Cryptic diversity and limited connectivity in octopuses: Recommendations for fisheries management

The market demand for octopus grows each year, but landings are decreasing, and prices are rising. The present study investigated (1) diversity of Octopodidae in the Western Indian Ocean (WIO) and (2) connectivity and genetic structure of Octopus cyanea and O. vulgaris populations in order to obtain baseline data for management plans. A fragment of the cytochrome C oxidase subunit 1 (COI) gene was sequenced in 275 octopus individuals from Madagascar, Kenya and Tanzania. In addition, 41 sequences of O. vulgaris from South Africa, Brazil, Amsterdam Island, Tristan da Cunha, Senegal and Galicia were retrieved from databases and included in this study. Five different species were identified using DNA barcoding, with first records for O. oliveri and Callistoctopus luteus in the WIO. For O. cyanea (n = 229, 563 bp), 22 haplotypes were found, forming one haplogroup. AMOVA revealed shallow but significant genetic population structure among all sites (ϕST = 0.025, p = 0.02), with significant differentiation among: (1) Kanamai, (2) southern Kenya, Tanzania, North and West Madagascar, (3) Southwest Madagascar and (4) East Madagascar (ϕCT = 0.035, p = 0.017). For O. vulgaris (n = 71, 482 bp), 15 haplotypes were identified, forming three haplogroups. A significant genetic population structure was found among all sites (ϕST = 0.82, p ≤ 0.01). Based on pairwise ϕST-values and hierarchical AMOVAs, populations of O. vulgaris could be grouped as follows: (1) Brazil, (2) Madagascar and (3) all other sites. A significant increase in genetic distance with increasing geographic distance was found (Z = 232443, 81 r = 0.36, p = 0.039). These results indicate that for O. cyanea four regions should be considered as separate management units in the WIO. The very divergent haplogroups in O. vulgaris from Brazil and Madagascar might be evolving towards speciation and therefore should be considered as separate species in FAO statistics.


Introduction
While the global market demand for octopus grows year after year, the supplies are becoming scarcer, and the prices are rising [1]. In 2017, China, Morocco and Spain were the largest a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 To maintain the sustainability of these socio-economic important octopus fisheries, management actions must be taken, such as the establishment of permanent Marine Protected Areas (MPAs) [19]. In these protected areas, extractive activities are limited or prohibited and are therefore effective in increasing the biomass of marine stocks, both within and outside the protected area [23]. When designing MPA networks, connectivity among populations through larval dispersal is a crucial factor, which should be taken into account, since connectivity determines gene flow as well as the ability to persist and recover from natural and anthropogenic stressors, such as overfishing [24]. Larval dispersal may be influenced by diverse factors, such as PLD, reproductive output, larval behaviour, direction and strength of ocean currents, geographic distance and oceanographic barriers [25]. Connectivity among populations can be determined using several genetic approaches. Mitochondrial DNA (mtDNA) is a widely used marker in population genetic studies since it possesses several advantageous characteristics. It is present in high numbers in the cell and is thereby easy to isolate [26]. Furthermore, mtDNA has a mosaic structure with fast evolving regions, resulting in a high variability, and more slowly evolving regions, allowing the use of universal primers. The latter makes mtDNA cheap and easy to amplify and sequence [26]. An example of a mtDNA marker is the cytochrome C oxidase subunit 1 (COI) gene, which has been applied as a DNA barcoding target gene in molluscs for species identification [27] and phylogeographic studies within a single species [28][29][30][31].
In order to obtain useful indications for the formulation of effective management plans, the present study investigated (1) the catch composition and diversity of Octopodidae in the Western Indian Ocean and (2) the connectivity and genetic structure of O. cyanea and O. vulgaris populations, using a fragment of the mitochondrial COI gene as marker. To the knowledge of the authors, this will be the first study dealing with population genetics of O. cyanea. Therefore, findings from this study are pioneering and can be used to further expand the knowledge base available for marine management decisions in the WIO.

