Marine Biology

, Volume 160, Issue 5, pp 1285–1296

Global population divergence of the sea star Hippasteria phrygiana corresponds to the onset of the last glacial period of the Pleistocene


    • Department of Biological SciencesLouisiana State University
  • S. D. Fatland
    • Department of Biological SciencesLouisiana State University
  • M. Eléaume
    • Département Peuplement et Milieux AquatiquesMuséum national d’Histoire naturelle
  • K. Markello
    • Department of Invertebrate Zoology and GeologyCalifornia Academy of Sciences
  • K. L. Howell
    • Marine Biology and Ecology Research Centre, Marine InstituteUniversity of Plymouth
  • K. Neill
    • National Institute of Water and Atmospheric Research
  • C. L. Mah
    • Department of Biological SciencesLouisiana State University
    • Department of Invertebrate ZoologyNational Museum of Natural History
Original Paper

DOI: 10.1007/s00227-013-2180-1

Cite this article as:
Foltz, D.W., Fatland, S.D., Eléaume, M. et al. Mar Biol (2013) 160: 1285. doi:10.1007/s00227-013-2180-1


Genetic structure and connectivity of populations of the globally distributed and eurybathic sea star Hippasteria phrygiana (Parelius 1768) were studied in 165 individuals sampled from three oceanic regions: the North Pacific Ocean, the South Pacific Ocean (considered to include the adjacent regions of the Southern Ocean and the southern Indian Ocean) and the North Atlantic Ocean. A nuclear gene region (ATP synthase subunit α intron #5, ATPSα) and a mitochondrial gene region (cytochrome oxidase subunit I, COI) were amplified and sequenced. Significant heterogeneity was detected in an AMOVA analysis among the three sampled oceanic regions for COI, but not for ATPSα. Neither gene showed significant genetic heterogeneity within the North Atlantic, as assessed by ΦST values. Significant heterogeneity was detected for COI (but not ATPSα) in the North Pacific, but the converse was true in the South Pacific. Coalescent simulations suggested that the three regions have been diverging with little or no gene flow for the past 50–75,000 years, a time frame that corresponds to the onset of the last glacial period of the Pleistocene. A possible genetic signature of recent population expansion (or non-neutrality) was detected for each gene in the North Pacific, but not in the other two oceanic regions.


Geological and climatic events in the Quaternary period (and the late Pliocene) have had major impacts on geographic ranges and levels of gene flow among populations of broadly distributed marine species. The opening of the Bering Strait and the closure of the Panamanian Seaway not only affected biotic interchange between the Atlantic and Pacific Oceans (e.g., Hardy et al. 2011 and references therein), but also affected currents, temperatures and nutrient levels in deeper waters (e.g., Marincovich and Gladenkov 1999; Lessios 2008). More recently, repeated glacial/interglacial cycles have likely promoted range shifts (contraction, expansion and/or fragmentation) and local extirpations for some marine species, particularly those in intertidal or shallow subtidal habitats (e.g., Wares and Cunningham 2001; Wares 2002; Addison and Hart 2005; Ilves et al. 2010; reviews by Jansson and Dynesius 2002; Maggs et al. 2008; Frasier et al. 2009; Norris and Hull 2011). These geological and climatic events interact with organismal life history and ecology (reviews by Bradbury et al. 2008; Dawson and Hamner 2008; Cowen and Sponaugle 2009; Weersing and Toonen 2009) to produce differing levels of connectivity and geographic range in various marine species. The effects of Quaternary/late Pliocene events on genetic structure and dispersal have been fairly well studied in trans-Arctic populations of various shallow-water marine invertebrate (Palumbi and Kessing 1991; Collins et al. 1996; Reid et al. 1996; Luttikhuizen et al. 2003; Väinölä 2003; Addison and Hart 2005; Harper et al. 2007; Nikula et al. 2007; Rawson and Harper 2009; review by Hardy et al. 2011) and vertebrate (Orti et al. 1994; Taylor and Dodson 1994; Dodson et al. 2007; review by Briggs 2003) species or species complexes. Much less attention has been paid to genetic structure and dispersal in deep-sea or eurybathic marine invertebrate species, with the exception of hydrothermal vent species (e.g., Vrijenhoek 2010) and zooplankton species (e.g., Goetze 2005; Durbin et al. 2008; Goetze 2011), whose ecological specializations provides few clues to the structure to be expected in more typical benthic marine invertebrate species.

