Reference: Biol. Bull. 201: 95–103. (August 2001)

Biogeography of Asterias: North Atlantic Climate Change and Speciation

JOHN P. WARES Duke University Zoology, Box 90325, Durham, North Carolina 27708

Abstract. Fossil evidence suggests that the seastar genus 1983; Vermeij, 1991). Today, two species are recognized in Asterias arrived in the North Atlantic during the trans- the North Atlantic: A. forbesi on the North American coast, Arctic interchange around 3.5 Ma. Previous genetic and primarily from Cape Hatteras to Cape Cod (Franz et al., morphological studies of the two species found in the At- 1981), and A. rubens on the American coast primarily from lantic today suggested two possible scenarios for the spe- Cape Cod northward (Franz et al., 1981), and on the Euro- ciation of A. rubens and A. forbesi. Through phylogenetic pean coast from Iceland to western France (Clark and and population genetic analysis of data from a portion of the Downey, 1992; Hayward and Ryland, 1995). American cytochrome oxidase I mitochondrial gene and a fragment of populations of A. rubens have been previously described as the ribosomal internal transcribed spacer region, I show that A. vulgaris, a junior synonymy (Clark and Downey, 1992). the formation of the Labrador Current 3.0 Ma was probably These species co-occur over a broad range of the North responsible for the initial vicariance of North Atlantic As- American continental shelf centered on Cape Cod (Gosner, terias populations. Subsequent adaptive evolution in A. 1978; Menge, 1979; Franz et al., 1981). forbesi was then possible in isolation from the European Two current hypotheses attempt to explain the recent species A. rubens. The contact zone between these two speciation between A. forbesi and A. rubens. Schopf and species formed recently, possibly due to a Holocene found- Murphy (1973) suggested that they were a germinate spe- ing event of A. rubens in New England and the Canadian cies pair formed by a late Pleistocene (0.02–2.5 Ma) vicari- Maritimes. ance event (i.e., a separation of populations) at Cape Cod, possibly due to lower sea levels during glacial maxima. Introduction There is some evidence for hybridization between these The North Atlantic Ocean is populated by hundreds of seastars, but the separation could be maintained by localized taxa which invaded from the North Pacific following the adaptation to the different thermal regimes north and south opening of the Bering Strait about 3.5 million years ago of Cape Cod (Franz et al., 1981). However, this thermal (Ma; Durham and MacNeil, 1967; Vermeij, 1991). Some of boundary was latitudinally unstable throughout the Pleisto- these species have maintained genetic contact with source cene (Cronin, 1988) and only in the past 20,000 years populations in the Pacific until recently (Palumbi and Kess- (Holocene) has it returned to its current state. If the geo- ing, 1991; van Oppen et al., 1995), but many of them have graphical isolation between these taxa was recent, as pro- subsequently differentiated from the source populations and posed in Schopf and Murphy (1973), then strong natural are now recognized as distinct species (e.g., Gosling, 1992; selection within each region has prevented widespread hy- Reid et al., 1996; Collins et al., 1996). Circumstantial bridization. evidence suggests strongly that the seastar genus Asterias The second hypothesis, based on morphological and pa- (Echinodermata: Asteroidea: Asteriidae: Asteriinae) partic- leoceanographic evidence, suggested a late Pliocene (ap- ipated in the trans-Arctic interchange (Worley and Franz, proximately 2.5–5 Ma) separation of Asterias into distinct North American and European species, followed by a Ho- locene recolonization of North America by the European Received 28 September 2000; accepted 10 May 2001. Current address: Dept. of Biology, University of New Mexico, Castetter species A. rubens (Worley and Franz, 1983). This hypoth- Hall, Albuquerque, NM 87131. E-mail: [email protected] esis would therefore suggest that the differentiation between

