Population differentiation or species formation across the Indian and the Pacific Oceans? An example from the brooding marine hydrozoan Macrorhynchia phoenicea

Abstract Assessing population connectivity is necessary to construct effective marine protected areas. This connectivity depends, among other parameters, inherently on species dispersal capacities. Isolation by distance (IBD) is one of the main modes of differentiation in marine species, above all in species presenting low dispersal abilities. This study reports the genetic structuring in the tropical hydrozoan Macrorhynchia phoenicea α (sensu Postaire et al., 2016a), a brooding species, from 30 sampling sites in the Western Indian Ocean and the Tropical Southwestern Pacific, using 15 microsatellite loci. At the local scale, genet dispersal relied on asexual propagation at short distance, which was not found at larger scales. Considering one representative per clone, significant positive FIS values (from −0.327*** to 0.411***) were found within almost all sites. Gene flow was extremely low at all spatial scales, among sites within islands (<10 km distance) and among islands (100 to >11,000 km distance), with significant pairwise FST values (from 0.035*** to 0.645***). A general pattern of IBD was found at the Indo‐Pacific scale, but also within ecoregions in the Western Indian Ocean province. Clustering and network analyses identified each island as a potential independent population, while analysis of molecular variance indicated that population genetic differentiation was significant at small (within island) and intermediate (among islands within province) spatial scales. As shown by this species, a brooding life cycle might be corollary of the high population differentiation found in some coastal marine species, thwarting regular dispersal at distances more than a few kilometers and probably leading to high cryptic diversity, each island housing independent evolutionary lineages.


| INTRODUCTION
In the context of biodiversity loss (Kolbert, 2014;Myers, Mittermeier, Mittermeier, da Fonseca, & Kent, 2000), assessing the degree of genetic connectivity (i.e., effective dispersal with consequences on gene flow; Carpenter et al., 2011;Jones et al., 2009;Schiavina, Marino, Zane, & Melià, 2014;Vellend & Geber, 2005) of marine populations is essential to establish effective marine protected areas. Such knowledge is valuable for determining the appropriate geographic span of their networks (Gerber et al., 2003), assuring the conservation of both evolutionary processes and alpha diversity (Christie et al., 2010). It is also indispensable for delineating relatively isolated populations or groups of populations more sensible to environmental variations, as they lack the capacity to acquire genetic variability from other populations (Cowen, Gawarkiewicz, Pineda, Thorrold, & Werner, 2007;Hellberg, 2009). Furthermore, connectivity studies can help determine whether individuals from a given species form a single randomly mating population or are members of different populations with various levels of genetic isolation. Indeed, genetic isolation fosters speciation opportunities (Audzijonyte, Baltrūnaitė, Väinölä, & Arbačiauskas, 2015) but may also cause extinctions of local populations (Underwood, Smith, van Oppen, & Gilmour, 2007). There is no consensus nor a general framework outlining the levels of connectivity at which populations should be considered as independent (Waples & Gaggiotti, 2006).
Genetic differentiation among populations may be observed in cases where an ancient separation is maintained with low migration rates, but also when a recent divergence arose without gene flow (Boissin, Hoareau, & Berrebi, 2011;Duda & Lessios, 2009). Indeed, geographic isolation alone is not sufficient to assess the degree of isolation (Donald, Keeney, & Spencer, 2011). Instead, population genetics provides a robust theoretical framework to estimate gene flows over multiple generations from which the degree of connectivity between pairs of populations can be assessed (Kool, Moilanen, & Treml, 2013;Wright, 1931). Traditional estimates of migration rates are based on population differentiation indices, notably Wright's F ST (Wright, 1931). However, the mathematical models linking genetic variance to migration rates make numerous assumptions that are often biologically unrealistic and violated (Hedgecock et al., 2007;Whitlock & McCauley, 1999), for example, no selection nor mutation within populations. Nevertheless, indices of population differentiation can still be used with confidence in comparative studies (Meirmans & Hedrick, 2010). Additionally, individual-based methods have been developed to highlight actual populations admixture (e.g., Jombart, Devillard, & Balloux, 2010;Pritchard, Stephens, & Donnelly, 2000), giving individual assignment probabilities to putative populations. Even though all these methods cannot estimate several population parameters (e.g., direction of migration, effective population size), identifying similar patterns of genetic structuring for multiple species with similar life history traits over the same geographic area is highly informative on the degree of population genetic connectivity, and consequently for optimizing and refining the design of marine protected areas networks.
For the majority of marine invertebrates, the adult phase is benthic with low mobility or fixed to the substrate and larvae represent the major dispersal phase ensuring population connectivity (Lopez-Duarte et al., 2012;Selkoe & Toonen, 2011;Treml, Halpin, Urban, & Pratson, 2007) and species cohesion (Knowlton & Jackson, 1993).
However, direct measures of larval dispersal are presently unfeasible as they depend on various factors that are often difficult to assess in the field, such as oceanic circulation, sea temperature, larval behavior, larval energetic resources, available habitats, and food resources [see Selkoe and Toonen (2011) for a review]. Yet this stage of the life cycle strongly influences the effective dispersal, which additionally encompasses the survival of larvae and adults, larval settlement on the substrate, and sexual reproduction with local conspecifics (Pineda, Hare, & Sponaungle, 2007). Pelagic larval duration (PLD; Shanks, 2009) is often used as a proxy for larval dispersal distance and connectivity, even though past biogeographic events affect genetic structure of marine populations (Faurby & Barber, 2012), modifying populations connectivity regardless of species PLD. Thus, the correlation between PLD and connectivity is not straightforward and must be considered cautiously (Paulay, 2006;Shanks, 2009). Indeed, if theory predicts that species presenting long PLDs will display high dispersal and thus low genetic structure, numerous examples of the contrary exist [see Weersing and Toonen (2009)