The genus Hippasteria Gray 1840 is a member of the Hippasterinae, a group of cold-water Goniasteridae which includes 16 nominal species that are widely distributed in the Atlantic, Pacific and Indian Oceans (Mah et al. 2010). Although many features of its life history are poorly known, Hippasteria has been found to be ecologically important (e.g., Birkeland 1974; Kreiger and Wing 2002). Here, we present nuclear and mitochondrial sequence data for specimens of the broadly distributed and eurybathic species Hippasteria phrygiana sampled from three oceanic regions: the North Pacific Ocean, the South Pacific Ocean (considered to include the adjacent regions of the Southern Ocean and the southern Indian Ocean) and the North Atlantic Ocean. Our objective was to compare the observed phylogeographic pattern to that reported in the previous studies of shallow-water marine invertebrate species cited above, particularly regarding connectivity of populations among these three regions. Although H. phrygiana has been collected at depths greater than 1,000 m (Mah et al. 2010), this species (as H. spinosa) has also been reported from shallow subtidal depths (10 m, e.g., Lambert 2000; present study), allowing the possibility that N. Atlantic and N. Pacific populations are connected by dispersal across the relatively shallow waters of the Bering Strait. In contrast to previous studies of shallow-water species and despite a sample size of approximately 50 specimens or sequences from each of these three regions, there was no sharing of mitochondrial sequence haplotypes among the three oceanic regions for H. phrygiana and estimated population divergence times among the regions ranged from 53–78 kya (1,000 years ago). The global population structure reported here for H. phrygiana is an alternative to the presence of cryptic species within a widely distributed genus (e.g., Vrijenhoek 2009) and has implications for understanding variation in ecology and feeding behavior across the species’ range.


Specimens were obtained from museum collections and from newly collected samples (see Supplemental Table S1 and Acknowledgements for details). DNA extraction, PCR and sequencing were carried out as in previous studies (Foltz et al. 2007; Foltz and Mah 2009). A 732-bp fragment of the cytochrome oxidase subunit I (COI) gene was amplified with the primers of Hoareau and Boissin (2010). A 278-bp fragment of intron #5 of the ATP synthase alpha subunit gene (ATPSα) was amplified with novel Hippasteria-specific primers:



These novel primers were based on initial Hippasteria sequences obtained using the Patiria miniata-specific primers (PMATPSF9 and PMATPSR669) of Keever et al. (2009). The forward primer started 6 bases downstream from the last base of primer PMATPSF9 and included all or part of codons 200–206 in the Strongylocentrotus purpuratus ATPSα gene (Y07895). The first base of the reverse primer overlapped the last base of primer PMATPSR669 and included the last 15 nucleotides of intron #5 and all or part of codons 215–218 in the S. purpuratus ATPSα gene. The primer pairs for each gene occasionally yielded spurious amplification products in the 100–200 bp range, in addition to a product of the expected length, which necessitated a clean-up step using PureLink PCR purification with high-cutoff binding buffer (Invitrogen). PCR products without low molecular weight products were usually sequenced after dilution with purified water.

Contigs were assembled in Sequencher V. 5 (Gene Codes Corp.), and sequences were aligned using Muscle V. 3.5 (Edgar 2004) within SeaView V. 4 (Gouy et al. 2010). The COI alignment was trimmed to 655 bp and the ATPSα alignment was trimmed to 181 bp (the included positions for ATPSα were all intronic). Intra-individual nucleotide site heterozygosity and length variation heterozygosity for ATPSα sequences were resolved as in Foltz (2007). Most sequence statistics, resolution of haplotypic phase for ATPSα sequences (via the PHASE V. 2.1 algorithm of Stephens and Donnelly 2003 with default parameters) and detection of putative ATPSα recombinant sequences were carried out in DnaSP V. 5.10.01 (Librado and Rozas 2009). AMOVA and pairwise ΦST analyses were carried out in Arlequin v. 3.11 (Excoffier et al. 2005) using pairwise differences. Network construction was done using the program TCS V. 1.21 (Clement et al. 2000), with a 95 % connection limit and outgroup probabilities calculated by the method of Castelloe and Templeton (1994), which uses coalescent theory (rather than simulations) and the observed genetic data to determine the probability that each haplotype is the ancestral haplotype. Putative recombinants were identified using the four-gamete test of Hudson and Kaplan (1985) in DnaSP V. 5.10.01 and were excluded from IMa and IMa2 analyses and from network construction in TCS, as these methods assume that sequences are non-recombining. Geographic coordinates of sampling localities were plotted using the map generator available at

Preliminary analyses in IMa and IMa2 using parameter-rich demographic models indicated that migration rates among contemporary and ancestral populations were low and not significantly different from zero (details not shown) and that the posterior probability distributions for population divergence times were not well-defined. Therefore, we estimated divergence times among the N. Pacific, S. Pacific and N. Atlantic regions using reduced pairwise models in IMa, for which migration rates were set to zero. The mutation rate for the IMa and IMa2 analyses was based on published sequence data for the COI gene averaged over presumed trans-Panamanian geminate species in the genus Astropecten (Zulliger and Lessios 2010): A. nitidus versus A. oerstedii, A. siderealis, A. erinaceus and A. verrilli. The resulting mutation rate (2.4 % per site per million years or 0.000015 mutations per locus per year) was very similar to one reported earlier for the COI gene in the genus Asterias based on presumed trans-Arctic dispersal (2.3 % per site per million years, recalculated from data in Wares 2001), but is about 33 % smaller than the rate for the presumed trans-Panamanian geminate sea star species Oreaster reticulatus and O. occidentalis (3.6 % per site per million years, recalculated from data in Hart et al. 1997). Although Oreaster is more closely related to Hippasteria than is Astropecten, the Oreaster-calibrated substitution rate is larger than other COI substitution rates previously calibrated from presumed trans-Panamanian geminate echinoid or asteroid species (Hart et al. 1997, Lessios 2008), and we chose to use the smaller rate rather than the outlier rate. Using a higher mutation rate would cause a proportionate reduction in the divergence times estimated here. Hart et al. (1997); Lessios (2008) have discussed the assumptions needed when calibrating a mutation rate with data for trans-Panamanian geminate species that are not closely related to the ingroup: (1) mutation rates are similar between Astropecten and Hippasteria, (2) mutation rates are not heterogeneous among H. phrygiana haplotypes and (3) the geminate species diverged at a time that coincided with the closure of the Isthmus of Panama. With no published ATPSα sequence data for geminate sea star species available, we had a mutation rate calibration only for the COI gene. However, both genes were includes in the IMa and IMa2 analyses.