Study area
For O. cyanea, the study focussed on the Western Indian Ocean. Because connectivity relies on dispersal of pelagic larvae, patterns are likely to be influenced by sea surface currents, moving larvae in a specific direction. In the WIO, the South Equatorial Current (SEC) flows from East to West across the Indian Ocean. Near the East coast of Madagascar, one component splits into the North-East Madagascar Current (NEMC) and South-East Madagascar Current (SEMC), while another one joins the NEMC at the Northern tip of Madagascar (Fig 1a). The latter continues West until it reaches the East African Coast, where it splits, creating on the one hand a number of southward propagating eddies on its way through the Mozambique Channel, and on the other hand 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) (Fig 1a) [32][33][34].  [35,36]. Pie charts show the fractional contribution of the defined haplogroups in the haplotype network (see d) for each of the sample sites. Abbreviations: AC: Agulhas Current, AE: Agulhas Eddies, Ai: Amsterdam Island, An: Andavadoaka, AR: Agulhas Retroflection, ARC: Agulhas were downloaded from the data portal of Barcode of Life Data Systems (BOLD Systems, http:// v3.boldsystems.org/, last consulted on 11/07/2018) and GenBank (https://www.ncbi.nlm.nih. gov/genbank, last consulted on 11/07/2018) and included in the genetic population structure analysis of this study. The complete list of used COI sequences in this study can be found in S1 Table. Since the number of samples per sample site derived from the databases was often very low (n < 3), it was decided to combine sampling sites based on small geographic distances, the presence of favourable currents for connectivity and results found in the original studies. Sequences (n < 3) from sites that could not be combined or sequences that were too short (< 482 bp) were only used in the phylogenetic analysis.

DNA extraction, PCR, and sequencing
A DNA extraction kit (E.Z.N.A. Tissue DNA Kit, Omega Bio-tek) was used for extracting DNA, following the manufacturer's protocol. Success of DNA extraction was measured with a NanoDrop 2000 spectrophotometer (Thermo Scientific). The universal primers HCO2198 (5'-TAA ACT TCA GGG TGA CCA AAA AAT CA-3') and LCO1490 (5'-GGT CAA CAA ATC ATA AAG ATA TTG G-3'), described by Folmer et al. [37], were used for the amplification of the COI gene fragment. PCR amplification was performed in a 25 μL reaction mixture containing 2 μL template DNA, 15.925 μL DNA free water, 1 μL (0.4 μM) of each forward and reverse primer, 2.5 μL (1 mM) PCR buffer, 0.5 μL (0.2 mM) dNTP, 0.875 μL (3.5 mM) MgCl 2 , 1 μl (10 mg/ml) BSA and 0.2 μL of (5 U/μL) Taq DNA polymerase, using a UNO96 VWR Thermal cycler. The following temperature profile was used: 5 min at 94˚C, followed by 40 cycles of 30 sec at 94˚C, 1 min at 51˚C and 1 min at 72˚C. Final extension was conducted at 72˚C for 10 min. After checking if the amplification was successful with gel electrophoresis on a 2% TBE agarose gel, fragment sequencing was carried out using an ABI 3730XL sequencer.

Genetic diversity
The sequences were edited using the software CHROMASPRO (v. 1.5, Technelysium Ltd, Leicester, UK), i.e. the correct representation of nucleotides was checked based on the chromatogram peaks. DNA barcoding for species identification was performed using the Identification System (IDS) for COI from BOLD systems (BOLD Systems, http://v3.boldsystems.org/, last consulted on 11/07/2018). In addition, a Neighbour Joining (NJ) tree [38] including bootstrap analysis (1000 replications) [39] was constructed using the software MEGA 7 (v. 7.0) [40]. For the latter, a few reference sequences of each identified species were downloaded from BOLD Systems (S1 Table).
Multiple and pairwise alignment of the sequences was carried out using Clustal W [41] as implemented in the software MEGA 7 and sequences were trimmed to the shortest common sequence length. The same program was further used to translate COI sequences into amino-acid sequences as quality control to exclude sequencing errors and/or to verify if functional mitochondrial DNA sequences were obtained and not nuclear pseudogenes. The online service FaBox DNA Collapser [42] (http://users-birc.au.dk/biopv/php-/fabox/dnacollapser.php#, last consulted on 12/07/2018) was used to reduce the sequence dataset to haplotypes. Nucleotide and haplotype diversities for all populations were calculated according to Nei [43] with the program Arlequin (v. 3.5) [44].

Historical demography
Unless otherwise stated, all following analyses were carried out with Arlequin. To insure suitability for population genetic analyses, the null hypothesis of neutral evolution of the marker was evaluated on the basis of Tajima's D [45] and Fu 's F S [46] tests, which also allows the detection of a recent population expansion or bottleneck. In order to discriminate between selective pressure and population expansion and investigate the historical demography, mismatch distribution analysis was performed [47] and the sum of square deviation [48] and Harpending's raggedness index [49] were calculated, thereby testing the model of sudden population expansion [50].

