Genetic Structure of the Star Sea Squirt, Botryllus Schlosseri, Introduced in Southern European Harbours

Molecular Ecology (2006) 15, 3957–3967 doi: 10.1111/j.1365-294X.2006.03087.x

BlackwellGenetic Publishing Ltd structure of the star sea squirt, Botryllus schlosseri, introduced in southern European harbours

SUSANNA LÓPEZ-LEGENTIL,* XAVIER TURON‡ and SERGE PLANES* *Centre de Biologie et d’Ecologie Tropicale et Méditerranéenne, EPHE–FRE CNRS 2935, University of Perpignan, 66860 Perpignan, France, ‡Department of Animal Biology (Invertebrates), Faculty of Biology, University of Barcelona, 645 Diagonal Ave, 08028 Barcelona, Spain

Abstract The introduction of new genetic variants or species is often caused by maritime transport between harbours. Botryllus schlosseri is a cosmopolitan ascidian species that is found in both harbours and open shore habitats. In order to determine the influence of ship traffic on the genetic structure and phylogeography of B. schlosseri in southern Europe, we analyzed the variability of a fragment of the mitochondrial gene cytochrome c oxidase subunit I (COI). We sampled seven Atlanto-Mediterranean harbour populations and three open- shore populations. In addition, we sequenced some colonies from the US-Atlantic coast and from other Mediterranean localities to perform phylogenetic analyses. Although the number of polymorphic sites recorded (25.8%) was within the range observed in other population studies based on ascidian COI sequences, the haplotypic diversity (16 haplotypes out of 181 sequences) was much lower. Moreover, a lack of intermediate haplotypes was observed. This pattern of high nucleotide diversity and low haplotype diversity was consistent with introduction events of a few divergent haplotypes. We found a strong genetic structure in the study populations. Gene flow was only appreciable between some harbour populations. Harbour- and open-shore populations were well differentiated, although there was no evidence for isolation by distance. A nested clade analysis pointed to long-distance colonization, possibly coupled with subsequent fragmentation, as the underlying process. Our results suggest that B. schlosseri entered the study area via harbour-hopping, possibly through recurrent introduction events. The haplotypes from North America and most of the European ones were grouped in the same phylogenetic clade. This suggests occasional gene flow between both continents, probably through ship transport. Keywords: ascidian, Atlanto-Mediterranean, introduced species, nested clade analysis, phylogeography, population genetics Received 29 March 2006; revision accepted 23 June 2006

factors make them more vulnerable to the introduction Introduction of new genetic variants or species, which are generally Harbours and marinas are often the entrance gates for new opportunistic forms. Several ascidians species are com- colonizers travelling on ship’s hulls, in ballast water and monly found inhabiting harbours and human-made floating sediments, or in biofilms (Carlton & Geller 1993). Harbour structures. Because their larvae have short planktonic life- communities generally present low species richness, strong spans (Olson 1985; Svane & Young 1989; Stoner 1994) and environmental stress and high levels of pollution. These most likely rely on anthropogenic transport for long-distance dispersal, ascidians are important indicators of between- harbour dynamics. †Present address: Center for Marine Science, University of North The presence of nonindigenous ascidian species or genetic Carolina at Wilmington, 5600 Marvin K. Moss Lane, Wilmington NC 28409, USA. variants has been recognized in many locations. However, Correspondence: Susanna López-Legentil, Fax: +1-910-9622410; in many cases they are confined to artificial substrates E-mail: [email protected] such as marina floats, boat hulls, pilings and aquaculture

3958 S. LÓPEZ-LEGENTIL, X. TURON and S . PLANES facilities (Brunetti 1979; Monniot 1981; Monniot et al. 1985; In this study, we analyzed the variation in mitochondrial Zibrowius 1991; Lambert & Lambert 1998, 2003; Castilla et al. cytochrome c oxidase subunit I (COI) gene fragments, both 2002; Lambert 2002, 2004; Turon et al. 2003). Eco-phenotypic to determine the genetic structure of harbour and open- variability and disruption of the functional integrity of an shore populations of B. schlosseri in southern Europe, and ecosystem may provide the first clue to the arrival of a new to ascertain the role played by ship traffic and natural variant or species in a region. Although new introductions dispersion in the dynamics and phylogeography of this may be common at first in marginal marine habitats (such species. Here we use the term ‘open-shore population’ as harbours), they can eventually spread to open habitats for populations inhabiting open littoral rocky shores, and (Turon et al., in press). For instance, substantial areas of ‘harbour population’ for those inside harbours or marinas. subtidal natural hard surfaces in California are now being In addition, we sequenced some colonies from the US- colonized by nonindigenous ascidians that are quickly Atlantic coast, and from the northwest Mediterranean, overgrowing almost every resident sessile species in the to perform phylogenetic analyses. The COI gene is highly process (Lambert 2002). However, if the colonizer is a polymorphic in different ascidian species (Tarjuelo et al. genetic variant that is morphologically indistinguishable 2001, 2004; Castilla et al. 2002; Turon et al. 2003; Turon & from the native population, its arrival may go undetected. López-Legentil 2004; López-Legentil & Turon 2006). Know- In such cases, only molecular tools can reliably identify ledge about the genetic structure and the phylogeographic and uncover a cryptic introduction. Not surprisingly, patterns of B. schlosseri may help track human-mediated instances of cryptic introduction and speciation have been dispersal pathways and evaluate the role of harbours in shown whenever ascidian species with variable morphology colonization dynamics. and wide distribution ranges have been studied with molecular markers (Aron & Solé-Cava 1991; Dalby 1997; Materials and methods Tarjuelo et al. 2001, 2004; Castilla et al. 2002; Turon et al. 2003; López-Legentil & Turon 2006). Samples Botryllus schlosseri Pallas 1766 is a polymorphic ascidian described in Falmouth harbour, England. Although no Ten populations of Botryllus schlosseri (Stolidobranchiata, firm evidence exists, B. schlosseri is believed to have spread F. Styelidae) were sampled from three open-shore areas in all around the world via ship traffic. Therefore, it is often northern Spain (from both, the Atlantic and the Mediterra- found in harbours and marinas (Lambert & Lambert 1998; nean sides): Cadaqués, Cubelles and Fornelos and seven Stoner et al. 2002). It was reported in Australia and New harbours in southern Europe: Estartit, Roses, Blanes, Canet, Zealand in 1928 (see Van Name 1945) and in Japan in 1929 Estaque (Marseille), Rochelle and Graña (Fig. 1; Table 1). (Tokioka 1953). In addition, it is believed to have been For phylogenetic purposes, we analyzed some colonies introduced to the east coast of North America prior to from Woods Hole marina (Massachusetts, US), Sainte the 1830s (Bancroft 1903; Ritter & Forsyth 1917; Van Name Marie la Mer harbour (Southeast France) and Palamós 1945; Stoner et al. 2002) and was first observed in San (Northeast Spain; Table 2). Two additional sequences from Francisco Bay in the mid-1940s (Cohen & Carlton 1995). Roscoff (Northwestern France) and Maine (Massachusetts; Since then, it has become established from San Diego to northwest US) were retrieved from GenBank (Table 2). Alaska (Lambert & Lambert 1998; Lambert 2001). Sampling was undertaken in 2005 by SCUBA diving or by