Roughly equal numbers of H. phrygiana from the N. Pacific Ocean (N = 57), S. Pacific/Southern/Indian Oceans (N = 49) and N. Atlantic Ocean (N = 55) were included in the analysis. In later sections, these three areas will be referred to as ‘major oceanic regions’ and unless otherwise noted below, the term ‘S. Pacific Ocean’ will refer to an area that includes adjacent regions of the Southern Ocean and the Indian Ocean. In the figures, these three areas were color-coded as follows: various shades of green were N. Atlantic locations (locations 1–3 in the tables), red shades were S. Pacific locations (locations 4–5) and blue shades were N. Pacific locations (locations 6–7). The global sample size was 165 (see Fig. 1 and Supplemental Table S1 for details) and included one specimen from the Solomon Islands (Central Pacific), one specimen from Japan, one specimen from Bouvet Island (South Atlantic), one specimen from the southern tip of Chile and additional small samples in the N. Pacific Ocean (two specimens from the Gulf of Alaska, one from President Jackson Seamount C, one from British Columbia, two whose only known locality from museum records was Alaska and one whose only known locality was California). These small samples were given their own colors in the figures and were excluded from most population genetics analyses (i.e., Tables 1, 2, 3, 4, 5), but were included in the IMa, IMa2, mismatch and Fu’s F analyses. This second set of analyses was intended to investigate possible genetic signatures of ancient events like population splitting or population expansion, so a region-wide perspective was appropriate.
Fig. 1

Map showing approximate sampling locations for 162 specimens of Hippasteria phrygiana. Two specimens from Alaska and one specimen from California lack exact geographic coordinates and are not shown here. Black arrows illustrate two colonization scenarios mentioned in the "Discussion". Numbers next to some circles are sample sizes; circles without adjacent numbers have sample sizes of one. The key in the lower left explains the color scheme. Colors for the seven sampling locations included in Tables 1, 2, 3, 4, 5 are numbered 1–7 for later reference; the remaining five unnumbered sampling locations each had too few specimens for inclusion in the analyses reported in Tables 1, 2, 3, 4, 5

Table 1

Summary statistics for 140 Hippasteria phrygiana COI sequences from seven locations

Sampling location

Average sampling depth (m)


No. of haplotypesa

Haplotype diversity

Nucleotide diversity (π)

1 N.W. Europe

171.1 ± 12.6


6 (0.60)

0.62 ± 0.08

0.18 ± 0.04

2 Newfoundland

728.3 ± 531.1


4 (0.33)

0.87 ± 0.13

0.37 ± 0.13

3 N.E. United States

613.8 ± 538.1


4 (0.43)

0.81 ± 0.13

0.39 ± 0.09

4 New Zealand

680.9 ± 214.6


4 (0.67)

0.58 ± 0.18

0.10 ± 0.04

5 Kerguelen Islands

266.5 ± 137.8


8 (0.41)

0.75 ± 0.05

0.28 ± 0.05

6 Bering Sea

397.0 ± 78.5


4 (0.33)

0.87 ± 0.13

0.19 ± 0.05

7 Aleutian Islands

195.9 ± 113.7


9 (0.86)

0.34 ± 0.09

0.06 ± 0.02

Diversity measures and average depth are shown ±1 SD; nucleotide diversity is expressed as a percentage. N number of individuals sampled. See Supplemental Table S1 for geographic coordinates

aFrequency of most common haplotype is shown in parentheses

Approximately 2/3 of the material was freshly collected material, non-museum research specimens or non-cataloged museum material from the Muséum national d’Histoire naturelle (MNHN, Paris, France); the rest was from the collections of the California Academy of Sciences (San Francisco, California), the US National Museum (Washington, District of Columbia) and the National Institute of Water and Atmospheric Research Ltd. (Wellington, New Zealand). Latitude and longitude coordinates, depth, collecting date, museum catalog numbers (where available) and GenBank accession numbers for each specimen included in the analysis are in Supplemental Table S1.

