Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Sea surface currents and geographic isolation shape the genetic population structure of a coral reef fish in the Indian Ocean

  • Filip Huyghe ,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Visualization, Writing – original draft

    filip.huyghe@vub.be

    Affiliation Marine Biology, Ecology and Biodiversity, Vrije Universiteit Brussel (VUB), Brussels, Belgium

  • Marc Kochzius

    Roles Conceptualization, Formal analysis, Funding acquisition, Investigation, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Marine Biology, Ecology and Biodiversity, Vrije Universiteit Brussel (VUB), Brussels, Belgium

Abstract

In this contribution, we determine the genetic population structure in the Skunk Clownfish (Amphiprion akallopsisos) across the Indian Ocean, and on a smaller geographic scale in the Western Indian Ocean (WIO). Highly restricted gene flow was discovered between populations on either side of the Indian Ocean using the control region as a mitochondrial marker (mtDNA). We verify this conclusion using 13 microsatellite markers and infer fine scale genetic structuring within the WIO. In total 387 samples from 21 sites were analysed using mtDNA and 13 microsatellite loci. Analysis included estimation of genetic diversity and population differentiation. A haplotype network was inferred using mtDNA. Nuclear markers were used in Bayesian clustering and a principal component analysis. Both markers confirmed strong genetic differentiation between WIO and Eastern Indian Ocean (EIO) populations, and a shallower population structure among Malagasy and East African mainland populations. Limited gene flow across the Mozambique Channel may be explained by its complex oceanography, which could cause local retention of larvae, limiting dispersal between Madagascar and the East African coast. Two other potential current-mediated barriers to larval dispersal suggested in the WIO, the split of the SEC at approximately 10° S and the convergence of the Somali Current with the East African Coast Current at approximately 3° S, were not found to form a barrier to gene flow in this species.

Introduction

Coral reefs are threatened worldwide by several human induced factors, such as overfishing, tourism, or pollution. They are at risk to convert to an alternative, macroalgae-dominated state that would alter their ecosystem characteristics. This process is further accelerated by human-induced environmental stresses, such as global warming and ocean acidification, making urgent conservation measures a necessity [1]. A very efficient and much used conservation instrument is the establishment of marine protected areas (MPAs), protecting certain coral reef zones from fishing pressure and other potentially harmful activities. Correctly established and managed MPAs serve several purposes: they fulfil a role as sanctuaries for both species and genetic diversity, but are also intended to strengthen ecosystem resilience and promote recovery after disturbances beyond their boundaries. To fulfil this dual role, the spacing of MPAs should take into account the trade-off between the protection of sites that are not connected by gene flow and have a unique genetic composition, and well-connected sites that promote ecosystem resilience better [2]. In addition, MPA spacing should respect the balance between promoting resilience through well connected MPAs and providing spill-over of new recruits to exploited adjacent areas and benefit fisheries [3].

Since gene flow is almost exclusively realised through larval dispersal in coral reef associated organisms, patterns of larval dispersal have been studied to evaluate connectivity in these organisms using various methods, such as population genetics and phylogeography, direct tagging of larvae, parentage analysis, otolith chemistry, and biophysical models [4, 5]. An overall pattern of larval dispersal has not been described yet, and empirical data obtained so far describe different degrees and configurations of population structure, indicating that larval dispersal may be influenced by a multitude of still unknown environmental and behavioural factors. Consequently, conservation managers are urged to plan MPA spacing in a precautious manner and scientists to continue gathering information of larval dispersal and connectivity among marine populations [6].

Results of population genetic studies of reef associated organisms spanning the whole Indian Ocean basin display a wide variety of population structure, with some revealing the existence of a barrier to gene flow between the Western Indian Ocean (WIO) and the Eastern Indian Ocean (EIO) [710], while others detect highly connected populations on both sides of the Indian Ocean [11, 12]. Because of the importance of the pelagic larval phase for dispersal, patterns are likely to be influenced by sea surface currents, moving drifting larvae in a specific direction. In the WIO, the prevailing South Equatorial Current (SEC) flows from East to West across the Indian Ocean until it reaches Madagascar, where one component splits into the North (NEMC) and South-East Madagascar Current (SEMC), while another one is joint by the NEMC at the Northern tip of Madagascar (Fig 1). This component of the SEC continues West until it reaches the East African Coast at approximately ten degrees South, close to the border between Tanzania and Mozambique. There, it splits into the southward Mozambique Current (MC), which creates a number of eddies on its way through the Mozambique Channel, and the northward East African Coast Current (EACC), which converges with the seasonal (November-April) southward Somali Current (SC) at the northern coast of Kenya, joining into the South Equatorial Counter Current (SECC) that flows to the East (Fig 1) [13, 14]. This current regime has the potential to create barriers to gene flow in three different regions of the WIO: First, at 10° S between reefs North and South, where the SEC splits into the MC and EACC. Limited gene flow across this barrier has been demonstrated, albeit the actual zone of differentiation was located more to the South, around Central Mozambique, in one study [15], and more to the North in another [10]. Second, at the northern coast of Kenya, where the EACC and SC meet and create an off-shore current flowing East. Limited gene flow across this second potential barrier to dispersal was found in two different reef fish species [16, 17]. Third, the westward direction of the SEC and the eddies in the Mozambique Channel could form a barrier to dispersal among populations along the East African Coast and populations at the various islands in the Indian Ocean. Indications of limited gene flow across this barrier have been encountered repeatedly in reef associated and coastal species [17, 18]. However, several other studies on coral reef associated species have revealed high connectivity and gene flow across all of these barriers, leading to meta-populations lacking genetic structure across the WIO [1922].

thumbnail
Fig 1. Sampling sites, haplotype network, PCA and structure bar plot.

(A) Map of the Indian Ocean and (B) Western Indian Ocean (WIO) showing the sample sites for Amphiprion akallopisos and prevailing currents during the Northeast Monsoon: EACC: East African Coast Current; MC: Mozambique Current; MCE: Mozambique Channel Eddies; NEMC: Northeast Madagascar Current; NMC: Northeast Monsoon Current; SC: Somali Current; SECC: South Equatorial Counter Current [13, 14]. Potential distance- or current-mediated barriers to larval dispersal indicated with black lines (1): large area of open ocean separating the EIO from the WIO; (2): split of the SEC at approximately 10° S; (3) Confluence of EACC and SC; (4) Mozambique Channel Eddies. Sampling site codes as in Table 1. (C) Haplotype network based on control region sequences. Black circles represent haplotypes from in the WIO and white circles from the EIO; size of the circles corresponds to the number of individuals as indicated, with the largest circle representing 217 individuals; lines represent one, small dots additional mutational steps. (D) Scatterplot of the Principal Component Analysis (PCA) based on microsatellite data; black circles represent individuals from in the WIO and white circles from the EIO;1 (vertical): first axis; 2 (horizontal): second axis. (E) Bar plot showing each individual’s estimated membership fraction to each of two clusters (K = 2) and the distribution of these clusters among sampling sites (codes as in Table 1) as revealed in Bayesian clustering implemented in STRUCTURE.