Table 1 Diversity measures for the populations of Botryllus schlosseri studied with sample locations (GPS position). Population code (Pc), Population sample size (n), number of haplotypes (Nh), haplotype codes, number of polymoprhic sites (Np), haplotype diversity (h), and nucleotide diversity (P). Standard deviations are given in parenthesis

Population Latitude Longitude Pc n Nh Haplotype code Np hP

Estartit Harbour 42:03:15N 3:12:16E 1 16 2 HA, HB 12 0.5250(0.0546) 0.0120(0.0067) Cadaques 42:17:55N 3:17:25E 2 25 2 HC, HD 73 0.1533(0.0915) 0.0214(0.0112) Roses Harbour 42:15:18N 3:10:47E 3 11 3 HA, HE, HF 83 0.4727(0.1617) 0.0636(0.0339) Canet Harbour 42:42:15N 3:02:06E 4 15 3 HA, HF, HG 80 0.6857(0.0683) 0.0792(0.0409) Graña Harbour 43:28:54N 8:15:36W 5 25 4 HD, HH, HI, HJ 79 0.7000(0.0649) 0.0414(0.0211) Blanes Harbour 41:40:30N 2:47:50E 6 13 2 HA, HJ 13 0.5385(0.0602) 0.0134(0.0075) Cubelles 41:11:53N 1:39:58E 7 14 2 HA, HK 15 0.4945(0.0876) 0.0142(0.0079) Fornelos 43:26:59N 8:18:37W 8 18 3 HL, HM, HN 89 0.6797(0.0556) 0.0796(0.0406) Rochelle Harbour 46:08:39N 1:10:13W 9 18 2 HB, HO 13 0.5229(0.0477) 0.0130(0.0072) Estaque Harbour 43:21:37N 5:18:48E 10 26 3 HA, HB, HP 12 0.4400(0.1017) 0.0076(0.0044)

GENETIC STRUCTURE OF BOTRYLLUS SCHLOSSERI IN SOUTHERN EUROPE 3959

Fig. 1 Map showing the 10 sampling sites of Botryllus schlosseri: Rochelle, Graña and Fornelos in the northeast Atlantic; Estartit, Canet, Roses, Estaque, Blanes, Cubelles and Cadaqués in the northwest Mediterranean. The three open-shore populations are under- lined. Haplotype frequencies for each population are also depicted.

Table 2 Additional sequences of Botryllus schlosseri used for the phylogenetic reconstruction, with sample locations (GPS position), size (n), number of haplotypes (Nh) and haplotypes codes. GenBank Accession nos are also indicated; a, this study; b, Stach & Turbeville (2002); c, Johnson et al. (unpublished)

Population Latitude Longitude n Nh Haplotype code GenBank Accession no.

Woods Hole 41:31:01N 70:40:06W 4 3 HQ, HR, HS DQ340222 to DQ340224(a) Vilanova 41:12:37N 1:43:28E 9 2 HT, HU DQ340218, DQ340219(a) Ste. Marie La Mer 42:43:24N 3:02:21E 1 1 HV DQ340221(a) Palamós 41:51:17N 3:08:36E 2 1 HW DQ340220(a) Roscoff 48:43:36N 3:59:11W 1 1 ST AY116601(b) Maine 44:24:48N 68:43:48W 1 1 BR DQ367525(c)

pulling up harbour ropes. The specimens were identified for 1 min; and extension at 72 °C for 1.5 min), and a final as B. schlosseri based on Van Name (1945) and Berrill (1950). extension at 72 °C for 5 min, in an Eppendorf AG Multicycler gradient PCR. PCR products were purified using the Qiagen PCR DNA extraction and sequencing purification kit. The sequencing reaction was carried out In order to maximize DNA extractions, we separated the with the BigDye TM terminator using the same primers as zooids from the tunic under a binocular microscope. The in the amplification step. Sequences were obtained on an zooids were then kept in absolute ethanol at −20 °C until ABI Prism 3730XL automated sequencer at MACROGEN used. Mitochondrial DNA (mtDNA) was extracted from Inc. (Seoul, Korea). the zooids using the Puregene kit (Gentra Systems). The universal primers HCO2198 and LCO1490, described in Population genetics Folmer et al. (1994), were used to amplify a segment of the COI mitochondrial gene. Amplification was performed in Sequences were aligned using bioedit version 7.0.5.2 (Hall a 25 µL total-reaction volume with: 1.25 µL of each primer 1999) and alignments were confirmed by eye. Tajima’s D (10 µm), 2.5 µL dNTP’s (2 mm), 2.5 µL 10x buffer, 2 µL statistic was estimated with dnasp version 4.10 (Rozas et al. µ µ MgCl2, 0.2 L Taq polymerase 5 U, and 1 L DNA. A single 2003), to test whether selective pressures were acting on soak at 94 °C for 2 min, was followed by 40 amplification the substitutions. Further analyses where performed using cycles (denaturation at 94 °C for 1 min; annealing at 51 °C arlequin version 2.000 (Schneider et al. 2000). Nucleotide

