| Edited by Etsuko Matsuura. Ryo O. Gotoh: Corresponding author. E-mail: rogotoh@gmail.com |
Oceanic islands, which have been geographically isolated from the main land since their formation, offer excellent opportunities for the evolutionary study of terrestrial species (Grant, 1998; Caccone et al., 2002; Calsbeek and Smith, 2003) due to the characteristics of limited gene flow, specialized environments and biota, and different selective pressure from the mainland leading to speciation. The most well-known evolutionary study on oceanic islands is Darwin’s finches in the Galapagos Islands, in which the size and shape of beaks were shown to have adapted to various food sources (Grant, 1986; Grant and Grant, 1989). Despite this example, little is known about the evolution of marine species in geographically isolated environments, because ocean currents tend to facilitate gene flow, particularly during the planktonic egg and larval stages, resulting in low levels of genetic differentiation even over large geographical ranges (Palumbi, 1994; Grant and Bowen, 1998; Avise, 2000). Recent studies using molecular genetics have reported that marine species inhabiting coral reefs have cryptic population structures even at relatively short ranges due to the effects of life-history traits on larval dispersion and oceanographic factors (Taylor and Hellberg, 2003; Purcell et al., 2006). These studies, however, have not been true analogues of the evolution on terrestrial islands for marine species, as the general conception of an island is a location entirely surrounded by a different environment. For example, coral reefs are surrounded by a medium that possibly includes eggs and larvae (Dawson, 2006). In an exception to this classic view, one marine habitat that clearly fits the definition of an island is the marine lakes on the Palau Islands (Dawson, 2006).
The Palau Islands are located in the western Pacific Ocean and consist of a large number of limestone islands in the central and southern areas. The marine lakes, which are small seawater bodies in the interior of the limestone islands, can be limnologically classified into two types: meromictic and holomictic lakes. Meromictic lakes have stratified water layers that do not intermix, resulting in the formation of a metalimnion (thermocline and nutricline) and the development of hypoxia (oxygen depletion) under the metalimnion (Saitoh et al., 2011). Holomictic lakes in contrast, experience a physical mixing of the surface and deep waters at least once each year. Approximately 70 marine lakes are located on the Palau Islands (Hamner and Hauri, 1981), which are thought to have been formed geologically through the following three steps. First, limestone was shaped and uplifted to form islands during the Miocene and Pleistocene (U.S. Geological Survey, 1956); second, the islands were gradually eroded by rain and wind resulting in the formation of numerous depressions; and finally, with the rising sea level at the end of last glacial age (~12,000 years ago), seawater flooded the depressions through fissures and tunnels to form lakes (Hamner and Hauri, 1981; Hamner and Hamner, 1998).
The marine lakes of Palau have several noteworthy characteristics that make them suitable ecosystems for the study of evolution. First, they have been isolated from adjacent lagoons to various degrees: numerous marine lakes are apparently connected to lagoons by tunnels, whereas other lakes are completely isolated. High and low tides occur even in highly isolated lakes, since seawater enters and exits through minute pores in the limestone (Humner and Humner, 1998). Second, the marine lakes formed in a chronological sequence: deeper lakes formed first (~12,000 years ago), followed by shallower lakes (~5,000 years ago) as seawater flooded the limestone depressions with rising sea level (Dawson and Hamner, 2005). Third, a number of marine species inhabiting adjacent lagoons are also found in marine lakes. This characteristic is particularly important because it enables the comparison of the likely source (ancestral) and colonizing populations. Finally, even marine lakes which are located in the identical island display different properties in size, depth, physical and chemical structure, fauna, and flora (Hamner and Hamner, 1998; Dawson and Hamner, 2005; Saitoh et al., 2011). In addition, Saitoh et al. (2011) reported that species diversity of copepods differs according to the limnological properties of the lakes, and demonstrated that these differences were associated with the selection of species inhabiting the lakes.
Previously, population genetic analyses for species inhabiting the Palauan marine lakes were conducted on only two taxonomic groups. Dawson and Hamner (2005) reported that golden jellyfish, Mastigias spp., are genetically and morphologically divergent among the lakes. Gotoh et al. (2009) have also concluded that the orbiculate cardinalfish, Sphaeramia orbicularis (Cuvier, 1828), has diverged genetically among the marine lake and lagoon populations within an extremely short geographical range (~150–250 m). Although these studies have demonstrated that marine lake populations are genetically differentiated from the lagoon populations that are considered to be the source population, the detailed characteristics of each marine lake population and the relationships between the marine lake and lagoon populations are unclear because the genetic polymorphisms used in these studies were insufficient to clarify their genetic structures.