95 96 J. P. WARES

Table 1 were stored at Ϫ80°C. PCR amplification of an approxi- Collection sites for individuals of each species in this study mately 700-bp portion of the mitochondrial cytochrome c oxidase I (COI) protein-encoding gene was performed using Species [Population] Location Sample size the primers LCO1490 and HCO 2198 from Folmer et al. (1994). Amplification was performed in 50-␮l reactions A. rubens [North America] Maine (44°N, 69°W) 12 containing 10–100 ng DNA, 0.02 mM each primer, 5 ␮l Nova Scotia (46°N, 62°W) 6 Newfoundland (50°N, 55°W) 5 Promega 10ϫ polymerase buffer, 0.8 mM dNTPs (Pharma- A. rubens [Europe] Iceland (64°N, 22°W) 2 cia Biotech), and 1 unit Taq polymerase (Promega). Reac- Norway (63°N, 10°E) 8 tions took place in a Perkin-Elmer 480 thermal cycler with Ireland (53°N, 10°E) 10 a cycling profile of 94° (60 s) Ϫ40° (90 s) Ϫ72° (150 s) for France (48°N, 3°E) 5 40 cycles. The internal transcribed spacer (ITS) region was A. forbesi North Carolina (34°N, 76°W) 5 Cape Cod (41°N, 70°W) 3 amplified under similar conditions, with an annealing tem- A. amurensis Sea of Japan (43°N, 131°E) 4 perature of 50°C and with primers ITS4 and ITS5 (White et Leptasterias sp. Iceland (64°N, 22°W) 1 al., 1990). For each individual, sequences were obtained for three to four clones, and the consensus sequence was ob- Voucher specimens are being maintained in the marine invertebrate collections of C. W. Cunningham at Duke University. tained to eliminate Taq error. PCR products were prepared for sequencing and were cycle-sequenced as in Wares (2001) using both PCR prim- the two Atlantic species is entirely due to long-term isola- ers. COI sequences representing each individual in this tion. Thus, subsequent physiological adaptations to warmer study have been deposited with GenBank (AF240022- water in A. forbesi (Franz et al., 1981) are independent of 240081); ITS sequences were only obtained for 10 individ- the speciation event. Essentially, the distinction between uals, representing each species and region, and are also these species reflects either primary divergence due to se- accessible in GenBank (AF346608-AF346617). Sequences lection or secondary contact following vicariance (Endler, were aligned and edited for ambiguities using complemen- 1977). tary fragments in Sequencher 3.0 (Genecodes Corp., Cam- In this study, mitochondrial and nuclear sequence data bridge, MA). No gaps or poorly aligned regions occurred in were collected from populations of A. forbesi and A. rubens the COI alignment, but missing characters were trimmed throughout North America and Europe, as well as from from the ends of the alignment to produce equal sequence populations of the Pacific sister taxon A. amurensis (Clark lengths for all individuals. In the ITS alignment, all missing and Downey, 1992). Phylogenetic and population genetic or ambiguous characters, including gaps, were removed. assays were used to test the hypotheses described above. It Consensus sequences were exported as a NEXUS file for appears that Worley and Franz (1983) were remarkably subsequent analysis in PAUP*4.0b4a (Swofford, 1998). accurate in suggesting a Pliocene speciation followed by a recent invasion of A. rubens from Europe, even in their Phylogenetic analysis prediction of details of timing, mechanisms, and effects. Although selection may have driven some of the diver- A heuristic search for the set of most-parsimonious trees gence, it now seems clear that the initial separation of A. based on the COI data was performed using PAUP*4.0b4a rubens and A. forbesi is due to late Pliocene changes in (Swofford, 1998). Trees were rooted using Leptasterias climate and ocean current flow, whereas North American polaris (Asteriinae) and individuals of A. amurensis. Start- populations of A. rubens are very recent arrivals. ing trees were obtained via stepwise addition, with simple addition sequence. Tree-bisection-reconnection was used Materials and Methods for branch swapping, and branches were collapsed if the maximum branch length was zero. Asterias specimens were collected from intertidal sites Maximum-likelihood (ML) phylogenies were also gener- listed in Table 1. Tube feet were immediately placed in 95% ated in PAUP*. The best-fit model for all likelihood anal- ethanol or DMSO buffer (0.25 M EDTA pH 8.0, 20% yses (HKY with ⌫-distributed rate variation; Hasegawa et DMSO, saturated NaCl; Seutin et al., 1991). Species were al., 1985; Yang, 1994) was determined by adding parame- identified on the basis of key morphological characters ters until the likelihood description of the neighbor-joining described in Clark and Downey (1992) and Hayward and tree did not significantly improve (Goldman, 1993; Cun- Ryland (1995). ningham et al., 1998), using the likelihood-ratio test of ModelTest (Posada and Crandall, 1998). A series of boot- DNA extraction and amplification strap replicates (100 ML replicates, heuristic search) using DNA was phenol-extracted from each specimen follow- PAUP* were performed to determine support for interspe- ing the protocol in Hillis et al. (1996). These extractions cific relationships in the clade. Estimates of the transition- BIOGEOGRAPHY OF ASTERIAS 97 transversion ratio for the HKY model, along with the other possible rooted trees was calculated using 107 simu- gamma-distributed parameter for among-site rate heteroge- lations in GeneTree. neity, were held constant for bootstrap replicates. A maxi- mum likelihood phylogeny of the ITS sequence data was Tests of rate constancy also generated using the appropriate best-fit model (F81: Likelihood-ratio tests (Felsenstein, 1988; Goldman, equal rates among sites, unequal base frequencies). 1993) were used to test the hypothesis that the data collected Estimates of speciation time within the North Atlantic were consistent with a constant-rate Poisson-distributed require an estimate of the mutation rate (␮). Because pale- process of substitution (molecular clock). This procedure ontological evidence suggests that Asterias arrived in the ensures that the data can be used to estimate the time of North Atlantic during the trans-Arctic interchange about 3.5 divergence between A. rubens and A. forbesi. The ML Ma (Worley and Franz, 1983; Vermeij, 1991), and because phylogeny was estimated using the best-fit model, and then climatic changes shortly thereafter would have prevented the likelihood of this phylogeny was recalculated while additional trans-Arctic migration, this date was used to constraining the estimate to fit the molecular clock model. calibrate the divergence between the Pacific species A. These likelihood (L) estimates were used to calculate the amurensis and the North Atlantic taxa. Other species, in- 2 ␹ -distributed test statistic ␦ ϭ 2[1n(L0) Ϫ 1n(L1)], with cluding the echinoderm Strongylocentrotus pallidus, have (n Ϫ 2) degrees of freedom where n is the number of taxa clearly maintained more recent connections across the Arc- in the tree. tic (Palumbi and Kessing, 1991). However, S. pallidus appears to be more tolerant of Arctic conditions than Aste- Neutrality tests rias (Worley and Franz, 1983; Palumbi and Kessing, 1991). The ML estimate of the internal branch length separating Because adaptive selection may have played a role in the the sister taxa (representing net nucleotide divergence d, divergence between A. rubens and A. forbesi (given a short Nei and Li, 1979) was used to estimate the appropriate divergence time; Schopf and Murphy, 1973), polymorphism mutation rate ␮ (Edwards and Beerli, 2000), where ␮ ϭ 0.5 data for each species were input to DNAsp v.3.5 (Rozas and d/(3.5 Ma). Estimates were obtained for the full COI data Rozas, 1999) to test for patterns of non-neutral evolution. set (first, second, and third codon positions), as well as third Within each species, Tajima’s (1989) test generates a beta- position only. Use of the third-position estimate circum- distributed parameter indicating the difference in two esti- vents problems with branch length estimation when there is mates (polymorphic sites and number of alleles) of diver- strong rate variation (Wares and Cunningham, in press), as sity. Significantly low statistics can indicate non-neutral well as problems with the potential influences of non- evolution (Tajima, 1989). Additionally, a McDonald-Kreit- neutral evolution. man test (McDonald and Kreitman, 