for a review].
Hydrozoans represent one of the oldest marine clades, and they have colonized all aquatic ecosystems across the globe since their appearance during the Cretaceous (Bouillon, Gravili, Pagès, Gili, & Boero, 2006;Park et al., 2012). They are among the first fixed organisms to colonize new habitats and provide shelter to a wide variety of invertebrate and microbial taxa (Boero, 1984;Gili & Hughes, 1995).
One of their key features is the variety of life history traits and reproductive strategies, notably including a medusa stage of variable duration depending on the taxon (Boero, Bouillon, & Piraino, 1992;Boero et al. 1995). The Aglaopheniidae Marktanner-Turneretscher, 1890 is one of the most species-rich families with more than 250 extant species found in all marine ecosystems (Millard, 1975); they are characterized by the absence of a medusa stage and by the incubation of larvae in dedicated structures, even if some species reacquired a temporally reduced medusa-like stage during their evolution (Leclère et al., 2009;Leclère, Schuchert, & Manuel, 2007). Aglaopheniids' genus-level taxonomy is mainly based on the morphology of the reproductive structures (Bouillon et al., 2006), but as many other characters, life cycles of hydrozoans are subject to convergent or reversible evolution across their phylogeny (Collins, 2002;Leclère et al., 2007Leclère et al., , 2009Marques & Collins, 2004;Miglietta & Cunningham, 2012) and the diversity of Aglaopheniidae is still under assessment (Moura et al. 2012;Postaire et al., 2016aPostaire et al., , 2016b. Active dispersal in this family is thought to be limited and only achieved via spermatozoids and mature larvae (Schuchert, 2014;Winston, 2012), an assumption that was confirmed using microsatellite data for a single morpho-species, Lytocarpia brevirostris (Busk, 1852), in a recent study centered on the Western Indian Ocean . Similar results were obtained in a study of the genetic connectivity of populations on the globally invasive hydrozoan Cordylophora Allman, 1844, using microsatellites but with a geographically and ecologically more limited sampling centered on the North American Great Lakes basin (Darling & Folino-Rorem, 2009). These two studies supported the idea of weak dispersal abilities in some hydrozoans due to a lack of a long dispersal phase, resulting in a pattern of high genetic differentiation among populations: One could consider each sampling site as hosting an independent biological species (Schuchert, 2014). However, more studies are needed to confirm these preliminary conclusions. Indeed, an important number of Aglaopheniidae morpho-species, as many other hydrozoans, contradict the postulate of limited connectivity: they present global distribution ranges and occur in a wide range of habitats and depths (Millard, 1975).
One of the first steps to conduct population genetic studies is to identify the species (Pante et al., 2015). In Aglaopheniidae, integrative taxonomy (Schlick-Steiner et al., 2010) and molecular-based species delimitation methods allowed the delineation of robust species hypotheses in this clade (Postaire et al., 2016a). Here, the clade formed by Macrorhynchia phoenicea (Busk, 1852) is a typical morpho-species presenting high morphological plasticity, asexual reproduction through stolon growth, a monophasic dioecious larviparous life cycle-that is, the larvae produced after internal fertilization are not released until competent-and an Indo-Pacific distribution on coral reefs (Di Camillo, Puce, & Bavestrello, 2009;Millard, 1975). Species delimitation methods based on DNA revealed that this morpho-species is actually composed of at least two sympatric cryptic species [sensu Bickford et al. (2007)], referred to as M. phoenicea morphotypes A and B in Postaire et al. (2016a) and henceforth named M. phoenicea α and β, respectively. They can be distinguished using a combination of general colony shape, color, microhabitats, and genetic data. The distribution ranges of both species differ, as M. phoenicea α is composed of two divergent lineages present in the Western Indian Ocean and the Tropical Southwestern Pacific [sensu Spalding et al. (2007)], whereas M. phoenicea β seems restricted to the Western Indian Ocean. The sexual dispersal abilities of M. phoenicea α are assumed to be limited in natural conditions. In laboratory conditions, larvae of M. phoenicea α settle in less than 24 hr (BP, pers. obs.), as found in other hydrozoan species (Sommer, 1990). Furthermore, hydrozoan sperm cells are reported to present a short planktonic life (4 hr; Yund, 1990).
To complement and confirm previous work on the population connectivity of marine hydrozoans (Darling & Folino-Rorem, 2009;Postaire et al., 2017), intensive sampling of M. phoenicea α populations was conducted in the Western Indian Ocean and the Tropical Southwestern Pacific. The aims were to (i) investigate the structure and connectivity of M. phoenicea α populations using microsatellites (Postaire et al. 2015), (ii) compare the results with the study of another Aglaopheniidae with a similar reproductive strategy , and (iii) discuss the distribution ranges of Aglaopheniidae species in light of our results.

