Large-scale biogeographic patterns are reflected in the genetic structure of a broadcast spawning stony coral

Countries in the Western Indian Ocean (WIO) and along the Red Sea are particularly vulnerable to coral reef degradation, and understanding the degree of connectivity among coral reefs is a first step toward efficient conservation. The aim of this study is to investigate the genetic diversity, population structure and connectivity patterns of the broadcast spawning coral Acropora tenuis, first at a large scale comparing the Red Sea and the WIO, and second at a smaller scale comparing sites within the WIO. In total 689 individual A. tenuis colonies were sampled on 28 locations in Saudi Arabia, Kenya, Tanzania, Mozambique and Madagascar and analysed with seven microsatellite markers. The sample site in the Red Sea was found to be differentiated from all other sites in the WIO, which confirms the hypothesised genetic break. High differentiation was found between the African mainland and Madagascar and within Madagascar. However, there is evidence for long-distance larval dispersal for A. tenuis in the North Mozambique Channel region, with exchange between northern Mozambique and northern Madagascar. The sites in the southwest of Madagascar show mixing with sites in northern Madagascar, as well as exchange with sites in northern Mozambique and Tanzania. Southern Mozambique forms a separate group in Bayesian clustering. High genetic connectivity was found for most sites along the East African mainland coast, with no indication for strict genetic barriers. These results support biophysical modelling studies, which propose Tanzania as a seeding source of larvae for downstream Kenya. These patterns of high genetic connectivity combined with contemporary dispersal barriers can be explained by the long larval duration of A. tenuis and the prevailing northbound East African Coast Current that facilitates higher genetic connectivity along the northern East African Coast, while eddies in the Mozambique Channel are causing larval retention in southern Mozambique and Madagascar.