There are 16 described extant species and many described subspecies of Hippasteria, some of which have been synonymized with H. spinosa or H. phrygiana on morphological grounds (Mah et al. 2010). Of these species, H. phrygiana has the broadest geographic distribution by far, with specimens recorded from the N. Atlantic Ocean, S. Pacific Ocean (principally New Zealand), Indian Ocean (principally the Kerguelen Islands) and Southern Ocean, with only isolated records from the S. Atlantic Ocean. There are described species of Hippasteria in the S. Atlantic Ocean (e.g., H. argentinensis Bernasconi 1961) which have been considered by some researchers to be subspecies of H. phrygiana due to close morphological overlap. However, because we have only one specimen from the S. Atlantic Ocean, the question of whether H. phrygiana is widely distributed in that region cannot be addressed in the present paper. We treated specimens previously identified (e.g., in museum catalogs or by Mah) as H. aleutica, H. armata, H. hyadesi, H. imperialis, H. kurilensis and H. spinosa as synonyms of H. phrygiana, if they shared a COI haplotype and/or an ATPSα haplotype with known specimens of H. phrygiana. An updated taxonomic treatment of Hippasteria based on molecular and morphological characters will be published elsewhere (Mah and Foltz, in preparation).


We obtained COI and/or ATPSα sequence data from 165 specimens, but not every specimen yielded useable data for both gene regions (Supplemental Table S1). There were 16 polymorphic sites in 140 diploid ATPSα sequences, and each of the 16 sites had only two nucleotide types. One individual was heterozygous at three of these sites, 14 were heterozygous at two sites, 41 were heterozygous at one site and the remaining individuals were heterozygous at none of the sites. Thus, about 90 % of the individuals with diploid ATPSα sequence data were heterozygous at fewer than two nucleotide sites and thus were unambiguous as to haplotype phase. After diploid ATPSα sequences were resolved into haplotypes and putative recombinants (14 recombinant alleles from 13 specimens) were removed from the data set, we analyzed 152 COI sequences and 266 ATPSα sequences. The TCS network for the COI gene (Fig. 2) revealed essentially no haplotype sharing among major oceanic regions, with the only exception being the central haplotype in the network, which was found in specimens from the S. Pacific Ocean plus single specimens from the Tierra del Fuego archipelago (Southern Chile), Bouvet Island (S. Atlantic) and President Jackson Seamount C (off the coast of Oregon). In the corresponding network for the ATPSα gene region (Fig. 3), the two most abundant haplotypes were shared among all three regions, and the third most abundant haplotype was shared between the S. Pacific and the N. Atlantic regions. The COI network (Fig. 2) indicated that N. Pacific populations were more closely related to S. Pacific populations than they were to N. Atlantic populations. Because each COI haplotype was essentially restricted to a single major oceanic region, the outgroup probabilities for the various haplotypes in the COI network could be used as probabilities of geographic origin. The probability of a N. Pacific Ocean origin was 43 %, followed by 29 % for a S. Pacific Ocean origin and 27 % for a N. Atlantic Ocean origin. Additionally, outgroup rooting in TCS V. 1.21 using the related species Hippasteria californica suggested that a N. Pacific Ocean and a S. Pacific Ocean origin were equally parsimonious, with a N. Atlantic Ocean origin being less parsimonious (Fig. 2). The outgroup sequence was nine mutational steps removed from the ingroup sequences, and other congeneric species like H. heathi were even more distant (details not shown). Other evidence in favor of a S. Pacific Ocean origin is the fact that the COI haplotype that was most abundant in the S. Pacific Ocean was also the most widely distributed (i.e., the central haplotype in Fig. 2). Finally, several lines of non-genetic evidence were consistent with a Pacific ancestry for Hippasteria phrygiana. The greatest diversity of Hippasteria—including H. californica, H. heathi and a new undescribed species—occurs exclusively in the Pacific (Mah and Foltz, unpubl. data). In contrast, only one species—H. phrygiana—truly occurs in the Atlantic Ocean. Of the nominal Atlantic species listed by Clark and Downey (1992: 246), H. caribaea was reassigned to Gilbertaster by Mah et al. (2010) and further records of Hippasteria falklandica (see Mah et al. 2010) imply that it occupies a subAntarctic distribution.
Fig. 2

Haplotype network for 33 COI haplotypes constructed with the TCS program. Small open circles are hypothetical intermediate haplotypes, and each line segment represents one inferred mutational step. Outgroup rooting in TCS using a COI sequence for Hippasteria californica (CASIZ 120067, GenBank JQ896456) produced two equally parsimonious roots, denoted by asterisk Sectors of each circle are color-coded to match the map in Fig. 1. Notes: a sector includes 1 haplotype with unknown geographic coordinates from Alaska; b circle represents 1 haplotype with unknown coordinates from Alaska
Fig. 3

Haplotype network for 16 ATPSα haplotypes. Details as in Fig. 2. Note: a sector includes 1 haplotype with unknown coordinates from California

