The study was conducted in East African coastal waters. This coastal zone is characterised by mangrove forests, fringing coral reefs, sand beaches, and rock outcrops [28]. Molluscs, mudskippers, sesarmid crabs, fiddler crabs, and giant mud crabs are commonly found in the mangrove forests [29,30]. The climate of the region has two alternating and distinctive seasons, influenced by the southern and the northern monsoons, which have a marked effect on winds, rainfall, as well as air and water temperature [1]. Ocean currents are driven by trade winds, which are greatly influenced by the movement of the thermal equator (intertropical convergence zone). The currents have a direct influence on nutrient transport and potentially on larval dispersal. The westward South Equatorial Current (SEC) splits at around 17°S in front of the East coast of Madagascar and flows northward as the Northeast Madagascar Current (NEMC) and southward as the Southeast Madagascar Current (SEMC) [31]. The extension of the SEC northwest of Madagascar reaches the African coast around 11°S, where it splits into the northward East African Coast Current (EACC) and the southward Mozambique Current (MC) [32]. Flow in the Mozambique channel is dominated by eddies which propagate southward into the Agulhas Current (AC) (Fig 1).

Sampling of giant mud crabs (S. serrata) was conducted between 2011 and 2015. Tissue samples of 230 individual giant mud crabs were collected from mangrove forests in Kenya, Tanzania, Mozambique, Madagascar, and South Africa (Fig 1 and Table 1). The mud crabs were collected at low tide with the help of local fishermen or bought at landing sites. A section of the pereopod tissue was collected from each animal and preserved in 99.9% ethanol for further analysis.

Permission to collect samples was provided by the Tanzania Commission for Science and Technology (COSTEC), Tanzania Ministry of Agriculture, Livestock and Fisheries, the University of Tuléar (Madagascar), and the School of Marine and Coastal Sciences, Eduardo Mondlane University.

The eight analysed microsatellite loci did not show significant evidence of large allele drop out or scoring errors due to stuttering. In addition, the analysed loci did not show significant evidence of null alleles in all sampled populations, except locus Scse43-1 at site 4. The loci Ssse96-1 and Scpa-INI-SSR showed significant deviation from the HWE at site 3. The locus Scse43-1 showed significant departure from the HWE at sites 4 and 7. Significant departure from the HWE was also shown by the loci SCY38 and SCY12 at sites 6 and 10, respectively. The locus Scse53-1 was monomorphic at all sites, except site 1 (Table 2). The total number of alleles ranged between 2 and 22. Expected heterozygosity ranged between 0.561 and 0.601 (Table 3). The within sub-population inbreeding coefficients (F_IS) were not significantly different from zero (p > 0.00057 (adjusted nominal level)).

A total of 230 COI sequences each with 535 base pairs were obtained. Diversity indices were calculated only for sites with at least 14 sequences. The analysed sequences showed 40 haplotypes. The highest haplotype diversity was observed at sites 2 and 8 (Table 4). The lowest haplotype diversity was measured in samples from site 11. The current nucleotide diversity was generally low as it ranged between 0.07% (site 11) and 0.32% (site 8). In addition, the current nucleotide diversity was generally low than the historical nucleotide diversity (θ_π < θ_w).

A multiple alignment of the 230 sequences obtained during this study and the 125 previously published COI sequences was performed. The sequences were collapsed with the program FaBox DNA collapser ver. 1.41 [51] to generate 57 haplotypes (Table 5). The haplotype sequences were submitted to GenBank (accession numbers for haplotypes 1–57 = MF496045—MF496101). Fu’s Fs and Tajima’s D test of the pooled samples showed significant deviation from the neutral evolution hypothesis (Tajima’s D = -2.36, p < 0.001: Fu’s Fs = -27.48, p < 0.001). Mismatch distribution of the pooled samples produced a unimodal distribution, supporting the null hypothesis of population expansion (Fig 2). The raggedness index and sum of squared deviations (SSD) showed that the null hypothesis of population expansion cannot be rejected (raggedness index = 0.025, p > 0.05: SSD = 0.00087; p > 0.05). Fu’s Fs and Tajima’s D test were also performed for each population and they indicated significant deviation from the hypothesis of neutral evolution for all sampled populations, except populations at sites 6, 7, 11, 13, 14, and 15 (Table 6). The raggedness index for each population was not significant except for the population at site 8.

