Is reproductive strategy a key factor in understanding the evolutionary history of Southern Ocean Asteroidea (Echinodermata)?

Abstract Life traits such as reproductive strategy can be determining factors of species evolutionary history and explain the resulting diversity patterns. This can be investigated using phylogeographic analyses of genetic units. In this work, the genetic structure of five asteroid genera with contrasting reproductive strategies (brooding: Diplasterias, Notasterias and Lysasterias versus broadcasting: Psilaster and Bathybiaster) was investigated in the Southern Ocean. Over 1,400 mtDNA cytochrome C oxidase subunit I (COI) sequences were analysed using five species delineation methods (ABGD, ASAP, mPTP, sGMYC and mGMYC), two phylogenetic reconstructions (ML and BA), and molecular clock calibrations, in order to examine the weight of reproductive strategy in the observed differences among phylogeographic patterns. We hypothesised that brooding species would show higher levels of genetic diversity and species richness along with a clearer geographic structuring than broadcasting species. In contrast, genetic diversity and species richness were not found to be significantly different between brooders and broadcasters, but broadcasters are less spatially structured than brooders supporting our initial hypothesis and suggesting more complex evolutionary histories associated to this reproductive strategy. Broadcasters' phylogeography can be explained by different scenarios including deep‐sea colonisation routes, bipolarity or cosmopolitanism, and sub‐Antarctic emergence for the genus Bathybiaster; Antarctic‐ New Zealand faunal exchanges across the Polar Front for the genus Psilaster. Brooders' phylogeography could support the previously formulated hypothesis of a past trans‐Antarctic seaway established between the Ross and the Weddell seas during the Plio‐Pleistocene. Our results also show, for the first time, that the Weddell Sea is populated by a mixed asteroid fauna originating from both the East and West Antarctic.