1991) was performed on Haplotype networks may be more appropriate represen- each pairwise set of species polymorphism data to deter- tations of genealogical relationships within species than are mine whether selection has played a role in the divergence outgroup-rooted phylogenetic trees, because ancestral hap- between A. rubens and A. forbesi. Also, DNAsp was used to lotypes are still present in the population (Crandall and calculate haplotype diversity (H, see eqn. 8.4 in Nei, 1987) Templeton, 1996). Methods associated with haplotype net- and sampling variance for each species or population. works were used to determine the root haplotype for A. rubens. Determination of the root haplotype prevents spu- Results rious conclusions about ancestry among populations. Net- The COI data set (60 individuals) includes 627 charac- works were created using a parsimony criterion in the ters, of which 484 are constant, 49 are parsimony-uninfor- program TCS (alpha version 1.01, Clement et al., 2000); at mative, and 94 are parsimony-informative. Base frequencies the same time, a Bayesian analysis of the likelihood that are 33.7% A, 19.6% C, 21.6% G, and 25.1% T for this parsimony is violated (Templeton et al., 1992) was per- fragment. Most of the substitutions (92.3%) are at third- formed to ensure that the data set was unlikely to be com- position sites; overall, 63% of all third-position characters plicated by homoplasy. are polymorphic. These third-position sites are heavily AT- The ML root was determined using GeneTree (Griffiths biased (39.0% A, 15.1% C, 11.8% G, and 33.9% T). and Tavare´, 1994); the likelihood of each possible rooted The best-fit model (HKY ϩ⌫) was used to estimate gene tree was determined under an infinite-alleles model. distances among individuals to determine whether there is This model assumes that there are no multiply substituted any evidence for saturation at third-position characters in nucleotide sites. The method allows for recoding of char- the COI coding region. A plot of pairwise genetic distances acters so that independent substitutions are analyzed sepa- versus number of third-position substitutions does not indi- rately, but this was not an issue with the A. rubens COI data. cate any pattern of saturation (data not shown); in fact, all of The relative likelihood of each tree in comparison with all the information within each species is based on third-posi- 98 J. P. WARES tion substitutions. Additionally, the best-fit model was re- to influence the age estimates of the COI (all positions) and estimated for this character partition; likelihood-ratio tests ITS data sets strongly. indicate that the HKY model with invariant sites (I ϭ A McDonald-Kreitman test (McDonald and Kreitman, 0.213) and no rate variation describes the third-position 1991) rejects a pattern of neutral substitution between A. data effectively. The Asterias data sets do not reject the rubens and A. forbesi (P Ͻ 0.01, Table 3). Despite branch molecular clock model, whether all positions are considered lengths that do not reject the molecular clock model, there (P ϭ 0.163), or only third positions (P ϭ 0.231). is an excess of amino acid replacement substitutions be- Maximum-likelihood analysis was used to determine the tween the Atlantic species. The replacement substitutions interspecific gene tree, using all codon positions and the between A. rubens and A. forbesi do not include any first- HKY ϩ⌫model (Tr:Tv 8.256, ␣ ϭ 0.0608, four rate position substitutions. Half (8/16) of the amino acid substi- classes). The ML tree (L ϭ 1472.87) is presented in Figure tutions do not involve a change in charge or polarity, 1A, including all individuals sampled within A. amurensis, whereas almost half (7/16) of the changes substitute a basic A. rubens, and A. forbesi. Bootstrap support is indicated on residue for an uncharged or nonpolar residue. However, the tree, with each species being fully resolved in 100% of there does not seem to be an obvious pattern to these replicates. The Pacific species A. amurensis is basal to a changes between A. rubens and A. forbesi. Other species strongly supported clade of Atlantic species in this phylog- comparisons do not reject the neutral model of substitution eny. (Table 3). Within each species, Tajima’s (1989) test is Following exclusion of missing and ambiguous charac- nonsignificant (A. amurensis, D ϭ 0.837, P Ͼ 0.10; A. ters in the ITS data set (length of fragment varies from 413 forbesi, D ϭϪ0.705, P Ͼ 0.10; A. rubens, D ϭϪ1.482, to 482 bases when gaps included), these data include 368 P Ͼ 0.10), indicating that there is no reason to suspect characters of which 343 are constant, 1 is parsimony-unin- non-neutral evolution in the intraspecific comparisons. formative, and 24 are parsimony-informative. Indels did not Additionally, Bayesian analysis (Templeton et al., 1992; vary within species and were removed (analysis with Clement et al., 2000) of the COI data within A. rubens gapped characters included produced nearly identical re- indicates greater than 95% confidence that the intraspecific sults). Parsimony analysis produced a single most-parsimo- gene tree is parsimonious. The ML root haplotype is found nious tree of 25 steps, and the ML phylogeny (best-fit model on both coasts of the Atlantic (Fig. 1A, Haplotype B), and F81, no rate variation) is shown in Figure 1B. Under a this haplotype is at least an order of magnitude more likely variety of mutational models, this phylogeny is statistically to be the ancestral haplotype than any other haplotype of A. indistinct from the COI phylogeny in Figure 1A. Likeli- rubens (likelihood index ϭ 0.857). All North American hood-ratio tests indicate that, in addition to a similar inter- haplotypes are also found in Europe; the unique haplotypes specific topology, branch lengths on the COI and ITS phy- found in Europe contribute to a significantly higher allelic logenies are proportional (P Ͼ 0.10), though the substitution diversity (P Ͻ 0.01, Table 4). The ITS data are consistent rate is significantly different (P Ͻ 0.05). Bootstrap replicates of with the COI data in that there is no allelic diversity among the ITS data also indicate strong support for differentiation North American and European individuals of A. rubens among these species. The ITS data do not reject a molecular (n ϭ 6). clock model. Divergence among these species is indicated in Table 2. Discussion HKY ϩ⌫distances in the COI fragment indicate that A. amurensis, A. forbesi, and A. rubens have been isolated Understanding the mechanisms that are responsible for from each other for a similar amount of time; assuming the divergence of Asterias rubens and A. forbesi first re- trans-Arctic isolation around 3.5 Ma, A. rubens and A. quires that the timing of their divergence be estimated. forbesi have been separated for at least 3.0 Ma. Although Estimates based on the molecular calibrations reported here the estimated divergence date is higher when all codon suggest that these species last shared a common ancestor at positions are included (Table 2), and these data do not reject least 3.0 Ma (Table 2), not long after the genus first arrived a molecular clock, neutrality tests (see below) suggest that in the North Atlantic (around 3.5 Ma; Worley and Franz, some second-position substitutions may be under selection. 1983; Vermeij, 1991). Note, however, that asterozoan skel- Therefore, third-position sites may be more appropriate for etons are rarely preserved in the fossil record, because they the divergence estimate. The estimated divergence time is lack rigidly articulated skeletons and rapidly disintegrate also higher when the ITS data are used; however, there is no (Barker and Zullo, 1980); indeed, fossils of A. forbesi have reason to believe that speciation predated the appearance of been reported only twice, each time in Pleistocene intergla- Asterias in the North Atlantic, and the long branch leading cial sediments. Thus, little direct evidence points to the first to A. forbesi is not easily explained since it appears in both appearance of Asterias in the North Atlantic (Durham and phylogenies (one using a protein-coding gene, one using MacNeil, 1967; Worley and Franz, 1983), and the biogeo- untranslated spacer region data). This longer branch appears graphic data used in this paper is therefore based on con- BIOGEOGRAPHY OF ASTERIAS 99