https://doi.org/10.1371/journal.pone.0193825.g001

Another factor capable of influencing the results of a phylogeographic study is the choice of genetic markers. Because of its matrilineal inheritance, mtDNA is likely to be more severely influenced by bottleneck events in populations compared to nuclear DNA due to lower effective population size. Consequently, mtDNA diversity is often very low in populations having suffered bottlenecks and may not have returned to equilibrium state yet after such an event. This will mask more recent genetic differentiation and means that mtDNA will often only provide evidence for historical gene flow [23]. The use of mtDNA in phylogeographic studies was further criticised, because the effect of bottlenecks on genetic diversity in populations of different animal species was found to be very variable, independent from species abundance. This must be seen as an indication that mitochondrial genetic markers are under selection pressure in many animal species [24]. Although it was later added that this phenomenon was stronger in certain taxa than in others [25], there is a risk that the neutrality assumption is not respected for phylogeographic studies based on mitochondrial markers. However, lower effective population size in mtDNA makes the effect of genetic drift stronger on this locus, which contributes to a shorter coalescence time [26]. For example, in a fish species with a large population size and a recent colonisation history like Pleuronectes platessa, a mtDNA marker enabled to define population structure more precisely than nuclear markers, i.e. microsatellite loci [27]. Microsatellite markers were criticised, because their high mutation rate, repetitive nature, and complex mutation process increase the risk of homoplasy and cause other interpretative problems, such as null alleles, lowered Fst-values and allele drop out [26, 28]. Microsatellite loci, however, are abundant in most genomes, making it possible to conduct multi-locus studies, and thus making analyses more robust against coalescent stochasticity and selective sweeps, for which single locus mitochondrial markers are more vulnerable [29]. Considering advantages and disadvantages of both types of markers, a combination of both can be seen as the ideal strategy to infer population genetic structure in most cases [30].

In studies where both marker types are simultaneously used, mito-nuclear discordance, or differences between results from mtDNA and nuclear markers, is common. Several explanations, such as introgression, sex-biased dispersal, selection, and different effective population sizes have been suggested [31]. More specifically, an analysis of 14 studies on reef fishes revealed that only in four data sets mtDNA and microsatellite markers were congruent. In six studies, microsatellites revealed a population structure while mtDNA did not, and the opposite occurred in four other studies [32]. In the WIO, microsatellite markers revealed population structure in the soldierfish Myripristis berndtii, while mtDNA did not [17], but in the snapper Lutjanus kasmira both markers were congruent [22]. A few studies using both marker types have been conducted on clownfish. In a study on A. ocellaris in the Indo-Malay Archipelago, results from both markers were congruent, but population structure inferred from mtDNA was more pronounced [33]. In three other studies, however, the opposite was true, with microsatellite results revealing genetic differentiation not detected using mtDNA [3436].

The population structure of the skunk clownfish has been studied using mtDNA, showing highly restricted gene flow between populations in the WIO and the EIO. However, only a weak genetic population structure was detected in the WIO and none in the EIO [9]. In this study, we complement this analysis using thirteen microsatellite loci. Furthermore, the existing WIO dataset was extended with samples from three different sites in Mozambique and one additional site in Madagascar in order to span all three potential current-mediated barriers to gene flow in this region. We aim to investigate whether: (1) the genetic break between WIO and EIO is also detected using microsatellite markers; (2) barriers to gene flow could be detected within the WIO using this extended dataset and a combination of different markers.

Materials and methods

Sampling and DNA extraction

A total of 387 A. akallopisos individuals were sampled between one and 25 m depth at 18 sites in the WIO and three in the EIO (Table 1 and Fig 1a and 1b). Fish were caught with two hand nets in their host anemone while SCUBA-diving.

thumbnail
Table 1. Diversity of mitochondrial control region sequences and microsatellites in Amphiprion akallopisos from the Indian Ocean.

https://doi.org/10.1371/journal.pone.0193825.t001

A small piece of the caudal fin, from 5 mm by 5 mm up to 20 mm by 20 mm, depending on the size of the individual, was removed after which the fish was returned into its host anemone. Samples were preserved in 96% ethanol immediately after the dive. Research permits were obtained from the following institutions: TANZANIA COMMISSION FOR SCIENCE AND TECHNOLOGY (COSTECH) Ali Hassan Mwinyi Road, Kijitonyama Area. P.O. Box 4302, Dar es Salaam, Tanzania; Zanzibar Research Committee, PO Box 239, Stone Town, Zanzibar, Tanzania; Kenya: National Council for Science and Technology, Nairobi, Research permit No: NCST/RRI/12/1/BS/250; Mozambique: Universidade Eduardo Mondlane, Escola de Ciências Marinhas e Costeiras (ECMC), Quelimane; Indonesia: Universitas Hasanuddin, Makassar, Research permit No: 55/PSTK/UH/XI/04; Madagascar: Ministère de l'Environnement, de l'Ecologie et des Forêts; Autorisation de recherche: N°287/15/MEEMF/SG/DGF/DAPT/SCBT. DNA from the samples from Kenya, Tanzania, Indonesia, and Madagascar (partly) was extracted with the QIAGEN (Düsseldorf, Germany) extraction kit, and DNA from samples from Madagascar (Sainte Marie) and Mozambique was extracted with the E.Z.N.A. Tissue DNA Kit (Omega Bio-tek, Norcross, Georgia, USA), following the manufacturer’s protocol.

Control region (CR)

A fragment of the CR was amplified by polymerase chain reaction (PCR) using the primers CR-A (5’-TTCCACCTCTAACTCCCAAAGCTAG-3’) and CR-E (5’-CCTGAAGTAGGAACCAGATG-3’) [37]. The PCR was conducted in an Eppendorf Ep S Mastercycler with a volume of 50 μl for each PCR that contained 2 μL DNA template, 10 mM Tris-HCl (pH 9), 50 mM KCl, 4 mM MgCl2, 0.4 μM of each primer, 0.2 mM dNTPs, and 1 U Taq polymerase. The following temperature profile was used: 94 °C for 5 minutes, followed by 35–40 cycles of 1 minute at 94 °C, 1.5 minutes at 45 °C and 1 min at 72 °C. Final extension was conducted at 72 °C for 5 minutes. Sequencing was done with an ABI 3770XL automated sequencer (Applied Biosystems, Foster City, USA). The sequences were edited using the software ChromasPro (v. 1.5, Technelysium Ltd, UK). A multiple alignment was done using Clustal W [38] as implemented in the software BioEdit v. 7.0.0.1 [39]. Haplotype and nucleotide diversity, F-statistics, as well as analysis of molecular variance (AMOVA) were calculated with the software Arlequin v. 3.5 [40]. Multiple testing can lead to the occurrence of false positives as a consequence of the multitude of tests. We corrected for this multiplicity problem using the False Discovery Rate method that sets a lower significance threshold than p<0.05 in function of the number of independent tests [41]. The program MIGRATE v. 3.11.6 [42] was used to test migration rates in either direction among populations. After initial testing, the following start parameters were chosen to do Bayesian inference with constant mutation rate (since there was only one locus) and Metropolis algorithm to generate posterior distribution. The prior distribution parameters were set at minimum = 0; mean = 500; maximum = 1.000. One long chain was run, with samples taken every 20 steps, and 100.000 samples taken per chain, with a burn-in of 50.000 samples. Five replicates were run per analysis. Four parallel chains were run under a static heating scheme at following start temperatures: 1–1.5–3–100.000. Tests for selective neutrality of the marker, Tajima’s D-test [43] and Fu’s Fs-test [44], as well as the sum of squared deviation and Harpending’s raggedness index [45] to test Rogers’ model of sudden population expansion [46] were also conducted with the software Arlequin. A haplotype network was constructed using the programme TCS v. 1.21 [47].

