Introduction

Understanding the processes sha** the distribution and population structure of freshwater-restricted fauna has become a popular topic in biogeographic study with significant implications for conservation management. Since these organisms complete their entire life cycle in a freshwater environment, inter-drainage dispersal through the sea and over land is not possible (Burridge et al., 2008). However, many of these organisms are wide-spread, inhabiting drainages separated by land and marine barriers. Their broad distribution has often been attributed to significant re-arrangements in the drainage. For instance, river capture is believed to effectively influence the dispersal among populations of freshwater fish, Galaxias sp., in previously isolated drainages in New Zealand (Burridge et al., 2008). While historical river capture events are not frequently documented, sea level fluctuation and climate change have been subjected to intensive study and could also shape the demography and connectivity of freshwater fish populations. Temperature reduction during the Last Glacial Maximum in the Pleistocene epoch, for example, is argued to have been responsible for population bottlenecks in many freshwater fish populations (for example, Iberian Squalius, Almada and Sousa-Santos, 2010). Simultaneously, the sea level regression associated with Last Glacial Maximum exposed a confluence of drainages that are currently isolated by marine barriers, and thus allowed the dispersal of fish among otherwise isolated drainages (Swartz et al., 2007). While these geological and climatic changes could facilitate inter-drainage dispersal of freshwater fauna, their effects are often intermittent; hence fine-scale population genetic structure is expected to be prevalent in obligate freshwater species today (Dudgeon et al., 2006). To date, the impact of past geological and climatic events on inter-drainage dispersal has mostly been investigated in major river systems over a broad geographic scale (for example, Huey et al., 2013), and such studies have been very limited in East Asia. A phylogeographic study on the catfish Hemibagrus guttatus suggests the historical sea-level fluctuations might have shaped the genetic structure and led to fragmentation among its populations in the Pearl River (Yang and He, 2008). However, without sufficient geological evidence illustrated in this study, the timing and processes of diversification of H. guttatus in this area remain unclear. In-depth studies on the fine-scale phylogeographic pattern of freshwater fish would shed light on the evolutionary history in this region, which harbours very high biodiversity, much of which is unexplored and yet under imminent threat due to urbanisation and anthropogenic activities. If the drivers and processes of fine-scale population differentiation are found to be common and operate at a similar magnitude as on a larger geographical scale, it would imply that the extent of freshwater fauna diversity in South China may be much higher than expected.

Hong Kong is a small territory (1140 km2) located in South China with a well-studied geological history. A high diversity of freshwater fish (more than 160 recorded species) can be found within a dense fluvial system, which presents an ideal system for fine-scale phylogeographic study aiming to assess the significance of paleo-climatic and geological events on inter-drainage dispersal of freshwater fish. The city consists of the ‘Mainland’ (Kowloon Peninsula and the New Territories), two large islands (Hong Kong Island and Lantau Island), and over 200 small islands (Figure 1). Most of the streams in Hong Kong, though closely located, have independent headwaters and form isolated drainage systems that are compact and short, with low stream order (Fyfe, 2000). As such, freshwater fauna is expected to exhibit low contemporary inter-drainage connectivity on a small geographic scale. Indeed, recent population genetic studies of freshwater shrimp, Caridina cantonensis, revealed strong genetic divergence and a low level of gene flow among populations within and among drainage systems (Tsang et al., 2016). The freshwater goby, Rhinogobius duospilus, which is a sedentary stream fish (Ho and Dudgeon, 2015), showed even stronger population differentiation and the pattern of population structure follows the Late Pleistocene paleodrainage pattern (Wu et al., 2016). However, with only two ecologically different and phylogenetically distant freshwater species examined in this region, it is unclear whether the observed patterns are common in multiple freshwater taxa in Hong Kong, and whether there were general historical causes that similarly shaped the population genetic structure of local stream inhabitants.

Figure 1
figure 1

Current drainage pattern (blue) and late Pleistocene paleodrainage pattern (grey) of Hong Kong showing the sampling localities of S. fasciolata and P. myersi. Localities are marked as squares (for both species), circles (for S. fasciolata only) and triangles (for P. myersi only). Two major geographical barriers, Tai Mo Shan and Ma On Shan, are indicated as dotted lines representing the ridge of the two mountain ranges. A full color version of this figure is available at the Heredity journal online.