Introduction
Understanding the degree of connectivity among coral reefs is a first step toward efficient conservation (Almany et al. 2009), as connectivity between reefs determines gene flow, genetic diversity and genetic structure of populations, as well as the capacity to persist under predicted climate change (Palumbi 2003). Coral connectivity, or the 'movement of individuals within and among local or subpopulations' (Cowen and Sponaugle 2009), is facilitated by dispersal of coral larvae. The Pelagic Larval Duration (PLD, i.e. the maximum time the larvae spend in the water column) and larval behaviour, but also environmental conditions, such as the suitability of the habitat for larval settling, currents and oceanographic barriers, determine the connectivity of coral reefs (Graham et al. 2008;Cowen Abstract Countries in the Western Indian Ocean (WIO) and along the Red Sea are particularly vulnerable to coral reef degradation, and understanding the degree of connectivity among coral reefs is a first step toward efficient conservation. The aim of this study is to investigate the genetic diversity, population structure and connectivity patterns of the broadcast spawning coral Acropora tenuis, first at a large scale comparing the Red Sea and the WIO, and second at a smaller scale comparing sites within the WIO. In total 689 individual A. tenuis colonies were sampled on 28 locations in Saudi Arabia, Kenya, Tanzania, Mozambique and Madagascar and analysed with seven microsatellite markers. The sample site in the Red Sea was found to be differentiated from all other sites in the WIO, which confirms the hypothesised genetic break. High differentiation was found between the African mainland and Madagascar and within Madagascar. However, there is evidence for long-distance larval dispersal for A. tenuis in the North Mozambique Channel region, with exchange between northern Mozambique and northern Madagascar. The sites in the southwest of Madagascar show mixing with sites in northern Madagascar, as Topic Editor Danwei Huang and Sponaugle 2009). In theory, pelagic larval duration is positively correlated with dispersal distance, and longer PLD increases the number of connected coral reef sites in the Western Indian Ocean (WIO) (Crochelet et al. 2016), but see Weersing and Toonen (2009).
Increased knowledge on connectivity is becoming crucial as coral reefs are degrading at an alarming rate, due to a variety of local and global anthropogenic stressors (Hoegh-Guldberg et al. 2017;Hughes et al. 2017), thereby threatening an essential habitat of at least a quarter of known marine species (Fisher et al. 2015). Coral reefs also sustain an estimated value of around $352,000 ha −1 per year of ecological goods and services, such as food, coastal protection and recreation for over 500 million people (Moberg and Folke 1999;Costanza et al. 2014). On the one hand, developing countries in the WIO and countries along the Red Sea are particularly vulnerable to coral reef degradation (e.g. overfishing and destructive fishing, pollution and coastal development), with very low economic adaptive capacity to recover from coral reef loss (Burke et al. 2011). On the other hand, coral reefs in the WIO are some of the least studied in the world. Marine Ecoregions of the World by Spalding et al. (2007), and Ecoregions as proposed by Obura (2012), describe two scenarios of regional biogeographic patterns. Within-species connectivity data form an essential tool to verify and explain biogeographic regions described for corals in the WIO and Red Sea. Variable but consistently high connectivity was found on a small spatial scale in Kenya and Tanzania for A. tenuis (van der Ven et al. 2016), but information on connectivity on a larger spatial scale within the WIO (Fig. 1a), and between the WIO and the Red Sea (Fig. 2), is lacking. Acropora tenuis (Dana (1846); Cnidaria; Scleractinia; Acroporidae) is a branching scleractinian coral, which grows on upper reef slopes and has a widespread distribution in the Indo-Pacific, commonly occurring in the Red Sea, Madagascar, and along the coast of East Africa in Kenya and Tanzania, as well as in the north and south of Mozambique (Veron 2000). This species is a broadcast spawning coral which releases buoyant eggs and sperm to the water column where fertilisation takes place. Information on the spawning of acroporids in East Africa is limited (Baird et al. 2021), but seems to take place in and around the time of the Northeast Monsoon (currents depicted in Fig. 1). Indications for synchronised mass-spawning between September and December was found in Vamizi (Mozambique), Andavadoaka (Madagascar) and in Reunion (Sola et al. 2016;Baird et al. 2021), while unsynchronised spawning over a time span of seven months (from October to April) was observed on some occasions in Mombasa, Kenya Harrison 2008, 2009). In Saudi Arabia in the Red Sea acroporids were observed to spawn between April and June (Bouwmeester et al. 2015;Baird et al. 2021).
The PLD, which potentially affects the capacity for longdistance dispersal, is highly variable in acroporids (e.g. between 54 and 209 days, Nishikawa and Sakai 2005;Graham et al. 2008). Acropora tenuis larvae are viable and can settle up to 69 days after spawning in laboratory experiments and have therefore the potential to disperse over long distances (Nishikawa et al. 2003). However, for all acroporids a high early mortality is found and most settlement occurs on the natal reef or adjacent reefs (Ayre and Hughes 2000;Graham et al. 2008;Underwood et al. 2009). Besides sexual reproduction through gametes, acroporid corals can also reproduce asexually (clonal reproduction) by colony fragmentation.
The aim of this study is to investigate the genetic diversity, population structure and connectivity patterns of the broadcast spawning coral A. tenuis first at a large scale, comparing the Red Sea and the WIO, and second at a smaller scale, comparing sites within the WIO. Based on multispecies distribution records of corals, six regions can be distinguished in the central WIO (Obura 2012;Fig. 1b): (1) Obura (2012), ocean currents and the life history of A. tenuis: (I) a genetic break between the Red Sea and the WIO, (II) a distinct NMC region with higher genetic diversity, (III) a distinct SMC region, (IV) distinct regions of EM and SM in Madagascar. The hypotheses are tested with seven microsatellite markers.

Coral reefs
Fringing reefs line the coast of the WIO, with the most extensive reefs found in the south of Tanzania and the north of Mozambique, and smaller and more fragmented reefs found in the north of Kenya and south of Mozambique (Obura 2012;McClanahan et al. 2014;Obura et al. 2017). The central Mozambique coastline is characterised by large-scale river outlets and mangrove forests and no coral reefs occur here, because of freshwater and sediment discharge (Obura et al. 2017). The Red Sea is a peripheral region of the Indo-Pacific and connected with the Gulf of Aden in the Indian Ocean through the shallow (130 m deep) and narrow strait of Bab-el-Mandab. The Red Sea has mainly fringing reefs along most of its shores, with atolls and pinnacle reefs found in the Southern and Central Red Sea (Wilkinson 2002).