Microsatellites

Primer candidates were identified in a literature study searching for microsatellite loci previously described for the genus Amphiprion (S1 Table). All found primer candidates were tested in a subset of the samples for the presence of the same locus in A. akallopisos. The 13 best amplifying primers were selected and divided into one set of six and one set of seven primer pairs based on the observed length of the PCR product (Table 2).

thumbnail
Table 2. Characterization of the 13 microsatellite loci in Amphiprion akallopisos with their respective motive, fluorescent label, PCR product length, number of alleles (Na), observed (Ho) and expected (He) heterozygosities.

The first six primers were used in one multiplex set, the next seven in another.

https://doi.org/10.1371/journal.pone.0193825.t002

Multiplex PCR was conducted using the four different fluorescent labels 6-FAM, VIC, PET and NED (Applied Biosystems, Foster City, CA, USA) with these two sets of primer pairs. The PCR was conducted in an Eppendorf Ep S Mastercycler with a volume of 12.5 μL for each PCR that contained 2.5 μL DNA template and 10 μL of a master mix containing 1.25 μL primer mix (containing 2 μM of each primer), 6.25 μL Multiplex PCR Master Mix (QIAGEN, Hilden, Germany), containing optimised concentrations of HotStarTaq® Plus DNA Polymerase, MgCl2, and dNTPs and Multiplex PCR Plus Buffer (with Factor MP), and 2.5 μL H2O. The following temperature profile was used: 95 °C for 5 minutes once, then 95 °C for 30 seconds, 1.5 minutes at 57 °C for annealing and 30 seconds at 72 °C for extension, repeated during 35 cycles. Finally, a last extension was done at 68 °C for 30 minutes. The PCR product was diluted 20 times and mixed with buffer (HiDi) and GeneScan-500 LIZ® size standard (Applied Biosystems, Darmstadt, Germany) and then analysed on an ABI 3100 Automated Sequencer (Applied Biosystems). The resulting electropherograms were scored manually with GeneMarker V2.6.3 (SoftGenetics, State College, PA, USA). The dataset was tested for the presence of null alleles with Microchecker v1.0 [48]. To test for Hardy-Weinberg Equilibrium (HWE), assess genetic diversity, observed and unbiased expected heterozygosity, as well as the number of private alleles per population, we used the program GenAlEx 6.5 [49]. Arlequin was used for F-statistics and to calculate AMOVA. The program MIGRATE v. 3.11.6 [42] was used to test migration rates in either direction among populations. After initial testing, Bayesian inference with constant mutation rate among loci and Metropolis algorithm to generate posterior distribution was run. The prior distribution parameters were set at minimum = 0; mean = 800; maximum = 8.000. One long chain was run, with samples taken every 100 steps, and 500.000 samples taken per chain, with a burn-in of 100.000 samples. Two replicates were run per analysis. Four parallel chains were run under a static heating scheme at following start temperatures: 1–1.5–3–100.000. The software STRUCTURE 2.3.4 [50], which uses a Bayesian clustering based on Markov chain Monte Carlo (MCMC) assignment method, was run without prior population information and under the admixture model to determine the number of genetic clusters (K). STRUCTURE was run for K = 1–21 for the whole dataset, for K = 1–18 for the WIO, and for K = 1–3 for the EIO, using 10 runs with a burn-in length of 100,000 and 1,000,000 MCMC replications. The most likely true number of clusters was determined with Evanno’s test using ΔK [51].

Finally, a Principal Component Analysis (PCA) and Discriminant Analysis of Principal Components (DAPC), which uses an algorithm that clusters individuals in groups optimising between group variance and minimising within-group variance, was done using the program AdeGenet in R [52].

Assessment of the influence of sampling size and confidence interval on genetic structure

When analysing population genetic structure, the accuracy of the analysis critically depends on the capacity of the sampled individuals to represent the gene pool of the population they originated from [53]. Especially with potentially highly polymorphic markers like microsatellites, there is a risk of overrepresentation of rare alleles, alleles occurring less than 1% in the real population, when the number of sampled individuals per population is low. Typically, when less than 20 individuals are sampled in a population, there is an increased risk that the allele frequencies used to infer population genetic structure do not reflect the real allele frequencies of the population [53]. This in turn can lead to artificially high Fst-values (or equivalent) and therefore false positive errors in population structure inference, although a higher number of loci seems to temper this effect [53, 54]. Our research involved sampling of natural populations of clownfish in sometimes remote areas. For three of the analysed populations (Lamu, Tofo, and Sainte-Marie), we were not able to analyse more than eight individuals, and for several others an ideal sampling size of 20 individuals could not be reached. The possible influence of reduced sample size on our results was therefore tested with reduced datasets of eight individuals per population, the lowest number of individuals per population encountered. We ran two tests. First, we withheld the first eight individuals of each population and discarded the others. Second, we randomly selected eight individuals per population, using a prize draw site (https://www.dcode.fr/tirage-au-sort-nombre-aleatoire, last accessed 26/12/2017) and discarded the other individuals. We then compared the ϕst, ϕct, Fst, and Fct-values obtained in these reduced datasets with the values obtained using the complete dataset. We calculated the difference between the value of the complete dataset and the reduced datasets and then tested whether the set of differences was statistically different from 0, using the z-test function in Excel.

Furthermore, it has been pointed out that low ϕ/Fst-values (< 0.1), even though significantly different from 0, can be misleading as estimates for gene flow [55]. To compensate for this ambiguity, we calculated the 95% confidence intervals of all pairwise ϕ/Fst-values and Fct-values of AMOVA significantly different from 0 in Arlequin using 20.000 permutations and only accepted the ϕ/Fst-values where the 95% confidence interval did not overlap with 0. Only when the 95% confidence interval does not overlap with 0, it can be accepted as an indication of limited gene flow among populations [56].

Results

Control region: Amplification and neutrality testing

Sequences from 74 individuals from three different sites in Mozambique as well as from 23 additional individuals from two different sites in Madagascar were successfully edited and aligned with the 263 sequences from Huyghe & Kochzius (2016) [9], resulting in an alignment of 360 sequences with a length of 337 base pairs. Sequence information was uploaded to the European Nucleotide Archive (accession numbers L824024-824092). Deviation from Hardy-Weinberg equilibrium in both the WIO and the EIO was detected in this enlarged dataset, confirming the results obtained in Huyghe & Kochzius (2016) (Fu’s F & Tajima’s D). For both regions, a recent population expansion was identified with the sum of squared deviation test and Harpending’s raggedness index, again confirming the results of Huyghe & Kochzius (2016).