Genetic population structure and connectivity
An analysis of molecular variance was performed to investigate the overall genetic population structure in the dataset (ϕ ST ) (AMOVA; [51]) and pairwise population differentiations (pairwise ϕ ST ) were determined. The pairwise ϕ ST P-values were adjusted using Holm's sequential Bonferroni correction [52]. Groups for hierarchical AMOVA were chosen to test for genetic population structure across potential oceanographic barriers. Geographic distances between sample sites were measured as the shortest connection via marine pathways using Google Earth. The Isolation by Distance Web Service ( [53]; http://ibdws.sdsu.edu, last consulted on 14/07/2018) was then used to assess the correlation between geographic and genetic distances (pairwise ϕ ST -values) of sampled populations by applying a Mantel test [54]. A haplotype network was drawn based on the Minimum Spanning Tree information provided by Arlequin in order to examine the relationship among haplotypes.  (Table 1).

COI
Historical demography. The null hypothesis of neutral evolution of the COI marker was rejected for five sample sites based on Tajima's D-test, whereas the results of Fu's Fs test rejected the null hypothesis for six sites (Table 1). These significant outcomes could indicate selection or departures from population equilibrium (e.g. population expansion). The latter is supported by non-significant values for SSD and HRI for all sample sites, indicating no significant deviation from a model of sudden demographic expansion.
Genetic population structure and connectivity. The evolutionary relationships among 22 O. cyanea haplotypes are presented in a haplotype network, showing one haplogroup (Fig  1c). AMOVA revealed a very low but significant overall genetic structure (ϕ ST = 0.025, p = 0.02), rejecting the null hypothesis of panmixia. Pairwise ϕ ST -values between sample sites showed homogeneity among most of them, with significant differentiation visible in only two out of the 91 pairwise population comparisons, after applying sequential Bonferroni correction (S2 Table). A hierarchical AMOVA was carried out, with all tested groupings being significant Cryptic diversity and limited connectivity in octopuses (Table 2). However, grouping (7)

Octopus vulgaris
Genetic diversity. Based on an alignment of 71 O. vulgaris COI sequences with a length of 482 base pairs, 15 haplotypes were defined. Haplotype diversity (h) within each population ranged from 0 in Ga, Se, Tc, Pe, and AI to 0.83 in NBr and Sb. Nucleotide diversity (π) ranged from 0 in Ga, Se, Tc, Pe, and AI to 0.013 in NBr (Table 1).   (Table 1). Departures from population equilibrium are supported by non-significant values for SSD (except Db) and HRI for all sample sites, indicating no significant deviation from a model of sudden demographic expansion.
Genetic population structure and connectivity. The evolutionary relationships among 15 O. vulgaris haplotypes are presented in a haplotype network, showing three haplogroups (Fig 1d). The distribution of haplogroups is shown in Fig 1(b). Haplotypes from the blue haplogroup were only present at Brazilian sites, while the yellow haplogroup was only present at sites in Madagascar and in Db. AMOVA showed a significant genetic structure in the overall sample (ϕ ST = 0.82, p � 0.01), rejecting the hypothesis of panmixia. The genetic differentiation between Brazilian, Malagasy and all other sites was confirmed by very high and significant pairwise ϕ ST -values (Table 3). Based on oceanography, hierarchical AMOVAs were carried out, with all tested groupings being significant (Table 4). However, grouping (1) had a slightly higher and highly significant fixation index: Brazil vs all other sites (ϕ CT = 0.902, p � 0.01). Since the population structure seems to be mainly driven by the samples from Brazil, with the blue haplogroup only occurring in that region and a minimum of nine mutation steps between the red and blue haplogroup (Fig 1d), the analysis was repeated without the samples from Brazil. However, the AMOVA still detected a high and significant level of genetic population structure (ϕ ST = 0.69, p � 0.01). Other hierarchical AMOVAs were carried out, with all tested groupings being significant (Table 5). However, grouping (1) had a slightly higher and highly significant fixation index: Madagascar vs all other sites (ϕ CT = 0.773, p � 0.01). Furthermore, analysis of isolation by distance revealed a significant increase in genetic distance with increasing geographic distance (Z = 232443, 81 r = 0.36, p = 0.039). All sample sites are located in the tropics, with SST ranging between 26.0˚C and 28.2˚C. However, La and Fd are located more south, in sub-tropical waters, with SST below 25.0˚C [57]. Egg laying and rearing under aquarium conditions was done at 18.2˚C-20.0˚C for C. luteus [58], at 21.2˚C for O. vulgaris [22], at an average of 25.1˚C for O. oliveri [59] and at an average of 25.5˚C for O. cyanea, with most rapid development at temperatures above 27.1˚C [21]. For C. ornatus no such data was available. From these experiments it seems that, even if ranges might overlap [15,[60][61], the optimal SST might be too low for O. cyanea in the southern located sites, while temperature might be too warm for the other octopus species in more northern located sites. Another explanation could be the differences in fishing time period. Traditionally, women and children catch octopus by gleaning at low spring tides on the exposed reef flats and probe octopus dens with spears [16]. Depending on when low spring tides occur, different species might be caught, since O. cyanea is active during the day [62], Cryptic diversity and limited connectivity in octopuses while C. ornatus [62], C. luteus [61], O. oliveri [63] and O. vulgaris [64] are nocturnal species. Also, stronger current and higher wave activity in South Madagascar [65] might be a plausible explanation for the observed distribution pattern. However, since samples were collected from February to October over several years, with southern locations sampled in May, a seasonal effect can be excluded.