Sampling
A total of 689 individual A. tenuis colonies were sampled (Fig. 2a, Table 1 and Appendix 1) in one location in Saudi Arabia, six locations in Mozambique and ten locations in Madagascar, while samples from five locations in Kenya and six locations in Tanzania were included from van der Ven et al. (2016). Colonies were randomly selected while maintaining at least two metres between them to avoid sampling clones. Each sample was obtained by breaking a 2-3 cm fragment from a branch. The fragments were stored in a plastic container with seawater until fixation in absolute ethanol. Samples were stored in the dark at 7 °C until extraction.

DNA extraction
The fragments, both skeleton and tissue together, were crushed manually with a bone cutter and incubated overnight at 56 °C with lysis buffer and Proteinase K. The NucleoSpin® Tissue DNA extraction kit (Macherey-Nagel, Düren, Germany) was used following the company's animal tissue protocol for 100 µL DNA extract. The purity and concentration of the DNA extract was determined with a NanoDrop 2000 spectrophotometer (Thermo-scientific).

Microsatellite genotyping
Samples were genotyped using seven microsatellite primer pairs designed for Acropora millepora (Appendix 2), but cross-amplifications showed that they can be applied to A. tenuis van der Ven et al. 2016) and several other acroporid species from the Pacific Ocean (Richards and van Oppen 2012). The primers were combined in one multiplex (PCR) with a volume of 12.5 µL containing 2.5 µL template DNA, 1.25 µL (2 µM) primer mix with fluorescently labelled primers 6-FAM, VIC, PET and NED (Applied Biosystems, Foster City, CA, USA), 2.5 µL H 2 O and 6.25 µL QIAGEN Multiplex PCR Master Mix (QIAGEN, Hilden, Germany). The PCR cycling parameters were: an activation step for 5 min at 95 °C, followed by 35 cycles of denaturation for 30 s at 95 °C, annealing for 90 s at 60 °C and 30 s at 72 °C. The final extension was done for 30 min at 68 °C. The PCR product was diluted 10 or 100 times and analysed together with a GeneScan tm -500 LIZ® size standard (Applied Biosystems, Foster City, CA, USA) on an ABI 3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA). The resulting electropherograms were scored manually with GeneMarker® (v. 2.4.0; SoftGenetics, State College, PA, USA).

Genetic and genotypic diversity analysis
The software GenAlex 6.5 (Peakall and Smouse 2012) was used to identify multilocus genotypes (MLGs) that were likely to be the result of fragmentation of the coral colony. Clonal richness (R) was calculated as R = (G-1)/(N-1) in which N is the total number of samples, and G is the number of MLGs (Dorken and Eckert 2001). The departure from Hardy-Weinberg equilibrium (HWE) was determined for each locus with FSTAT 2.9.3 (Goudet 1995) and presented as F IS (calculated as smallF; Weir and Cockerham 1984), with significance levels based on 210,000 permutations. In all FSTAT analyses, the number of permutations was automatically determined based on the number of loci and populations in the experimental design. Populations with heterozygote deficiency were further analysed with Inest 2.0 (Chybicki and Burczyk 2009) utilising a Bayesian approach for estimating both null alleles and inbreeding simultaneously (Campagne et al. 2012). The model was run with 50,000 burn-ins and 500,000 cycles. Linkage disequilibrium (LD) was calculated with FSTAT, and the P value for genotypic disequilibrium was based on 6,300,000 permutations. Number of alleles per population, corrected for sample size and expressed as allelic richness, was also calculated with FSTAT. Private alleles and observed, expected and unbiased (sample size corrected) heterozygosity were calculated with GenAlex.