Microsatellites: Characterisation and neutrality testing

One of the 16 initially selected primer pairs (AC137; S1 Table) did not produce a PCR product in several individuals and was discarded from further analysis. Of the resulting 15 loci, two (AM10 and AM6; S1 Table) showed signs of null alleles in all populations, and were therefore also removed from the dataset. For the 13 remaining loci (Table 2), there was no indication of linkage disequilibrium and there was no indication for departure from HWE for any of the populations across loci. The number of alleles per locus for the 13 markers used in the analysis varied between 5 (locus 120 & Am7) and 41 (locus 61). Observed heterozygosity ranged between 0.261 (locus 120) and 0.890 (D103) and was very close to expected heterozygosity for all markers. No significant heterozygosity deficit or excess was detected for any of the loci. These results allow us to conclude that the dataset is appropriate for the proposed analysis.

Genetic diversity: Control region and microsatellites

In the dataset of 360 individuals, 98 haplotypes were identified, of which none occurred simultaneously in the WIO and the EIO (Fig 1c). Just like in Huyghe & Kochzius (2016), however, the WIO and EIO individuals did not form monophyletic groups. Haplotype diversity was much higher in the EIO (1.000 to 0.985) than in the WIO (0.786 to 0.199), as was the haplotype per individual ratio per population (EIO: 1.000–0.800; WIO: 0.625–0.105). Nucleotide diversity was on average an order of magnitude higher and at least more than double as high in the EIO populations (2.64–3.29%) compared to the WIO populations (0.06–1.06%). Within the WIO, the Dar es Salaam population consistently registered the lowest diversity values, and the Sainte Marie population the highest (Table 1). The higher diversity in the EIO can also be seen in the higher divergence, i.e. more mutational steps, among haplotypes found in EIO individuals than in WIO individuals (Fig 1c). In the WIO, 217 out of 312 individuals (69%) belonged to the same dominant haplotype and most other individuals belonged to haplotypes that differed by only one mutational step from this central haplotype, giving the section of the network to which most African individuals belong a typical star like appearance.

As for the microsatellite data, mean observed heterozygosity over all loci was moderate in both the EIO populations (0.769–0.646) and the WIO populations (0.743–0.520). Allelic richness was higher in the EIO (11.15–8.85) than in the WIO (9.62–4.69).

Genetic structure

Control region.

The threshold to accept ϕst-values as significantly different from 0 applying the False Discovery Rate approach was set at p<0.0241. All values significantly different from 0 were tested for overlap of the 95% confidence interval with 0 (S2 Table). Only the values without overlap were accepted. The results from Huyghe & Kochzius (2016) regarding population structure in the Indian Ocean were confirmed with an AMOVA (overall ϕst = 0.24; p < 0.001) and pairwise ϕst-values (Table 3). Strongly restricted gene flow between EIO and WIO populations, already suggested by the absence of shared haplotypes, was detected in this enlarged dataset. Within the WIO, significant population structure was detected by AMOVA (overall ϕst = 0.029; p < 0.001). Population structure was clearly caused by the differentiation of the population of Sainte Marie at the East coast of Madagascar from the other WIO populations (all pairwise ϕst-values significantly different from 0, except with Lamu, Misali, Jambiani, and Tofo (Table 3)). Hierarchical AMOVA, testing for population structure across potential oceanographic barriers, indicated population structure between EIO and WIO (Barrier 1; Table 4). Within the WIO, there was no indication of population structure across the other three potential oceanographic barriers when explicitly tested (Table 4).

thumbnail
Table 3. Pairwise Fst (Microsatellites: above diagonal)/ϕst(mtDNA: under diagonal) of Amphiprion akallopisos populations from the Indian Ocean with adapted significance levels.

https://doi.org/10.1371/journal.pone.0193825.t003

thumbnail
Table 4. ϕct- and Fct–values of hierarchical AMOVAs based on mtDNA (CR) and microsatellite data of Amphiprion akallopisos populations from the Indian Ocean across potential oceanographic barriers to dispersal.

https://doi.org/10.1371/journal.pone.0193825.t004

The results from the Migrate analysis confirmed the very limited gene flow between the WIO and the EIO. The 95% posterior distribution interval for migration in both directions between WIO and EIO overlapped with 0. The distribution interval gives the probability that the actual value of the assessed parameter falls within the interval. Within the WIO, 95% posterior distribution also overlapped with 0, except for estimates of migration from West-Madagascar to East-Madagascar. In the EIO, gene flow was equal and relatively high in both directions (Table 5).

thumbnail
Table 5. Bayesian estimates of migration rates among populations of Amphiprion akallopisos using the software Migrate.

https://doi.org/10.1371/journal.pone.0193825.t005

Microsatellites.

The same False Discovery Rate threshold (p<0.0241) as for mtDNA analysis was applied, and Fst—values significantly different from 0 were equally checked for 95% confidence interval overlap with 0 (S1 Table). Differentiation between WIO and EIO populations was also detected with microsatellite markers. Hierarchical AMOVA grouping WIO against EIO yielded a significant Fct-value (Fct = 0.11; p < 0.001; Table 4) and all pairwise Fst-values between WIO and EIO populations were also significant, ranging from 0.070 to 0.151 (Table 3). This was confirmed by a Principal Component Analysis (PCA), where the first axis clearly distinguished the EIO populations from the WIO populations (Fig 1d). Low levels of gene flow between both regions were also estimated with the software Migrate, although the 95% posterior distribution was higher than 0, with a higher migration from the WIO towards the EIO (Table 5).

The highest ΔK in Evanno’s test was reached for K = 2 (ΔK = 1312.862). In a STRUCTURE analysis conducted under this scenario the WIO populations were almost exclusively assigned to cluster one and EIO populations almost exclusively assigned to cluster two (Fig 1e). When a DAPC was conducted without prior definition of clusters, but forcing the dataset into 2 clusters, the EIO individuals were also clearly separated from the WIO individuals on the first axis (Fig 1d). These results form a strong indication of the pronounced genetic differentiation between the EIO and the WIO populations.

Among WIO populations, a shallow but significant population structure was detected (overall Fst = 0.012; p < 0.001). Three distinct groups could be derived from the pairwise Fst-values. One population from the North of Kenya, Watamu, had a high number of significant pairwise Fst-values (12/17 pairwise comparisons) with the other populations of the WIO, as well as the three populations from Madagascar: Nosy Bé, Tuléar and Sainte Marie (9/17, 7/17 and 2/17, respectively; Table 3). This suggests the existence of three geographically distant, differentiated groups which are separated by oceanographic barriers 3 and 4. Hierarchical AMOVA confirmed differentiation between populations on either side of barrier 4, between Madagascar and East-Africa, but not of barrier 3, between northern Kenya and populations further South (Table 4).

