Introduction

The ascidian Ciona intestinalis has served as a developmental model species for over a century (Chabry 1887; Morgan 1904; Minganti 1948; Sordino et al. 2000). Both C. intestinalis and its congener C. savignyi have invaded benthic communities throughout the temperate zone, necessitating expensive eradication programs (Lambert 2001; Lambert and Lambert 2003). Recently, the genome sequences of C. intestinalis and C. savignyi have become available (Dehal et al. 2002; Vinson et al. 2005), providing tools for comparative genomics and insights into vertebrate evolution (Boffelli et al. 2004; Johnson et al. 2004; Missal et al. 2005). Comparisons of the mitochondrial genomes of C. intestinalis, C. savignyi, and other tunicate species also have answered questions about chordate evolution (Gissi and Pesole 2003; Yokobori et al. 2003, 2005). Despite the historical and current interest in C. intestinalis and its congeners, we know very little about genealogical and phylogenetic relationships within and among members of this genus.

The genus Ciona comprises 12 known species, 4 found in shallow water and 8 found exclusively in deep water (Harant and Vernieres 1933; Van Name 1945; Monniot and Monniot 1977, 1983, 1989, 1990; Hoshino and Nishikawa 1985; Sanamyan 1998). Many of the deep-water species have been described from only one specimen and most are difficult to collect. Here we focus on relationships among the common shallow-water species. Ciona intestinalis is thought to be endemic to northeast Atlantic waters, where it was described (Linnaeus 1767). It is widely distributed in the temperate zone presumably through transport in ballast water and on ship bottoms and now has a range that includes the Pacific Coasts of Japan and North America, New Zealand, Australia, and Chile, the Atlantic Coasts of North America, Europe, and South Africa, the Black Sea and the Mediterranean Sea (Van Name 1945; Kott 1952). Ciona savignyi (Herdman, 1882) is endemic to Japan, but has spread to the Pacific Coast of North America and the Atlantic Coast of Argentina (Hoshino and Nishikawa 1985). Ciona roulei (Lahille, 1887) is found only in the Mediterranean Sea, where it has only been described from five sites along the coasts of France and Spain (Harant and Vernieres 1933). The fourth shallow water species, C. edwardsi (Roule, 1886), is rare and was not included in this study. Although the morphology of several Ciona species has been compared (Hoshino and Nishikawa 1985), no study thus far has examined phylogenetic relationships among these species.

Within C. intestinalis, two “types” have been recognized (Suzuki et al. 2005). Type A inhabits the Pacific Ocean and the Mediterranean Sea, while Type B inhabits the Atlantic Ocean. Type A and Type B are working names agreed upon by the community of Ciona biologists. The two types of C. intestinalis are partially reproductively isolated. Crosses using British Type B C. intestinalis eggs and Japanese Type A C. intestinalis sperm result in normal rates of fertilization, but the reciprocal cross yields many fewer fertilized eggs (Suzuki et al. 2005). While C. roulei and C. intestinalis are clearly distinct on morphological grounds (Harant and Vernieres 1933), crosses between these two species are also known to produce viable offspring. In particular, Mediterranean Type A C. intestinalis sperm × C. roulei eggs produced viable F1 tadpole larvae. However, fertilization fails in the reciprocal cross (Lambert et al. 1990). Ciona intestinalis and C. savignyi are completely reproductively isolated (Byrd and Lambert 2000).

Here we use mitochondrial DNA (mtDNA) sequence data to investigate the genealogical and phylogenetic relationships within and among Ciona species. In particular, we focus on the relationships and amounts of divergence between Type A and B C. intestinalis, and among C. roulei, C. savignyi, and C. intestinalis.