Population structure
To investigate population structure among populations, pairwise F ST values were estimated using the method of Weir and Cockerham (1984) (θ) with FSTAT and tested for significance with 55,000 permutations. In GenAlex D EST (Jost 2008) was calculated and tested for significance with 9999 permutations. Pairwise F ST values as well as D EST values were utilised in a Mantel test in GenAlex to test for linear association with geographic distances measured as the shortest distance by sea (isolation by distance: IBD) with 9,999 permutations. The pairwise geographic distance was measured as the shortest path possible by water between two sample sites in Google Earth Pro (www. google. com/ earth). An analysis of molecular variance (AMOVA) was conducted with GenAlex, estimating overall F ST as well as testing two potential population structures: Marine Ecoregions of the world as described by Spalding et al. (2007), and Ecoregions as proposed by Obura (2012), Fig. 1a and 1b, respectively. Significance was tested with 9,999 permutations. SPAGEDI (Hardy and Vekemans 2002) was used to test if the assumption of step-wise mutations was met (in this case R ST would reflect population differentiation better), as well as to investigate the relative importance of mutation rates versus migration rates on population differentiation in distance intervals, tested with 20,000 permutations. Principal coordinates analyses (PCoA) were conducted in GenAlex to examine the spatial variation based on pairwise F ST . The software STRU CTU RE 2.3.4 (Pritchard et al. 2000), which uses a Bayesian clustering based on the Markov chain Monte Carlo (MCMC) assignment method, was run under the admixture model to determine the number of genetic clusters (K) and was  Table 1. b Pie charts on the map show the distribution of five genetic clusters (K = 5) at sample sites as revealed in Bayesian clustering implemented in STRU CTU RE (Pritchard et al. 2000) for sample sites in the Western Indian Ocean (WIO) only. c K = 8 bar plot for individuals from the Red Sea and WIO. The colours in the bar plot visualise the individual's estimated membership fraction in each of the K clusters inferred by Bayesian clustering with the programme STRU CTU RE, with each vertical bar representing one individual ◂ run both with and without prior population information. STRU CTU RE analysis was performed in the Par-allelStructure package (Besnier and Glover 2013) in R (https:// www.r-proje ct. org/). The STRU CTU RE analysis was run for all data K = 1-28, as well as for the Western Indian Ocean excluding the Red Sea (K = 1-27). Analyses included 10 runs with a burn-in length of 100,000 and 1,000,000 MCMC replications. All analyses were implemented in the CIPRES Science Gateway portal (Miller et al. 2015). HARVESTER 0.6.94 (Earl and Vonholdt 2012) was used to visualise and analyse the STRU CTU RE output by plotting log probability L (K) and ΔK (Evanno et al. 2005).
Additionally, all STRU CTU RE analyses were repeated in Instruct to verify if the population structure found in STRU CTU RE is affected by inbred populations (Gao et al. 2007). Lastly, a one-sided comparison (9,999 permutations) of H O , A r, F IS and F ST was made (FSTAT) to test whether these parameters were significantly different between the regions in the WIO.

Genotypic diversity
Of the 689 sampled individuals, 629 were found to be unique multilocus genotypes (MLGs). Clonal richness (R) was mostly found to be the maximum value of 1 (Table 1), which indicates absence of clonality, with all different samples analysed corresponding to distinct lineages. Lower values of R were found in the south of Madagascar in Fa, Lv and FD (0.3, 0.3 and 0.5, respectively). As the probability of identity (PI) was low (< 0.001 for all sample sites, below the value of 0.01 set for adequate population studies (Waits et al. 2001)), the chance of colonies accidentally sharing the same genotype is very low. Therefore, the colonies sharing the same genotype were considered as potentially belonging to the same individual. Further analysis was performed with a reduced dataset, with all but one of the duplicated individuals from every identical MLG, in total 60 samples, excluded.