Within both the WIO and EIO regions, relatively high values of migration were measured in all directions with Migrate (Table 5). Within the WIO, estimated levels of migration were threefold higher from East to West than in the other direction (Table 5).

Under Evanno’s test, K = 2 had the highest ΔK (542.896) followed by K = 3 (89.823). When a STRUCTURE analysis was conducted with both scenarios, however, clustering of individuals did not correspond to two or three geographically distinct areas, but rather failed to assign individuals to either of the two or three clusters (S1 Fig).

In PCA analysis, individuals belonging to the Northern Kenyan populations (Watamu and Lamu), Madagascar, and the rest of the WIO populations, were not differentiated (S2 Fig). However, a DAPC, designed to optimise differentiation among individuals, did distinguish between Malagasi and East African populations when two clusters were set out on a single axis with Malagasi individuals mostly clustering within the same group. Such a distinction was not detected between the population of Northern Kenya (Watamu and Lamu) and the rest of the WIO populations (S3 Fig).

Assessment of limited sample size bias

In the two datasets with reduced sample size we composed and tested, the number of pairwise ϕst-values that were significantly different from 0 dropped from 81 to 56 (69%) and 55 (68%) in Control Region analysis (S3A Table). The number of pairwise Fst-values in microsatellite analysis that was significantly different from 0 dropped from 108 to 76 (70%) and 64 (59%) in the reduced datasets (S3B Table). Furthermore, ϕ/Fst-values were on average lower when the reduced datasets were used, and the difference between reduced dataset values and complete dataset values was not significantly different from 0 (S3A and S3B Table). Based on the Control Region sequences, the population structure derived from AMOVA was similar in the reduced datasets, with lower ϕct-values (S3D Table). In the microsatellite analysis, Fct-values were also lower in the reduced datasets, and the differentiation between the population from Watamu and the rest of the WIO was no longer supported, with 95% Confidence Intervals overlapping 0 and non-significant Fct-values. Genetic differentiation between the EIO and WIO, and between Malagasy and East-African populations, was still detected (S3E Table).

Genetic diversity was not higher in the populations of which we analysed only small sample sizes than in the populations in which more samples were analysed (Table 1). We therefore find no indication that the populations in our dataset for which a relatively small (less than 20) number of samples were analysed contain an excess number of rare alleles that could distort our assessment by artificially inflating differentiation among populations. Rather than increasing the risk of false positive results, a reduced sample size in our dataset seems to cause a decrease in differentiation among populations. The most plausible explanation for this is that decreasing the sample size simply decreases genetic diversity.

Discussion

EIO-WIO genetic break confirmed

Genetic differentiation between the WIO and EIO populations of the skunk clownfish, already demonstrated with mtDNA [9], was confirmed with microsatellite markers. These findings are congruent with earlier studies on algae, fishes, and invertebrates that indicate limited gene flow between the EIO and the WIO [7, 8, 10]. Both the Fst-values derived from AMOVA and the pairwise Fst-values from the microsatellite analysis were two to four times lower than the same values derived from mtDNA analysis. This is not surprising and can be attributed to two separate phenomena. First, because mtDNA is maternally inherited, effective population size is smaller, which results in higher genetic drift and faster differentiation among populations [26]. Second, within population diversity is relatively high when measured with microsatellite markers compared to mtDNA and this deflates Fst-values, without therefore affecting their significance [57].

Nuclear and mitochondrial markers were also congruent in detecting higher genetic diversity in the EIO compared to the WIO. Higher genetic diversity in the EIO in combination with the fact that EIO and WIO individuals do not form separate clades in our haplotype network supports the hypothesis that the EIO and WIO populations are composed of descendants from a pan-Indian Ocean population that originated in the EIO [9]. The haplotype network forms multiple, reticulate, links between the WIO and EIO groups that somewhat contradict the overall results of high differentiation between both regions. We think, however, that these reticulate links are merely artefacts caused by the high genetic diversity in the EIO population. Within the latter, the haplotype network contains a high number of mutational steps between haplotypes. It is possible that some of these mutational steps are actually intermediate haplotypes we have not been able to sample or which became extinct. Only additional sampling in the EIO could solve this issue. Another possibility for the rather unusual haplotype network could be that a former population in the central Indian Ocean became extinct and intermediate haplotypes between the WIO and EIO haplotypes got lost.

Population structure in the WIO

East African mainland versus Madagascar.

There was a sign of genetic differentiation between Malagasy populations on one side and African mainland populations on the other side. These results are congruent with those of two other studies in the WIO that also detected genetic differentiation between populations on both sides of the Mozambique Channel in a soldierfish [17] and a mangrove crab [18], and a third study where genetic differentiation between populations on the Mascarene Plateau and eastern Madagascar on one hand and populations within the Mozambique Channel and in East Africa on the other hand was discovered in a grouper [58]. They do not concur with studies on the blue starfish [10], the Dory snapper [19], the Kashmir snapper [22], and the crown-of-thorns starfish [21], where no genetic differentiation between populations on both sides of the Mozambique Channel was detected. It is, however, difficult to compare the results of these studies, because samples were collected at different sample sites and often only partly cover the Mozambique Channel. Furthermore, not all of these studies use identical genetic markers. In the future, a more integrated approach among research groups, using identical sampling sites and genetic markers across species when possible, would greatly improve our insight in marine connectivity in the WIO.

The Mozambique Channel sea surface currents are characterised by a series of anticyclonic mesoscale eddies interspersed by smaller cyclonic eddies that move water slowly southward through the channel and are capable of trapping particles for several months [59], creating an effective barrier to gene flow between Madagascar and the African mainland. However, frontal zone transport also seems to exist between the edges of these eddies, occasionally creating connectivity between Madagascar and the African mainland. Experimental drifters have been recorded to cross the Mozambique Channel using these frontal zones in as little as 15 days [59], which corresponds exactly to the estimated PLD of A. akallopisos larvae [60]. Such a combination of overall larval retention mediated by mesoscale eddies preventing continuous gene flow and occasional connectivity through frontal zone transport offsetting complete differentiation may perfectly explain the pattern of moderate genetic differentiation we encountered. In such a scenario, gene flow from Madagascar towards the East African coast would be more frequent than in the other direction, which corresponds to our findings. Geographically intermediate populations of skunk clownfish occur on several islands and atolls in the Mozambique Channel. Recently, significant population structure was detected among populations of the skunk clownfish on four of these islands, indicating two barriers to gene flow for this species within the Mozambique Channel, separating northern, central, and southern populations. The complex oceanography and more specifically eddies in the Mozambique Channel were suggested as causes for these barriers to gene flow [61]. The genetic relatedness of these island populations with the populations studied here, however, is not known. However, populations of several other reef fishes did not show significant genetic structure across [17, 22, 58] or within [61] the Mozambique Channel.