Correlation analysis showed that the association between the microsatellite based genetic differentiation (G_ST) and the within subpopulation expected heterozygosity (Hs) is not significant (G_ST = 0.0093Hs—0.0023: r = 0.327, p > 0.05). The analysis of molecular variance of the microsatellite data showed that the variation among sites was not significant (F_ST = 0.00424, p > 0.05, 100172 permutations). STRUCTURE analysis did not detect meaningful genetic clusters (Fig 1). Apart from that, the analysis of molecular variance (AMOVA) of the COI sequences revealed significant genetic differentiation among sites (F_ST = 0.158, p < 0.001; Φ_ST = 0.238, p < 0.001). Pairwise comparison of F_ST-values showed variable connectivity among the sample sites. With the exception of site 8, populations from Seychelles, Kenya, Tanzania, Mozambique, South Africa, and the West Coast of Madagascar did not show significant genetic differentiation (Table 7). Apart from that, a significant genetic break was observed for populations on the East coast of Madagascar (ECM) (sites 10 and 11), which were significantly differentiated from the other sample sites. This was confirmed by hierarchical AMOVA, which showed significant genetic differentiation between populations on the ECM and other sampled populations (F_CT = 0.361, p < 0.01: Φ_CT = 0.564, p < 0.01). The observed pattern of genetic differentiation was also revealed in the haplotype network (Fig 1). The network showed a star-like shape, with two main clusters of haplotypes joined to the main haplotype by few mutations. Haplotypes of populations on the ECM formed a separate cluster, which contained very few shared haplotypes (4 shared haplotypes and 4 private haplotypes). Sites from Kenya, Tanzania, and the Mozambique Channel showed high effective population size compared to other sites (Table 8). Estimates of the immigration rate showed high rate of immigration to Kenyan and Tanzanian mangroves and the lowest immigration rate to mangroves at the ECM.

The East African S. serrata populations showed high mitochondrial DNA haplotype diversity and low nucleotide diversity (Table 4). This is due to the fact that most haplotypes differed from each other by very few mutations (one to five mutations, S1 Table). Similar observations were previously reported in S. serrata in the WIO [26,27] and the Western Pacific [33]. Similar findings were also reported in other mangrove fauna in the WIO [19,24]. The high haplotype diversity and low nucleotide diversity might indicate genetic bottleneck events, where most haplotypes became extinct, followed by population expansion [57]. The measured haplotype diversity (h = 0.75) and nucleotide diversity (θ_π = 0.29%) are low compared to the reported levels of genetic diversity in the mangrove crabs Uca hesperiae (h = 0.80 ± 0.02, θ_π = 0.25 ± 0.16%), Perisesarma guttatum (h = 0.85 ± 0.02, θ_π = 0.42 ± 0.25%), and Neosarmatium africanum (h = 0.82 ± 0.02, θ_π = 0.46 ± 0.26%) [58] from the WIO. Nevertheless, the measured indices of genetic diversity are higher than the reported levels of genetic diversity in the mangrove crab Uca occidentalis (h = 0–0.679, θ_π = 0–0.13% [24]: h = 0.19 ± 0.03, θ_π = 0.03 ± 0.04% [58]) from the WIO, but comparable to the reported levels of genetic diversity in S. serrata in the WIO [26,27,33,58]. The measured indices of microsatellite diversity are also comparable to reported levels of microsatellite diversity in Scylla paramamosain from the East China Sea [36]. Since genetic diversity is the raw material for evolution [59], these findings suggest that the studied population is genetically robust. However, the fact that current genetic diversity was low compared to historical genetic diversity indicates that the studied population experienced periods of overexploitation or historical bottlenecks.

The Fu’s Fs and Tajima’s D test of the pooled sample showed significant deviation from the neutral evolution hypothesis (Tajima’s D = -2.36, p < 0.001: Fu’s Fs = -27.48, p < 0.001). When the hypothesis of neutral evolution was tested for each population, significant departure from the hypothesis were observed at all sampled populations, except the populations at sites 6, 7, 11, 13, 14, and 15 (Table 7). This indicates selection or demographic expansion of the S. serrata populations in the study area. Mismatch distribution of the pooled sample produced a unimodal distribution, supporting the null hypothesis of population expansion (Fig 2). The raggedness index and sum of squared deviations (SSD) showed that the null hypothesis of population expansion cannot be rejected (raggedness index = 0.025, p > 0.05: SSD = 0.00087; p > 0.05). The constructed haplotype network also support the null hypothesis of recent population expansion. The network produced a star like structure, with the central haplotypes surrounded by several haplotypes that show little base pair differences (Fig 1). This suggests that most haplotypes originated recently and it is indicative of recent population expansion from a small number of founders [60]. The time of expansion was estimated from the expansion parameter tau (τ), using the equation t = τ/2 μ, where μ is the rate of mutation. Using the estimated τ of 1.10625 and the COI mutation rate of 1.15% per million years [33], the time at which population expansion began was estimated to be about 90 thousand years ago. This time corresponds with the last glacial period which spanned from 125 to 14.5 thousands of years ago. The observed population expansion was probably due to sea level oscillations during this time [61].