The most probable value of the estimated divergence time of the N. Pacific and S. Pacific samples was 53 kya with 95 % highest posterior density (HPD) interval of 31–93 kya. For N. Atlantic versus S. Pacific samples, the estimated divergence time was 78 kya (95 %HPD interval was 42–133 kya). For N. Pacific versus N. Atlantic samples, the estimated divergence time was 55 kya (95 %HPD interval was 29–144 kya). Fu’s Fs statistic was significantly negative (after adjustment for multiple tests) in the N. Pacific for both the COI gene (Fs = −10.5) and the ATPSα gene (Fs = −5.12), but none of the tests was significant in the other two oceanic regions. Nucleotide diversity and haplotype diversity for the COI gene were low in the Aleutian Islands sample (Table 1), compared to the other sampling locations with five or more sequences, but the same pattern did not hold for the ATPSα gene (Table 2). There was no consistent connection between nucleotide or haplotype diversity and average depth of collection across the seven sampling locations. For example, the average depth of collection for the New Zealand collection was over twice as large as for the Kerguelen Islands collection, and while there was greater haplotype and nucleotide diversity in the New Zealand collection for the ATPSα gene as compared to the Kerguelen Islands collection, the opposite pattern was observed for the COI gene. Haplotype and (to a lesser extent) nucleotide diversity were largely determined by whether the most common sequence occurred in high frequency (>50 %) or low frequency (see Tables 1, 2 for frequency information). The AMOVA results from Arlequin were similar for the two gene regions, in that there was significant (P < 0.05 after adjustment for multiple testing) heterogeneity among individuals within sampling locations and among sampling locations within major oceanic regions. However, there was also significant heterogeneity among major oceanic regions for the COI gene (Table 3), but not for the ATPSα gene (Table 4). Also, the percentage of the total genetic variation attributable to the within sampling locations level was much greater for the ATPSα gene (70 %) than for the COI gene (34 %). The hierarchical model for the AMOVA analyses was based on the network (Fig. 2) showing very limited COI haplotype sharing among regions. However, the regional heterogeneity for COI remained significant even after adjusting the significance level for an a posteriori test. Further insight into population structure at the level of sampling locations was obtained from the pairwise ΦST results from Arlequin (Table 5). For both genes, there was no heterogeneity among sampling locations within the N. Atlantic Ocean. In contrast, there was significant heterogeneity among sampling locations within the N. Pacific Ocean for the COI gene but not for the ATPSα gene, even though sample sizes and geographic distances were roughly comparable between the two oceanic regions. The heterogeneity for the COI gene in the N. Pacific Ocean primarily reflected the existence of three COI haplotypes found only in the Bering Sea (Figs. 1, 2). For most pairwise comparisons, ΦST values were larger for the COI gene than for the ATPSα gene. The most striking exception to this generalization was the Kerguelen Islands–New Zealand comparison, for which ΦST was significantly greater than zero for ATPSα, but not for COI. Frequency-based (FST rather than ΦST) AMOVA and pairwise FST analyses that included 14 putative recombinant sequences for ATPSα produced variance components and FST values that were generally smaller than those shown in Tables 3 and 5, but the associated probability values were mostly the same (details not shown).
Table 2

Summary statistics for 250 Hippasteria phrygiana ATPSα sequences from seven locations

Sampling location


No. of haplotypesa

Haplotype diversity

Nucleotide diversity (π)

1 N.W. Europe


6 (0.48)

0.70 ± 0.04

0.77 ± 0.05

2 Newfoundland


4 (0.40)

0.73 ± 0.10

0.81 ± 0.13

3 N.E. United States


6 (0.33)

0.81 ± 0.06

0.81 ± 0.12

4 New Zealand


4 (0.54)

0.70 ± 0.12

1.06 ± 0.19

5 Kerguelen Islands


4 (0.61)

0.50 ± 0.04

0.32 ± 0.04

6 Bering Sea


3 (0.73)

0.47 ± 0.16

0.46 ± 0.18

7 Aleutian Islands


9 (0.41)

0.69 ± 0.04

0.51 ± 0.06

Details as in Table 1, except that N = number of haploid sequences resolved from diploid genotypes, excluding putative recombinant sequences

aFrequency of most common haplotype is shown in parentheses

Table 3

AMOVA results for COI, where oceanic regions (N. Pacific, S. Pacific and N. Atlantic) are as defined in the text





Fixation index


Among major oceanic regions






Among sampling locations within regions






Among individuals within sampling locations











Variance components (σ2) in bold are significantly (P < 0.05) greater than zero. DF degrees of freedom, SS sum of squares. See Table 1 and Supplemental Table S1 for the definition of sampling locations and sample size per location

Table 4

AMOVA results for ATPSα





Fixation index


Among major oceanic regions






Among sampling locations within regions






Among individuals within sampling locations











Details as in Table 3, except that sample size per location is from Table 2

Table 5

Pairwise ΦST results from Arlequin for seven sampling locations with adequate sample sizes (see Table 1 and Fig. 1 for more information on these locations)


























































Results for the COI gene region are shown below the main diagonal and ATPSα results are above the main diagonal. Values in bold are significantly (P < 0.05) greater than zero, after adjusting for multiple testing. Sampling locations are numbered as in Table 1 and Supplemental Table S1