Both marker types detected genetic differentiation across the Mozambique Channel, but mtDNA only identified the Sainte Marie population in East Madagascar as differentiated, whereas microsatellites indicated genetic differences among Madagascar and the African mainland. Furthermore, with both marker types there was an indication of stronger gene flow from Madagascar to the East African coast than in the other direction, but in mtDNA the signal was restricted to reduced gene flow from eastern Madagascar to the West. One possible explanation for the discordance in this context is the higher resolving power of several microsatellite markers combined over mtDNA as a single locus [29]. The Sainte Marie population is separated from the East African mainland by a larger geographic distance than the other Malagasy populations included in this study. In addition to this, larval dispersal from Sainte Marie to the other reefs may not only be hampered by the sea surface current configuration in the Mozambique Channel as described earlier, but further restricted by the SEC that splits into the NEMC and the SEMC at approximately the location of Sainte Marie, i.e. 17° South [13]. Therefore, Sainte Marie might be isolated more strongly from the East African coast than the populations from West Madagascar and mtDNA markers might be able to detect this signal of genetic differentiation along with microsatellites, but not the weaker signal between western Madagascar and East Africa.

However, after corrections for false discovery and confidence interval uncertainty, Sainte Marie was only significantly differentiated from two African mainland populations (Stone Town and Mafia) based on microsatellites. This could be a consequence of the low sample size at this site. As illustrated with tests using a reduced number of samples for all populations (S3A and S3B Table), a lower sample size seems to decrease statistical power in our dataset. With microsatellite markers, where intra-population diversity can be high when many alleles exist per locus, the loss of statistical power can be more explicit than with less variable markers like mtDNA [53].

Another possible explanation can be derived from the fact that mtDNA is maternally inherited. The skunk clown fish, like all anemone fishes, lives in size-based hierarchical groups with only one female per group of four to eight individuals [62, 63]. Effective population size is therefore strongly reduced for mtDNA, which can cause founder effects and promotes genetic drift in an isolated population like Sainte Marie. Yet another explanation could be that repetitive microsatellite sequences evolve more rapidly than mtDNA and therefore the latter provides a more historical picture of gene flow. Estimates of migration rates suggest lower levels of migration based on mtDNA than on microsatellites, both between East Africa and Madagascar and among Malagasy populations. This could indicate that connectivity in the WIO has changed over time and both marker types reflect gene flow of a different time frame. In any case, this discordance between mtDNA and microsatellites further highlights the need to combine both marker types to obtain an optimal estimate of gene flow among populations of marine organisms [32].

Northern Kenya versus southern regions.

Contrary to the mtDNA analysis, pairwise Fst-values based on microsatellite markers suggested a moderate genetic break between the North Kenyan population of Watamu and other populations in the WIO. Geographically, genetic differentiation between the population at Watamu and populations further South corresponds to a suggested barrier to larval dispersal mediated by the confluence of the Somali Current and the EACC (barrier 3), which can prevent larval dispersal across this zone [13, 16, 17]. However, only the population of Watamu was differentiated from the rest, and not the population of Lamu, located further to the North of Watamu. If sea surface currents form a barrier to dispersal in South to North direction along the Kenyan coast, then we would expect to see differentiation between Lamu and more southern populations as well, but hierarchical AMOVA did not support this hypothesis. The genetic differentiation between Watamu and other populations could also be explained by the stochastic nature of larval dispersal, which could have created a pattern of chaotic genetic patchiness that can be responsible for the emergence of genetic differentiation among populations [64]. Genetic differentiation between Watamu and other populations, however, was not inferred by the analyses based on confidence intervals and clustering methods. When sample sizes were reduced in all populations, genetic differentiation was also no longer detected between Watamu and the other populations. Overall, statistical power dropped to a level that no longer permitted the detection of population structure within the WIO with a sample size of only eight individuals per population. This could suggest that population structure might be present in the WIO, but could not be detected due to the limited number of analysed samples in some populations. Only the analysis of additional samples from these sites could provide an answer to this question.

Conclusions

This study confirms earlier findings on a barrier to gene flow and larval dispersal separating the WIO and EIO populations of the skunk clownfish. It also identifies fine scale population structure within the WIO. Two of the three suggested current-mediated barriers to larval dispersal, the split of the SEC at approximately 10° S, and the convergence of the Somali Current with the EACC at approximately 3° S, were not found to form a barrier to gene flow in this species. The third, caused by the complex oceanographic nature of the Mozambique Channel, could promote genetic isolation among Malagasy and African mainland populations. The exact geographic location of this differentiation, however, shifted depending on the genetic markers used. Estimated gene flow within the WIO was found to be stronger from East to West than in the other direction, which corresponds to the direction of the dominant sea surface current, but was also different between marker types. The latter could be an indication of variability of gene flow over time. Furthermore, significant population structure has been detected among populations of the skunk clownfish within the Mozambique Channel, confirming the possible restricting influence of eddies on gene flow in the Mozambique Channel, at least in some reef fish species [61].

Supporting information

S1 Table. Overview of microsatellite markers tested for A. akallopisos.

https://doi.org/10.1371/journal.pone.0193825.s001

(DOCX)

S2 Table. 95% Confidence Intervals ϕst and Fst-values.

https://doi.org/10.1371/journal.pone.0193825.s002

(DOCX)

S3 Table.

(A) Pairwise ϕst-values among populations using reduced datasets. (B) Pairwise Fst-values among populations using reduced datasets. (C) Composition of reduced populations with eight randomly chosen individuals. (D) ϕct–values of hierarchical AMOVAs. (E) Fct–values of hierarchical AMOVAs.

https://doi.org/10.1371/journal.pone.0193825.s003

(DOCX)

S2 Fig. PCA of the WIO using 3 predefined groups.

https://doi.org/10.1371/journal.pone.0193825.s005

(DOCX)

S3 Fig. DAPC of WIO individuals defining 2 clusters.

https://doi.org/10.1371/journal.pone.0193825.s006

(DOCX)

Acknowledgments

We would like to thank the institutions and foundations that have made our work possible: Vlaamse Interuniversitaire Raad—Universitaire Ontwikkelingssamenwerking (VLIR-UOS) for a travel grant and VLADOC PhD scholarship to F. H.; Fonds Wetenschappelijk Onderzoek Vlaanderen (FWO) for funding the COCOMA-WIO project (Connectivity of Coral Reefs and Mangroves in the Western Indian Ocean; grant no. 1501612N); German Federal Ministry of Education and Research (BMBF) for funding the SPICE project (Science for the Protection of Indonesian Coastal Marine Ecosystems; grant no. 03F0390B); the competent authorities for permits; H. A. Ratsimbazafy for providing samples from Madagascar; R. M. van der Ven (Vrije Universiteit Brussel), M. Sheikh and M. S. Mohammed (State University of Zanzibar), Janne Timm (University of Bremen), as well as Jelvas M. Mwaura (Kenya Marine and Fisheries Research Institute) for assistance during fieldwork; K. De Winter, L. Otwoma, H. A. Ratsimbazafy, T. Sierens, and N. Vanbekbergen (Vrije Universiteit Brussel) for assistance during laboratory work.