| INTRODUC TION
The diversity of marine life in the Southern Ocean (SO) has long been underestimated, and many taxa could be overlooked or misidentified due to the reliance on systematics based solely on morphological characters (Clarke, 2008;Gutt, Sirenko, Smirnov, & Arntz, 2004).
The growing awareness of these issues and the lack of representative sampling (Griffiths, Putte, & Danis, 2014) have prompted Antarctic marine biologists to undertake a series of comprehensive census surveys covering a wide range of taxa in the last two decades.
International efforts such as the International Polar Year (IPY 2007(IPY -2008 and the Census of Antarctic Marine Life (CAML 2005(CAML -2010 have been the launching pads for a better assessment of SO biodiversity and its underlying ecological processes. Recent studies that have applied molecular techniques to these SO specimens, exploring diversity, systematics, and phylogeography, have significantly increased our understanding of Antarctic benthic ecosystems (Sands, O'Hara, Barnes, & Martín-Ledo, 2015). These international SO sampling expeditions have achieved several major objectives, such as the creation of a baseline census of biodiversity (De Broyer & Danis, 2011;Griffiths, Danis, & Clarke, 2011), proposing how evolution has been influenced by the regional geological, climatic, and oceanographic histories (Fraser, Nikula, Spencer, & Waters, 2009;González-Wevar et al., 2018), and disentangling phylogeographic patterns at lower taxonomic levels to better understand relationships among populations and species on a case-by-case strategy (Brasier et al., 2017;Dömel, Melzer, Harder, Mahon, & Leese, 2017). This extensive work led to the discovery and the description of many new species (d'Udekem d' Acoz & Verheye, 2017;Janosik & Halanych, 2010). More importantly, these studies show the frequent discordance existing between traditional (morphology-based) and molecular (DNA-based) methods for assessing species diversity (Dömel et al., 2017;Janosik, Mahon, & Halanych, 2011). Cryptic speciation is a documented source of species diversity underestimation, but it is not the only one. A recent study on the most studied sea star species in the SO, Odontaster validus, Peck, Clark, and Dunn (2018) showed that polymorphism in some morphological characters could lead to misidentification of this frequently encountered species. Incorrect taxonomic assignments due to the lack of clear identification keys (Allcock & Griffiths, 2014), descriptions of nominal species based on distribution only (Díaz, Féral, David, Saucède, & Poulin, 2011;Saucede et al., 2015), and descriptions based on juvenile specimens (Roberts, Hopcroft, & Hosie, 2014) or on deteriorated specimens due to inappropriate conservation practices are all common limitations in meaningful biodiversity assessments (Meyer, 2016).
Processes that have and continue to drive complex diversity patterns in the SO are far from being fully understood, but the role of certain drivers has been demonstrated in a number of molecular studies. Processes can be extrinsic (e.g., paleogeographic, climatic, oceanographic) or intrinsic (life history traits) (Allcock & Strugnell, 2012;Thatje, 2012), and their combined effects are cumulative, making identification of explanatory processes a difficult endeavor. Life history traits such as reproductive strategies have been proven to shape the genetic structure of species in contrasting ways (Modica, Russini, Fassio, & Oliverio, 2017). Most marine benthic organisms show low to zero mobility during the adult stage. The dispersal capacity of larvae is thus expected to condition population genetic structure (Bowen, Bass, Muss, Carlin, & Robertson, 2006;Purcell, Cowen, Hughes, & Williams, 2006). Larval development (e.g., planktotrophic, lecithotrophic), parental care (brooders vs. broadcasters), and pelagic larval duration (from days to several months; Selkoe & Toonen, 2011) are intrinsic factors that can combine with extrinsic factors, such as oceanographic currents or geological history, to determine the population genetic structure (Hoffman, Clarke, Linse, & Peck, 2011;Jossart et al., 2017).
As a consequence of their high diversity and differentiated functional roles, benthic invertebrates have been the subject of many studies on genetic diversity and connectivity (see Riesgo, Taboada, & Avila, 2015 for a review). Sea stars (Asteroidea, Echinodermata) represent a diversified, abundant, and common ecological group in SO benthic habitats (Jossart, Moreau, Agüera, De Broyer, & Danis, 2015). To date, around 300 species have been recorded from the SO Moreau et al., 2018) but it is likely that many more remain to be described. Recent studies based on species morphological identification tried to disentangle distribution patterns and biogeographic relationships in sea stars at the scale of the entire SO (Moles et al., 2015;Moreau et al., 2017). These studies described the spatial structure of asteroid assemblages and stressed the major influence of life history traits, among which reproductive strategy appeared to have the most notable effect (Moreau et al., 2017).
Few molecular studies have been performed on SO asteroids, and they are nearly exclusively focused on the common Odontaster genus (Janosik & Halanych, 2010;Janosik et al., 2011). The phylogeny and evolutionary history of the Asteroidea in the SO have, however, been discussed in several studies (Mah & Foltz, 2011a, 2011b, 2014Mah et al., 2015). In their comprehensive studies on the molecular phylogeny of the superorders Valvatacea and Forcipulatacea, Mah and Foltz (2011a) describe some diversification processes in several Antarctic and sub-Antarctic families (e.g., Odontasteridae, Antarctic Asteriidae) but they did not analyze the lower taxonomic levels (i.e., genera and species).
Brooders and broadcasters have contrasting dispersal capabilities (low-range dispersal in brooders vs. high range in broadcasters), and brooding taxa are usually hypothesized to display (a) higher genetic diversity, (b) greater species richness, and (c) more genetic differentiation through space than broadcasters (Modica et al., 2017;Purcell et al., 2006). In the present study, we tested these assumptions by investigating the genetic diversity and the phylogeography of five widely distributed asteroid genera across the SO.

| Comprehensive sampling and DNA sequence compilation
Studied specimens were sampled during several expeditions in the SO (Figure 1; Appendix S1); all were preserved in 96% ethanol or frozen and identified at species or genus level (either by Christopher Mah or Camille Moreau). Genomic DNA was extracted using either a salting-out protocol (modified from Sunnucks & Hales, 1996, with larger volumes and incubation at 70°C for 10 min to inhibit protein activity after digestion) or Qiagen DNeasy extraction kits. A fragment (612 nucleotides) of the mitochondrial gene cytochrome c oxidase subunit I (COI) was then amplified using the specific forward primer LCOech1aF1 for the class Asteroidea and the universal HCO2198 reverse primer (Folmer, Black, Hoeh, Lutz, & Vrijenhoek, 1994). COI sequences amplified using the same primers were also obtained through The Barcode of Life Data System (BOLD, Ratnasingham & Hebert, 2007) in both public and private datasets, accounting for 460 of the 1,416 sequences analyzed in this study (Appendix S1). As the Astropectinidae genera are also recorded outside the SO (under different nominal species), all available sequences for these from outside the SO were included within this study (Appendix S1).
Reverse and forward sequences were edited and assembled using CodonCode Aligner v6.0.2 and translated using the echinoderm mitochondrial genetic code to ensure the absence of a stop codon.