In the present study, we focused on the striped silverside, Atherinomorus endrachtensis (Quoy and Gaimard, 1825), as this species has a large population size in several marine lakes and lagoons of Palau. To elucidate the genetic structure of marine lake and lagoon populations of A. endrachtensis, we conducted population genetic analyses using two mitochondrial DNA (mtDNA) markers differing in evolutionary rate: the cytochrome b (cyt b) gene and the control region. The evolutionary rate of the control region is two- to five-fold higher than that of mitochondrial protein-coding genes (Meyer, 1993). Here, we first investigated the genetic structure and demographic history of lagoon populations to reveal unique characteristics of marine lake populations. We then examined the level of genetic differentiation and evolutionary characteristics of marine lake populations. Finally, we discuss the possibility that the marine lakes of Palau are excellent environments for the evolutionary study of marine species.
A total of 203 individuals of A. endrachtensis were collected by angling and using a casting net. The sampling was conducted in the following five marine lakes in Palau between 2004–2009 (Table 1 and Fig. 1): Jellyfish Lake (JFL), Big Crocodile Lake (BCM), North Cassiopea Lake (NCM), South Cassiopea Lake (SCM), and Ongael Lake (ONG). The first four lakes are located on Mecherchar Island, while ONG is located on Ongael Island. Sampling was also conducted in the following four locations in lagoons: Jellyfish Lake outside (JFOS), Ngeteklou Lake outside (NTOS), Ngeruktabel Lake 1 outside (NGOS), and under Koror-Babeltuab Bridge (KBB). Collected specimens were identified according to the classification scheme of Kimura et al. (2001). In addition, three individuals collected in Sulawesi, Indonesia were included as outgroups. Small pieces of skeletal muscle were excised from specimens and preserved in 99% ethanol. The preserved tissues were transported to our laboratory at room temperature and subjected to DNA extraction.
 View Details | Table 1 Marine lake and lagoon sampling locations, sample sizes, and physical characteristics |
 View Details | Fig. 1 Sampling localities of A. endrachtensis used in this study. a: A map of the western Pacific, b: Sampling locations on the Palau Islands, c: A map of Ongael Island and the surrounding area, d: A map of Mecherchar Island. |
Preserved skeletal muscle samples were finely chopped in DNA extraction buffer (Asahida et al., 1996) containing proteinase K (10 mg/ml), digested overnight at 37°C, and then subjected to standard phenol-chloroform extraction (Sambrook et al., 1989). Total DNA was precipitated by adding an equal volume of 99.5% 2-propanol, washed, resuspended in TE buffer, and then stored at –20°C until needed.
In the population genetic analyses, we used two mtDNA markers that differed in evolutionary rate: the cyt b gene and the control region. The cyt b region was amplified from 187 individuals by the polymerase chain reaction (PCR), which was performed in a final reaction volume of 25 μl, containing 50 ng template DNA, 2.5 μl 10× PCR buffer, 2 μl 2.5 mM dNTP mixture, 10 pmol of each primer: AJG15 (Akihito et al., 2000) and newly designed Actmtcy-1R (5’-GCT TAC AAG ACC GGC GCT CT-3’), 0.2 μl Ex Taq Polymerase (TaKaRa), and 16.5 μl distilled water. The amplification conditions consisted of an initial denaturation step at 94°C for 5 min, followed by 30 cycles of denaturation (94°C, 45 sec), annealing (57°C, 30 sec), extension (72°C, 1 min), and a final extension step at 72°C for 7 min.
The mitochondrial control region was also amplified from 205 individuals using identical PCR reagents, quantities, and amplification conditions as for the cyt b region, with the exception that the primers L15927-Thr (Miya and Nishida, 2000) and Simt 12-Fb (Kuriiwa et al., 2007) were used. For a few individuals, the region from the cyt b to the control region was amplified using the primers AJG15 and Simt 12-Fb and the identical PCR reagents and quantities as for control region. The amplification conditions were also identical, except that the extension step was performed at 72°C for 1 min 30 sec.
