The Antarctic Circumpolar Current isolates and connects: Structured circumpolarity in the sea star Glabraster antarctica

Abstract Aim The Antarctic Circumpolar Current (ACC) connects benthic populations by transporting larvae around the continent, but also isolates faunas north and south of the Antarctic Convergence. We test circumpolar panmixia and dispersal across the Antarctic Convergence barrier in the benthic sea star Glabraster antarctica. Location The Southern Ocean and south Atlantic Ocean, with comprehensive sampling including the Magellanic region, Scotia Arc, Antarctic Peninsula, Ross Sea, and East Antarctica. Methods The cytochrome c oxidase subunit I (COI) gene (n = 285) and the internal transcribed spacer region 2 (ITS2; n = 33) were sequenced. We calculated haplotype networks for each genetic marker and estimated population connectivity and the geographic distribution of genetic structure using ΦST for COI data. Results Glabraster antarctica is a single circum‐Antarctic species with instances of gene flow between distant locations. Despite the homogenizing potential of the ACC, population structure is high (ΦST = 0.5236), and some subpopulations are genetically isolated. Genetic breaks in the Magellanic region do not align with the Antarctic Convergence, in contrast with prior studies. Connectivity patterns in East Antarctic sites are not uniform, with some regional isolation and some surprising affinities to the distant Magellanic and Scotia Arc regions. Main conclusions Despite gene flow over extraordinary distances, there is strong phylogeographic structuring and genetic barriers evident between geographically proximate regions (e.g., Shag Rocks and South Georgia). Circumpolar panmixia is rejected, although some subpopulations show a circumpolar distribution. Stepping‐stone dispersal occurs within the Scotia Arc but does not appear to facilitate connectivity across the Antarctic Convergence. The patterns of genetic connectivity in Antarctica are complex and should be considered in protected area planning for Antarctica.


| INTRODUC TI ON
Population connectivity in benthic marine species depends on extrinsic environmental and oceanographic factors and biological features of behavior and development. Pelagic larval stages allow long-range dispersal in benthic organisms; however, indirect estimates of dispersal ability such as planktonic larval duration (PLD) can be surprisingly unreliable in predicting the geographic distribution of benthic adults (Lester, Ruttenberg, Gaines, & Kinlan, 2007;Paulay & Meyer, 2006;Shanks, 2009). Dispersal ability can be modified behaviorally, for example, by rafting (Helmuth, Veit, & Holberton, 1994;Highsmith, 1985;Nikula, Fraser, Spencer, & Waters, 2010), which can result in differences between PLD-predicted and realized dispersal. Genetic proxies, such as F ST , show consistent but only moderate correlation with PLD (Selkoe & Toonen, 2011;Weersing & Toonen, 2009), however, aggregate studies may underestimate this correlation (Dawson, 2014;Dawson, Hays, Grosberg, & Raimondi, 2014). Larval type and PLD are strongly correlated with local temperatures and productivity (Marshall, Krug, Kupriyanova, Byrne, & Emlet, 2012), and connectivity at high latitudes may be constrained by strong seasonality in reproduction and primary productivity.
Organisms in the Southern Ocean have evolved reduced feeding in planktonic larval stages compared to species at lower latitudes (Marshall et al., 2012;Thorson, 1950), but the effect of this evolutionary tendency on dispersal remains unknown in many taxa.
Currents, water mass isolation, nutrient dynamics, and distance between suitable habitats are important in establishing species ranges. The Southern Ocean is an extreme environment in many of these respects, and drivers of dispersal and connectivity in the Antarctic fauna remain poorly understood. The Antarctic Circumpolar Current (ACC) is an unusually strong barrier to the north-south exchange of organisms owing to its thermal and density-driven isolation of water masses, as well as strong eastward flow.
Estimates of the timing of ACC formation range from 20 to 41 Ma, corresponding with the opening of Drake Passage and the Tasman Seaway (Barker & Thomas, 2004;Ladant, Donnadieu, & Dumas, 2014;Lagabrielle, Goddéris, Donnadieu, Malavielle, & Suarez, 2009;Scher & Martin, 2006;Sijp et al., 2014). Intensification of the ACC at the Miocene-Pliocene boundary is correlated with the timing of genetic separation of several benthic marine species pairs from the Magellanic region and the Antarctic continent, suggesting that water mass isolation generated by the ACC has had an important impact on Antarctic marine biotic isolation (Poulin, González-Wevar, Díaz, Gérard, & Hüne, 2014).
Most of the Antarctic continental shelf is narrow and isostatically depressed, with a mean depth of about 200 m (http://www.gebco. net/), and is subject to high disturbance via seasonal ice scour and iceberg groundings (Gutt, 2001). During the last glacial maximum (LGM), grounded ice shelves are estimated to have covered much of the continental shelf, significantly decreasing available habitat for the shelf fauna (Anderson, Shipp, Lowe, Wellner, & Mosola, 2002;Huybrechts, 2002). However, some refugial habitat on the shelf and in deeper water must have persisted through glacial periods Anderson et al., 2002;Thatje, Hillenbrand, & Larter, 2005). Differential use of refugia would result in populations varying in signals of expansion. Glacial survival in deep sea refugia is supported by eurybathy in several groups of Antarctic invertebrates (Brey et al., 1996) and genetic signatures of long-term population stability in many benthic species .