3960 S. LÓPEZ-LEGENTIL, X. TURON and S . PLANES

diversity and haplotype diversity (Nei 1987) were calculated geographic (Dc and Dn, respectively) distances were then for each population. The exact test of population differ- interpreted in terms of biological processes using the entiation (Raymond & Rousset 1995) was used to test the inference key described in Templeton (2004) and updated null hypothesis of a random distribution of the different in November 2005 (http://darwin.uvigo.es/). haplotypes among geographic locations. Its significance was estimated from 10 000 random permutations of the Phylogenetic analysis original data matrix. We calculated pairwise FST values and their significance through permutation tests (10 000 Relationships between haplotypes (Tables 1 and 2) were replicates) among pairs of populations. In addition, we established by maximum-likelihood methods. We used the used t-tests to compare FST values and other statistics of the COI sequence of another botryllid species, Botryllus tyreus harbour and the open-shore populations. We performed a from Palau (Micronesia, GenBank Accession no. DQ365851) Mantel test with 10 000 permutations to test for isolation identified following Monniot & Mooniot 1987, 2001) as an by distance between populations (Rousset 1997). Analysis outgroup for the phylogenetic analyses. Because nucleotide of molecular variance (amova; Excoffier et al. 1992) was variation was mainly restricted to the third base position performed to examine the hierarchical population structure: of the codon, two maximum-likelihood trees (Felsenstein either by pooling the populations in harbour- and open-shore 1981), based on nucleotide sequences and translated environments; in Atlantic and Mediterranean harbours; or aminoacid sequences, were inferred with treeﬁnder version with no population grouping. October 2005 (Jobb 2005). The use of translated sequences provided an alternative dataset less prone to potential saturation problems. We used modeltest 3.0 (Posada & Demographic analyses Crandall 1998) for the first tree, to select the best-fit model To detect population growth and assess the history of of nucleotide substitution for our data set. We used the effective population size we computed the raggedness program ProtTest version 1.2.7 (Abascal et al. 2005) for index, based on the mismatch distribution (Harpending 1994), the second tree, to determine the best model of protein using dnasp version 4.10 (Rozas et al. 2003). In addition, evolution to input in treeﬁnder. Confidence in the nodes to obtain a wider view of the evolutionary scenario, and was assessed by 10 000 bootstrap replicates Felsenstein (1985). given the fact that mismatch distributions have been found to be very conservative (Ramos-Onsins & Rozas 2002), we Results also calculated Fu’s Fs (Fu 1997) and the R2 statistic (Ramos- Onsins & Rozas 2002) for all populations. Population genetics We obtained 181 partial sequences for the COI mitochondrial Phylogeographic analysis gene of the ascidian Botryllus schlosseri from 10 populations Nested clade analysis (NCA) aims to distinguish recurrent along the western Mediterranean and eastern Atlantic gene flows from historical events, which are determinants coasts (Table 1). After alignment and trimming, a final of the present-day haplotype distribution (Templeton 1998). length of 524 bp was used. All sequences have been We estimated a haplotype network using the program tcs deposited in the GenBank data base (Accession nos: from version 1.21 (Clement et al. 2000). This method estimates DQ223766 to DQ223768, and from DQ340205 to DQ340217, an unrooted tree and provides a 95% plausible set for respectively). Sixteen haplotypes were found in the parsimonious relationships between the haplotypes. The 181 specimens, with a total of 135 (25.8%) variable sites. network was then nested by hand, using the procedures Nucleotide variation was scattered across the entire described in Templeton et al. (1987) and extended in sequenced region. However, it was mainly restricted to Crandall (1996). Once the nesting design was complete, the third base position of the codons. Thus, 130 (99.24%) we performed an NCA (Templeton et al. 1987) on the nucleotide substitutions out of the 131 analyzed (the clades, using geographic and genetic information and the remaining four were found in complex codons and were program geodis version 2.4 (Posada et al. 2000). A matrix not analyzed) yielded synonymous changes. Only one of geographic distances (in km by sea) between the localities resulted in a nonsynonymous substitution (a transition was input. Two measures were computed from distance G–A). Tajima’s D statistic was not significant (D = 0.1693; > data: clade distances (Dc, a measure of the geographic P 0.1) for the entire data set. Thus, there was no evidence dispersion of individuals bearing haplotypes of a given of selection acting on this locus and a neutral model could clade) and nested clade distances (Dn, a measure of how not be rejected. far individuals with haplotypes belonging to a given clade The main parameters describing variability within popu- are from individuals bearing haplotypes of the next-higher lations: the number of haplotypes (Nh); the number of level clade). Significant associations between genetic and polymorphic sites (Np); haplotype diversity (h); and

GENETIC STRUCTURE OF BOTRYLLUS SCHLOSSERI IN SOUTHERN EUROPE 3961

< Table 3 Pairwise FST values between populations of Botryllus schlosseri. Significant values at P 0.05 are indicated with an asterisk