Figure 1. Phylogenetic trees for Asterias generated using the best-fit maximum likelihood model in each data set (COI: HKY ϩ⌫; ITS: F81). (A) Cytochrome c oxidase I phylogeny of inter- and intraspecific Asterias relationships. Here all characters (first, second, and third position) are included; an identical topology is found using parsimony or distance methods, or looking at third-position characters alone. Bootstrap support for each species is indicated by the numbers below each branch. These data do not reject a molecular clock model. The divergence across the Arctic (between A. amurensis and the Atlantic species) is considered to be 3.5 Ma; this generates an estimate of about 3.0 Ma for the divergence between A. rubens and A. amurensis (see Table 2 and Discussion). Haplotypes A–D of A. rubens are found on both the North American and European coasts (A: Maine (n ϭ 8), Nova Scotia (n ϭ 2), Newfoundland (n ϭ 2), Iceland (n ϭ 1), Norway (n ϭ 2), Ireland (n ϭ 1); B: Maine (n ϭ 2), Nova Scotia (n ϭ 4), Newfoundland (n ϭ 3), Iceland (n ϭ 1), Ireland (n ϭ 2), France (n ϭ 2); C: Maine (n ϭ 1), Norway (n ϭ 3), Ireland (n ϭ 2), and France (n ϭ 1); D: Ireland (n ϭ 1), and Maine (n ϭ 1)). Amphi-Atlantic haplotype B is the maximum likelihood root (index ϭ 0.857). (B) Internal transcribed spacer (ITS) phylogeny of inter- and intraspecific Asterias relationships. Likelihood ratio tests do not reject a hypothesis of proportional branch lengths (P Ͼ 0.10) suggesting that, aside from substantial differences in substitution rate, the two phylogenies are equivalent representations of interspecific differentiation. A nearly identical phylogeny is reconstructed when indels are included in the ITS data. sistent fossil evidence from other cold temperate species around 5°–6°C warmer in the North Atlantic and Arctic, that participated in the trans-Arctic exchange. Nevertheless, permitting the initial trans-Arctic passage of temperate spe- there is reason to believe that Asterias also spread from the cies (Berggren and Hollister, 1974; Vermeij, 1991), but then Pacific to the Atlantic at about 3.5 Ma (Worley and Franz, two dramatic changes were initiated around 3.0 Ma that 1983). Miocene and early Pliocene temperatures were appear to play a role in speciation within the North Atlantic. 100 J. P. WARES

Table 2 Internal branch lengths (based on best-fit likelihood model) separating Asterias species (lower triangle*, all 3 matrices)