Despite isolation between South America and the Antarctic
Peninsula at the Antarctic Convergence, there remains a great deal of faunal overlap between these regions. The Scotia Arc is a chain of islands, seamounts and ridges spanning the Antarctic Convergence, providing areas of shallow shelf habitat in the Scotia Sea. These habitats may act as "stepping-stones", allowing dispersal across isolated water masses. Most recent studies on genetic connectivity in Antarctic marine invertebrates have generally focused sampling in the Scotia Arc region (e.g., Hoffman, Peck, Linse, & Clarke, 2011;Hunter & Halanych, 2010;Janosik, Mahon, & Halanych, 2011); however, there are few explicit tests of this stepping-stone hypothesis (but see Wilson et al., 2007), and only recent studies include circumpolar sampling for comparison (Galaska, Sands, Santos, Mahon, across the Antarctic Convergence, abundance, and high dispersal potential make it an excellent model system to investigate genetic connectivity in the Antarctic and subantarctic region on several geographic scales.
Here, we use evidence from two genetic markers to assess three hypotheses: 1. Glabraster antarctica is a single panmictic circum-Antarctic species.

This connectivity is facilitated by the Antarctic Circumpolar
Current.

3.
Populations of G. antarctica maintain genetic connectivity across the Antarctic Convergence via "stepping-stone" dispersal along the Scotia Arc.

| Data analyses
To provide a hypothetical framework for population structure and to facilitate discussion of results, sampling sites were assigned to a pri-  Figure 2).
Statistical parsimony networks were calculated in TCS (Clement, Posada, & Crandall, 2000) for each genetic marker with a 95% connection limit and gaps treated as missing data. Likelihood model calculations were performed for COI data in jModelTest (Posada, 2008) with five substitution schemes. The best fitting model was chosen using the Akaike Information Criterion implemented in jModelTest (Guindon & Gascuel, 2003). Both uncorrected and model-corrected genetic distances between COI haplotypes were calculated in PAUP* (Swofford, 2002).
In order to test for genetic structure in the COI dataset (Hypothesis 1), an Analysis of Molecular Variance (AMOVA) was performed in Arlequin 3.5 (Excoffier & Lischer, 2010)  Antarctic Peninsula Antarctic Peninsula Antarctic Peninsula 10,000 iterations for the simulated annealing process and repeated to evaluate consistency. To identify the appropriate grouping scheme (K-value) for these data, the pseudo-F criterion (Caliński & Harabasz, 1974), and the relative differences in Φ CT were calculated for each K. Enderby eighteen of these haplotypes were private (found in one sampling site), and 106 were singletons (Figure 3). Uncorrected COI genetic distances ranged from 0.15% to 3.36%, and the GTR + I + G AIC best-fit model-corrected distances ranged from 0.15% to 3.73%. ITS2 sequencing (n = 33) recovered 17 haplotypes, of which 13 were private and 11 singletons ( Figure 4). Population statistics are given for each sampling site in Table 1.

| RE SULTS
Although morphological variation in size and the presence of abactinal spines exists among the sampled areas (Figure 1), these variants do not correspond to distinct genetic entities and G. antarctica appears to constitute a single morphologically variable species.
COI data formed a single, diffuse haplotype network, with some regional clustering of haplotypes (Figure 3). There were some un-   SAMOVA results show strong population structure (Φ CT ) for each K from 2 to 7 (p < 0.005 in all cases), and the population groupings for each K are given in Table 3. The pseudo-F criterion value is maximized at K = 5, however, the rate of change between Φ CT values is maximized at K = 4 (Figure 6). At both K = 4 and K = 5, geographically  Enderby (East Antarctica). These results support the hypothesis of circumpolar connectivity despite overall strong population structure.
The inclusion of Shag Rocks and two other sites south of the Antarctic

Convergence within the Magellanic group indicates that the Antarctic
Convergence is an incomplete barrier to gene flow in this species.

