene glaciations.
During unfavourable periods of marine invasion, local adaption or migration for survival even beyond atypical migratory range (approximately 500 km as measured along the ancient Siam River System) was the only means of escape for this C. striata population leading to the pattern of genetic diversification which still persist until today. Though the populations from east Peninsular harboured higher genetic variation, the lower genetic variation detected in the KP population might be due to demographic bottleneck. Further investigation on samples from Sumatra should be undertaken to elucidate the phylogenetic relationships of this group.

Conclusions
Genetic patterning among populations of the striped snakehead, C. striata in the Sundaland was highly influenced by the interplay of several factors. The most prominent finding of this study was the segregation of this obligate freshwater species into three highly structured and significant phylogenetic groups, limited by effective geographical barriers and thus low gene flow between each population and phylogenetic region. Anthropogenic activity may also have played a major part via the probable translocation of this species from central west Peninsular Malaysia to as far away as Malaysian Borneo as detected here using a maternal lineage marker. Furthermore, ancient dispersal through the palaeo river system running across two presently isolated regions was also apparent indicating the typical Pleistocene glacial signature of the genetic structure of freshwater fishes in the Sundaland. This study has indirectly revealed the dispersal power of C. striata when dispersal limitation was broken down and its high mobility and rapid adaptability into a newly colonized area.

Supporting Information
Table S1 Haplotype frequency and nucleotide polymorphic sites encoded by 27 haplotypes. Non-synonymous amino acid substitution is indicated by * (DOCX)