| Phylogenetic reconstruction
Due to relatively high genetic distances, phylogenetic relationships were reconstructed independently within the Asteriidae and the Astropectinidae. Coscinasterias muricata and Thrissacanthias penicillatus were used as outgroups, respectively, following previous phylogenetic studies (Mah & Foltz, 2011a, 2011b. Maximum likelihood (ML) and Bayesian analyses (BA) were used to construct the trees using only unique haplotypes. ML reconstructions were generated F I G U R E 1 Maps of the SO indicating sample location for each target group (red stars). Sampling locations are labeled: red-South Georgia; pink-South Sandwich Islands; yellow-Bouvet Island; dark blue-Kerguelen Islands; light blue-East Antarctica; light green-Amundsen Sea; orange-Antarctic Peninsula; dark green-Burdwood Bank; and purple-Weddell Sea. Projection: South Pole Stereographic using a codon partitioned model and the GTR + G substitution model in RAxML v 8.1.2 (Stamatakis, 2014) through the raxmlGUI interface (Silvestro & Michalak, 2012). To assess branch support, 10 runs were realized with 1,000 thorough bootstraps each. The PartitionFinder v2 software (Lanfear, Frandsen, Wright, Senfeld, & Calcott, 2016) was used for the BA reconstructions on the CIPRES Science Gateway (Miller, Pfeiffer, & Schwartz, 2010) to select for best-fit partitioning schemes and models of evolution. An XML file was created with

| Species delineation
Several single-locus methods of species delineation were used to delineate and explore diversity among the studied genera. Two of these methods, the Generalized Mixed Yule Coalescent (GMYC- Pons et al., 2006;Fujisawa & Barraclough, 2013) and the multirate Poisson Tree Processes (mPTP- Kapli et al., 2017), are tree-based methods requiring an ultrametric tree for the former and a maximum likelihood tree for the latter. Both single (sGMYC)-and multiple-threshold (mGMYC) models were investigated using the R package SPLITS (Ezard, Fujisawa, & Barraclough, 2009), and the ultrametric tree was obtained using BEAST during the phylogenetic reconstruction. The online web service (available at http://mptp.hits.org) was used for mPTP with the ML tree constructed using RAxML. We also performed a distance-based analysis using the Automatic Barcode Gap Discovery (ABGD-Puillandre, Lambert, Brouillet, & Achaz, 2012) on the online server (http://wwwabi.snv. jussi eu.fr/publi c/abgd/abgdw eb.html) with default settings for the prior range (0.001, 0.1), K80-corrected distances, and a value of 1.5 for the relative gap width (X). We also used the newly developed ASAP method (http://wwwabi.snv.jussi eu.fr/publi c/asap/; N. Puillandre et al., in prep.) with default settings and K80-corrected distances. For subsequent analyses, we have selected the method giving the smallest number of delineated entities as recommended to avoid false interpretation (Carstens, Pelletier, Reid, & Satler, 2013). All molecular diversity indexes were calculated using the DnaSP v6 software (Rozas et al., 2017) for each previously delineated species. Differences in haplotype and nucleotide diversity between brooders and broadcasters were tested using a Wilcoxon-Mann-Whitney test (R Core Team, 2018). Only calculations for sample sizes ≥25 individuals were discussed as suggested by Goodall-Copestake, Tarling, and Murphy (2012). Intra-and interspecific genetic distances within each genus were calculated using MEGA v7.0.18 (Kumar, Stecher, & Tamura, 2016) and the Kimura 2-parameter model. Haplotype networks were generated using a TCS network method (Clement, Snell, Walker, Posada, & Crandall, 2002)

| RE SULTS
In total, 1,416 sequences of 612bp were included in the analyses (Appendix S1, Table 1, Figure 1), varying from 224 for Lysasterias to 318 for Notasterias. Neither stop codons nor significant evidence for base heterogeneity were observed.
Haplotype and nucleotide diversity are not significantly different between brooders and broadcasters (Wilcoxon-Mann-Whitney tests; p-values: 0.1709 and 0.9433, respectively).