Estartit Roses Canet Graña Blanes Rochelle Estaque Harbour Cadaques Harbour Harbour Harbour Harbour Cubelles Fornelos Harbour Harbour Population 1 2 3 4 5 6 7 8 9 10

10 2 0.42087* 0 3 0.24572* 0.20866* 0 4 0.38903* 0.41662* 0.05319 0 5 0.03441 0.23461* 0.08564 0.23517* 0 6 −0.02544 0.40497* 0.22017* 0.36587* 0.00918 0 7 0.49359* 0.45377* 0.30306* 0.43542* 0.29777* 0.48006* 0 8 0.61193* 0.60729* 0.35104* 0.23374* 0.49634* 0.59114* 0.61906* 0 9 0.42988* 0.63595* 0.45490* 0.46888* 0.27944* 0.43865* 0.65945* 0.64418* 0 10 0.21801* 0.34628* 0.24797* 0.48438* 0.15171* 0.21993* 0.51240* 0.68223* 0.65465* 0

Table 4 Analysis of molecular variance for the cytochrome c oxidase subunit I (COI) sequences of Botryllus schlosseri. Analyses are presented for the whole area studied without grouping, pooling populations in harbour and natural habitats, and pooling harbour populations in Mediterranean and Atlantic groups. Significant values at P < 0.05 are indicated with an asterisk. Va, Vb and Vc are the associate covariance components. FCT, FSC and FST are the F-statistics

Sum of Variance Percentage Fixation Source of variation d.f. squares components of variation indices

Among populations without grouping 9 1112.925 6.42071 Va 43.07

Within populations 171 1451.379 8.48760 Vb 56.93 FST: 0.43068*

Between harbours and open shore groups 1 181.988 0.67258 Va 4.41 FCT: 0.04409 Among populations within groups 8 930.936 6.09533 Vb 39.95 FSC: 0.41798* Within populations 171 1451.379 8.48760 Vc 55.64 FST: 0.44364* − − − Between Atlantic and Mediterranean harbours 1 63.203 0.21096 Va 1.94 FCT: 0.01942 Among populations within groups 5 313.320 3.26234 Vb 30.03 FSC: 0.29455* Within populations 117 914.178 7.81349 Vc 71.92 FST: 0.28085* nucleotide diversity (π), as a measure of within-population There was no evidence of isolation by distance between variation (Nei 1987) are summarized in Table 1. Haplotype the populations studied (Mantel test, P = 0.143). In fact, frequencies per sampling site are depicted in Fig. 1. There whereas the Graña harbour (northeast Spain; Atlantic were no significant differences between open-shore and side) and Blanes harbour populations (northwest Spain; harbour populations in terms of haplotype (t-test; P = 0.336) Mediterranean side) were not significantly different, the or nucleotide (t-test; P = 0.801) diversity. reverse was true for the closest populations, at Fornelos The results of the exact test for population differentia- and Graña harbour (3.27 km apart). tion, based on haplotype frequencies between the different The hierarchical amova analysis revealed that 57% of the locations, revealed significant heterogeneity in the distri- genetic variance was found among populations. A slightly bution of haplotypes (P < 0.05). Permutation tests showed smaller amount of variance (43%) was explained by com- that almost all pairwise FST values corresponded to signi- parisons within populations (Table 4). Both components ficant genetic divergence. Some exceptions were: Estartit were highly significant in a permutation test (P < 0.0001). harbour compared with Graña and Blanes harbours; Roses The overall FST value (0.43068) was larger than those obtained harbour compared with Graña and Canet harbours; and from random permutations of haplotypes between popu- Graña harbour compared with Blanes harbour (Table 3). lations. That indicates that there was strong genetic struc- All pairwise comparisons involving open-shore populations ture in the study area. The amovas that considered harbour yielded significant FST values. The mean FST value among and open-shore populations, and Atlantic and Mediterra- harbour populations was significantly lower than among nean harbours as different groups, yielded similar results. open-shore populations. This indicates that there was stronger The highest variation was found among populations within genetic differentiation among the latter (0.27 ± 0.03 vs. groups (39.95% and 30.03%, respectively) and within popu- 0.56 ± 0.53, respectively; mean ± SE; t-test, P = 0.01). lations (55.64% and 71.92%, respectively). The comparison

Table 5 Results of the demographic parameters for the 10 populations studied: the raggedness index (r), Fu’s Fs statistic (Fs), and Ramos-Onsins & Rozas statistic (R2). Significant values at P < 0.05 are indicated with an asterisk

Population rFs R2

Estartit Harbour 0.7769 11 255 0.2625 Cadaqués 0.7639 20 066 0.0767* Roses Harbour 0.4922 16 864 0.1928 Canet Harbour 0.4297 24 341 0.2595 Graña Harbour 0.2599 18 220 0.1359 Blanes Harbour 0.7929 10 827 0.2692 Cubelles 0.7446 11 671 0.2470 Fornelos 0.6330 28 191 0.2344 Rochelle Harbour 0.7744 12 513 0.2614 Estaque Harbour 0.3648 7 225 0.1651

between harbour and open-shore environments explained only 4.41% of variance. A negative and nonsignificant variance component was found among harbour populations separated into Mediterranean and Atlantic groups (−1.94%; Table 4). Therefore, there was no evidence of any genetic partition between harbours and open-shore populations or between Atlantic and Mediterranean harbours.