After checking the size of amplified fragments on a 1% agarose gel, PCR products were precipitated with an equal volume of 99.5% 2-propanol. Direct sequencing of purified PCR products was performed with the DYEnamic ET Terminator Cycle Sequencing Kit (GE Healthcare) in 10 μl reactions (0.4 μl DYE, 5 pmol of primer, 1 μl PCR products, 0.9 μl buffer, and distilled water) that were subjected to 30 cycles at 96°C for 20 sec, 50°C for 15 sec, and 60°C for 1 min, followed by a final holding step at 4°C. The following primers were used for the sequencing reactions: AJG15, L15369-CYB (Miya and Nishida, 2000), H15149-CYB (Inoue et al., 2000), Actmtcy-1R, L15927-Thr, AthemtD-F (5’-CTT GGC ATT TCA CAG TGC AT-3’), and AthemtD-R (5’-GAC CAA GCC TTT GTG CCT AC-3’). After the generated PCR products were purified using 99% ethanol, their nucleotide sequences were determined using an ABI Prism 3100 Genetic Analyzer (Applied Biosystems).
The PCR sequences of the cyt b gene and control region were aligned using the computer program Clustal X (Thompson et al., 1997) and the aligned sequences were checked by eyes.
The haplotype diversity h (Nei, 1973) and nucleotide diversity π (Nei and Tajima, 1981), which show the degree of genetic diversity, were estimated using the computer program Arlequin ver. 3.5.1.2 (Excoffier and Lischer, 2010). To determine the level of genetic differentiation between each examined population, the pairwise Φst (Excoffier et al., 1992), which includes information on mitochondrial haplotype frequency and genetic distance, was also implemented using Arlequin ver. 3.5.1.2. The significance of the Φst value was evaluated by performing a randomization test with 1000 replications.
For phylogenetic analyses, the software TCS 1.21 was used to construct a statistical parsimony haplotype network (SPN), considering gaps as the fifth state (Clement et al., 2000). First, a distance matrix was generated from all pairwise comparisons of haplotypes, for which the maximum number of mutations that did not exceed the probability of parsimony by 0.95 was determined (Templeton et al., 1992). All haplotypes satisfying this parsimony criterion were connected to a single network, while those with probabilities exceeding 0.95 were resolved as separate networks.
To test for departure from mutation-drift equilibrium, we conducted Tajima’s D (Tajima, 1989a) and Fu’s Fs (Fu, 1997) tests of neutrality in Arlequin ver. 3.5.1.2 based on sequence variations of the control region. Neutrality tests, which were originally developed to test neutral hypotheses, are now widely used to detect changes in population size because they are based on the expectation that a population of constant size is at mutation-drift equilibrium (Mousset et al., 2004). If D values are significant, the population would not reach mutation-drift equilibrium, and a tendency for Tajima’s D value to be negative in an expanding population would exist (Tajima, 1989b). Fu’s Fs test is also sensitive to population demographic expansion, which generally leads to large negative Fs values in expanding populations.
The demographic history of the A. endrachtensis populations was investigated based on sequences of the mtDNA control region by mismatch distribution analyses (Rogers and Harpending, 1992). The observed distribution was tested for goodness-of-fit to a model of sudden population expansion using parametric bootstrapping with 1000 replicates in Arlequin ver. 3.5.1.2.
The parameters of demographic expansion τ, θ0, and θ1, were estimated by a generalized non-linear least-squares approach (Schneider and Excoffier, 1999). The absolute time of population expansion (t) was calculated through the relationship t = τ/(2u), where τ measures the time in units of 1/2 μ generations (μ is the mutation rate per sequence under study per generation), and u represents the mutation rate per sequence under study per year (Rogers and Harpending, 1992). Mutation rates were calculated from the overall mean p-distance of JFL population and the estimated formation age of the lake (8,000 to 12,000 years ago).
The complete cyt b gene sequences (1141 bp) were determined for a total of 187 individuals of A. endrachtensis, and 25 different haplotypes substituted at 34 base sites were identified (Accession No. AB295500–AB295524). All of the base substitutions were transitions, including 28 synonymous and 6 non-synonymous substitutions. Notably, only four haplotypes (haplotypes A, D, F, and I) were shared by more than one population. Specifically, haplotype A was detected in all populations in Palau; haplotype D was shared between the JFL and NGOS populations; haplotype F was shared among all lagoon and two marine lake populations (NCM and SCM); and haplotype I was detected in the JFOS, NCM, and SCM populations. Sixteen haplotypes were singletons, and one and two unique (population-specific) haplotypes were identified in the ONG and JFL populations, respectively.