References

  1. 1. Hoegh-Guldberg O, Mumby PJ, Hooten AJ, Steneck RS, Greenfield P, Gomez E, et al. Coral reefs under rapid climate change and ocean acidification. Science. 2007; 318: 1737–1742. pmid:18079392
  2. 2. Almany GR, Connolly SR, Heath DD, Hogan JD, Jones GP, McCook LJ, et al. Connectivity, biodiversity conservation and the design of marine reserve networks for coral reefs. Coral Reefs.2009; 28: 339–351.
  3. 3. Gaines SD, White C, Carr MH, Palumbi SR. Designing marine reserve networks for both conservation and fisheries management. Proc Natl Acad Sci U S A. 2010; 107(43): 18286–18293. pmid:20200311
  4. 4. Hellberg ME. Footprints on water: the genetic wake of dispersal among reefs. Coral Reefs. 2007; 26: 463–473.
  5. 5. Jones GP, Almany GR, Russ GR, Sale PF, Steneck RS, van Oppen MJH, et al. Larval retention and connectivity among populations of corals and reef fishes: history, advances and challenges. Coral Reefs. 2009; 28: 307–325.
  6. 6. McCook LJ, Almany GR, Berumen ML, Day JC, Green AL, Jones GP, et al. Management under uncertainty: guide-lines for incorporating connectivity into the protection of coral reefs. Coral Reefs. 2009; 28: 353–366.
  7. 7. Ridgway T, Sampayo EM. Population genetic status of the Western Indian Ocean: What do we know? Western Indian Ocean Journal of Marine Science. 2005; 4: 1–9.
  8. 8. Hui M, Kraemer WE, Seidel C, Nuryanto A, Joshi A, Kochzius M. Comparative genetic population structure of three endangered giant clams (Tridacnidae) throughout the Indo-West Pacific: implications for divergence, connectivity, and conservation. J Molluscan Stud. 2016; 82: 403–414.
  9. 9. Huyghe F, Kochzius M. Highly restricted gene flow between disjunct populations of the Skunk Clownfish (Amphiprion akallopisos) in the Indian Ocean. Mar Ecol. 2016; 38: e12357.
  10. 10. Otwoma LM, Kochzius M. Genetic population structure of the coral reef sea star Linckia laevigata in the Western Indian Ocean and Indo-West Pacific. PLoS ONE. 2016; 11(10): e0165552. pmid:27798700
  11. 11. Craig T, Elbe JA, Bowen BW, Robertson DR. High genetic connectivity across the Indian and Pacific Oceans in the reef fish Myripristis berndti (Holocentridae). Mar Ecol Prog Ser. 2007; 334: 245–254.
  12. 12. Horne JB, van Herwerden L, Choat JH, Robertson DR. High population connectivity across the Indo-Pacific: Congruent lack of phylogeographic structure in three reef fish congeners. Mol Phylogenet Evol. 2008; 49: 629–638. pmid:18804542
  13. 13. Schott FA, McCreary JP. The monsoon circulation of the Indian Ocean. Prog Oceanogr. 2001; 51: 1–123.
  14. 14. Schouten MW, de Ruijter WPM, van Leeuwen PJ, Ridderinkhof H. Eddies and variability in the Mozambique Channel. Deep-Sea Research II. 2003: 50; 1987–2003.
  15. 15. Silva IC, Mesquita N, Paula J. Genetic and morphological differentiation of the mangrove crab Perisesarma guttatum (Brachyura: Sesarmidae) along an East African latitudinal gradient. Biol J Linn Soc Lond. 2010: 99: 28–46.
  16. 16. Visram S, Yang M-C, Moothien Pillay R, Said S, Henriksson O, Grahn M, et al. Genetic connectivity and historical demography of the blue barred parrotfish (Scarus ghobban) in the western Indian Ocean. Mar Biol. 2010; 157: 1475–1487.
  17. 17. Muths D, Tessier E, Gouws G, Craig M, Mwale M, Mwaluma J, et al. Restricted dispersal of the reef fish Myripristis berndti at the scale of the SW Indian Ocean. Mar Ecol Prog Ser. 2011; 443: 167–180.
  18. 18. Ragionieri L, Cannicci S, Schubart CD, Fratini S. Gene flow and demographic history of the mangrove crab Neosarmatium meinerti: A case study from the western Indian Ocean. Estuar Coast Shelf Sci. 2010; 86: 179–188.
  19. 19. Dorenbosch M, Pollux BJA, Pustjens AZ, Rajagopal S, Nagelkerken I, van der Velde G, et al. Population structure of the Dory snapper, Lutjanus fulviflamma, in the western Indian Ocean revealed by means of AFLP fingerprinting. Hydrobiologia. 2006; 568: 43–53.
  20. 20. Hoareau TB, Bosc P, Valade P, Berrebi P. Gene flow and genetic structure of Sicyopterus lagocephalus in the south-western Indian Ocean, assessed by intron-length polymorphism. J Exp Mar Bio Ecol. 2007; 349: 223–234.
  21. 21. Vogler C, Benzie J, Barber PH, Erdmann MV, Ambariyanto , Sheppard C, et al. Phylogeography of the Crown-of-Thorns Starfish in the Indian Ocean. PLoS ONE. 2012; 7(8): e43499. pmid:22927975
  22. 22. Muths D, Gouws G, Mwale M, Tessier E, Bourjea J. Genetic connectivity of the reef fish Lutjanus kasmira at the scale of the western Indian Ocean. Can J Fish Aquat Sci. 2012; 69: 842–853.
  23. 23. Fauvelot C, Bernardi G, Planes S. Reductions in the mitochondrial DNA diversity of coral reef fish provide evidence of population bottlenecks resulting from Holocene sea-level change. Evolution. 2003; 57(7): 1571–1583. pmid:12940362
  24. 24. Bazin E, Glémin S, Galtier N. Population size does not influence mitochondrial genetic diversity in animals. Science. 2006; 312: 570–572. pmid:16645093
  25. 25. Mulligan CJ, Kitchen A, Miyamoto MM. Comment on “Population size does not influence mitochondrial genetic diversity in animals”. Science. 2006; 314: 1390a.
  26. 26. Zink RM, Barrowclough GF. Mitochondrial DNA under siege in avian phylogeography. Mol Ecol. 2008; 17: 2107–2121. pmid:18397219
  27. 27. Hoarau G, Piquet AM-T, van der Veer HW, Rijnsdorp AD, Stam WT, Olsen JL. Population structure of plaice (Pleuronectes platessa L.) in northern Europe: a comparison of resolving power between microsatellites and mitochondrial DNA data. J Sea Res. 2004; 51: 183–190.
  28. 28. Brito PH, Edwards SV. Multilocus phylogeography and phylogenetics using sequence-based markers. Genetica. 2009; 135: 439–455. pmid:18651229
  29. 29. Edwards S, Bensch S. Looking forwards or looking backwards in avian phylogeography? A comment on Zink and Barrowclough 2008. Mol Ecol. 2009; 18: 2930–2933. pmid:19552688
  30. 30. Barrowclough GF, Zink RM. Funds enough, and time: mtDNA, nuDNA and the discovery of divergence. Mol Ecol. 2009; 18: 2934–2936.
  31. 31. Toews DPL, Brelsford A. The biogeography of mitochondrial and nuclear discordance in animals. Mol Ecol. 2012; 21: 3907–3930. pmid:22738314