All characters A. amurensis A. rubens A. forbesi A. amurensis ␮ ϭ 1.954 ϫ 10Ϫ8 Ϯ 8.63 ϫ 10Ϫ9 ␮ ϭ 2.665 ϫ 10Ϫ8 Ϯ 9.59 ϫ 10Ϫ9 A. rubens 0.13678 Ϯ 0.06044 3.59 Ma A. forbesi 0.18658 Ϯ 0.06715 0.16576 Ϯ 0.04595

3rd position only A. amurensis A. rubens A. forbesi A. amurensis ␮ ϭ 6.689 ϫ 10Ϫ8 Ϯ 3.36 ϫ 10Ϫ8 ␮ ϭ 9.751 ϫ 10Ϫ8 Ϯ 3.74 ϫ 10Ϫ8 A. rubens 0.48084 Ϯ 0.2352 2.96 Ma A. forbesi 0.68254 Ϯ 0.26168 0.49270 Ϯ 0.15661

ITS-1 A. amurensis A. rubens A. forbesi A. amurensis ␮ ϭ 5.142 ϫ 10Ϫ9 Ϯ 2.04 ϫ 10Ϫ9 ␮ ϭ 7.188 ϫ 10Ϫ9 Ϯ 2.40 ϫ 10Ϫ9 A. rubens 0.0361 Ϯ 0.0143 3.84 Ma A. forbesi 0.0500 Ϯ 0.0168 0.0470 Ϯ 0.0163

The calibration date of 3.5 Ma is used to obtain the mutation rate ␮ for comparisons between A. amurensis and the Atlantic species. The estimated divergence time between A. rubens and A. forbesi is based on the mean of this calibrated mutation rate (cytochrome c oxidase I [COI] all positions, top; COI 3rd position only, middle; internal transcribed spacer (ITS) 1, bottom). * In each matrix, the lower triangle containing the internal branch lengths is made up of the matrix cells below the diagonal line of empty cells representing comparisons within the same value. The upper triangle contains the estimated mutation rates and estimated divergence data.

At that time, warm North Atlantic currents were dis- shelf off Nova Scotia and the rest of New England (Berg- placed by the formation of the cold-water Labrador Current. gren and Hollister, 1974; Worley and Franz, 1983; Cronin, This event created a significant thermal gradient in the 1988). As Northern Hemisphere glaciation began, the North Atlantic, and tropical-temperate faunas were abruptly present-day latitudinally controlled faunal provincialization replaced with polar and subpolar faunas on the continental was established as well (Berggren and Hollister, 1974). This dramatic cooling of the northwestern North Atlantic prob- ably initiated the separation of North Atlantic Asterias into Table 3 European and North American populations with very little McDonald-Kreitman tests on each Asterias species pair using genetic contact (Worley and Franz, 1983). Subsequent cytochrome c oxidase I (COI) translated data Pleistocene glaciation would have prevented the long-term

Species pair Fixed differences Polymorphisms Table 4 A. rubens-A. forbesi Synonymous 39 19 Comparisons of haplotype diversity (H, see eqn. 8.4 in Nei 1987, Nonsynonymous 16 0 calculated in DNAsp 3.50, Rozas and Rozas 1999) for the cytochrome c P Ͻ 0.001 oxidase I fragment in each species and population of A. rubens A. rubens-A. amurensis Synonymous 36 21 Species/Population Haplotype diversity (H) ␴2 Nonsynonymous 12 1 P Ͼ 0.05 Asterias rubens 0.793 0.00138 A. forbesi-A. amurensis North America 0.597 0.00395 Synonymous 44 15 Europe 0.893 0.00143 Nonsynonymous 14 1 Asterias forbesi 0.964 0.00596 P Ͼ 0.15 Asterias amurensis 0.999 0.03125