Demographic analyses The demographic parameters estimated to detect population growth indicated that a model of constant size could not be rejected for the populations studied (Table 5). A significant R2 value was only found for the natural population of Cadaqués. However, its corresponding Fu’s Fig. 2 Statistical parsimony cladogram obtained from the tcs Fs and r-values were nonsignificant (P = 1 in both cases). program. Each line represents a mutational step. Missing inter- Fu’s Fs, r and R2 values follow the same trend under mediates are indicated by ‘o’. The eight groups that could not be the hypothesis of population expansion. However, their connected unambiguously are designated letters A–H. The nesting design adopted is shown with boxes. Boxes with dotted margins behaviour can differ under constant-size populations that indicate the assignment of symmetrically stranded clades. are well structured (J. Rozas, personal communication), and this may be the case for the Cadaqués population.

limit of 10. Group A had a four-step nesting design; group Phylogeographic analysis B had a three-step one and the other groups were one- The parsimony haplotype network is presented in Fig. 2. level clades (C, D, E, F, G and H). The results of the NCA The maximum number of steps for a 95% parsimonious analysis showed that most associations were significant. connection between haplotypes (i.e. with a 95% confidence Again this indicates that the data had a fairly strong level that no multiple substitution has occurred at some geographic structure (Table 6). According to the inference site) is 10 mutations. We obtained eight groups of haplo- key, the process explaining the pattern found for Clade types (A– H) that were separated by more than 10 changes. 1-1 (including haplotypes found in the harbours of Estartit, No loop was found connecting the haplotypes within Rochelle, Estaque, Graña and Blanes), was a restricted gene groups. The eight groups found in the parsimony network flow with isolation by distance. For Clade 1-2 (including were subsequently treated as independent cladograms haplotypes found in the harbours of Estartit, Roses, Canet, (Fig. 2). No attempt was made to group these terminal Estaque, Graña and Blanes and in the open-shore popu- units, as the number of changes in the shortest connections lation of Cubelles), a restricted gene flow/dispersal, but between them was generally well above the parsimony with some long-distance dispersal, was substantiated.

Table 6 Results of the nested clade analysis (NCA). Only clades with significant geographic associations as detected by the permutation test are included in the table. For each clade we list the members, the corresponding clade (Dc) and nested clade (Dn) distances, and the Interior-Tip contrasts. We indicate significantly small (S) or large (L) distances. The ‘Inference’ column describes the steps followed in the inference key applied to that clade and the biological process inferred. For the whole cladogram, the older clades of the maximum-likelihood (ML) tree obtained in Fig. 3 were considered as interior clades

Clade Clade members Dc Dn Inference

1-1 HB 1542.85L 1462.48L 1-2-3-4-NO HJ 1358.56 1376.78 Restricted gene flow with isolation by distance I-T 184.29L 85.71L 1-2 HA 170.26S 407.53S 1-2-3-4-5-6-7-YES HI 0S 2090.31 Restricted gene flow/dispersal but with some HP 0 390.99L long-distance dispersal I-T 163.58 −1683.43S 1-5 HH 0 2.91L 1-19-20-NO HN 0S 1.51S Inadequate geographic sampling I-T 0 1.40L 2-1 1-1 1423.49L 1400.35 1-2-11-12-NO 1-4 0S 1149.63L Contiguous range expansion I-T −1423.49S −250.72 3-3 1-5 1.99S 467.52S 1-19-20-2 1-6 0 2046.20L Inconclusive outcome 4-1 3-1 1376.84L 1208.58L 1-2 3-2 609.64S 802.38S Inconclusive outcome Whole cladogram 1-7 0S 1892.58L 1-2-3-5-6-13-YES 1-8 0S 1575.85L Long distance colonization possibly coupled 1-9 0 817.87 with subsequent fragmentation 1-10 0S 852.56S 1-11 0S 1567.03L 1-12 28.73S 847.68S 3-3 760.34 1448.39L 4-1 967.81S 1087.70S

The inference of Clade 1-5 yielded inadequate geographic phylogenetic analyses. Comparisons between the different sampling, although the two populations found in this likelihood scores showed that the TIM + G model (Transi- clade (Fornelos and Graña harbour) are separated by tional model; Posada & Crandall 1998) was the best-fit only 3.27 Km. In Clade 2-1, the NCA suggested that a model for nucleotide substitution. Blosum62 + I + F (Henikoff contiguous range expansion was responsible for the distri- & Henikoff 1992) was the best model for protein evolution. bution of nearly all the haplotypes in harbours, except for As the TIM model is not implemented in treeﬁnder, Roses. For Clade 3-3 and Clade 4-1 (the clade that links we used the slightly more complex GTR + G model, which the whole of group A), the tip/interior status could not was the second-best model suggested by modeltest, and be determined, yielding an inconclusive outcome. Finally, the closest (in terms of number of parameters) available at the level of the whole cladogram, the differentiation in treeﬁnder. The topology of the maximum-likelihood between interior and tip clades was based, as suggested by trees obtained assuming both Blosum62 and GTR + G the inference key, on the tree obtained in the phylogenetic models was the same. Therefore, only the former is shown analysis (see below). The NCA indicated that the overall in Fig. 3. Five major clades appeared in both trees. The first distribution of B. schlosseri in the area studied in southern clade is formed by the haplotypes included in Clades B, H Europe might be explained by long-distance colonization, and E inferred in the NCA, and a single sequence obtained possibly coupled with subsequent fragmentation. for Palamós (HW). The second group (G in the NCA) has a haplotype from Fornelos (HL) and another one from Roscoff (ST). The third clade (corresponding to Clade D in Phylogenetic analysis the NCA) groups another haplotype from Fornelos (HM) The sequence BR from Maine (northeast USA) was identical and one of the two haplotypes obtained from Vilanova to our haplotype HO from Rochelle harbour (northwest (HT). Clade 4 has the second haplotype from Vilanova France). Therefore, only the latter was included in the (HU). Clade 5 (formed by the nested clade groups F, A and

Fig. 3 Maximum likelihood (ML) tree obtained with the Blosum62 model of protein evolution (likelihood = 3812.6). Haplotype codes as in Tables 1 and 2. Clades 1, 2, 3 and 5 correspond to the nested clade groups: B-H-E; G; D and F-A-C, respectively (Fig. 2). Numbers above branches indicate bootstrap support percentages (only when ≥ 50%) obtained from treeﬁnder.