Full size image

The unique geological settings and past sea-level changes may have played a significant role in sha** the population structure of freshwater fish, with some enhanced but some suppressed inter-drainage dispersal. The South China Sea level rose continuously from the early Miocene (23.03–15.97 mya) to the late Pliocene (3.60–2.58 mya) (Wang and Li, 2009; Ding et al., 2013; Zhou et al., 2015). Hence, the two main islands, Hong Kong and Lantau, were isolated from the Mainland for a prolonged period of time during a high sea level stand (Fyfe, 2000; Ding et al., 2013; Zhou et al., 2015), and so were their freshwater fauna on the islands from the Mainland counterpart. During the late Pleistocene (0.126–0.0177 mya), the sea level fluctuated rapidly with several major regressions recorded (Zong et al., 2009). The sea level in the South China Sea dropped from −80 to −130 m, and a large part of the continental shelf near the Pearl River mouth to the northern South China Sea was intermittently exposed during the late Quaternary glacial period (Eroglu et al., 2012). The two main islands thus were connected for only a short time to the Mainland, potentially permitting dispersal between the Mainland and the island populations. Furthermore, on the Mainland, the regressions allowed isolated drainages in Pat Sin Range and Ma On Shan to form a common confluence at the now submerged northeast basin (Tolo Harbour) before draining to the South China Sea (Figure 1). Therefore, the freshwater fish populations from the northeast possibly exhibit a higher level of historical gene flow than those from other regions of Hong Kong.

This study aims to understand the association between the evolutionary history of benthic freshwater fish and past climatic and geological changes in Hong Kong. The target species are two loaches, Schistura fasciolata and Pseudogastromyzon myersi. They are small benthic balitoroid loaches (TL<10 cm, Ho and Dudgeon, 2015) that widely coexist from middle to upper streams in Hong Kong and obligate to freshwater (Dudgeon, 1992). There are, however, subtle differences in their ecological traits: S. fasciolata is commonly found in East Asia streams on both stony and sandy substrates and mainly feeds on benthic micro-invertebrates and algae, while P. myersi is an algivorous loach endemic to the Guangdong province of mainland China that prefers streams with a stony substratum and abundant algal supply (Lee et al., 2004; Yang and Dudgeon, 2009). A recent mark-recapture study of these species in Hong Kong suggests that these species are highly sedentary, with a high recapture rate (62% of P. myersi and 67% of S. fasciolata) comparable to that of the freshwater goby R. duospilus (70%; Ho and Dudgeon, 2015). By analysing the phylogeographic patterns based on sequence variation in the mitochondrial control region (CR), the nuclear interphotoreceptor retinoid binding protein gene 2 (IRBP), and an exon-primed intron-crossing marker from ribosomal protein S13 (R13), we tested (1) if past climatic and geological events have exerted similar effects on the phylogeography of these sympatric species, and (2) if their highly sedentary nature confers a similar level of fine-scale genetic differentiation in these species.

Materials and methods

Specimens for S. fasciolata and P. myersi were collected from 33 sites in Hong Kong between 2013 and 2016 using hand nets and fish cages from 27 and 13 sampling localities, respectively (Figure 1). Sample sizes were not uniform due to the difficulties in obtaining specimens in some localities with a small population size. All collected individuals were>1.5 cm (total length). Specimens were stored in an icebox during transportation then killed by freezing and preserved in 95% ethanol before DNA extraction.

Total genomic DNA was extracted from skeletal muscle or fin clip from one side of the body using QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. The primer pairs used for PCR were ThrF and 12SR (Miya and Nishida, 1999; Cheng et al., 2012) for the complete CR sequence, IRBP109F and IRBP1001R for the IRBP fragment (Chen et al., 2008), and 36298E1_F and 36298E1_R for R13 fragment (Li et al., 2010). PCR amplification of CR was performed in a 40 μl reaction volume containing 27.2 μl sterile distilled water, 4 μl of 10 × PCR reaction buffer, 3.2 μl of dNTP (2.5 mM each), 2.0 μl each of primers (10 μM), 0.2 μl Taq polymerase and 2 μl DNA template. The thermal profile was as follows: initial denaturation at 95 °C for 30 s followed by 34 cycles of denaturation for 30 s, annealing at 55 °C for 30 s and extension at 72 °C for 1.5 min, and a final extension at 72 °C for 5 min. The PCR mixture and profile for the two nuclear markers followed the same procedures as CR amplification except 4 μl DNA template were added to the reaction mix and annealing temperature was set at 48 °C for the thermal cycling conditions. PCR products were visualised using electrophoresis in 1.0% agarose gel using 0.5 × Tris/Borate/EDTA (TBE) buffer and purified using the Millipore PCR plate (Merck Millipore, Billerica, USA) for products with a clear single band. Where there were multiple bands present, the desired DNA band was cut and eluted using the QIAquick PCR purification Kit (Qiagen). Sequencing was performed using the same set of primers as in PCR amplification on an Applied Biosystems (ABI) (Foster City, CA, USA) 3710xl automated sequencer, following the standard cycle sequencing protocol with the ABI Big-dye Ready-Reaction mix kit.