Among a total of 205 A. endrachtensis individuals, the complete control region sequences varied from 874 to 877 bp in length, and several indels were identified. Seventy-two different haplotypes substituted at 67 base sites were detected from the sequences (Accession No. AB295525–AB295596). The base substitutions included 63 transitions and 6 transversions. Three indels were found excluding CT repeats. CT repeat variation (4–5 repeats) was found at the 3’ terminal side of the control region. This variation was treated as an offspring by a one-step mutation event in the subsequent analyses because the repeat number typically changes by one unit. As in many other teleost fishes, most substitutions were observed in the 5’ terminal side, which is known as a hypervariable region (Saccone et al., 1987; Lee et al., 1995). Only 11 of the observed control-region haplotypes were shared by more than one population.
In the analysis of the complete cyt b gene sequences, the level of haplotype diversity (h) of the marine lake populations was low to moderate, with observed values ranging from 0.000 (BCM) and 0.593 (NCM) (Table 2). Although the NGOS population displayed a relatively low level of haplotype diversity (h = 0.371), the other lagoon populations showed much higher diversity, with observed values between 0.688 (KBB) and 0.747 (NTOS) (Table 2). In contrast, the nucleotide diversity (π) was very low in both the marine lake and lagoon populations, although the values for the lagoon populations were higher than those of the marine lake populations, except for NGOS (Table 2). High pairwise Φst values were observed between the NCM or SCM and the other marine lake populations, while the other calculated values among the marine lake populations were low and not significant (Table 3). The pairwise Φst values between the marine lake and lagoon populations were moderate to high with significant p-values, except those between the marine lake and NGOS populations (Table 3).
 View Details | Table 2 Haplotype diversity (h) and nucleotide diversity (π) for each population of striped silverside, A. endrachtensis |
 View Details | Table 3 Population pairwise Φst values for the mtDNA cyt b gene (above the diagonal) and control region (below the diagonal) |
In the analysis of the complete mtDNA control region sequences, the haplotype diversity (h) of marine lake populations was low to moderate, with observed values ranging from 0.151 (ONG) and 0.593 (NCM) (Table 2). In contrast, the values of four lagoon populations were extremely high and comparable to the levels reported for the hardyhead silverside (A. lacunosus) in the Red Sea (h = 0.97; Bucciarelli et al., 2002) and other pelagic fishes (Grant and Bowen, 1998; Zardoya et al., 2004; Atarhouch et al., 2006; Liu et al., 2006; Zane et al., 2006). Similar to the cyt b gene, the mtDNA control region nucleotide diversity (π) of marine lake populations was low, with observed values between 0.026 (ONG) and 0.290 (NCM) (Table 2). In contrast, although the π values observed in lagoon populations were higher than that of marine lakes, they were relatively low compared with those of other pelagic fishes (Grant and Bowen, 1998). The pairwise Φst values between the marine lake populations ranged from 0.044 to 0.864, while no significant values were obtained between the lagoon populations, with the exception of the JFOS and other lagoon populations. All pairwise Φst values between the marine lake and lagoon populations were relatively high (0.134 to 0.536) and significant (Table 3).
In the mtDNA cyt b gene sequence analyses, we obtained a single statistical parsimony network (Fig. 2). Nearly all of the minor haplotypes were radically generated from common haplotype A by a one-base substitution. The two haplotypes detected from the Sulawesi outgroup individuals were connected to common haplotype A by 4 and 11 single-base substitutions. Haplotype Y detected from KBB was more closely related to the two haplotypes of Sulawesi than to those of Palau.
 View Details | Fig. 2 Statistical parsimony networks of the mtDNA cyt b (above) and control regions (below) based on the 95% connection limit. The size of circles represents the frequency of each haplotype. Open circles represent a missing haplotype. Solid rectangles in the network of the control region indicate a change in the number of CT repeats in the 3’ terminal side. Each haplotype is labeled with a letter in the network based on the cyt b gene, while numbers are used in the networks based on the control region. |
In the complete mtDNA control region sequence analyses, we identified three separated networks, since the haplotypes were highly differentiated and exceeded the probability of parsimony by 0.95 (Fig. 2). The first network was constructed with only the haplotypes detected from Palau, while the second only consisted of two haplotypes from Sulawesi, which were connected with each other by four base substitutions. The third network included two haplotypes, one from Palau and the other from Sulawesi, and was connected to each other by five base substitutions. In the first network, nearly all of the haplotypes displayed complicated (bush-like) connections with each other. The lagoon populations in particular displayed a more complex network pattern with homoplasy, whereas the network of marine lake populations demonstrated a relatively simple pattern. Notably, the relationships among the haplotypes detected from JFL and BCM in Mecherchar Island showed a “starburst” network.