Only the comparison between A. rubens and A. forbesi indicates a European populations of A. rubens have significantly higher allelic significant departure from neutral evolution. A two-tailed Fisher’s exact diversity than North American populations (P Ͻ 0.01); this finding is test was used for each set of comparisons. supported by nonparametric haplotype sampling in Wares (2000). BIOGEOGRAPHY OF ASTERIAS 101 establishment of populations in New England, as most of 1997). However, the North American colonization is diffi- the North American coast from Long Island Sound north- cult to date because there are no unique haplotypes in North ward was covered by a kilometer of ice during glacial America; ancestral allelic polymorphism tends to inflate maxima (Kelley et al., 1995). indirect estimates of population size and age (Kuhner et al., Pacific and Atlantic populations of other species appear 1998; Edwards and Beerli, 2000). The lack of unique di- to have had more recent trans-Arctic genetic contact than versity in North America also prevents the meaningful use the estimates above would suggest for Asterias (Palumbi of other phylogeographic methods; for instance, statistics of and Kessing, 1991; van Oppen et al., 1995). Moreover, the geographic dispersion of haplotypes (for review see rapid climatic fluctuations (Cronin, 1988; Roy et al., 1996) Templeton, 1998) are uninformative (Wares, unpubl. data). during the Pleistocene could have permitted large-scale This is primarily because even closely related individuals changes in the geographic range of cold temperate species. (identical haplotypes) are distributed across the entire geo- However, both the sea urchin Strongylocentrotus pallidus graphic range of A. rubens. It is possible that the multiple (Palumbi and Kessing, 1991) and the red alga Phycodrys shared alleles between Europe and North America represent rubens (van Oppen et al., 1995) appear to have greater a multiple-invasion history; Asterias larvae are planktotro- tolerance for Arctic waters than Asterias does. Worley and phic and disperse in the water column for 6 or more weeks Franz (1983) report that expansion of Asterias populations (Clark and Downey, 1992). into habitats as far north as Greenland only occurs period- There is evidence that natural selection has played some ically, and that these populations cannot tolerate colder role in the overall divergence between these species. A waters (Franz et al., 1981). However, the indirect morpho- significant number of amino acid replacement substitutions logical and paleontological evidence is bolstered by the distinguish A. rubens from A. forbesi (Table 3), all of them molecular evidence, which strongly suggests that A. rubens reflecting second- or third-position nucleotide substitutions. and A. forbesi diverged shortly after their ancestral lineage There is no obvious pattern to the amino acid replacements, separated from the Pacific A. amurensis. The estimates of as most of them involve substitutions among uncharged or mutation rate presented here are very similar to other esti- nonpolar amino acids. Two of the three species in the genus mates for both the COI fragment (Knowlton and Weigt, Asterias are found in cold-temperate waters, while A. 1998; Schubart et al., 1998; Wares, 2001; Wares and Cun- forbesi is found in the warmer mid-Atlantic region (Schopf ningham, in press) and the ITS fragment (Schlo¨tterer et al., and Murphy, 1973; Franz et al., 1981). Many of the phys- 1994; van Oppen et al., 1995). Thus these data strongly iological differences between A. rubens and A. forbesi support earlier inferences of a late Pliocene trans-Arctic (Franz et al., 1981) reflect this latitudinal distribution. How- passage and subsequent speciation within the Atlantic. ever, the possibility that these amino acid substitutions are An analysis of genealogical patterns within A. rubens related to physiological differences in the warm-temperate confirms that the North American populations of this spe- A. forbesi has never been tested. The difference in temper- cies are descendants of a recent colonization from Europe ature between the habitats of A. rubens and A. forbesi is that probably followed the most recent glacial maximum unlikely to contribute to differences in metabolic rate that (about 20,000 BP, Holder et al., 1999). The genealogical could accelerate the mutation rate (for review see Rand, data presented here fit several important patterns that sug- 1994). Nevertheless, this hypothesis is worth examination, gest a recent range expansion (Wares, 2000). All North because A. forbesi is supported by relatively long branches American haplotypes are identical to the most-common in both the COI and the non-coding ITS region (Table 2, European haplotypes (Fig. 1A). Generally, invading haplo- Fig. 1B). If natural selection is playing a role in the amino types are the most deeply nested haplotype in the European acid divergences of the mitochondrial COI gene between A. (putative source) population. This is to be expected, because rubens and A. forbesi, there is no reason why a noncoding deeply nested ancestral haplotypes are often the most com- nuclear sequence should reflect the same increase in diver- mon (Castelloe and Templeton, 1994), and therefore have a gence rate. higher probability of participating in long-distance dispersal In conclusion, the biogeographic response of Asterias to events. Haplotype B (Fig. 1A) is a good illustration of this late Pliocene climatic and oceanographic change fits a pat- expectation—it is closely related to each other haplotype tern predicted by Worley and Franz (1983). Following the and has a high copy number in both European and American arrival of Asterias in the North Atlantic around 3.5 Ma populations. These observations contribute to the high like- (Worley and Franz, 1983; Vermeij, 1991), populations were lihood (85.7%, more than an order of magnitude greater established on both the European and North American likelihood than any other haplotype) that this is the ancestral coasts during a period when the North Atlantic was as much allele in A. rubens. as 5–6°C warmer (Berggren and Hollister, 1974). The for- Additionally, allelic diversity is significantly lower in mation of the Labrador Current 3.0 Ma rapidly changed the North American A. rubens than in Europe (Table 4), a signal faunal composition of the intertidal Canadian Maritimes and of recent range expansion (Hewitt, 1996; Austerlitz et al., New England coast, and Asterias populations in this region 102 J. P. WARES probably went extinct. An American population survived Brown. 