Genetic diversity and historical demography
All populations of O. cyanea and O. vulgaris showed a moderate level of haplotype diversity (h) and a low level of nucleotide diversity (π). This pattern has also been reported in other studies in the WIO [30,66]. The high h and low π might indicate a genetic bottleneck that has caused stochastic extinction of most haplotypes, followed by population expansion [67]. This is congruent with the results of Tajima's D-test, Fu's Fs test, the mismatch distribution analysis and Rogers' test for sudden population expansion, that suggest demographic growth. This finding is similar to the population expansion reported in the WIO for the African giant mud crab Scylla serrata [66], the skunk clownfish Amphiprion akallopisos [68] and the mangrove whelk Terebralia palustris [30].

Genetic population structure and connectivity
Octopus cyanea. The results from this study indicate that the WIO population of O. cyanea is characterised by a dominant haplotype that is present at all sample sites (Fig 1c). The haplotype network has a star-like structure, with a central haplotype surrounded by several smaller ones that show only little base pair differences. This might be an indication of recent population expansion from a small number of founders, supporting our previous findings. Even though all sites share the most common haplotype and the network suggests panmixia, a very weak but significant genetic population structure was found across all sample sites in the WIO (ϕ ST (4) southern Madagascar [30]. For the giant mud crab Scylla serrata a significant genetic divergence was found between sites in eastern Madagascar and sites on the West coast of Madagascar, mainland East Africa, as well as the Seychelles [66]. At a smaller scale, a weak genetic structure was detected among populations of Acropora tenuis in Tanzania and Kenya [69]. On the contrary, some studies in the WIO regarding invertebrates found high connectivity and no genetic structure [70,71]. This lack of genetic population structure could be attributed to the absence of samples from Madagascar in both studies.
No significant correlation was found between genetic differentiation and geographical distance. Therefore, the main factor that restricts gene flow and thus dispersal seems to be the direction and strength of ocean currents. During the Northeast Monsoon, the East African Coast Current (EACC) and the Somali Current (SC) converge into the eastward South Equatorial Counter Current (SECC) [32], which may potentially prevent dispersal from southern locations to the northernmost sample site Kanamai (Ka). The South Equatorial Current (SEC) may transport larvae from North Madagascar to Tanzania and Kenya, by splitting of the EACC, and to the northwest coast of Madagascar by the southward split of the SEC. In addition, islands like the Comoros, Mayotte, and Îles Glorieuse could serve as stepping-stones between western Madagascar and East Africa [30]. An experiment with satellite-tracked drifters demonstrated that eddies in the Mozambique Channel (MCE) are capable of trapping drifting objects for periods of several weeks to months [33]. Differentiation between southwestern sites and other localities may therefore be due to trapping, isolation and retention of the larvae inside southward propagating eddies. The splitting of the SEC into the southward Southeast Madagascar Current (SEMC) and the northward Northeast Madagascar Current (NEMC) at the East coast of Madagascar might be a reason for restricted gene flow between sites in Northeast Madagascar and southern sites. A similar pattern was found for the skunk clownfish, Amphiprion akallopisos, for which the population from Sm was significantly differentiated from the other WIO populations [68]. The southernmost location in Madagascar (Fd) is very likely under the influence of the SEMC, what might prevent exchange and gene flow with sample sites at the West coast of Madagascar or African mainland. Such a separation was also observed in Terebralia palustris [30]. However, a recent paper suggests that mtDNA sequence data are often not suitable to test for IBD because mtDNA often fails to reliably detect such pattern when other markers, like microsatellites, can. This failure of mtDNA seems to be the result of a reduction in genetic diversity due to selection, which can obscure spatial genetic differentiation [72].