In addition, we determined that the marine lake populations had numerous population-specific haplotypes derived from common haplotype 1 (JFL, haplotype 2–8; BCM, haplotype 9–11, Fig. 2). The ONG population in Ongael Island also displayed a similar pattern, consisting of one major haplotype and a few population-specific haplotypes that were connected by one or two substitutions to the major haplotype. The NCM and SCM populations demonstrated a different pattern than the other marine lake populations, as a major haplotype was not identified. Rather, several haplotypes with varying frequencies were observed, and the relationships between these haplotypes were not close: the maximum number of substitutions among the haplotypes was eight, and no population-specific haplotypes were detected.
In addition to the phylogenetic analyses, we also performed neutrality tests and mismatch distributions on each population based on the complete mtDNA control region sequences (Table 4 and Fig. 3). All examined populations displayed negative Tajima’s D values, with the exception of the NCM population; however, only the D values of the two marine lake populations (BCM and ONG) were significantly negative. The calculated Fu’s Fs values were also highly negative with significance, except for the NCM and SCM populations. The sudden expansion model could not be rejected in all populations (Table 4). Histograms of the mismatch distribution of haplotypes in the lagoon populations presented unimodal distributions, with the exception of KBB, which showed a bimodal distribution (Fig. 3). The histograms for the marine lake populations could be classified into two types: those with simple L shape distributions (JFL, BCM, and ONG), and those with distributions composed of discontinuous peaks (NCM and SCM). Furthermore, neutrality tests and mismatch distribution were performed on pooled samples, labeled “LAG”, which consisted of all lagoon individuals except for one individual from KBB with a haplotype closely related to one from the Sulawesi outgroup (Fig. 2). The pooled samples presented negative Tajima’s D values and high Fu’s Fs values with significance (Table 4). In addition, the sudden expansion model was not rejected, and the mismatch distribution was clearly unimodal (Fig. 3). The overall mean p-distance of JFL population was 0.001. Mutation rates of control region of A. endrachtensis calculated from mean p-distance and the estimated formation age of Jellyfish Lake (8,000 to 12,000 years ago) was ranging from 8.3%/Myr to 12.5%/Myr.
 View Details | Table 4 Results of the Tajima’s D and Fu’s Fs tests, and parameters of the sudden expansion model and goodness-of-fit test with respective significance for each population based on the control region |
 View Details | Fig. 3 Mismatch distribution analysis based on haplotypes of the mtDNA control region in each examined population of A. endrachtensis. The vertical bars are the observed distribution of mismatches, and the lines represent the expected distribution under the sudden-expansion model. |
The analyses of genetic variability for the examined lagoon populations of Palau indicated moderate to high haplotype diversity (h) and low to moderate nucleotide diversity (π), both in the cyt b gene and mtDNA control region (Table 2). Grant and Bowen (1998) have proposed that such a genetic feature, namely moderate to high h and low to moderate π, is characteristic of a population that experienced rapid growth and accumulated novel mutations after a bottleneck event. The rapid growth of the lagoon populations is also suggested from the star-shaped haplotype network based on the cyt b gene (Fig. 2). This type of network arises in populations that have experienced recent expansion in population size from a smaller number of founders (Avise, 2000). In addition, the results of the mismatch distribution analyses also support the recent expansion of the lagoon populations (Table 4).