C) has haplotypes found in all the harbours studied (HA, structure (significant FST values) among the studied popu- HB, HD, HG, HI, HJ, HO, HP); Cadaqués (HD); Cubelles lations, with the exception of some between-harbour (HA, HK); and all haplotypes from North America: from comparisons. There was no evidence for isolation by Woods Hole marina (HR, HS, HQ); and Maine (BR = HO). distance acting on the scale studied. In addition, the Therefore, the phylogenetic results showed an agreement raggedness index and the Fu’s Fs and R2 statistics were with the NCA groups, but no clear geographic or ecological consistent with a constant size of the study popula- separation of clades. tions over time. The NCAs showed that the haplotypes present are highly divergent. Several clades could not be connected parsimoniously, and many intermediate Discussion haplotypes were lacking. The NCA revealed that there was Analysis of mtDNA sequences of the Atlanto-Mediterranean restricted gene flow in nearly all the clades inferred. Finally, ascidian Botryllus schlosseri demonstrated a strong genetic for our overall data set, the inference key suggested

© 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd GENETIC STRUCTURE OF BOTRYLLUS SCHLOSSERI IN SOUTHERN EUROPE 3965 long-distance colonization, possibly coupled with sub- difference found between open-sea populations and their sequent fragmentation. neighbouring harbours, and by the constant population The Mediterranean is connected by considerable sea trade size inferred in the demographic studies. with most global regions; and therefore organisms found We found similar or identical haplotypes within the North in harbours may be introduced variants or species. B. schlosseri Atlantic (both on the European and American sides) and was reported in an English harbour as early as 1766, and the Mediterranean specimens. Indeed, all the haplotypes it has now a worldwide distribution, most frequently from Maine, Woods Hole marina and most of the European inhabiting harbours and marinas (Lambert & Lambert 1998; sites were grouped in the same clade of the phylogenetic Stoner et al. 2002). When a colonization event occurs, analyses (Clade 5, Fig. 3). Finding identical or closely related founding populations usually undergo a population haplotypes overseas is not surprising if we consider that bottleneck, resulting in low genetic diversity (Holland the North Atlantic contributes to over 55% of the vessel 2000). In the present study, only 16 haplotypes out of 181 entries in the Mediterranean (CIESM 2002). This significant sequences were found. This haplotype diversity is surpris- ship traffic could easily support occasional gene flow between ingly low, given the fact that the COI gene has been found harbours. to be very variable at the intraspecies level in ascidians (e.g. Our results indicated that the genetic structure of Tarjuelo et al. 2001, 2004; López-Legentil & Turon 2005, Botryllus schlosseri in southern Europe was due to multiple 2006). The number of polymorphic sites (25.8%) found colonization events from several source populations, while for B. schlosseri sequences was, however, very similar to exchange of genes between the study localities is restricted that found for the genus Cystodytes (26%; López-Legentil to some harbour populations. The similarity found between & Turon 2005) and higher than that found for Clavelina European and northeast American haplotypes suggests lepadiformis (8.7%; Turon et al. 2003) and Pseudodistoma that there is occasional gene flow between both continents, crucigaster (3.8%; Tarjuelo et al. 2004). Low haplotype which is attributable to maritime transport. Therefore, ship diversity coupled with high nucleotide variability pointed transport and harbour-hopping play an important role in to a lack of intermediate haplotypes. This together with the the population structure of B. schlosseri in southern Europe. NCA results, supports the hypothesis that the populations Finally, B. schlosseri is believed to have originated in in the study area originated from the arrival of only a few Europe (Berrill 1950) and to have consequently spread well-differentiated haplotypes of B. schlosseri (probably through European shipping to the east coast of North by ship transport). No significant partitioning of genetic America (Bancroft 1903; Ritter & Forsyth 1917; Van Name variance (amova) was found between open-shore and 1945; Stoner et al. 2002). However, Carlton (2005) recently harbour populations, indicating that the former also result proposed the alternative idea that, B. schlosseri may have from introductions. In addition, FST values among the been introduced to Europe as early as the 1500s from the three open-shore populations were high and significant, Pacific Ocean; the centre of botryllid diversity. Indeed, irrespective of whether the comparison involved popu- only 16 species of Botryllus have been reported for the lations from the Mediterranean or the Atlantic. The greater Atlantic, whereas 38 have been reported for the Indo- divergence found between open-shore populations Pacific (see Sanamyan 2002). An Indo-Pacific origin of this (even those that were geographically closer together) with species, with recent introduction to Europe, would explain respect to harbour populations, is probably due to the the low haplotypic diversity found in the present study fact that the former rely entirely on natural dispersion for and the lack of intermediate haplotypes observed even genetic interchange. in open-shore populations. However, the Indo-Pacific The presence of unique haplotypes (e.g. in Fornelos, species included here as an outgroup was quite divergent Cubelles and Cadaqués populations), and the differentia- from the ingroup. Further sampling including more Pacific tion of the Estaque and Rochelle harbours from nearly species and a larger geographic scale would be required to all of the rest, suggest that multiple introductions occurred confirm the evolutionary and phylogeographic relationships at different times, probably from different geographic of Botryllus species. sources. The lack of differentiation among some of the other harbours suggests that some gene flow occurs. Acknowledgements This indicates that the harbour dynamics is punctuated by occasional movement of haplotypes among both distant M. Rius (University Barcelona), C. Menniti (University Perpignan) and close localities. The larvae of colonial ascidians have and Sandrine Ruitton-Roussel (UMR-CNRS 6540) helped with the a short, planktonic lifespan (Svane & Young 1989), and sampling. Lori J. Bell (Coral Reef Research Foundation, Palau), and Alex Lorente (Fundación Fondos Marinos, Spain) provided harbours and marinas are enclosed environments. There- some specimens. J. Rozas (University Barcelona) and D. Posada fore, natural interchange of larvae is unlikely, and any (University Vigo) provided advice with the analyses. Sheri gene flow is attributable to ship transport or related human Johnson (University Maine) kindly shared sequence DQ367585 activities. This fact is sustained in our study by the significant and J.T. Carlton (Williams College) his knowledge about Botryllus