The correlation between microsatellite genetic differentiation (G_ST) and the within subpopulation heterozygosity was not significant (p > 0.05). This shows that single loci G_ST-values are marker independent and that G_ST is the best estimate of genetic differentiation in the study area [47]. Apart from that, AMOVA of the microsatellite data did not detect significant genetic differentiation among sites (F_ST = 0.00424, p > 0.05, 100172 permutations). In contrast, COI showed significant genetic differentiation among sites (F_ST = 0.158, p < 0.05; Φ_ST = 0.238, p < 0.05). The contrasting patterns of genetic differentiation between nuclear and mitochondrial DNA are reported in several other species [62,63] and they can be due to a complex array of conditions that include selection, fluctuations in populations size, variations in sex ratio, and introgressive hybridisation [62,64]. Introgressive hybridisation from related species, does not account for the observed patterns, since S. serrata is the only Scylla species occurring in the study area [1]. In addition, the observed discrepancy might not be due to selection, because the neutrality tests showed no evidence of selection in the mtDNA (Table 7). While nuclear DNA is less likely to be affected by bottlenecks and rapid population expansions, mtDNA is more susceptible to these evolutionary forces due to its smaller effective population size [65,66]. Evidence of sudden expansion of the East African S. serrata populations was revealed by the mismatch analysis (Fig 2). This suggests that the observed discordance might be due to the varying effects of genetic drift on mitochondria and nuclear DNA. In addition, if there is no variations in sex ratio, the index of genetic differentiation is expected to be four times higher in mtDNA than nuclear DNA [67]. The ratio of mtDNA to nuclear DNA differentiation in the present study was 37 (0.157/0.00424). This suggests that the observed discrepancy in population differentiation between mitochondrial and microsatellite DNA is probably due to variation in sex ratio. Generally, in areas without sex-biased fishery, males giant mud crabs can outnumber females by three folds [68]. This is in line with what was observed during fieldwork, since males showed high abundance. High abundance of male S. serrata in East African mangroves was also reported in a previous study [11]. Therefore, the observed discordance is probably due to variation in sex ratio.

The analysis of molecular variance did not detect significant genetic differentiation among sites in the Seychelles, Kenya, Tanzania, Mozambique, and South Africa. The observed lack of genetic differentiation between these sites is in line with the findings of previous studies [27,58]. A similar pattern of genetic differentiation in this region was also reported in other mangrove fauna [24,25,58]. In contrast to these studies, this study detected significant genetic differentiation between sites at the East coast of Madagascar (ECM) and sites in mainland East Africa and the Seychelles (F_CT = 0.361, p < 0.05: Φ_CT = 0.564, p < 0.05). The fact that previous studies did not include samples from the ECM [24,25,58] can explain why no genetic differentiation was detected in previous studies. The fact that significant genetic differentiation between Mauritius Island and mainland East Africa was not detected in two previous studies [27,33] can be attributed to consequences of low sample size. The studies used only five sequences, representing only one haplotype from Mauritius. The observed genetic break between populations on the ECM and other sample sites is probably due to the influence of ocean circulation on larval transport and dispersal. Oceanic circulations in the region are influenced by trade and monsoon winds [1]. Because S. serrata has a planktonic larval stage, the SEC is expected to transport and disperse larvae from the ECM to East Africa through the EACC, MC and AC. The observed genetic differentiation indicate that there is limited larval exchange between sites in the ECM and sites in mainland East African and the Seychelles. This is supported by the measured mutation scaled immigration rate, which showed lowest rate of immigration to mangroves at the ECM (Table 8). The observed pattern in genetic differentiation is also supported by the haplotype network, which showed haplotypes from the ECM in a separate cluster, containing very few shared haplotypes (Fig 1). The patterns of currents can also account for the lack of genetic differentiation among sites on the coastline of East Africa. Circulation in East African coastal waters are influenced by the northward EACC, as well as the MC and eddies in the Mozambique channel, which propagate southward into the south-bound AC [31]. These currents are probably responsible for the dispersal of larvae among adjacent populations and thus accounting for the observed connectivity. This argument is supported the haplotype network, which showed that mangrove forests in the Mozambique channel, Kenya, Tanzania, South Africa and the Seychelles share the most common haplotypes (Fig 1).

The study showed that the current genetic diversity is low compared to historical genetic diversity (θ_π < θ_w). This shows that the studied population experienced bottlenecks in its recent history. Considering that indications of mud crab overexploitation and mangrove degradation are reported in the study area [2,11], measures aimed at enhancing sustainable use of resources should be strengthened. The observed limited gene flow between ECM and other sites indicate that protected mangroves and MPAs in the west coast of Madagascar and mainland East Africa cannot help to protect the biodiversity of mangroves in the ECM. Since Madagascar is planning to triple the extent of its MPAs by 2020, the observed patterns provide useful information for establishment of MPA networks around the island. Since giant mud crabs in the WIO are heavily exploited for food and trade, the observed low population size in the ECM and the Seychelles suggest that these areas require immediate attention. The mangroves in Kenya and Tanzania showed a high effective population size, which is maintained by a high rate of immigration from other mangroves in mainland East Africa. Estimates of migration rate showed highest number of immigrants to mangroves in this region (Table 8), indicating that overexploitation and degradation of some ecosystems is likely to affect recruitment and stock structure of adjacent ecosystems. Therefore, management efforts should strive to maintain connectivity among mangroves in this region. Since female giant mud crabs migrate offshore to spawn, management efforts should focus on both intertidal and offshore ecosystems.