Sequences were manually edited by using Seqtrace v0.9.0 (Stucky, 2012) followed by alignment using MUSCLE implemented in MEGA v7.0.18 (Kumar et al., 2016). Alleles with heterozygous SNPs for the two nuclear markers were inferred by PHASE v2.1 (Stephens and Scheet, 2005) implemented in DnaSP v5.0 (Librado and Rozas, 2009) using default settings. Population genetic diversity indices, including haplotype diversity (h), nucleotide diversity (π), Tajima’s D and Fu’s FS, were calculated for each site in each species using DnaSP v5.0. Neutrality tests (Tajima’s D and Fu’s FS) were conducted to infer the demographic history of each species to determine if the mutations were selectively neutral. Demographic expansion events were also examined by calculating Harpending’s raggedness index (r) and mismatch distribution analysis performed using Arlequin v3.5.2.2 (Excoffier and Lischer, 2010). The pairwise genetic distance was calculated with the Taimura–Nei model using the ape package in R v.3.2.2 (Paradis et al., 2004). To reveal the genealogical relationships among haplotypes, haplotype networks at 95% confidence level were constructed for all markers using TCS v.1.13 (Clement et al., 2000). The maximum connection unit was set to 40 in order to connect distantly related haplotype groups. Haplotype networks were visualised using Gephi v0.8.2 (Bastian et al., 2009).

We used SAMOVA v2.0 (Dupanloup et al., 2002) to delineate the best clustering of population-groups in terms of genetic distinctiveness in the three markers. The default setting specified for DNA sequence was used, with T92 model as the molecular substitution model and an initial configuration of groups at 100. The optimal number of groups (K) was selected based on the highest ΦCT value and configuration with no single population-group included (except Lantau Island populations, see Results). Analysis of Molecular Variance (AMOVA) and pairwise genetic differentiation values (ΦST) were then calculated for the partitions defined by SAMOVA v2.0.

Aligned CR sequences of both species were subjected to phylogenetic analysis and molecular dating. Phylogenetic analysis was not performed on nuclear markers because of the low number of informative sites for both species. The most appropriate model of nucleotide substitution was selected by JModelTest v2.1.7 (Posada, 2008) using the Akaike Information Criterion. The optimum model of substitution was HKY+I for the CR gene in both species. The CR sequence of S. thai (GenBank accession no.: DQ105264) was used as an outgroup in the phylogenetic analyses of S. fasciolata, while Pseudogastromyzon jiulongjiangensis (DQ105293) was used as an outgroup in the analysis of P. myersi. In the absence of fossil records and direct molecular clock calibration for the CR of S. fasciolata and P. myersi, we set the CR substitution rate, with reference to the mitochondrial cytochrome b (cyt b) clocks reported from their sister family Cobitidae (Liu et al., 2012), since cyt b and CR are linked during inheritance and have similar proportions of variable sites. Two comparable molecular clocks of cyt b were reported: 0.42% per lineage per myr for Cobitis taenia (Perdices and Doadrio, 2001) and 0.34% per lineage per myr for the genus Cobitis (Doadrio and Perdices, 2005). The relative mutation rate of the two genes, cyt b and CR, have been reported to be genus specific (Ruokonen and Kvist, 2002), and in the superfamily Cobitoidae, the mutation rate of CR is slower than that of cyt b (CR/ cyt b=0.78) in closely related species (Figures 4d and f). Yet, the numbers of mutation steps between haplotype groups were lower than those in the CR network due to the more conservative nature of the nuclear markers (see Supplementary Table 9). Both CR and nuclear marker networks reveal pronounced population genetic structure in this species.

SAMOVA defined three population-groups of P. myersi which are equivalent to the three lineages in BI tree in all three loci. Pairwise ΦST analyses of the three loci reveal a high level of genetic differentiation across all population-groups; all the values between the three groups were high (> 0.7) and significant (P<0.01) (Table 4). Most of the values between different localities within groups were high (> 0.7) and significant (P<0.01) in the CR marker (Supplementary Tables 5 and 6). A high global ΦCT value of 0.83 (P<0.01) was calculated from AMOVA analyses on CR (Supplementary Table 8). The global ΦCT values for the two nuclear markers were higher than 0.90 (P<0.01). The genetic variations among groups were much larger than the variations within groups or within localities.