The only prior study of nucleotide diversity for the ATPSα intron #5 in echinoderms is that of Keever et al. (2009) in Patiria miniata, who found higher haplotypic and approximately tenfold greater nucleotide diversity for intron #5 than reported here. However, the sequences for intron #5 were essentially unalignable between the two species and the lengths were very different (roughly 552 bp in Patiria miniata versus 222 bp in H. phrygiana: Foltz, unpubl. data). The shared coding regions just outside the inferred splice sites showed only synonymous substitutions between the two species (4 inferred substitutions in 12 codons compared), and the 5 nucleotides immediately downstream from the upstream splice site and the 4 nucleotides immediately upstream from the downstream splice site were also completely conserved in interspecific comparison. These results make it unlikely that ATPSα intron #5 sequences analyzed here are the result of paralogous or spurious amplification and instead suggest a high rate of sequence evolution. Apart from the length difference for ATPSα intron #5 between Patiria and Hippasteria, nucleotide site heterozygosity and haplotypic phase were resolved differently between the two studies and may have contributed to the different levels of diversity. The nucleotide diversity observed here for ATPSα intron #5 is comparable to values reported for other introns in sea stars (Foltz et al. 2007). Similarly, the nucleotide diversity observed for the COI gene region in H. phrygiana (global average was 0.40 ± 0.15 %) is within the range of diversity values for COI sequence data for 20 other asteroid species compiled by Ward et al. (2008, mean 0.53 ± 0.13 %, range 0–1.85 %) and for COI sequences collected for 11 other asteroid species by Corstorphine (2010, mean 0.87 ± 0.17 %, range 0.20–1.87 %).

The COI network in Fig. 2 showed no shared haplotypes between the N. Pacific and N. Atlantic regions and two or more mutational steps separating haplotypes from the two regions, which was inconsistent with human transport or other contemporaneous event facilitating colonization of the N. Atlantic (Waters and Roy 2003; Rawson and Harper 2009). Our COI data (e.g., Table 5) suggested greater connectivity of H. phrygiana populations in the Southern Hemisphere (at least at latitudes higher than 40°S, where most of the southern samples came from) than in the Northern Hemisphere. The same cosmopolitan COI sequence haplotype was found in New Zealand, the Southern Ocean south of New Zealand, the Kerguelen Islands in the Indian Ocean, Bouvet Island in the S. Atlantic and off the coast of S. Chile, whereas Northern Hemisphere haplotypes were more geographically restricted. Reconstructed ATPSα haplotypes were more cosmopolitan, with the two most abundant haplotypes being found in all three of the major oceanic regions that we sampled and the next two most abundant haplotypes being found in two regions (Fig. 3). The more cosmopolitan nature of the ATPSα haplotypes could explain the failure to detect significant genetic variation among oceanic regions in the AMOVA analysis reported in Table 4. About 87 % of the total variance for ATPSα occurred at the level of oceanic region or below, leaving less variance apportioned to the between-oceanic region level. In contrast, only 41 % of the total variance for COI occurred at the level of oceanic region or below. Despite the failure of the AMOVA analysis to detect significant among-region variation for ATPSα, 10 of 16 pairwise ΦST values for between-oceanic region comparisons were significantly greater than zero, after adjustment for multiple tests (Table 5), indicating some inter-region heterogeneity for this gene. The difference in degree of geographic restriction of sequence haplotypes between the two genes could reflect greater ancestral polymorphism for the ATPSα gene, because of the larger effective population size expected for nuclear genes compared to mitochondrial genes. Other possible explanations for conflicting geographic patterns between mitochondrial and nuclear genes have been discussed by Toews and Brelsford (2012), including human introduction, hybridization and adaptive introgression and demographic disparities including sex-biased dispersal. About 90 % of their examples of nuclear-mitochondrial discordance involve secondary contact and introgression of mitochondrial genomes between previously described and closely related species or sub-species, often with the inference of adaptive differences between mitochondrial genomes. The greater geographic restriction of COI haplotypes compared to ATPSα haplotypes observed here and the fact that the ATPSα haplotypes were intronic rather than exonic make adaptive introgression an unlikely explanation, and we have already presented mitochondrial evidence against human introduction as a factor influencing geographic variation between oceanic regions in the Northern Hemisphere. Nothing is known about sex-biased dispersal or other demographic asymmetries in Hippasteria.

There are no Arctic records for H. phrygiana in the Canadian Museum of Nature (Ottawa) catalog (J.-M. Gagnon, pers comm to Mah) or in published accounts (Grainger 1966; D’iakonov 1968; Anisimova 1989). The reasons for the apparent inability of H. phrygiana to disperse into the Arctic Ocean are unknown. Adult specimens have been collected in water as shallow as 10 m; for example, the depth of collection in Supplemental Table S1 varies from 10 to 1,400 m. H. phrygiana has pelagic lecithotrophic larvae (Strathmann 1987), but nothing is known about the duration of the larval period at ambient temperatures. Possibly the relatively shallow waters of the Chukchi shelf and/or the lower temperatures of the Arctic Ocean compared to the N. Pacific Ocean prevent or retard dispersal of larvae and adults of H. phrygiana from the Bering Sea to the Arctic Ocean, similar to the pattern observed in a variety of marine organisms (Dunton 1992; see also Darling et al. 2007; Nelson et al. 2009; Hardy et al. 2011). Temperature is important to larval development in asteroids and has been shown to have a significant influence on processes such as developmental rate (Hoegh-Guldberg and Pearse 1995). H. phrygiana’s presence on the continental shelf but absence from deeper-water settings (e.g., Alton 1966; Franz et al. 1981) may indicate low-temperature restriction of either adults and/or larvae. High temperature is also a possible constraint, given H. phrygiana’s close proximity to shelf settings near cold to temperate water currents and scarcity in tropical or warmer water settings.