Genetic diversity
The seven microsatellite loci amplified successfully and were polymorphic for all 28 sample sites with 2-12 alleles per locus (Appendix 3). Allelic richness (A r ) ranged between 1.5 and 1. The pairwise F ST and D EST did not vary linearly with the geographic distance, and the mantel test showed no significant correlation between population differentiation and geographic distance for all sites in the Red Sea and Western Indian Ocean, Madagascar or the sites along mainland East Africa. However, there was isolation by distance when tested within the WIO (IBD F ST ; R 2 = 0.0583, P < 0.05; IBD D EST ; R 2 = 0.088, P < 0.05; Appendix 5). No significant contribution of step-wise mutations (R ST > F ST ) was found when testing the Red Sea and the WIO (interval 1-4000 km and interval 4001-6900 km), but significant contribution of stepwise mutations was found for distances beyond 1300 km within the WIO (Appendix 6).
Bayesian clustering in STRU CTU RE which included a priori population information of the six ecoregions defined in the WIO (Fig. 1b) plus the Red Sea, resulted in a cluster distribution congruent with some of the predefined groups (especially the NMC and SMC, Appendix 7). However, this approach does come with the possibility that real population differences (within the defined groups) may be overlooked (Pritchard et al. 2000). Indeed, analyses without a priori population information revealed additional and alternative genetic structuring. For these latter analyses, a plateau for the log probability for the values of K ranging from 7 to 9 was found, of which K = 8 was selected (Appendix 8 and Fig. 2c), but the additional K's were also included in the investigation. Within the WIO clustering revealed a plateau for the values of K ranging from 5 to 9, of which K = 5 was selected (Appendix 8, Fig. 2b). All clusters occurred at all sample sites, but based on the occurrence and proportion of the different genetic clusters within each sample site, groups of sample sites could be identified. A separate group can be distinguished in Saudi Arabia (RM), although some of the individuals might belong to different genetic clusters. Based on analysis including the WIO and Madagascar (K = 5, Fig. 2b), four groups can be distinguished (1) sites in Kenya, Tanzania, northern Mozambique and western Madagascar (La, Wa, Di, Mo, Ki, Mi, St, Ja, DS, Mf, Mk, Va, Pe, Na, Nb, Sa, To, An) forming a diverse group of different genetic clusters, although based on cluster distribution, substructures can be identified, (2) sites in southern Mozambique (Vi, Ba, In), (3) north and northeast Madagascar (Ra, Nt, Sm), (4) south of Madagascar (Fa, Lv, FD). Within these groups evidence of mixing of genetic clusters was found at several sample sites: Va, Ja, Ki, Mo (with signature of group  (Weir & Cockerham 1984) and level of significance after 435,000 permutations based on seven microsatellite loci in Acropora tenuis.
Above the diagonal: pairwise D EST values (Jost 2008) and level of significance after 9,999 permutations based on seven microsatellite loci in A. tenuis.

Groupings based on the ecoregions defined by
Obura (2012)   NA: test not available due to small sample size; P values have been corrected for multiple tests; for abbreviations of sample sites see Table 1; for geographic location of samples sites see Fig. 2a and Appendix 1 1, but additionally also group 4) and Nb (with signature of group 1, but additionally also 3).
In the PCoA based on pairwise-F ST of all sample sites, the first axis divides RM, Mo, Ki, Ja, Va, Ra, Nt, Sm, Fa and Lv from all others. The second axis separates RM, Ki, Ra, Nt and Fa from this segregation, as well as divides La, Wa, Di, Mi, St, DS, Mf, Mk, Na, Pe, Nb, An and FD from Vi, Ba, In, Sa and DS, together explaining 40% of the variation (Fig. 3a). In the PCoA including only sites in the WIO the first axis divides Mo, Ki, Ja, Va, Ra, Nt, Sm, Fa and Lv from all others, and the second axis separates Mo, Ki, Va, Ra and Sm from this segregation, as well as dividing La, Wa, Di, Mi, DS, Mf, Mk, Pe, Na, Ba and In, from St, Vi, Nb, Sa, To, An and FD, together explaining 38% of the variation (Fig. 3b).
In the hierarchical AMOVA, the largest component of variance was found within populations (85%-86% in de WIO; Table 3). Both the grouping based on Marine Ecoregions of the world (Spalding et al. 2007) and the biogeographical ecoregions proposed by Obura (2012), gave identical results in explaining the genetic variance among groupings in the analysis of hierarchical AMOVA in the WIO (component variance among groupings 3%, with F RT = 0.031 for both groupings). Lastly, no significant differences were found in the one-sided comparisons of H O , A r, F IS and F ST between different regions in the WIO.  (Weir and Cockerham 1984) for Acropora tenuis, the colours representing the Ecoregions as suggested by Obura (2012) see Fig. 1b: a the Red Sea and the Western Indian Ocean, the first two axes explain 40.14% of the variation (the first axis explains 24.79%, the second axis explains 15.35% of the variation); b Western Indian Ocean, the first two axes explain 37.54% of the variation (the first axis explains 25.12%, the second axis explains 12.42% of the variation). For sample site codes see Table 1   Table 3 Components of variance from AMOVA in GenAlEx 6.5 (Peakall & Smouse 2012) of seven microsatellite loci in Acropora tenuis in the Western Indian Ocean (WIO) Groupings based on the ecoregions defined by a Spalding et al. (2007); b Obura (2012) (Fig. 1)