Past demographic changes were a result of complex interactions of biology, geography, and climatic shifts (Hewitt, 2000). Most of the species inhabiting lagoons are considered to have been exposed to severe bottlenecking during the last glacial age, as the lower sea level would have remarkably limited the number of lagoon habitats. It has been shown that such a reduction in population size and subsequent rapid growth of fishes in lagoons were caused by Holocene sea-level changes ~18,000 years ago (Fauvelot et al., 2003) in the case of the butterflyfish, Chrysiptera glauca (Cuvier, 1830), and damselfish, Dascyllus aruanus (Linnaeus, 1758) in French Polynesia. Although the Palau Islands also experienced a low sea level at this time (Easton and Ku, 1980; Kayanne et al., 2002), the time since expansion for the pooled LAG samples, which displayed significantly negative Tajima’s D and Fu’s Fs values, was estimated to be 20,086~30,251 years ago from the sudden expansion model (τ = 4.404). This estimation suggests that expansion of the lagoon populations in Palau occurred before the Holocene sea-level change. Since A. endrachtensis inhabits other areas besides lagoons, such as bays, inlets, and shallow coastal areas (Kimura et al., 2001; Ivantsoff and Crowley, 2000), this species may have been able to maintain its population size through the last glacial age. Therefore, a more plausible explanation for the low population size and subsequent expansion of the lagoon populations is the founder effect, meaning that the current lagoon population originated from a small number of individuals that initially colonized the Palau archipelago or was the result of a bottleneck event that occurred before the Holocene sea level change.
The high level of nucleotide diversity (π) and bimodal distribution from the mismatch analysis that were detected in the KBB population (Table 2 and Fig. 3) is mainly attributable to the presence of a single unique haplotype which is more closely related to the Sulawesi haplotypes than the other haplotypes of Palau. Since A. endrachtensis typically inhabits lagoons or the inner areas of bays and inlets (Kimura et al., 2001), gene flow seldom occurs on a large scale, which is similar to most fish species inhabiting lagoons (Warner et al., 2000; Cowen, 2002). However, our results suggest that gene flow rarely occurs over a large geographical scale. Although migration by human activity (e.g. accidentally carried by ballast water or intentionally transferred for commercial use) also caused in the same results, this is unlikely because A. endrachtensis is too weak to tolerate accidental long transfer in ballast water and has no value for commercial use suggesting that the species is never intentionally transferred.
Based on our analysis of the cyt b gene and mtDNA control region, the level of genetic differentiation was very low among nearly all of the examined lagoon populations (Table 3), implying that frequent gene flow has occurred among these populations. Despite this apparent genetic exchange, the JFOS population has slightly differentiated from the other lagoon populations (Table 3), and also displayed approximated half of the nucleotide diversity (π) in the control region as was observed in the other populations (Table 2). These results suggest that the Palau lagoon populations likely consist of several sub-populations. Although the specific factor that prevents frequent gene flow among lagoon populations is unclear, it may be due to the geographical characteristics surrounding Mecherchar Island, which form an inlet, and the ecological characteristic that atherinids spawn adhesive demersal eggs (Takemura et al., 2004) and prefer inlets to open coastal areas. This speculation may appear incongruent with the discussion concerning large-scale dispersal, but we suggest that the bilateral character of large-scale dispersal and limited gene flow, even at a small range, is the characteristic of marine species inhabiting coasts or lagoons. Future studies using an increased number of samples and nucleic genetic markers such as microsatellites will help clarify the genetic structure of lagoon populations in Palau.
The marine lakes surveyed in the present study are divided into two types according to limnological conditions: meromictic (JFL and BCM) and holomictic lakes (NCM, SCM, and ONG; Table 1). Holomictic lakes are often connected to an outer lagoon and typically contain more fish species compared with meromictic lakes. Although the connectivity of holomicitc lakes and lagoons are limited and vary among individual lakes, the rich fauna of holomictic lakes, which is often similar to that of an adjacent outer lagoon, has led to the assumption that lagoon populations might move in and out of holomictic lakes. Our results, however, support the conclusion that marine lake populations, even those of holomictic lakes, are completely isolated from lagoon populations because the lake and lagoon populations have completely different genetic structures (Fig. 2). If individuals constantly move in and out marine lakes, the haplotypes existing in lagoons would be expected to be more frequently observed in marine lakes.
All of the examined marine lake populations displayed low genetic diversity and a simple network pattern of haplotypes compared to those of the lagoon populations (Table 2 and Fig. 2), although a number of differences in genetic variations were identified among the populations. From the viewpoint of genetic structure, we classified the marine lake populations into two groups: JFL-BCM-ONG and NCM-SCM. The first group was characterized by a star-shaped network consisting of one dominant and a few population-specific haplotypes connected by one or two base substitutions (Fig. 2), distributions close to zero in the mismatch analysis (Fig. 3), and negative Tajima’s D and Fu’s Fs values (Table 4). These characteristics suggest that the effective population size of this group decreased within the recent thousands or tens of thousands years and subsequently increased exponentially (Grant and Bowen, 1998; Avise, 2000; Frankham et al., 2002). In contrast, the second group did not exhibit a “starburst” haplotype network pattern, and was composed of several haplotypes that occurred with moderate frequency. These features suggest that this group has undergone strong genetic drift since the decline of the population size.