For numerous marine invertebrate species, genetic comparisons of N. Pacific, Arctic and N. Atlantic populations (most typically involving COI sequences) have found either sharing of haplotypes among at least two of these ocean regions (e.g., the copepods Calanus glacialis [Nelson et al. 2009], Pleuromamma xiphias [Goetze 2011], Podon leuckarti [clade II only, Durbin et al. 2008], Pleopis polyphemoides [clade I only, Durbin et al. 2008] and Eucalanus spinifer [Goetze 2005], the echinoderms Strongylocentrotus pallidus [Palumbi and Kessing 1991] and S. droebachiensis [Addison and Hart 2005] and the bivalve Mytilus trossulus [Rawson and Harper 2009]) or else the haplotypes found in the N. Pacific and N. Atlantic populations are closely related with typically <1 % sequence divergence (e.g., the copepods Eucalanus hyalinus s.s. [Goetze 2005], Pseudevadne tergestina and Penilia avirostris [Durbin et al. 2008], the echinoderms Solaster endeca, Crossaster papposus and Pteraster militaris [Corstorphine 2010], the bivalve Macoma balthica [clade III only, Luttikhuizen et al. 2003; clade D only, Nikula et al. 2007] and the crustacean Mysis segerstralei [Audzijonyte and Väinölä 2006]). Sharing of identical or similar haplotypes is plausibly interpreted as the result of trans-Arctic dispersal, particularly when the species in question also occurs in the Arctic Ocean, as in S. pallidus (and to a lesser extent S. droebachiensis), S. endeca, C. papposus and M. trossulus. Few of the studies cited above have estimated when the presumed trans-Arctic dispersal event occurred, although several studies have estimated late Pleistocene dispersal dates (S. pallidus, S. droebachiensis) or multiple dispersal events beginning in the early Pleistocene/late Pliocene and ending in the late Pleistocene or early Holocene (M. balthica, M. trossulus).

Available data are equivocal as to a N. Pacific Ocean versus a S. Pacific Ocean origin for Hippasteria phrygiana. Castelloe and Templeton (1994) have discussed the difficulty of rooting intraspecific trees or networks using a distantly related haplotype as outgroup, as with our COI data (see section "Results"). Given this uncertainty about the ancestral range of H. phrygiana, nothing definitive can be said about dates and routes by which the species colonized oceans on a global scale (excluding the Arctic Ocean), except that the dispersal event[s] presumably occurred during or before the last interglacial period (110–130 kya). The population splitting times estimated in IMa refer to the times when the species’ panmictic global population structure was replaced by a genetically subdivided structure with little or no connectivity. The most probable estimates of the inter-region divergence times (53–78 kya) suggest that connectivity was lost sometime after the onset of the last glacial period of the Pleistocene (approximately 110 kya: Ehlers and Gibbard 2007). Sea surface temperature data summarized by Rahmstorf (2002); Kaiser et al. (2005); Dubois et al. (2011) and others show pronounced cooling and the development of cold thermocline waters in the N. Pacific Ocean during the 60–70 kya period.

There is correspondence between the modern day distribution of H. phrygiana and ice cover recorded during the Last Glacial Maximum or LGM (Ehlers and Gibbard 2007). In the Northern Hemisphere, glaciers covered an extensive part of the North Pacific including the area near the Okhotsk Sea, the southeastern Bering Sea and the Aleutians to northern Canada during the LGM (Ehlers and Gibbard 2007; Barron et al. 2009), as well as much of the North Atlantic coast adjacent to Greenland and Northern Europe. In the Southern Hemisphere, ice covered Antarctica, the west coast of the tip of South America and part of New Zealand during the LGM. Gutt (2001) has documented benthic recolonization by various invertebrates and algae following withdrawal of ice cover. We propose that, after the LGM, H. phrygiana expanded its range from one or more refugia in each of the three ocean regions sampled here, when suitable substratum became available for larval settlement following withdrawal of glacial cover over nearshore and continental shelf settings. Retreat of glacial ice would have provided an extended colonizable region for H. phrygiana to exploit, and the results of Fu’s test suggested a possible signature of recent population expansion (or other departure from neutrality) for the N. Pacific Ocean samples. However, attempts to date the timing of an assumed sudden population size expansion for each oceanic region analyzed separately using Schneider and Excoffier’s (1999) mismatch analysis in Arlequin v.3.1.1 produced expansion dates (τ) with wide confidence intervals that not only overlapped the point estimates of population splitting reported above, but also for two regions overlapped the present time (details not shown). This suggestion of post-LGM range expansion is similar to that outlined by Frasier et al. (2009) for in the widespread Southern Bull Kelp (Durvillaea antarctica). We speculate that post-LGM colonization may have been facilitated by optimal temperature for larval settlement (as indicated above), by reduction in the numbers or diversity of competing species, or perhaps by the opportunity to feed on sessile cnidarians, such as corals or alcyonarians, which have been recorded on post-ice benthos (Gutt 2001). Pelagic lecithotrophic larvae would be carried by cold-water currents to the shelf regions, where settlement conditions would be optimal. In the N. Pacific, larvae would be associated with the California, Alaska and Kamchatka currents. In the Southern Hemisphere, larval transport would be facilitated by the Peru, Cape Horn, Falkland and Antarctic circumpolar currents, which includes the New Zealand, South American (Patagonian), South African and South Indian Ocean (Kerguelen Islands, etc.) regions. Distribution data for several subAntarctic/high latitude echinoderms from New Zealand, South America and South Africa have long suggested the Antarctic Circumpolar Current (i.e., the West Wind Drift) as a likely larval transport mechanism across the Southern Hemisphere (Fell 1962). N. Atlantic cold-water currents include the Canary, Greenland and Labrador Currents and the North Atlantic Drift.