Discussion
This study describes comprehensive research of the genetic diversity, population structure and connectivity patterns of the broadcast spawning coral Acropora tenuis using microsatellite markers. First the large-scale patterns comparing the Red Sea and the Western Indian Ocean (WIO) are described, followed by the smaller-scale patterns within the WIO and within the southern Mozambique Channel.

Isolation of the Red Sea
The sample site in the Red Sea (RM) was found to be differentiated from all other sites in the WIO, except some sites with low sample size. This confirms the hypothesised genetic break based on Marine Ecoregions of the World, in which the Red Sea and the WIO belong to different provinces (Spalding et al. 2007), which was also recently found for the brooding coral Seriatopora hystrix ) and the giant clam Tridacna maxima and T. squamosa (Hui et al. 2016). This isolation can be explained by both historical and contemporary barriers for larval dispersal between the Red Sea and the WIO. The Red Sea was historically isolated from the Indian Ocean by land barriers as a direct result of lower sea levels caused by glacials, and during this time the Red Sea experienced periods of extreme desiccation and hypersalinity (Siddall et al. 2003). Nevertheless, it remains under discussion whether corals survived these conditions in refugia or recolonised the Red Sea from the Gulf of Aden (DiBattista et al. 2016). Contemporary barriers for dispersal between the Red Sea and Indian Ocean are eddies within the central Red Sea (~ 20° N) and limited exchange through the shallow and narrow Bab-el-Mandab Strait. Furthermore, cold-water upwelling in the Gulf of Aden, northern Somalia and Yemen during the summer monsoon, limits the formation of coral habitat in this area (Furnas 2011;Kemp 1998), but see DiBattista et al. (2016). Additionally, the northbound East African Coast Current (EACC) is diverted under the influence of the Somali Counter Current (SC) and larvae are subsequently transported away from the East African coast by the South Equatorial Counter Current (SECC) during the Northeast Monsoon (Gamoyo 2018).