  32. 32. DiBattista JD, Rocha LA, Craig MT, Feldheim KA, Bowen BW. Phylogeography of two closely related Indo-Pacific Butterflyfishes reveals divergent evolutionary histories and discordant results from mtDNA and microsatellites. J Hered. 2012; 103(5): 617–629. pmid:22888133
  33. 33. Timm J, Planes S, Kochzius M. High similarity of genetic population structure in the False Clown Anemonefish (Amphiprion ocellaris) found in microsatellite and mitochondrial control region analysis. Conserv Genet. 2012; 13: 693–706.
  34. 34. Van der Meer MH, Hobbs J-PA, Jones GP, van Herwerden L. Genetic connectivity among and self-replenishment within island populations of a restricted range subtropical reef fish. PLoS ONE. 2012; 7(11): e49660. pmid:23185398
  35. 35. Van der Meer MH, Jones GP, Hobbs J-PA, van Herwerden L. Historic hybridization and introgression between two iconic Australian anemonefish and contemporary patterns of population connectivity. Ecol Evol. 2012; 2(7): 1592–1604. pmid:22957165
  36. 36. Dohna TA, Timm J, Hamid L, Kochzius M. Limited connectivity and a phylogeographic break characterize populations of the pink anemonefish, Amphiprion perideraion, in the Indo-Malay Archipelago: inferences from a mitochondrial and microsatellite loci. Ecol Evol. 2015; 5(8): 1717–1733. pmid:25937914
  37. 37. Lee WJ, Conroy J, Huntting-Howell W, Kocher TD. Structure and evolution of teleost mitochondrial control regions. J Mol Evol. 1995; 41: 54–66. pmid:7608989
  38. 38. Thompson JG, Higgins DG, Gibson TJ. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignments through sequence weighting, position specific gap penalties and weight matrix choice. Nucleic acids Res. 1994; 22: 4673–4680. pmid:7984417
  39. 39. Hall TA. BIOEDIT: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucleic Acids Symp Ser (Oxf). 1999; 41: 95–98.
  40. 40. Excoffier L, Fischer HEL. Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Mol Ecol Resour. 2010; 10: 564–567. pmid:21565059
  41. 41. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: a practical and powerful approach to multiple testing. J. R. Statist. Soc. B. 1995; 57: 289–300.
  42. 42. Beerli P, Felsenstein J. Maximum likelihood estimation of a migration matrix and effective population sizes in n subpopulations by using a coalescent approach. Proc Natl Acad Sci U S A. 2001; 98: 4563–4568. pmid:11287657
  43. 43. Tajima F. Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics. 1989; 123: 585–595. pmid:2513255
  44. 44. Fu YX. Statistical tests of neutrality of mutation against population growth, hitchhiking and background selection. Genetics. 1997; 147: 915–925. pmid:9335623
  45. 45. Harpending HC. Signature of ancient population growth in a low-resolution Mitochondrial DNA Mismatch Distribution. Hum Biol. 1994; 66: 591–600. pmid:8088750
  46. 46. Rogers AL. Genetic evidence for a Pleistocene population explosion. Evolution. 1995; 49: 608–615. pmid:28565146
  47. 47. Clement M, Posada D, Crandall KA. TCS: a computer program to estimate gene genealogies. Mol Ecol. 2000; 9: 1667–1669.
  48. 48. Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P. MICRO-CHECKER: software for identifying and correcting genotyping errors in microsatellite data. Mol Ecol Notes. 2004; 4: 535–538.
  49. 49. Peakall R, Smouse PE. GenAlEx 6.5: genetic analysis in Excel. Population genetic software for teaching and research-an update. Bioinformatics. 2012; 28: 2537–2539. pmid:22820204
  50. 50. Pritchard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data. Genetics. 2000; 155: 945–959. pmid:10835412
  51. 51. Evanno G, Regnaut S, Goudet J. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol. 2005; 14: 2611–2620. pmid:15969739
  52. 52. Jombart T, Devillard S, Balloux F. Discriminant analysis of principal components, a new method for the analysis of genetically structured populations. BMC Genet. 2010; 11: 94. pmid:20950446
  53. 53. Ryman N, Palm S, André C, Carvalho GR, Dahlgren TG, Jorde PE et al. Power for detecting genetic divergence: differences between statistical methods and marker loci. Mol Ecol. 2006; 15: 2031–2045. pmid:16780422
  54. 54. Hale ML, Burg TL, Steeves TE. Sampling for Microsatellite-Based Population Genetic Studies: 25 to 30 Individuals per Population Is Enough to Accurately Estimate Allele Frequencies. PLoS ONE 2012; 7(9): e45170. pmid:22984627
  55. 55. Landguth EL, Fedy BC, Oyler-McCance SJ, Garey AL, Emel SL, Mumma M et al. Effects of sample size, number of markers, and allelic richness onthedetection of spatialgenetic pattern. Mol Ecol Resour. 2012; 12: 276–284.
  56. 56. Palsboll PJ, Peery MZ, Bérubé M. Detecting populations inthe‘ambiguous’ zone: kinship-based estimation of population structure at low genetic divergence. Mol Ecol Resour. 2010; 10: 797–805. pmid:21565091
  57. 57. Meirmans PG, Hedrick PW. Assessing population structure: Fst and related measures. Mol Ecol. 2011; 11: 5–18.
  58. 58. Muths D, Tessier E, Bourjea J. Genetic structure of the reef grouper Epinephelus merra in the West Indian Ocean appears congruent with biogeographic and oceanographic boundaries. Mar Ecol. 2015; 36: 447–461.
  59. 59. Hancke L, Roberts MJ, Ternon JF. Surface drifter trajectories highlight flow pathways in the Mozambique Channel. Deep Sea Res A. 2014; 100: 27–37.
  60. 60. Dhaneesh KV, Nanthini Devi K, Ajith Kumar TT, Balasubramanian T, Tissera K. Breeding, embryonic development and salinity tolerance of Skunk clownfish Amphiprion akallopisos. Journal of King Saud University—Science. 2012; 24: 201–209.
  61. 61. O’Donnell JL, Beldade R, Mills SC, Williams HE, Bernardi G. Life history, larval dispersal and connectivity in coral reef fish among the Scattered Islands of the Mozambique Channel. Coral Reefs. 2017; 36: 223–232.
  62. 62. Buston PM. Size and growth modification in clownfish. Nature. 2003; 424: 145–146. pmid:12853944
  63. 63. Fautin DG, Allen GR. Field Guide to Anemone Fishes and Their Host sea Anemones. 1992; Western Australian Museum, Perth, Australia: 168 pp.
  64. 64. Selkoe KA, Gaines SD, Caselle JE, Warner RR. Current shifts and kin aggregations explain genetic patchiness in fish recruits. Ecology. 2006; 87: 3082–3094. pmid:17249233