| Sampling and DNA extraction
Thirty sampling sites were explored within two marine provinces (Spalding et al., 2007): the Western Indian Ocean and the Tropical Southwestern Pacific Ocean, presenting seven islands/archipelagoes (Table 1). At each site, individuals (feather-shaped units) were collected haphazardly using scuba during a single dive (ca. 60 min) and were placed in sequentially numbered individual bags to approximate distances between individuals; we preferentially collected individuals several centimeters apart to limit clone sampling. Macrorhynchia phoenicea α (see Supplementary Material 1 in Postaire et al., 2016a) was commonly found on outer reef slopes exposed to strong currents, often associated with Pocillopora colonies, suggesting ecological preferences in this species. Large individuals with visible reproductive structures were preferentially sampled, and all samples were stored in 95% ethanol before DNA extraction. Preliminary species identification was performed in the field (as explained in Postaire et al., 2016a) and later confirmed by detailed inspection of morphological characters (Millard, 1975) (Table 1).
Considering the whole dataset, over the 26 available loci for M. phoenicea sp., 15 amplified correctly M. phoenicea α individuals, that is, presented less than 10% of missing data, and were considered for all analyses.  Table 1).

| Summary statistics
One representative of each MLG per site was used for further analyses. All tests in this study were corrected for false discovery rate (FDR) in multiple tests (Benjamini & Hochberg, 1995 (Kalinowski, 2005) software to correct for uneven sample sizes by rarefaction. The software sampled 11 individuals at random from each site to match the smallest sample size (i.e., MAY1; Table 1).

| Population differentiation
We investigated population differentiation and structure using four  (Raymond & Rousset, 1995a) were performed in Genepop v.4.6 (Raymond & Rousset, 1995b). To understand the mechanisms that may be responsible for the observed patterns of population structure, we compared estimates of genetic differentiation to geographic distances among sites. We used a Mantel test (Mantel, 1967) (Table 2). This relationship is expected to be positive and linear in the context of a twodimensional Isolation by distance (IBD) model (Rousset, 1997). All Mantel tests were performed using the program GENODIVE v.2.0 (Meirmans & van Tienderen, 2004) with 10 4 random permutations to assess significance.