The results of this study suggest there is high genetic exchange of O. cyanea in the WIO region due to the long-distance dispersal capacity of O. cyanea paralarvae. However, currents are one of the main factors responsible for restricted gene flow to a certain extent between the central part (Kenya, Tanzania, north-and western Madagascar), Kanamai (convergence of EACC and SC into SECC), eastern Madagascar (NEMC and SEMC) and southwestern Madagascar (EACC retroflection and MCE). It is important to remark that the small number of samples for certain sites may have influenced the results that were found. Furthermore, we hypothesise that the contribution of adult migration to population connectivity is negligible. Adult O. cyanea may use one den for three to five weeks [62,73]. Even though no records are available of the average distance between old and new dens for O. cyanea, it is known for other octopus species that this distance is around 15 m [74]. If these distances would also apply for O. cyanea then the species appears to be relatively resident and migration of adult O. cyanea is negligible. However, research needs to be done to confirm this hypothesis.
Octopus vulgaris. A strong and significant population structure (ϕ ST = 0.82, p � 0.01), was found among all sample sites. Based on very high and significant pairwise ϕ ST -values and hierarchical AMOVA, restricted gene flow could be detected between populations from Brazil, Madagascar and all other sampling sites. The strong genetic differentiation of the Brazilian and Malagasy sites can be attributed to the exclusive occurrence of haplotypes from the blue (Brazil) and yellow (Madagascar) haplogroup (Fig 1b and 1d). In Durban (Db), the yellow haplogroup is also present. Another study found that specimens of O. vulgaris from Db were genetically more different from other South African sites than specimens from Senegal or Tristan Da Cuhna in the southern Atlantic Ocean [60]. The authors suggested that the dominance of one haplogroup along the coast of South Africa, with another genetic lineage so far only found in Durban, might be the result of a recent introduction of the latter, in which case identical haplotypes should be found in other regions [60]. Our study can confirm that one of the two haplotypes belonging to the yellow haplogroup in Durban is also present in Madagascar (La and Fd). This might indicate that paralarvae of O. vulgaris may be transported from Madagascar to Durban, in which case the latter might act in the future as a stepping stone for dispersal of paralarvae between Madagascar and South Africa. An experiment with satellitetracked drifters in the Mozambique Channel showed that cross-channel transport between Madagascar and Mozambique was possible, with time spans of 19-30 days [33]. If the yellow haplogroup has its origin in Madagascar, this might be the result of a colonisation event of Madagascar, after which the Malagasy subpopulation started to evolve towards insular speciation (followed by possible dispersal to Durban, African mainland), as can be observed in many of the terrestrial fauna and flora on the island [75]. However, another possibility is that the ancestral haplotype of the yellow haplogroup has its origin in Mozambique and that larvae from that region are transported south to Madagascar and Durban by Mozambique Channel Eddies. This was found for T. palustris, in which southern Mozambique was postulated as potential origin of the populations elsewhere in the WIO [30]. In this case, gene flow between Madagascar and South Africa might not take place. Restricted gene flow between the southern African mainland and Madagascar is found by other studies. For the spiny lobster Palinurus delagoae, a genetic break was found between populations from South Africa and Walters Shoals, a submerged seamount on the Madagascar Ridge [76]. For the mangrove whelk Terebralia palustris, significant differentiation was found between southern Madagascar and southern Mozambique [30]. Giant mud crab (Scylla serrata) populations from South Africa and the West Coast of Madagascar did not show significant genetic differentiation among each other, but a significant genetic break was observed with populations from East Madagascar [66]. Due to low sample size in the South African localities and the limited number of sites investigated, further sampling and research needs to be conducted in order to confirm either of both hypothesis. Anyhow, the uniqueness of populations from the WIO compared with populations elsewhere in the Indian Ocean was also found for other species such as the skunk clownfish [68] and the mud crab Scylla serrata [77].