Genetic connectivity in northern Madagascar and the East African mainland
High differentiation was found between the East African mainland and Madagascar and within Madagascar. Similar differentiation between the East African mainland and Madagascar was also found in other coral reef taxa, such as the stony coral S. hystrix , anemonefish Amphiprion akallopisos (Huyghe and Kochzius 2018) and Octopus cyanea (Van Nieuwenhove et al. 2019). Also mangrove taxa showed congruent patterns of differentiation, such as the mangrove tree Rhizophora mucronata (Triest et al. 2021), snails Terebralia palustris (Ratsimbazafy and Kochzius 2018), Littoraria subvittata and L. pallescens (Nehemia et al. 2019), as well as the giant mud crab Scylla serrata (Rumisha et al. 2017). However, in this study there is evidence for long-distance larval dispersal for A. tenuis in the NMC region, with exchange through the Mozambique Channel between northern Mozambique (Va, Pe, Na) and northern Madagascar (Ra, Nb). This is congruent with the hypothesis that coral reefs within the Northern Mozambique Channel are connected with each other through oceanographic conditions in the channel and share genetic input from the South Equatorial Current (SEC), causing them to have a more similar coral fauna (Obura 2012).
But this larval exchange is not exclusive within the NMC as there is an indication for exchange between site Ra (NMC) and the sites Nt and SM (Eastern Madagascar, EM), which are upstream of the NMC in the Northeast Madagascar Current (NEMC). Also, there is exchange from Nb and Ra with sites downstream of the SEC, such as Va and other sites downstream of the EACC along the coast to Tanzania and Kenya (Mo, Ki, Ja). Long-distance travel of larvae released in northern Mozambique up the coast of Africa (Tanzania, Kenya and up to Somalia) is supported by dispersal models for the WIO for corals with long PLD such as A. tenuis (Crochelet et al. 2016); PLD = 50 days). However, long-distance exchange of larvae is likely infrequent considering the levels of differentiation found in this study based on pairwise F ST and D EST values between Mozambique and northern Madagascar, particularly for Ra, while Nb depicts lower levels of differentiation and shows a mixed signature of both mainland sites as well as sites in southwest Madagascar (Sa, To, An). High genetic connectivity was found for most sites along the East African mainland coast, with no indication for strict genetic barriers between the NMC and EAC region. These results reconfirm the high connectivity found for a smaller study on A. tenuis in Tanzania and Kenya (van der Ven et al. 2016) and are congruent with the lack of distinct dispersal barriers found for corals with a substantial lower dispersal capacity (S. hystrix, van der Ven et al. 2021). High connectivity along the coast is supported by biophysical modelling studies, which propose Tanzania as a seeding source of larvae for downstream Kenya (Crochelet et al. 2016;Mayorga-Adame et al. 2016).

Substructure of reefs in the southern Mozambique channel
The sites within the SMC cluster together, although within the SMC there is a substructure based on both clustering and pairwise F ST and D EST values between sites in Mozambique (Vi, Ba) and Madagascar (Sa, To, An). This barrier was hypothesised based on the species distribution of corals, for which the SMC is described as a distinct area, but with a potential barrier in the Mozambique Channel (Obura 2012). These results also support a larval dispersal modelling study in the WIO, in which sites in southern Mozambique and southeast Madagascar are isolated from all other sites as well as each other (Crochelet et al. 2016); PLD = 50 days).
The sites in southern Mozambique (Vi, Ba) are significantly differentiated from most other sites in the WIO, except for In (southern Mozambique), which is supported by larval dispersal models that consider this area to be selfrecruiting, although connected with sites further south down to South Africa (Crochelet et al. 2016). This differentiation in southern Mozambique can be explained by the isolation of these smaller and patchier reefs that are separated from the northern reefs by an area devoid of coral reefs due to freshwater discharge and sediments from large-scale river outlets in central Mozambique. This pattern is supported by research on the stony coral Platgyra daedalea, in which Bazaruto (in this study Vi) and Inhaca (in this study In) were clustered together and separate from reefs in northern Mozambique (Montoya-Maya et al. 2016), as well as by a studies on Acropora austera, in which panmixia was found between Bazaruto and Inhaca, as well as between some of the Mozambique and South African sites (Macdonald et al. 2011;Montoya-Maya et al. 2016). Also, mangrove taxa, such as the mangrove tree R. mucronata (Triest et al. 2021), as well as the snails L. subvittata (Nehemia et al. 2019) and T. palustris (Ratsimbazafy and Kochzius 2018), show a genetic separation of population in southern Mozambique from other regions in the WIO.
The sites in the southwest of Madagascar show, besides mixing with sites in northern Madagascar (Nb), exchange of larvae with sites in northern Mozambique and Tanzania, as revealed by individual-based cluster analysis. This exchange was also found in drifter experiments (Ramanantsoa et al. 2018), and it supports a larval dispersal modelling study, in which strong southward dispersal and subsequently transport in the Mozambique Channel eddies were found for larvae released from sites south of Pe (Pembe, Mozambique; Gamoyo (2018)). Connectivity between southwest Madagascar and northern Mozambique is also observed in the stony coral S. hystrix . Lastly, southern Madagascar (SM) forms a separate group in Bayesian clustering, but with high differentiation of Lv, potentially caused by eddies and ocean currents directed away from land, high wave action and high asexual reproduction (discussed below).