| Clustering analyses
Population structuring was also assessed without a priori stratification of samples. We first performed a discriminant analysis of principal components (DAPC) using the package adegenet (Jombart, 2008;Jombart et al., 2010) DAPC is a non-model-based method that maximizes the differences among groups while minimizing variation within groups without prior information on individuals' origin. In addition, the method does not assume HWE or absence of LD. We used the function find.clusters() to assess the optimal number of groups with the Bayesian in- is ideally the optimal number of clusters). Note that BIC values may keep decreasing after the true K value in case of genetic clines and hierarchical structure (Jombart et al., 2010) and that retaining too many discriminant functions with respect to the number of populations may lead to overfitting the discriminant functions, resulting in spurious discrimination of any set of clusters. Therefore, the rate of decrease in BIC values was visually examined to identify values of K after which BIC values decreased only subtly (Jombart et al., 2010); we tested values of K = 1-30. The dapc() function was then executed using the best grouping, retaining axes of PCA sufficient to explain ≥70% of total variance of data, and coloring individuals according to their sampling site.
The population clustering was also explored using the software Structure v.2.3.2 (Pritchard, Wen, & Falush, 2010;Pritchard et al., 2000), with the admixture model and correlated allele frequencies (Falush & Pritchard, 2003). This analysis assumes that within the analyzed dataset reside K populations, and individuals are assigned probabilistically to each population in order to maximize HWE and minimize LD. Due to the important size of our dataset and following the recommendations of Rosenberg et al. (2002) and Jakobsson et al. (2008), we studied our dataset using a hierarchical approach. For each group of sites ( Figure 2) and each tested value of K (K varying from 1 to 10), three independent runs were conducted with a burn-in period of 5 × 10 4 steps followed by 5 × 10 5 Markov chain Monte Carlo iterations. We used the statistic proposed by Evanno, Regnaut, and Goudet (2005), implemented in Structure Harvester v.1.0 (Earl & von-Holdt, 2012), to estimate the best number of K for each group of sites.
The software CLUMPP v.1.0 (Jakobsson & Rosenberg, 2007) was used to summarize results, and they were formatted with DISTRUCT v.1.1 (Rosenberg, 2004). The software Arlequin v.3.5 (Excoffier & Lischer, 2010) was then used to perform hierarchical analyses of molecular variance using clusters identified by Structure as populations, which mostly corresponded to islands/archipelagoes, and provinces as groups.
Finally, network analyses were performed on individuals and sites. The pattern of genetic relationship among individuals was illustrated by networks built with two measures integrating genetic information in terms of time and divergence history: the Rozenfeld Distance index (RD) and the Shared Allele Distance index (SAD). RD has been developed from the Goldstein distance index. It provides a parsimonious representation of the genetic distance between individuals based on the difference of the microsatellites allele lengths (Rozenfeld et al., 2007). On the other hand, SAD provides the genetic distance between individuals based on the proportion of shared alleles (Chakraborty & Jin, 1993). RD helps to resolve ancestral polymorphism through allele lengths impinged on slow evolutionary processes, while SAD helps to understand recent gene flow characterized by direct allelic exchange. The global pattern of genetic relationships among sites was illustrated by networks built with two different measures: the Goldstein distance index (GD) and F ST fixation index (F ST ). The GD groups sites considering their historical origin, while F ST takes into account the site structure. Once the matrices of genetic distances between individuals or sites were estimated, different networks were built considering individuals/sites and genetic distances as nodes and links between them, respectively. For the network construction, links were included for all distances and were removed in decreasing order until the percolation threshold (Dpe) was reached (Rozenfeld et al., 2007), threshold below which the network fragmented into small clusters. The average clustering coefficient < C > of the whole network was estimated for each of the four built networks. These analyses were performed using EDENetworks software (Kivelä, Arnaud-Haond, & Saramäki, 2015).

| Multilocus genotyping and asexual reproduction
Using the 15 loci that amplified correctly (see Section 2.2), our analysis of 1,257 individuals yielded 1,081 MLGs, indicating the presence of asexual reproduction in some sites (Table 1). Individuals sharing the same MLG were always found within the same site (i.e., no MLGs were shared among sites) and were found close to one another (i.e., small difference in sampling numbers). The clonal richness R ranged from 0.449 to 1.