The speculated population size reduction applies equally to the JFL-BCM-ONG and NCM-SCM groups. Such population size fluctuations and reductions of genetic diversity were likely caused by the founder effect. The founder effect clearly affected the marine lake populations of A. endrachtensis following its initial colonization of the marine lakes. In addition, there is evidence that bottleneck events occurred in the marine lakes due to environmental fluctuations such as El Niño. For example, Dawson et al. (2001) reported that high salinity and water temperature in marine lakes on the Palau Islands coincides with the 1997/1998 El Niño and the disappearance of 1.5 million Mastigias medusae. Therefore, it is thought that the bottlenecking occurred in the marine lake populations of A. endrachtensis due to drastic climatic changes, such as a recent El Niño. Alternatively, the marine lake populations may have maintained their sizes under harsh environmental conditions because atherinids have a remarkable ability to survive a wide range of salinities (0.1–100‰; Potter et al., 1986). If the bottlenecking has occurred repeatedly since the formation of the marine lakes, the number of population-specific haplotypes would be small and nearly equal among all of the marine lake populations of A. endrachtensis due to the effects of strong genetic drift, which would have purged most of the pre-existing haplotypes. However, in our data, the number of population-specific haplotypes clearly differs among the examined populations and a tendency for older (deeper) marine lake populations to have a greater number of population-specific haplotypes than those of younger (shallower) populations in the control region analyses was observed (Table 1 and Fig. 2). Thus, we conclude that genetic diversity of the marine lake populations was strongly affected by the founder effect, rather than bottlenecking, after the formation of the lake populations.
As mentioned above, the results of our genetic analyses of the A. endrachtensis populations in four marine lakes on Mecherchar Island suggest that these lakes can be classified into two groups: JFL-BCM and NCM-SCM. In the analysis of the control region sequences, all of the population-specific haplotypes detected in JFL and BCM appear to have derived from dominant haplotype 1 (Fig. 2). From the observed relationships, it is likely that these population-specific haplotypes were generated in the marine lakes after the populations were isolated from lagoons. This assumption is further supported by the presence of unique genetic variations in the marine lake populations, as most of the haplotypes detected in the lagoon populations had four CT repeats in the 3’ side of the control region, while all haplotypes detected in the marine lake populations had five CT repeats (Fig. 2). Although it is unclear whether the dominant haplotype has existed since the initial establishment of the marine lake and lagoon populations, it is apparent that the population-specific haplotypes were generated in these marine lakes. Taken together, these results strongly suggest that the marine lake populations of Mecherchar Island have been completely isolated from the lagoon populations, and have undergone distinct evolutionary processes.
Interestingly, the JFL and BCM populations share a common dominant haplotype, in both the cyt b gene and control region, which is unlikely to have occurred by chance, since lagoon populations maintain an extremely large number of haplotypes. If each marine lake population experienced genetic drift independently, the probability that the identical haplotype would become dominant in different populations is extremely low. Dawson and Hamner (2005) have suggested that the time required to form marine lakes depends on their depth, as deeper marine lakes would have been flooded with seawater earlier. Following this assumption, JFL (< 30 m depth) is thought to have been formed thousands of years earlier than BCM (< 22 m depth). Based on these assumptions, we propose the following hypothesis to account for the presence of a common haplotype in two marine lakes. The JFL population was isolated first, and the dominant haplotype was then fixed as haplotype 1 in the population. Following the formation of BCM thousands of years later, individuals from JFL presumably colonized this new lake, which explains why both marine lake populations have the identical dominant haplotype. Today, the migration of individuals between the lakes is likely prevented, since the deeper layers of these two lakes are extremely anaerobic environments and contain an abundance of hydrogen sulfide produced by anaerobic microorganisms. From the control region analyses, it is suggested that low levels of genetic differentiation have already occurred between the JFL and BCM populations (Φst = 0.059, P < 0.05; Table 3). To clarify the evolutionary process of JFL/BCM populations, the further study using nuclear DNA markers should be conducted.