Our COI sequence data support the synonymy of various species and subspecies of Hippasteria in the N. Pacific with H. phrygiana (see list in the section "Methods"), but we have only one specimen from the S. Atlantic. On morphological grounds, subspecies recorded from the S. Atlantic, including H. phrygiana argentinensis and H. p. capensis, are almost certainly synonyms of H. phrygiana (Mah, unpubl. data). Our single specimen from the S. Atlantic Ocean was identified morphologically as H. hyadesi but shared a COI sequence haplotype with other widely separated Southern Hemisphere populations of H. phrygiana. In addition to our data, an H. phrygiana specimen from Chile has been previously recorded (USNM 1121154), and H. hyadesi has been recorded from Marion Island and Prince Edward Island in the southwestern Indian Ocean (Branch et al. 1993; Stampanato and Jangoux 2004). Given that Hippasteria populations in the S. Atlantic are similar genetically (this study) and morphologically (Clark and Downey 1992) to N. Atlantic H. phrygiana, it is likely that the S. Atlantic Ocean and southwestern Indian Ocean populations of Hippasteria provide a conspecific link between N. Atlantic and S. Pacific populations of H. phrygiana, in which case colonization of the Atlantic Ocean from the Pacific Ocean could have occurred either via the southern tip of S. America, or via the Southern Ocean/southwestern Indian Ocean to the Atlantic coast of Africa, or via both routes (see Fig. 1 for more details). More genetic data from S. Atlantic Ocean populations of Hippasteria are needed to distinguish among these possibilities.

Molecular studies of the wide-ranging tropical corallivore Acanthaster planci (i.e., the crown of thorns starfish) have shown four cryptic species where only one was previously known (Vogler et al. 2008). H. phrygiana shows an opposite example where multiple nominal species have been shown to be one distinct lineage. H. phrygiana is also one of several hippasterines which have been identified as predators on deep-sea corals and other cold-water cnidarians, at least in the Pacific Ocean (Kreiger and Wing 2002; Mah et al. 2010). One of its major prey species in the Pacific Ocean (the sea pen Ptilosarcus gurneyi, see Birkeland 1974) is absent from the Atlantic Ocean (Williams 2011), which suggests intraspecific variation in feeding ecology. Hippasteria is one of over three dozen cold-water asteroid genera with globally distributed members (other examples include Pteraster, Henricia and Solaster: Clark 1996). Current DNA barcoding projects for these taxa (e.g., Corstorphine 2010) tend to sample specimens from only a limited portion of the entire range. Broader sampling could potentially uncover additional examples of a single widely distributed species showing morphological and ecological variation across its range, with important implications for biodiversity assessment, ecology and conservation.


We thank the following persons for sampling and/or providing access to specimens: David Clague and Lonny Lundsten (Monterey Bay Aquarium Research Institute, Moss Landing, CA), Roger N. Clark (Insignis Biological Consulting, Eagle Mountain, UT), Prof. G. Duhamel and Noémie Vasset (MNHN, Paris, France), Annie Mercier and J.F. Hamel (Ocean Sciences Center, Memorial University, Newfoundland, Canada), Martha Nizinski (National Marine Fisheries Service, NOAA, Washington DC), Neil McDaniel (British Columbia, Canada), Robert J. Van Syoc and Christina Piotrowski (California Academy of Sciences, San Francisco, CA), Amber York (Woods Hole Oceanographic Institution, Woods Hole, MA). Collection of North Atlantic specimens was facilitated by J. Drewery (Marine Scotland-Science, Aberdeen, Scotland, AB11 9DB UK). Specimens provided by the National Institute of Water and Atmospheric Research (NIWA) Invertebrate Collection were collected from various cruises funded by the New Zealand Ministry for Primary Industries (Fisheries, MPI), Land Information New Zealand (LINZ), NIWA and New Zealand Department of Conservation; the Scientific Observer Program funded by the New Zealand MPI; the New Zealand International Polar Year–Census of Antarctic Marine Life with project governance provided by the MPI Science Team and the Ocean Survey 20/20 CAML Advisory Group (LINZ, MPI, Antarctica New Zealand, Ministry of Foreign Affairs and Trade and NIWA); New Zealand Foundation for Research, Science and Technology, and CSIRO’s Division of Marine and Atmospheric Research. Michael W. Hart and Sam White provided help with the IMa2 analysis; Taylor M. Bass assisted in the laboratory. Funding was provided by NSF award DEB-1036358 to Foltz and Mah, as well as by two REU supplements to the above award.

Supplementary material

227_2013_2180_MOESM1_ESM.xlsx (25 kb)
Supplementary material 1 (XLSX 25 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2013