| Genetic clusters
Both DAPC and Structure analyses indicated significant structuring of sites, with MLGs clustering according to their geographic origin.
The first round of Structure analyses identified two clusters, each

| Network analysis
The topology of the network built with the SAD index at the percolation threshold (Dpe = 0.92) showed that individuals of one island/ archipelago remained linked and were more closely related to each other than to individuals from other islands/archipelagoes. In contrast, the network built with the RD index (Dpe = 5.62) resulted in less geographic structure among MLGs (Fig. S1). The average clustering coeffi-  (Table S3). The genetic variation explained by differences between provinces was higher than the genetic variation explained by differences among islands within provinces (23.51%* and 17.52%***, respectively), but the highest amount of genetic variation was found among sites within islands (58.97%***). to 0.558*** between GDT4 and GDT9 (mean ± SE = 0.240 ± 0.007; Table 4). The lowest differentiation values recorded in our sampling were measured among sites from the Chesterfield Islands (mean ± SE = 0.064 ± 0.004), with a maximum of 0.111*** between CHE3 and CHE8. Overall, the differentiation among sites from Grande Terre or among those from the Loyalty Islands was approximately half that of the differentiation that existed among the Western Indian Ocean sites: Within Grande Terre, values ranged from 0.064*** between GDT1 and GDT2 to 0.558*** between GDT4 and GDT9

| DISCUSSION
We explored the population genetic structuring and connectivity of  and associated islands). Our results revealed a high level of genetic differentiation among sites across the Indo-Pacific at all spatial scales, with strong isolation by distance, and with genetic clusters mostly corresponding to islands. Our findings are in accordance with a growing body of literature highlighting the extreme spatial structuring of marine hydrozoans that lack a medusa dispersal stage Schuchert, 2005Schuchert, , 2014.

| Life history traits affect genetic diversity
Macrorhynchia phoenicea α showed departures from HWE in almost all sites, generally with significant heterozygote deficit (revealed by high positive F IS values). However, this result could be explained by the presence of null alleles which may occur due to some mutations in the flanking regions of microsatellite loci (Callen et al., 1993). Yet, while these are important to reveal, null alleles have little effect on structuring analyses when populations are strongly differentiated (Carlsson, 2008;Putman & Carbone, 2014), as observed here. Biological processes, such as nonrandom mating between individuals, inbreeding and/or Wahlund effects, probably also contribute to the heterozygote deficit within sites. Dioecious Aglaopheniidae species, like M. phoenicea α, are generally larviparous and several life history traits (supposed limited larval dispersal abilities and reproduction between spatially proximate individuals) intuitively enhance self-recruitment and minimize emigration out of settled populations: larvae that settle quickly should remain close to the mother individual if they encounter suitable environmental conditions, thus forming patches of related individuals over several generations. This assumption, however, has not been tested yet.
Heterozygosity deficiencies could also be due to a temporal Wahlund effect resulting from (i) different cohorts at each site or (ii) different breeding units among sampling sites, as proposed to explain the high heterozygosity deficiencies in Caribbean sponges (Chaves-Fonnegra, 2014;Duran, Pascual, & Turon, 2004). Indeed, the availability of food and oxygen are the main limiting resources for growth, sexual reproduction, and gamete production in hydrozoans (reviewed in Gili & Hughes, 1995). Thus, local conditions (water flow, temperature, planktonic productivity, sedimentation) could result in desynchronized reproduction among individuals in the population, favoring inbreeding. This is a plausible hypothesis as M. phoenicea α appears to reproduce throughout the year (BP pers. obs.), similarly to the tropical aglaopheniid hydrozoan Lytocarpia brevirostris . Inbreeding might also be fostered by a spatial Wahlund and cohort studies are necessary to resolve these issues, but this is particularly difficult in hydrozoans due to their relative small size and numerous hidden stages (Gili & Hughes, 1995 (Bouillon et al., 2006;Gili & Hughes, 1995), long periods of asexual reproduction can lead to negative F IS values (Balloux, Lehmann, & de Meeus, 2003;Stoeckel & Masson, 2014).
Thus, life history traits seem to profoundly affect genetic diversity at the site scale (<200 m).

| Small-scale spatial genetic structure and diversity
Macrorhynchia phoenicea α is distributed on many reefs in the Western  (Table S2), as well as various loci failing to amplify in individuals from several sites (potentially because of null alleles). Additionally, Bayesian clustering, PCA and network analyses identified a highly geographically structured dataset, populations grouping according to islands or archipelagoes. Furthermore, private alleles were present within all sites (but with a higher frequency in the Western Indian Ocean) and the number of alleles per loci was extremely variable among sites (sometimes even monoallelic).
The population structuring described here is comparable to the pattern uncovered in the brooding Aglaopheniidae L. brevirostris α in the Western Indian Ocean and the Tropical Southwestern Pacific . The similar, but not identical geographic coverage of the sampling, due to the absence of the considered species at some sampling sites (Postaire et al., 2016a)  in several other marine species, ranging from kelps to teleost fishes, for which a high pairwise differentiation was measured when habitat patches were isolated (Alberto et al., 2010;Billot et al., 2003;Riginos & Nachman, 2001). Habitat continuity might thus be an important predictor of genetic connectivity of coral reef species, having important implications for marine conservation planning, but also on macroevolutionary processes.