| Phylogenies and divergence time estimates
In total, 578 sequences in broadcasters and 838 in brooders were used for ML and BA reconstructions. Both methods give congruent results for broadcasters with similar taxonomic groupings and high node supports ( Figure 2). Node support was higher using the BA method for brooders. Branching patterns were, however, identical in all reconstructions.
The monophyly of the genus Psilaster is not supported by our analysis, but SO Psilaster representatives are monophyletic (i.e., the nominal species P. charcoti). The genus Bathybiaster is monophyletic and includes specimens from the SO and from the Northern Hemisphere  Figure 2). This is also in line with the time range computed for broadcasters.

| Phylogeographic patterns
All brooders display clear geographic patterns, with a distinction between the East Antarctic and the Antarctic Peninsula

| Species richness, genetic diversity, and taxonomic implications
We found that genetic diversity did not significantly differ between At best, these methods should be considered as a first step toward subsequent and more integrative taxonomic works (Kekkonen et al., 2015), and therefore, a conservative approach is recommended (Carstens et al., 2013). Genetic diversity among the delineated spe- integrative taxonomic approach at both species and genus levels.

| Dispersal capacity and biogeographic patterns
Several mechanisms have been proposed to account for the unusually high proportion of brooding species in the SO, and the evolutionary success of Antarctic brooders is now widely recognized within the community of Antarctic biologists (Poulin, Palma, & Féral, 2002). While adaptation to polar environmental conditions is not considered a determining mechanism (Pearse, Mooi, Lockhart, & Brandt, 2009;Poulin et al., 2002), reproductive strategy has been hypothesized as the main factor structuring the evolution and the diversity of SO benthic life (Raupach et al., 2010;Thatje, 2012). Whether they are based on morphology (Moreau et al., 2017) or genetic markers (Hoffman et al., 2011), most studies agree that biogeographic structures are more pronounced in brooders than in broadcasters due to contrasting dispersal capacities. In contrast to the marked genetic structures described above, exceptions have been in certain brooders due to the effect of unusual dispersal vectors reported (Cumming, Nikula, Spencer, & Waters, 2014;Díaz et al., 2011). This is also the case in this study, as the brooding species D. meridionalis shares genetic units between South

Most
Georgia and the Kerguelen Islands, which suggests an ongoing connectivity between the two distant regions. This pattern could be explained by kelp rafting via the Antarctic Circumpolar Current (ACC), which has previously been found in other SO brooding echinoderms (O'hara, 1998) and taxa such as in isopods and mollusks (González-Wevar et al., 2018;Leese, Agrawal, & Held, 2010). Kelp rafting was suggested as a viable dispersal vector for shallow-water species living in macro-algal beds (O'hara, 1998), which is the case of D. meridionalis.
Genetic structures in broadcasters under study are quite different from the patterns observed in brooders. The studied broadcasters show circumpolar structures that suggest the prevalence of gene flow across the SO promoted by higher larval dispersal capacities. Geographic patterns in P. charcoti, with several circumpolar entities, suggest the existence of past refugia during past glacial maxima and subsequent dispersal centers (Hemery et al., 2012). Unfortunately, little is known about asteroid larval biology in the SO, apart from a handful of well-studied species (Agüera, Collard, Jossart, Moreau, & Danis, 2015;Pearse, McClintock, & Bosch, 1991;Peck, Souster, & Clark, 2013;Souster, Morley, & Peck, 2018). Major differences exist between the genetic structure of Psilaster and Bathybiaster: (a) Clades of P. charcoti are found to have a greatly fragmented pattern, potentially as a result of lower dispersal capacity than representatives of B. loripes (found at continental shelf depths), but this could also reflect a longer evolutionary history as these clades are older (Figure 2), (b) some specimens of P. charcoti occur on both sides of the PF, and (c) Bathybiaster sp2 shows a potential bipolar distribution (sensu Darling et al., 2000) and likely corresponds to the nominal species B. vexillifer.
The occurrence of shared haplotypes between the Northern and Southern Hemispheres in Bathybiaster sp2 suggests the existence of recent gene flow between the two hemispheres. Morphological similarities between North American and South African specimens had already been indicated in previous work on the echinoderm fauna of South Africa (Clark, 1923). This could be the result of a deep-sea dispersal route, a scenario coined the "thermohaline expressway" by Strugnell, Rogers, Prodöhl, Collins, and Allcock (2008), but a wide, cosmopolitan distribution of the species is probably the most reasonable hypothesis. Unfortunately, no tropical deep-sea samples were available for study. Bathybiaster sp2 is recorded on the Kerguelen Plateau around the Kerguelen Islands at shallow depths, giving credit to a possible sub-Antarctic emergence scenario and colonization of the Kerguelen Plateau from the deep sea (Aronson et al., 2007;Díaz et al., 2011).
Interpretation of the present results is limited by the use of a single mtDNA locus. The observed patterns could be due to mechanisms such as adaptive introgression, demographic disparities, or sex-biased asymmetries arising from the sole use of mtDNA (Toews & Brelsford, 2012). Furthermore, if data derived from COI analyses reflect a long-term effect of contrasting dispersal capacities, other fast-evolving nuclear markers such as microsatellites or SNPs (RAD-Seq data) will be needed in the future to investigate these recent and ongoing processes. Nevertheless, the causal relationship between genetic patterns and dispersal capacities of asteroids can be linked to past climatic and geological events and give some clues to the upstream drivers of species evolution.