1996. Evolutionary history of northern hemisphere Nucella under the conditions of the mid-Atlantic coast and Gulf (Gastropoda, Muricidae): molecular, morphological, ecological, and Stream waters (A. forbesi), and the European population (A. paleontological evidence. Evolution 50: 2287–2304. Crandall, K. A., and A. R. Templeton. 1996. Applications of intraspe- rubens) has recently recolonized the cold-temperate shores cific phylogenetics. Pp. 81–99 in New Uses for New Phylogenies, P. H. of New England and the Canadian Maritimes. Thus, the Harvey, A. J. Brown, J. Maynard Smith, and S. Nee, eds. Oxford zone of sympatry between these two species appears to be a University Press, Oxford. zone of secondary contact. Hybridization is considered rare Cronin, T. M. 1988. Evolution of marine climates of the U. S. Atlantic between these species (Schopf and Murphy, 1973; Worley coast during the past four million years. Philos. Trans. R. Soc. Lond. B and Franz, 1983), but whether behavioral mechanisms 318: 661–678. Cunningham, C. W., and T. M. Collins. 1998. Beyond area relation- (Franz et al., 1981) or gametic recognition mechanisms ships: extinction and recolonization in molecular marine biogeography. (Hellberg and Vacquier, 1999; Pernet, 1999) are responsible Pp. 297–321 in Molecular Ecology and Evolution: Approaches and is unclear. Applications, B. Schierwater, B. Streit, G. Wagner, and R. DeSalle, The genetic data presented here illustrate a strong con- eds. Birkhauser Verlag, Basel. cordance between paleoceanographic changes and indirect Cunningham, C. W., H. Zhu, and D. M. Hillis. 1998. Best-fit maxi- estimates of speciation between the North Atlantic Asterias mum-likelihood models for phylogenetic inference: Empirical tests with known phylogenies. Evolution 52: 978–987. species. The species boundaries are phylogenetically quite Durham, J. W., and F. S. MacNeil. 1967. Cenozoic migrations of distinct, and the divergence estimates based on these genetic marine invertebrates through the Bering Strait region. Pp. 326–349 in data appear to support a late Pliocene, rather than late The Bering Land Bridge, D. M. Hopkins, ed. Stanford University Press, Pleistocene or Holocene, separation. A better understanding Stanford. of the balance between oceanographic and climatic changes Edwards, S. V., and P. Beerli. 2000. Perspective: Gene divergence, in the late Pliocene and Pleistocene, and of the response of population divergence, and the variance in coalescence time in phylo- geographic studies. Evolution 54: 1839–1854. species based on varying life-history characters to these Endler, J. A. 1977. Geographic Variation, Speciation, and Clines. changes, will enable us to predict the responses of other taxa Princeton University Press, Princeton, NJ. (Cunningham and Collins, 1998; Wares and Cunningham, Felsenstein, J. 1988. Phylogenies from molecular sequences: inference in press). and reliability. Annu. Rev. Genet. 22: 521–565. Folmer, O., M. Black, W. Hoeh, R. Lutz, and R. Vrijenhoek. 1994. Acknowledgments DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates. Mol. Mar. Biol. Bio- I thank G. Manchenko, A. Ingo´lfsson, J. Maunder, B. technol. 3: 294–299. O’Connor, D. Garbary, D. M. Rand, and C. Damiani for Franz, D. R., E. K. Worley, and A. S. Merrill. 1981. Distribution patterns of common seastars of the middle Atlantic continental shelf of assistance in the field collecting seastars. The manuscript the northwest Atlantic (Gulf of Maine to Cape Hatteras). Biol. Bull. was greatly improved thanks to discussions with T. Turner 160: 394–418. and the suggestions of two anonymous reviewers. These Goldman, N. 1993. Statistical tests of models of DNA substitution. J. analyses were done in the laboratory of C. W. Cunningham, Mol. Evol. 36: 182–198. whose aid during this and other projects was invaluable. A Gosling, E. M. 1992. Systematics and geographical distribution of Myti- National Science Foundation Dissertation Improvement lus. Pp. 1–20 in The Mussel Mytilus: Ecology, Physiology, Genetics, and Culture, E. M. Gosling, ed. Elsevier, Amsterdam. Grant (NSF DEB-99-72707) to J. P. W. funded this study. Gosner, K. L. 1978. A Field Guide to the Atlantic Seashore From the Bay of Fundy to Cape Hatteras. Houghton Mifflin, Boston. Literature Cited Griffiths, R. C., and S. Tavare´. 1994. Sampling theory for neutral alleles in a varying environment. Philos. Trans. R. Soc. Lond. B 344: Austerlitz, F., B. Jung-Muller, B. Godelle, and P.-H. Gouyon. 1997. 403–410. Evolution of coalescence times, genetic diversity and structure during Hasegawa, M., H. Kishino, and T. Yano. 1985. Dating of the human- colonization. Theor. Popul. Biol. 51: 148–164. Barker, L., and V. A. Zullo. 1980. Asterias forbesi (Desor) (Asterozoa, ape splitting by a molecular clock of mitochondrial DNA. J. Mol. Evol. Asteroidea) from the Pleistocene “Coquina” at Fort Fisher, New 21: 160–174. Hanover County, North Carolina. J. Elisha Mitchell Sci. Soc. 96: Hayward, P. J., and J. S. Ryland. 1995. Handbook of the Marine 39–44. Fauna of North-West Europe. Oxford University Press, Oxford. Berggren, W. A., and C. D. Hollister. 1974. Paleogeography, paleobio- Hellberg, M. E., and V. D. Vacquier. 1999. Rapid evolution of fertil- geography, and the history of circulation in the Atlantic Ocean. Soc. ization selectivity and lysin cDNA sequences in teguline gastropods. Econ. Paleontol. Mineral. 20: 126–186. Mol. Biol. Evol. 16: 839–848. Castelloe, J., and A. R. Templeton. 1994. Root probabilities for in- Hewitt, G. M. 1996. Some genetic consequences of ice ages, and their traspecific gene trees under neutral coalescent theory. Mol. Phylogenet. role in divergence and speciation. Biol. J. Linn. Soc. 58: 247–276. Evol. 3: 102–113. Hillis, D. M., B. K. Mable, A. Larson, S. K. Davis, and E. A. Zimmer. Clark, A. M., and M. E. Downey. 1992. Starfishes of the Atlantic. 1996. Nucleic acids IV: Sequencing and cloning. Pp. 321–384 in Chapman & Hall, London. Molecular Systematics, D. M. Hillis, C. Moritz, and B. K. Mable, eds. Clement, M., D. Posada, and K. A. Crandall. 2000. TCS: a computer Sinauer, Sunderland, MA. program to estimate gene genealogies. Mol. Ecol. 9: 1657–1660. Holder, K., R. Montgomerie, and V. L. Friesen. 1999. A test of the Collins, T. M., K. Frazer, A. R. Palmer, G. J. Vermeij, and W. M. glacial refugium hypothesis using patterns of mitochondrial and nu- BIOGEOGRAPHY OF ASTERIAS 103