| Large scale isolation by distance and speciation opportunities
In this study, we detected population IBD over relatively large spatial scales (several hundreds of km, i.e., archipelago scale or higher) and  (Table 2).
We used Euclidian distances to measure distances between sites, ignoring the presence of landmasses and the general direction of marine currents, although they are known to influence the connectivity of marine organisms (Schiavina et al., 2014;White et al., 2010). Oceanic circulation models of the studied regions are still under development and we could not meaningfully adjust our dispersal distance estimates.
Both the observed IBD pattern and the large distribution range of M. phoenicea α are related to its life history traits. Similar to many other hydrozoans (Gili & Hughes, 1995) Considering these results, the actual number of hydrozoans species may be considerably higher than previously thought. While their rafting ability has been proposed earlier to explain the apparent global distribution of several hydrozoan morpho-species (Cornelius, 1981(Cornelius, , 1992, inferring distribution ranges of hydrozoans species based on morphology alone might be erroneous as morpho-species that comprise multiple cryptic species and allopatric lineages are common (e.g., Leclère et al., 2007;Moura et al., 2012;Postaire et al., 2016a).

| CONCLUSIONS AND IMPLICATIONS FOR MARINE BIODIVERSITY CONSERVATION
Our study revealed that Macrorhynchia phoenicea α is composed of multiple, highly genetically isolated metapopulations, with low genetic diversity and high consanguinity (or traces of population functioning mainly via asexual reproduction). The simplest explanation for the observed genetic structuring and low connectivity is larviparity: limited planktonic dispersal capacity induces small effective population size by reducing gene flow between populations, accelerating genetic drift. This reproductive strategy combined with the inferred capacity to successfully disperse through rafting can account for their apparent extended distribution but these traits also enhance speciation opportunities. From an evolutionary point of view, each island hosts a species (sensu Samadi & Barberousse, 2006) and our study highlights the preeminent role of allopatrism and vicariance in the diversification of coastal brooding species (Paulay & Meyer, 2002). Rather than real cosmopolitan species, hydrozoans and many other marine organisms are likely mosaics of morphologically similar independent metapopulations, or even species (depending on the criterion used), and thus should be studied accordingly (Pante et al., 2015). These results highlight that speciation in the sea can occur at small spatial scales, contributing to the accumulation of species in marine biodiversity hotspots.
The observed geographic structuring does not correspond to defined biogeographic ecoregions (Spalding et al., 2007), exemplified by the Western and Northern Madagascar ecoregion comprising three clusters and New Caledonia, at least four (i.e., the Chesterfield Islands, West and East coasts of Grande Terre, and the Loyalty Islands). Similar discrepancies have been observed in several organisms from the Western Indian Ocean, such as scleractinians (Ridgway & Sampayo, 2005;Ridgway et al., 2008), coastal fishes (Muths et al., 2011), marine turtles (Bourjea et al., 2006), and hydrozoans , highlighting the disjunction between the northern and southern parts of the Mozambique Channel and the isolation of Juan de Nova Island, probably due to the presence of oceanic gyres. Our results underline that the hierarchical three-level classification (i.e., realm, province, and ecoregions) proposed by Spalding et al. (2007) is too coarse to encompass the genetic diversity of larviparous hydrozoans and potentially many other marine species. For marine brooding organisms with low PLDs, each island/archipelago could potentially represent an evolutionary hotspot (Hoareau et al., 2013;Vandergast et al., 2008), underlining the need of a network of marine protected areas to ensure the conservation of marine organisms as well as the maintenance of evolutionary mechanisms across oceans, rather than delimiting a limited number of extended marine sanctuaries.

ACKNOWLEDGMENTS
We gratefully acknowledge the Laboratoire d'Excellence CORAIL for financial support. Hydrozoan sampling in New Caledonia (HM) was Sampling in Mayotte (HM) was supported by the project SIREME (FED).
We gratefully acknowledge the Plateforme Gentyane of the Institut

CONFLICT OF INTEREST
None declared.