| The role of past climate change
The SO has been partially isolated for nearly 40 million years since the first opening of the Drake Passage, which led to the onset of the ACC and subsequent cooling of the ocean when the ACC intensified.
However, isolation of the SO was reduced several times during periods of climate warming and the decreasing ACC intensity (Dalziel et al., Antarctic forcipulataceans (Mah & Foltz, 2011b). The origin of the Antarctic Asteriidae has been dated to 21 Ma in this study, suggesting a diversification of the family starting in the Miocene after the initial opening of the Drake passage (c. 34 Ma) and before the full establishment of the ACC (c. 14 Ma) (Lawver & Gahagan, 2003 as hypothesized by the "Antarctic diversity pump" hypothesis (Clarke & Crame, 1989, 1992 and even promoting evolutionary radiations of the benthic fauna (Allcock, 2005;Raupach, Malyutina, Brandt, & Wägele, 2007;Wilson, Schrödl, & Halanych, 2009). These major environmental changes could have also resulted in cryptic speciation in brooding species and broadcasting species with limited larval dispersal capabilities, by the maintenance of genetic differentiation at local scales (Thatje, 2012). Finally, the ACC hypothesis predicts the "existence of many species in clades of varied divergence times, at a wide range of depths, but with highest diversity downstream of the Drake Passage, in the Scotia Arc and Weddell Sea" (sic. Pearse et al., 2009).
This hypothesis could explain some of the diversification patterns we have observed in asteroids. Nevertheless, the apparent correlation with past geological and climatic events should be taken with great care due to the lack of fossil records and uncertainties in calibration of the molecular clock also known to be taxon-and climate regimedependent (Gillooly, Allen, West, & Brown, 2005).

| Conclusion and future prospects
The observed genetic patterns of SO sea stars highlight the relevance of considering life history traits to understand spatial patterns of genetic diversity. Our results suggest that reproductive strategy could have conditioned the spatial structuring of genetic diversity, but with no apparent effect on the level of genetic diversity. Molecular results also indicate that an in-depth taxonomic revision of the group is needed based on an integrative taxonomy approach (combining genetic and morphological data). The increasing development of genetic surveys in biodiversity conservation and management plans (Goodall-Copestake et al., 2012) stresses the need for robust estimates of species diversity. Species complexes are frequent in the SO (Hemery et al., 2012), and estimates of species diversity based on morphology can lead to under-or alternatively, overestimating "true" biological diversity. In that context, multilocus approaches are essential to detect evolutionary processes within species. Finally, the apparent bipolar or cosmopolitan distribution of the species B. vexillifer should be further investigated using additional specimens and multiple molecular markers. This will bring new insights on a potential deep-sea route connecting the two polar oceans and the relevance of biogeographic scenarios such as the thermohaline expressway (Strugnell et al., 2008) and the sub-Antarctic emergence (Díaz et al., 2011). The presence of B. loripes on the continental shelf could be the result of a past colonization from the deep sea (Strugnell et al., 2011), which potentially challenges the established paradigm of an isolated SO benthic diversity (Clarke, Aronson, Crame, Gili, & Blake, 2004).

ACK N OWLED G M ENTS
The work was supported by a "Fonds pour la formation à la Recherche