| A single circum-Antarctic species
Morphological variants of G. antarctica do not reflect distinct genetic entities that would support the presence of cryptic species.
The Antarctic Peninsula morphotype (Figure 1a) is small, with strong abactinal spination, and shares haplotypes with the large Scotia Arc samples lacking abactinal spines (Figures 2-4), the latter corre- Our results support the synonymy of these subspecies (Mah & Foltz, 2014), and the view that G. antarctica represents a single, morphologically variable, circum-Antarctic species.
Glabraster antarctica is characterized by high genetic diversity, many private haplotypes, and substantial population structure; yet, there are low genetic distances among haplotypes, and subpopulations are distributed across broad spatial scales. Although we reject the hypothesis of panmixia in this species (Hypothesis 1), these results demonstrate enough intermittent genetic connectivity to maintain a structured circumpolar species.

| Structure in circumpolarity
The larvae of G. antarctica are highly buoyant, with a PLD of 60 days (Bosch, 1989), allowing dispersal in the fastest upper portion of the Antarctic Circumpolar Current (Ivchenko & Richards, 1996) and potentially driving the long-distance connectivity patterns seen in part of the haplotype network. The dispersal potential over a single generation in G. antarctica is unknown, but larval behavior is important to realized dispersal ability in marine species (reviewed by Levin, 2006), and complexities in the population structure of this species indicate multiple factors at play. SAMOVA results indicate genetic connectivity among extremely distant sampling sites, and several geographically proximate sites were grouped separately in the SAMOVA analysis with significant pairwise Φ ST in the AMOVA analysis. In East Antarctica, Shelf Break is genetically isolated from another continental site Enderby ( Figure 5)  Antarctic Antarctica grouped with the Aurora Bank Heard Island site in the SAMOVA analysis (Table 3). This pattern may be driven by entrainment of Shelf Break larvae in the Prydz Bay gyre (Heywood, Sparrow, Brown, & Dickson, 1999;Nicol, Pauly, Bindoff, & Strutton, 2000). Our Enderby site is well outside the western limit of this gyre (45°E, Figure 2) and is grouped with Scotia Arc sites in the SAMOVA analysis, while the Shelf Break site is located within the region influenced by the Prydz Bay gyre. A Prydz Bay-Scotia Arc connection was noted in the octopus Pareledone turqueti , and a similar pattern of long-distance connectivity between East Antarctica and the Antarctic Peninsula was found in the amphipod Eusirus giganteus, contrasting with strong genetic structure within East Antarctica in Eusirus perdentatus (Baird, Miller, & Stark, 2011 Kelp-rafting peracarid crustaceans have shown a similar pattern of circumpolar dispersal via ACC transport (Nikula et al., 2010).
Corroboration of COI results with further nuclear data is necessary to verify patterns of connectivity in G. antarctica.

| The "stepping-stones" of the Scotia Arc
Stepwise dispersal across the shelf habitats of the  Figure 5); Magellanic haplotypes are shared with Shag Rocks but no other Scotia Arc samples (Figures 3 and 4).
The unique current processes at play in the vicinity of South Georgia may explain the genetic isolation of Shag Rocks from South Georgia in G. antarctica. Currents immediately between Shag Rocks and South Georgia (~38°W) of Antarctic Intermediate Water show a relatively strong southward flow of 20 cm/s (Arhan, Naveira Garabato, Heywood, & Stevens, 2002). Furthermore, the ACC deflects northward on the east side of South Georgia before returning to eastward flow (Orsi, Whitworth, & Nowlin, 1995;Thorpe, Heywood, Brandon, & Stevens, 2002), and the combination of these factors may entrain short-lived larvae from South Georgia and contribute to its isolation from Shag Rocks despite geographic proximity.
Shag Rocks is situated in an area with relatively little seasonal and annual variation in the location of the Antarctic Convergence (Moore, Abbott, & Richman, 1999), and it is therefore unlikely that  (Leese, Kop, Wägele, & Held, 2008). Shag Rocks is both physically distant from other Magellanic sites and thermally separated by the Antarctic Convergence (Moore et al., 1999;Smith, Stevens, Heywood, & Meredith, 2010), and the Subantarctic Front to the east of Burdwood Bank (Smith et al., 2010), challenging the idea that pelagic dispersal is limited across these frontal zones.
Overall, our study demonstrates that localized current regimes and water mass isolation may drive fine-scale regional genetic isolation, for example, of the Scotia Arc region from the Magellanic region, and affect even those species with planktonic development and broad depth distributions. Several instances of connectivity among geographically distant regions demonstrate the dispersive influence of the ACC on planktonic developers. The striking diversity, genetic structure, and complex pattern of circumpolarity in G. antarctica is driven by the unique oceanographic and historical features of Antarctica and these complexities should be carefully considered in conservation planning.

ACK N OWLED G M ENTS
We thank the crew of the RVIB Nathaniel B. Palmer and science participants on cruises NBP11-05 and NBP13-03, and the Australian Antarctic Division for access to specimens, particularly Felicity McEnnulty, Glenn Johnstone and Jonny Stark (Lamb, Robertson, & Welsford, 2006-2015. Samples were provided by the Ministry