© 2006 The Authors Journal compilation © 2006 Blackwell Publishing Ltd 3966 S. LÓPEZ-LEGENTIL, X. TURON and S . PLANES introductions. This study was funded by a Ministry of Education Fu YX (1997) Statistical test of neutrality of mutations against and Science/Fulbright grant to S.L-L, and the project CTM2004- population growth, hitchhiking and biological selection. Genetics, 05265 of the Spanish Government. 147, 915–925. Hall TA (1999) BioEdit: a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nuceic References Acids Symposium Series, 41, 95–98. Harpending H (1994) Signature of ancient population growth in a Abascal F, Zardoya R, Posada D (2005) ProtTest: Selection of best- low-resolution mitochondrial DNA mismatch distribution. fit models of protein evolution. Bioinformatics, 21, 2104–2105. Human Biology, 66, 591–600. Aron S, Solé-Cava AM (1991) Genetic evaluation of the taxonomic Henikoff S, Henikoff JG (1992) Amino acid substitution matrices status of two varieties of the cosmopolitan ascidian Botryllus from protein blocks. Proceedings of the Natural Academy of Sciences, niger (Ascidiaceae: Botryllidae). Biochemical Systematics and USA, 89, 10915–10919. Ecology, 19, 271–276. Holland BS (2000) Genetics of marine bioinvasions. Hydrobiologia, Bancroft FW (1903) Variation and fusion of colonies in compound 420, 63–71. ascidians. Proceedings of the Californian Academy of Science, 3, 137– Jobb G (2005) treeﬁnder version of October 2005. Munich, 186. Germany [WWW document]. URL http://www.treefinder.de Berrill NJ (1950) The Tunicata. With an account of the British spe- [accessed on 9 January 2006] cies. Ray Society, London. Lambert G (2001) A global review of ascidian introduction and Brunetti R (1979) Polyandrocarpa zorritensis (Van Name, 1931) a their possible impact on the endemic fauna. In: The Biology of colonial ascidian new to the Mediterranean record. Vie Milieu, Ascidians (eds Sawada R, Yokosawa H, Lambert CC). Springer- 29, 647–652. Verlag, Heidelberg, New York, Tokyo. Carlton JT (2005) Setting ascidian invasions on the global stage. Lambert G (2002) Nonindigenous ascidians in tropical waters. Proceedings of the International Invasive Sea Squirt Conference. Pacific Science, 56, 291–298. Woods Hole Oceanographic Institution, Woods Hole, Massa- Lambert G (2004) The South temperate and Antarctic ascidian chusetts. Corella eumyota reported in two harbours in north-western Carlton JT, Geller JB (1993) Ecological roulette: the global trans- France. Journal of the Marine Biology Association of India, 84, 239– port of nonindigineous marine organisms. Science, 261, 78–82. 241. Castilla JC, Collins AG, Meyer CP, Guíñez R, Lindberg DR (2002) Lambert CC, Lambert G (1998) Non-indigenous ascidians in Recent introduction of the dominant tunicate, Pyura praeputialis southern California harbours and marinas. Marine Biology, 130, (Urochordata, Pyuridae) to Antofagasta, Chile. Molecular Ecology, 675–688. 11, 1579–1584. Lambert CC, Lambert G (2003) Persistence and differential distri- CIESM (2002) Alien marine organisms introduced by ships in the bution of nonindigenous ascidians in harbours of the Southern Mediterranean and Black Seas. CIESM Workshop Monographs California Bight. Marine Ecology Progress Series, 259, 145–161. 20. Monaco. [WWW document]. URL http://www.ciesm.org/ López-Legentil S, Turon X (2005) How do morphoytpes and online/monographs/Istanbul.html [accessed on 22 July 2006] chemotypes relate to genotypes? The colonial ascidian Cystodytes Clement M, Posada D, Crandall KA (2000) TCS: a computer (Ascidiacea: Polycitoridae). Zoologica Scripta, 34, 3–14. program to estimate gene genealogies. Molecular Ecology, 9, 1657– López-Legentil S, Turon X (2006) Population genetics, phylo- 1660. geography and speciation of Cystodytes (Ascidiacea) in the western Cohen AN, Carlton JT (1995) Nonindigenous aquatic species in a Mediterranean Sea. Biological Journal of the Linnaean Society, 88, United States estuary: a case study of the biological invasions of 203–214. the San Francisco Bay and delta. National Technical Information Monniot C (1981) Apparition de l’ascidie Microcosmus exasperatus Service, Springfield, VA. dans les ports méditerranéens. Tethys, 10, 59–62. Crandall KA (1996) Multiple interspecies transmissions of human Monniot C, Mooniot F (1987) Les ascidies de Polynesie française. and simian T-cell Leukemia/Lymphoma virus type I sequences. Memoires Du Museum National d’Histoire Naturelle, 136, 1–155. Molecular Biology and Ecology, 13, 115–131. Monniot C, Monniot F, Laboute P (1985) Ascidies du port de Dalby JE Jr (1997) Reproductive and electrophoretic evidence Papeete (Polynésie française): relations avec le milieu naturel et for genetic maintenance of dimorphism in the ascidians aports intercontinentaux par la navigation. Bulletin Du Muséum Pyura stolonifera near Melbourne, Australia. Ophelia, 47, 227– National d’Histoire Naturelle, 3, 481–495. 243. Monniot F, Monniot C (2001) Ascidians from the tropical western Excoffier L, Smouse PE, Quattro JM (1992) Analysis of molecular Pacific. Zoosystema, 13, 201–383. variance inferred from metric distances among DNA haplotypes: Nei M (1987) Molecular Evolutionary Genetics. Columbia University application to human mitochondrial DNA restriction data. Press, New York. Genetics, 131, 479–491. Olson RR (1985) The consequences of short-distance larval dispersal Felsenstein J (1981) Evolutionary trees from DNA sequences: a in a sessile marine invertebrate. Ecology, 66, 30–39. maximum likelihood approach. Journal of Molecular Evolution, Posada D, Crandall KA (1998) modeltest: testing the model of 17, 368–376. DNA substitution. Bioinformatics, 14, 817–818. Felsenstein J (1985) Confidence limits on phylogenies: an approach Posada D, Crandall KA, Templeton AR (2000) Geodis: a program using the bootstrap. Evolution, 39, 783–791. for the cladistic nested analysis of the geographical distribution Folmer O, Hoeh W, Black M, Lutz R, Vrijenhoek R (1994) DNA of genetic haplotypes. Molecular Ecology, 9, 487–488. primers for amplification of mitochondrial cytochrome c oxidase Ramos-Onsins SE, Rozas J (2002) Statistical properties of new subunit I from diverse metazoan invertebrates. Molecular Marine neutrality tests against population growth. Molecular Biology and Biology and Biotechnology, 3, 294–299. Evolution, 19, 2092–2100.