clear DNA sequence variation in rock ptarmigan (Lagopus mutus). hybridizing seastars Asterias forbesi and Asterias vulgaris and impli- Evolution 53: 1936–1950. cations for their evolution. Biol. Bull. 145: 589–597. Kelley, J. T., A. R. Kelley, and S. Appollonio. 1995. Landforms of the Schubart, C. D., R. Diesel, and S. B. Hedges. 1998. Rapid evolution to Gulf of Maine. Pp. 19–38 in From Cape Cod to the Bay of Fundy: an terrestrial life in Jamaican crabs. Nature 393: 363–365. Environmental Atlas of the Gulf of Maine, P. W. Conkling, ed. MIT Seutin, G., B. N. White, and P. T. Boag. 1991. Preservation of avian Press, Cambridge, MA. blood and tissue samples for DNA analyses. Can. J. Zool. 69: 82–90. Knowlton, N., and L. A. Weigt. 1998. New dates and new rates for Swofford, D. L. 1998. PAUP*. Phylogenetic Analysis Using Parsimony divergence across the Isthmus of Panama. Proc. R. Soc. Lond. B 265: (* and Other Methods). Version 4. Sinauer Associates, Sunderland, 2257–2263. MA. Kuhner, M. K., J. Yamato, and J. Felsenstein. 1998. Maximum like- Tajima, F. 1989. Statistical method for testing the neutral mutation lihood estimation of population growth rates based on the coalescent. hypothesis by DNA polymorphism. Genetics 123: 585–595. Genetics 149: 429–434. Templeton, A. R. 1998. Nested clade analyses of phylogeographic data: McDonald, J. H., and M. Kreitman. 1991. Adaptive protein evolution testing hypotheses about gene flow and population history. Mol. Ecol. at the Adh locus in Drosophila. Nature 351: 652–654. 7: 381–397. Menge, B. A. 1979. Coexistence between the seastars Asterias vulgaris Templeton, A. R., K. A. Crandall, and C. F. Sing. 1992. A cladistic and A. forbesi in a heterogeneous environment: a non-equilibrium analysis of phenotypic associations with haplotypes inferred from explanation. Oecologia 41: 245–272. restriction endonuclease mapping and DNA sequence data. III. Cla- Nei, M. 1987. Molecular Evolutionary Genetics. Columbia University dogram estimation. Genetics 132: 619–633. Press, New York. van Oppen, M. J. H., S. G. A. Draisma, J. L. Olsen, and W. T. Stam. Nei, M., and W.-H. Li. 1979. Mathematical model for studying genetic 1995. Multiple trans-Arctic passages in the red alga Phycodrys variation in terms of restriction endonucleases. Proc. Natl. Acad. Sci. rubens: evidence from nuclear rDNA ITS sequences. Mar. Biol. 123: USA 76: 5269–5273. 179–188. Palumbi, S. R., and B. D. Kessing. 1991. Population biology of the Vermeij, G. 1991. Anatomy of an invasion: the trans-Arctic interchange. trans-Arctic exchange: mtDNA sequence similarity between Pacific Paleobiology 17: 281–307. and Atlantic sea urchins. Evolution 45: 1790–1805. Wares, J. P. 2000. Abiotic influences on the population dynamics of Pernet, B. 1999. Gamete interactions and genetic differentiation among marine invertebrates. Ph.D. dissertation, Duke University, Durham, three sympatric polychaetes. Evolution 53: 435–446. NC. Posada, D., and K. A. Crandall. 1998. Modeltest: testing the model of Wares, J. P. 2001. Patterns of speciation inferred from mitochondrial DNA substitution. Bioinformatics 14: 817–818. DNA in North American Chthamalus (Cirripedia: Balanomorpha: Rand, D. 1994. Thermal habitat, metabolic rate, and the evolution of Chthamaloidea). Mol. Phylogenet. Evol. 18: 104–116. mtDNA. Trends Ecol. Evol. 9: 125–131. Wares, J. P., and C. W. Cunningham. Phylogeography and historical Reid, D. G., E. Rumbak, and R. H. Thomas. 1996. DNA, morphology ecology of the North Atlantic intertidal. Evolution (in press). and fossils: phylogeny and evolutionary rates of the gastropod genus White, T. J., T. Bruns, S. Lee, and J. Taylor. 1990. Amplification and Littorina. Philos. Trans. R. Soc. Lond. B 351: 877–895. direct sequencing of fungal ribosomal RNA genes for phylogenetics. Roy, K., J. W. Valentine, D. Jablonski, and S. M. Kidwell. 1996. Pp. 315–322 in PCR Protocols: A Guide to Methods and Applications, Scales of climatic variability and time averaging in Pleistocene biotas: M. A. Innis, D. H. Gelfand, J. J. Sninsky, and T. J. White, eds. implications for ecology and evolution. Trends Ecol. Evol. 11: 458–463. Academic Press, San Diego. Rozas, J., and R. Rozas. 1999. DnaSP version 3: an integrated program Worley, E. K., and D. R. Franz. 1983. A comparative study of selected for molecular population genetics and molecular evolution analysis. skeletal structures in the seastars Asterias forbesi, A. vulgaris, and A. Bioinformatics 15: 174–175. rubens with a discussion of possible relationships. Proc. Biol. Soc. Schlo¨tterer, C., M. Hauser, A. von Haeseler, and D. Tautz. 1994. Wash. 96: 524–547. Comparative evolutionary analysis of rDNA ITS regions in Drosoph- Yang, Z. 1994. Maximum likelihood phylogenetic estimation from DNA ila. Mol. Biol. Evol. 11: 513–522. sequences with variable rates over sites: Approximate methods. J. Mol. Schopf, T. J. M., and L. S. Murphy. 1973. Protein polymorphism of the Evol. 39: 306–314.