Inbreeding within all sites and genetic diversity
Significant F IS values indicating heterozygote deficits were found for all sample sites. Heterozygote deficits are common in studies on scleractinian corals and were found in A. tenuis in northern Western Australia (Underwood 2009) and for A. austera in Mozambique and South Africa (Montoya-Maya et al. 2016). Heterozygote deficiencies can be caused by null alleles, but because the deficit was found for all loci in all locations, the contribution of null alleles is unlikely, which is partly supported by the analysis with Inest, indicating that inbreeding rather than null alleles is causing the heterozygote deficits at most sample sites. Additionally, the heterozygote deficits can be attributed to the sampling of different genetic cohorts, also known as the Wahlund effect, or a genetic patchiness caused by biological factors concerning temporal and spatial admixture, which can be amplified by high asexual reproduction (as seen in Fa, Lv, FD, see below). This fits the pattern of long-distance dispersal found in this study and was also found for A. tenuis from northern Western Australia (Underwood 2009).
Asexual reproduction caused by coral fragmentation is common in branching acroporids (Baums et al. 2006;Baums 2008), including A. tenuis (Underwood 2009). High clonal richness was found in three locations in southern Mozambique (Fa, Lv, FD), with 48% to 67% of the colonies formed by clonal fragmentation, compared to an average of 2% in all other sites combined. Since the sampling method in this study was designed to avoid sampling clones, the observed clonal richness might underestimate the total contribution of asexual reproduction in A. tenuis at these sites. The contribution of colonies formed by fragmentation was variable in northern Western Australia, with an average of 10% of sampled colonies for A. tenuis (Underwood 2009). A low contribution of asexual reproduction was also found for A. digitifera in Japan, where 1% of sampled individuals was formed by clones (Nakajima et al. 2010). The reefs on the coast of south Madagascar are influenced by extreme wave action, notably altering the coral's branched growth form towards a more robust and sturdier colony. This wave action is likely also causing fragmentation and subsequently the growth of clonal colonies, which can explain the high clonal richness in this area.
Patterns of genetic connectivity can be used to verify and explain proposed biogeographic regions such as Marine Ecoregions of the World (Spalding et al. (2007) and Ecoregions based on coral diversity (Obura 2012). Here we aimed to infer dispersal patterns to identify potential barriers in the WIO and Red Sea and to verify the proposed Ecoregions. The broadcast spawning coral A. tenuis displays genetic connectivity over long distances within the WIO, but the Red Sea is isolated confirming hypothesis I, which can be attributed to both historical and maintained contemporary barriers for larval dispersal. In the WIO, high genetic connectivity was found in the northern sites along the coast of Kenya and Tanzania and northern part of Mozambique, but sites in the South of Mozambique and offshore sites in the north and south of Madagascar appear less connected. Neither strong genetic barriers, nor differences in genetic diversity could be found within the WIO, rejecting hypothesis II-III based on biogeographic regions as proposed by Obura (2012). However, there is an indication that regions can be defined based on species diversity in Madagascar confirming hypothesis IV. Remarkable is the genetic connectivity found between northeast Madagascar and northern Mozambique and Tanzania, as well as between western Madagascar and the northern Mozambique sites, although this genetic connectivity is probably low or infrequent based on the high differentiation found among these sites. These patterns of high genetic connectivity combined with contemporary dispersal barriers can be explained by the long larval duration of A. tenuis and the prevailing northbound East African Coast Current that facilitates higher genetic connectivity along the northern East African Coast, while eddies in the Mozambique Channel are causing larval retention in south Mozambique and Madagascar.