The exclusive presence of the blue haplogroup in Brazilian populations, together with very high and significant pairwise ϕ ST -values and the hierarchical AMOVA, indicate very high differentiation with individuals from the other sample sites. This is congruent with previous studies that found morphological [11,78] and molecular evidence [8,11] that O. vulgaris from Brazil might be a distinct species [11]. Furthermore, other studies found significant differentiation with microsatellite markers along the southern Brazilian coast [79].
The isolation-by-distance analysis revealed a significant correlation between genetic and geographical distances. Therefore, geographic distance seems to be one of the main factors that restricts gene flow and hence dispersal. Paralarvae of O. vulgaris settle and become benthic between day 47 and 54 after hatching [22]. However, time of settlement is independent of age, but related to reaching a critical size (173 mg and 7.5 mm mantle length, at an average temperature of 21.2˚C) and the duration of the planktonic life stage is temperature dependent [22]. Another study recorded a paralarvae duration of 22-30 days at 24.3-26.9˚C [80]. The species has a higher growth rate at high water temperatures, due to an increased food intake rate [81]. Following that reasoning, paralarvae have a higher dispersal distance in temperate waters than in (sub)tropical waters. The non-significant pairwise ϕ ST -values and the presence of the red haplotypes in the East Atlantic Ocean, South Africa and the Southern Indian Ocean suggest high connectivity and are not in line with the results of the isolation-by-distance analysis. The most plausible explanation for these results is that due to the low sampling size other haplotypes are overlooked and thereby gene flow is overestimated. Another explanation might be that there is high dispersal of O. vulgaris paralarvae among the East Atlantic Ocean, South Africa and Amsterdam Island in the Southern Indian Ocean. The Canary and Benguela Currents may transport paralarvae in the East Atlantic Ocean, which are expected to have high dispersal capacities reasoned by the cold character of the currents [36,82]. Water between the WIO and the Atlantic Ocean can be exchanged through the shedding of Agulhas Eddies (AE) from the Agulhas Retroflection (AR). Paralarvae may then be transported towards Amsterdam Island by the Agulhas Return Current (ARC) [36,83]. Even though genetic differentiation was found for O. vulgaris along the Senegalese and Mauritanian coast [84], such high connectivity levels between different oceans were observed for sea urchins [85] and bigscale soldierfish [86]. Further sampling and research need to be conducted to confirm either hypothesis.
The complete analysis was also conducted with sequences of 462 bp, which included samples from Greece (Accession numbers: JX500627-JX500639) and more samples of Galicia (Accession numbers: JX500655-JX50065577) (S1 Table). Since we lost 4% of the genetic data compared with the present analysis (482 bp) it was decided to use less but longer sequences. Samples from Greece belonged to the red haplogroup with most sequences belonging to the dominant/ancestral haplotype. The samples from Galicia were also part of the red haplogroup, but most sequences belonged to a haplotype that differed three mutation steps from the dominant/ancestral haplotype.

Future research
Over the past couple of years, concern was raised about the exclusive use of mtDNA markers, since they only reflect one locus, only showing the maternal genetic history [87] and they might be under selection in many animal species [88]. To be sure that the observed genetic structure patterns that were found with COI are not influenced by these shortcomings, nuclear loci should be analysed additionally in the future [87]. By comparing the results of mtDNA and nuclear markers such as microsatellites, the suitability of mtDNA can be validated and additional information might be obtained [66,68].

Conclusions
More species of the family Octopodidae occur in the WIO than previously thought. Furthermore, this study revealed high levels of connectivity among O. cyanea populations in the WIO, with low but significant differentiation between the central part (Kenya, Tanzania, north-and western Madagascar), Kanamai, eastern Madagascar and southwestern Madagascar. For O. vulgaris highly restricted gene flow could be detected between populations from Brazil, Madagascar and all other sampling sites due to the large geographical distance. Therefore, this study suggests that four separate management units should be considered for O. cyanea in the WIO. The very divergent haplogroups in O. vulgaris from Brazil and Madagascar might be evolving towards speciation and therefore should be considered as separate species in FAO statistics.
Supporting information S1