The control region analyses also revealed that the NCM and SCM populations share the identical haplotypes, except for haplotype 15, and showed little genetic differentiation in both the cyt b gene and control region (Table 3 and Fig. 2). These results strongly suggest that these two marine lake populations underwent similar historical processes, at least until a certain point in the recent past. It is therefore probable that the NCM and SCM lakes are currently physically divided because the species confirmed in each marine lake differ slightly. For example, we could only observe the raccoon butterfly, Chaetodon lunula (Lacepède, 1802), narrow-lined puffer, Arothron manilensis (Marion de Procé, 1822), and white-spotted puffer, Arothron hispidus (Linnaeus, 1758), in NCM (Gotoh et al., unpublished data). To reveal the relationships of the two marine lakes in more detail, we will conduct detailed faunal surveys in a future study.
Interestingly, marine lake populations in the same island contained detectable differences in their genetic structure and establishment time in spite of the short distances between neighboring lakes.
The present study clearly showed high genetic differentiation between several marine lake and lagoon populations of Palau. We conclude that the observed genetic structures of the marine lake populations have been shaped by the following steps. Initially, a small number of founders that inhabited lagoons and possessed high levels of genetic variation were isolated in marine lakes in chronological sequence of their formation during the past 12,000 years. The abundant genetic variety then decreased by genetic drift, and would have been fixed to one haplotype in each of the marine lakes JFL, BCM, and ONG, and nearly fixed in the marine lakes NCM and SCM. Subsequently, novel population-specific haplotypes accumulated as the populations expanded, because such rapid population growth would have enhanced the retention of new mutations (Avise et al., 1984). Eventually, genetic differentiation occurred between the marine lake and lagoon populations (Table 3), in spite of the short geographical distance between the marine lakes and lagoons (85 to 316 m, see Table 1). To date, the occurrence of high level of genetic differentiation between such extremely close populations has scarcely been reported in other marine fish species.
As marine lakes are completely isolated and differ in environmental conditions and biota from lagoons, they have a significantly higher possibility to speciate. Barluenga et al. (2006) have reported rapid speciation in the Midas chichlid species complex (Amphilophus spp.) in young (< 10,000 years) and small (diameter, ~5 km; max. depth, ~200 m) volcanic crater lakes in Nicaragua. Although a newly identified species, Amphilophus zaliosus (Barlow, 1976), and an ancestral species, A. citrinellus (Günther, 1864), in these lakes share identical haplotypes in the mtDNA control region, the two species are reproductively isolated and eco-morphologically distinct. Similar examples of speciation in extremely small lakes have been also reported for chichlids in the two volcanic crater lakes Barombi Mbo and Bermin (4.15 and 0.6 km2) in Cameroon, which harbor 11 and 9 endemic species, respectively (Schllewen et al., 1994). Beheregaray et al. (2002) have also identified the rapid radiation of the silverside Odonthesthes perugiae complex associated with the last sea-level changes in southern Brazil. Additionally, Vamosi (2003) has proposed that absence of predators could have affected speciation in the threespine stickleback Gasterosteus aculeatus Linnaeus, 1758 inhabiting lakes that formed ~12,000 years before present in the Strait of Georgia region of south-western British Columbia. As no predators for A. endrachtensis exist in JFL and BCM, this situation could alter selection pressure and the behavior of this species. In fact, we have previously reported that marine lake populations of S. orbicularis, differ in feeding and escaping behaviors from lagoon populations (Gotoh et al., 2009).
In conclusion, the present study has revealed that the examined marine lake populations of A. endrachtensis have been completely isolated and have differentiated from adjacent lagoon populations, and are experiencing different evolutionary processes. These findings clearly show that the marine lakes of Palau represent excellent environments for the evolutionary study of marine species.
We thank Drs. Y. Hara and J. Yokoyama of Yamagata University for their helpful advice and discussions. We also thank the Ministry of Resources and Development, Republic of Palau, for permitting us to collect fish samples, Marino, Urui, Vitk, Baste, I. Kishigawa, and other members of the Carp Corporation for supporting our sampling in Palau, and S. Kimura, Mie University, for providing Sulawesi samples. This work was conducted as a part of Grant-in-Aid for Science Research (B) from the Japan Society for the Promotion of Science (No. 16405012).
|
Index