Table 4 Matrix of pairwise ΦST for (a) mitochondrial CR, (b) nuclear R13 (above diagonal) and IRBP (below diagonal) of P. myersi (bold, P<0.0001)
Full size table

Discussion

Phylogeographic inference

We present herein the first phylogeographic study of balitorid loaches on a small geographical scale in Hong Kong, which reveals pronounced population structure in both studied species. Six and three deeply divergent lineages are recovered in S. fasciolata and P. myersi, respectively. A high level of differentiation is not only observed among lineages, but also between adjacent localities, suggesting limited gene flow between neighbouring streams separated by as little as 10 km. The high ΦST values are in congruence with results from previous studies on two other freshwater organisms in Hong Kong, which show high population differentiation between individual streams (Tsang et al., 2016; Wu et al., 2016). Moreover, by comparing and contrasting the phylogeographic structure of the two species studied, our study infers that climatic changes, sea level fluctuation and biological traits of organisms all play important roles in sha** the evolutionary history of loaches, on even a small geographical scale.

Our study reveals very similar temporal patterns and generally congruous spatial patterns of lineage divergence for the two loach species, in Hong Kong. Temporally, most of the lineages are estimated to have diverged during the late Miocene to the late Pliocene, which coincides with the marked climatic change in the late Miocene. East China had generally been humid from the Late Oligocene, until significant aridity developed in East Asia in the mid-Miocene and intensified during the late Miocene (Liu et al., 2009). Drought reduces hydrological connectivity among streams and is a major cause of freshwater habitat fragmentation (Covich et al., 2003). The prolonged aridity would have been able to significantly limit the headwater connectivity and reduce opportunities of flooding in the lower stream reaches, and thus facilitate lineage diversification of the two loach species in Hong Kong. This proposition is supported by other studies that suggested that low flow conditions weakened connectivity in freshwater fish species (for example, Etheostoma cragini, Labbe and Rausch, 2000). As there were no significant tectonic events recorded in this period, diversification due to drainage rearrangement caused by major geological changes is deemed less likely.

Spatially, our study shows that the degree of isolation among lineages on the major islands (Hong Kong Island and Lantau Island) and the Mainland is closely associated with the paleodrainage pattern and the location of common confluences. Both Lantau Island and Hong Kong Island are located very close to the Mainland, but the populations on the two islands show a distinct level of divergence with respect to the Mainland populations. The Lantau Island lineages diverged during the late Pliocene to the early Pleistocene in both species, coinciding with the rapid rise in sea level of the South China Sea during the late Pliocene (Wang and Li, 2009; Ding et al., 2013). The high sea level stand that separated the Mainland and Lantau Island might have acted as a strong geographic barrier to gene flow for the loaches. During the late Quaternary glacial cycle, the sea level fluctuated extensively, and a large part of the northern South China Sea was repeatedly exposed when the sea level was –120 m (Wang and Li, 2009). As a result, the Ma Wan Channel that separates Lantau Island and the Mainland would have been exposed and acted as a land bridge between the two. The paleodrainage in the western Mainland and Lantau Island converged far away at the Pearl River estuary which is situated around 15 km west of Lantau Island and the Mainland (Fyfe, 2000). This temporary connection between Lantau Island and the Mainland may have been temporally too short or spatially too distant for effective gene flow among freshwater fish populations on the two landmasses, especially for the loaches that inhabit the mid-stream and upper stream. By contrast, the confluence of Hong Kong Island and the Mainland palaeodrainage was close (< 1 km) to the Hong Kong Island, whereby population connectivity during periodical Quaternary low sea levels were possible. This might explain why S. fasciolata populations on Hong Kong Island share some haplotypes with those in the Mainland. Hence, the populations on Hong Kong Island were probably separated from those in the Mainland very recently by the last Holocene transgression. The effect of exposure of palaeodrainage confluence in facilitating gene flow is also evidenced in the mixing of S. fasciolata Central and East lineages in the northeast Mainland: the middle and east paleodrainages should have converged at the northeast lowland (that is, the current location of Tolo Habour) close to the shoreline (Fyfe, 2000).