Raymond M, Rousset F (1995) An exact test for population differ- Templeton AR (1998) Nested clade analyses of phylogeographic entiation. Evolution, 49, 1280–1283. data: testing hypotheses about gene flow and population history. Ritter WE, Forsyth RA (1917) Ascidians of the littoral zone of Molecular Ecology, 7, 381–397. southern California. University of California Publications in Templeton AR (2004) Statistical phylogeography: methods of Zoology, 16, 439–512. evaluating and minimizing inference errors. Molecular Ecology, Rousset F (1997) Genetic differentiation and estimation of gene 13, 789–810. flow from F-statistics under isolation by distance. Genetics, 145, Templeton AR, Boerwinkle E, Sing CF (1987) A cladistic analysis 1219–1228. of phenotypic associations with haplotypes inferred from Rozas J, Sánchez-DeI, Barrio C, Messeguer X, Rozas R (2003) restriction endonuclease mapping. I. Basic theory and analysis DnaSP, DNA polymorphism analyses by the coalescent and of alcohol dehydrogenase activity in Drosophila. Genetics, 117, other methods. Bioinformatics, 19, 2496–2497. 343–351. Sanamyan K (2002) ZooBase 2.20. Ascidians. [WW document]. Tokioka T (1953) Ascidians of Sagami Bay. Iwanami Shoten, Tokio. URL http://lithopssoft.com/zoo [accessed on 22 July 2006] Turon X, Nishikawa T, Rius M (in press) Spread of Microcosmus Schneider S, Roessli D, Excoffier L (2000) Arlequin ver 2000. A squamiger (Ascidiacea: Pyuridae) in the Mediterranean Sea. Jour- software for population genetics data analysis. Genetics and nal of Experimental Marine Biology and Ecology. Biometry Laboratory, Department of Anthropology, University Turon X, López-Legentil S (2004) Ascidian molecular phylogeny of Geneva, Geneva. inferred from mtDNA data, with emphasis on the Aplouso- Stach T, Turbeville JM (2002) Phylogeny of Tunicata inferred from branchiata. Molecular Phylogenetics and Evolution, 33, 309–320. molecular and morphological characters. Molecular Phylogenetics Turon X, Tarjuelo I, Duran S, Pascual M (2003) Characterizing and Evolution, 25, 408–428. invasion processes with genetic data: an Atlantic clade of Clavelina Stoner DS (1994) Larvae of a colonial ascidian use a non-contact lepadiformis (Ascidiacea) introduced into Mediterranean harbours. mode of substratum selection on a coral reef. Marine Biology, Hydrobiologia, 503, 29–35. 121, 319–326. Van Name WG (1945) The north and south American ascidians. Stoner DS, Ben-Shlomo R, Rinkevich B, Weissman IL (2002) Bulletin of the American Museum of Natural History, 84, 1–463. Genetic variability of Botryllys schlosseri invasions to the east and Zibrowius H (1991) Ongoing modification of the Mediterranean west coasts of the USA. Marine Ecology Progress Series, 243, 93–100. marine fauna and flora by the establishment of exotic species. Svane I, Young CM (1989) The ecology and behaviour of ascidian Mésogée, 51, 83–107. larvae. Oceanography and Marine Biology: An Annual Review, 27, 45–90. Tarjuelo I, Posada D, Crandall KA, Pascual M, Turon X (2001) Susanna López-Legentil research focuses on biology of marine Cryptic species of Clavelina (Ascidiacea) in two different invertebrates from molecular, ecological and chemical perspectives. habitats: harbours and rocky littoral zones in the north-western Xavier Turon works on ascidian taxonomy and on biology of Mediterranean. Marine Biology, 139, 455–462. benthic invertebrates, including chemical ecology, reproductive Tarjuelo I, Posada D, Crandall KA, Pascual M, Turon X (2004) biology and population genetics. Serge Planes has focused his Phylogeography and speciation of colour morphs in the colonial work on population genetics and evolutionary biology of marine ascidian Pseudodistoma crucigaster. Molecular Ecology, 13, 3125– organisms, with special interest on coral reef fish. 3136.