Our results show that the evolution of the paleodrainage system might have been critical in sha** the population structure in S. fasciolata, suggested by the dissimilar haplotypes found on the two main islands (Figures 2 and 4a). This is in agreement with the findings of a recent phylogeographic study on a freshwater goby, R. duospilus in Hong Kong (Wu et al., 2016). Unfortunately, we were unable to sample P. myersi on Hong Kong Island in this study for comparison. More sampling of populations on Hong Kong Island is needed to confirm our hypothesis. Nonetheless, our results suggest that divergence of freshwater fish may also occur among islands that are extremely close to the coast. This implies high genetic diversity of aquatic fauna in the myriad islands along the southeastern coast of China. The population structure of other freshwater fish probably also follows the paleodrainage pattern to some extent, with the degree of similarity varying with species-specific life history traits.

Both extrinsic (for example, climatic and geological history) and intrinsic (for example, ecological specialisation, dispersal and reproductive behaviours) factors influence the population dispersal processes (Tibbets and Dowling, 1996). Many studies have reported the effects of intrinsic factors on evolutionary processes of closely related species and demonstrated the importance of species-specific ecological characteristics on dispersal (Koblmuller et al., 2007). Our analyses reveal a subtle difference in the degree of population structure between the two loach species. Generally, P. myersi exhibits a shallower genetic structure than S. fasciolata, most obviously among the Mainland populations. The difference might be attributed to their specific biological traits, particularly their mobility. S. fasciolata and P. myersi are well adapted to fast-flowing streams, but they may be swept downstream during a downpour in the rainy season. A recent mark-recapture study in Hong Kong reveals a marked difference in their mobility: rheophilic behaviour is common in balitoroid loaches, and P. myersi shows a particularly strong tendency to move upstream (Ho and Dudgeon, 2015). In contrast, S. fasciola is sedentary with limited movement along the stream. In addition, we observed gregarious behaviour in S. fasciolata in many localities which may restrict their mobility. It would be less likely for S. fasciolata swept down to lower stream reaches to migrate back to their original habitats than P. myersi. Therefore, we believe that the difference in mobility is a major factor leading to the more pronounced population structure in S. fasciolata.

We also observed a difference in the location of the major genetic break among the Mainland populations in the two loach species. The population subdivision of S. fasciolata is congruent with the late Pleistocene palaeodrainage pattern, which roughly comprises four drainages flowing west, central, east and northeast, with the Mainland populations primarily separated into two groups by the Ma On Shan ridge. Our results show that the populations located on the two sides of the ridge are genetically distinct from each other and do not share common haplotypes. This east–west separation is also observed on Hong Kong Island. However, for P. myersi, analyses of both mitochondrial and nuclear markers reveal that Tai Mo Shan ridge, instead of the Ma On Shan ridge is the major genetic break. This discrepancy in population structure between the two species remains to be explored.

Conservation implications

Located in the Indo-Burma biodiversity hotspot, Hong Kong has a high diversity of freshwater fauna, which are, unfortunately, severely threatened by rapid urban development. Although S. fasciolata and P. myersi inhabit upstream and mid-stream areas that are relatively distant from the highly developed coastal area, several habitats with populations of high levels of genetic differentiation are under serious threat of imminent urbanisation. The populations in Lantau Island (IWL and ITF) are under threat due to a proposed land development plan that aims to change the island into a major recreation and tourism zone. In addition, P. myersi warrants more management attention since it is uncommon and geographically more restricted. Furthermore, this species was historically found in both Guangdong and Hong Kong, but not recorded in a recent comprehensive survey in Guangdong (Zhao and Cui, 2011), probably due to habitat destruction there. In this study, we were only able to sample 13 P. myersi populations, most of which are found in the north Tai Mo Shan drainage. A more detailed survey is urgently needed to assess the extinction risk of P. myersi.

Conclusion

This study reveals marked population structure and similar temporal-spatial patterns of lineage divergence in two stream loaches, S. fasciolata and P. myersi, and provides insights into the evolutionary history of freshwater fish in South China. By comparing the phylogeographical pattern of the two species, we suggest that two historical events, namely the climatic change during the late Miocene and the major sea-level transgression during the Pliocene, played important roles in sha** the evolutionary history of freshwater fish in Hong Kong. The subtle discordance in their population structure is mainly attributed to their difference in mobility. The finding that considerable population structure exists on such a fine scale across multiple species implies that the overall freshwater biodiversity is higher than expected and warrants more investigation. Considering the current geographic isolation and strong genetic differentiation of Lantau Island populations of S. fasciolata and P. myersi, we identified them as populations with a high priority in biodiversity conservation, especially for P. myersi which has a restricted distribution.

Data archiving

Sequence data have been submitted to GenBank: accession numbers KY421988 - KY423282.