Materials and methods

Sampling

Individuals from three described species (C. intestinalis, C. roulei and C. savignyi) and one unknown species (Ciona sp.) were collected by SCUBA or from docks (Table 1). Ciona intestinalis individuals were collected from five sites on the California Coast of North America, 12 sites on the Atlantic Coast of North America, and 1 site in the Northwestern Mediterranean (Banyuls-sur-mer, France). Additional individuals were obtained from two sites in the Northeast Atlantic (Fiskebäckskil, Sweden and Breskens Harbor, The Netherlands). Ciona roulei individuals were sampled from one site in the Northwestern Mediterranean (Banyuls-sur-mer, France). Individuals of a previously unidentified species (Ciona sp.) were also collected in the harbor at Banyuls-sur-mer. Ciona savignyi individuals were sampled from four sites on the California Coast of North America. Specimens of C. roulei and Ciona sp. were deposited at the American Museum of Natural History. A sequence from Halocynthia roretzi (GenBank Accession No. NC_002177; Yokobori et al. 1999) was used as the outgroup. Sequence data were obtained from 106 of the collected individuals, and the data set contained 107 sequences including the outgroup.

Table 1 Collection sites for Ciona species for which COI sequences were determined
Full size table

DNA extraction, amplification and sequencing

Ovaries were dissected from freshly-collected individuals, cut into several pieces, immediately preserved in dimethyl sulfoxide (DMSO), and ultimately (within 12 days) stored at −80°C until needed. We used ovaries because they are easily removed from the body and yield high-quality DNA. Samples from Sweden and The Netherlands were preserved and shipped in ethanol soon after collection. Upon arrival, the ovaries from these individuals were dissected and preserved in DMSO. Total DNA was extracted from the ovaries using the Qiagen DNeasy® Tissue Kit (Qiagen). Primers to amplify 856 bp of the mitochondrial cytochrome c oxidase subunit I (COI) gene were developed from the consensus sequence of the published C. savignyi and C. intestinalis mitochondrial genomes (Consensus F: 5′GAGTAAGAACTGGRTGRACAGTTTAYCCTCC 3′, Consensus R: 5′ATTAAAACTTAATCTAGTAAAAAGAGGRRATCAATGG 3′). PCR amplification was performed in a 20-μl total reaction volume with 2 mM MgCl2, 0.2 mM dNTPs, 2 μl of 10× buffer, 0.2 μM of each primer, 0.8 U of Taq Polymerase (Gibco-BRL) and 2 μl of template DNA. The PCR protocol was as follows: 35 cycles (95°C for 50 s, 48°C for 1 min and 72°C for 1 min) and a final extension step at 72°C for 7 min, on an OmniGene (Hybaid) thermal cycler. PCR products (diluted up to 1:20 depending on the intensity of the PCR band on an agarose gel) were incubated with 1 μl each of Exonuclease I and Antarctic Phosphatase at 37°C for 30 min, followed by 90°C for 10 min. The products were purified on Sephadex® columns (Sigma-Aldrich). The purified product was sequenced with a Big Dye Terminator Cycle sequencing kit and an ABI-3700 automated sequencer (Applied Biosystems) using the primers listed above. All unique haplotypes (56 sequences) have been submitted to GenBank (Accession Numbers EF209056–EF209110).

Alignment

Sequences were edited and trimmed to 744 bp with SeqMan (DNASTAR) and aligned using the Clustal-W algorithm in MEGA 3.1 (Kumar et al. 2004). Alignments were confirmed visually, and no gaps were present in the final alignment. There is no evidence for nuclear pseudogenes in our dataset. Direct sequencing of PCR products yielded clean sequence and no stop codons were found in the inferred amino acid sequence.

Phylogenetic analysis

Phylogenetic trees were constructed using Maximum Parsimony (MP), Maximum Likelihood (ML), and Neighbor Joining (NJ) methods. Analyses were performed using a data set comprised only of the unique haplotypes (56 sequences).

Parsimony analyses were performed using PAUP* 4.0 (Swofford 2003). Tree space was explored using a heuristic search, with 1,000 replicates of random stepwise addition and TBR branch swapping. A strict consensus was obtained from this heuristic search. Support was assessed using bootstrap analysis (1,000 replicates) with 10 replicates of random sequence addition for each of the bootstrap replicates.

The best-fit model of nucleotide substitution, determined by the hierarchical Likelihood Ratio Test in Modeltest 3.7 (Posada and Crandall 1998), was TIM+G. The TIM+G model was used to generate a ML tree in PAUP* 4.0 (Swofford 2003). Bootstrap support was determined with 250 replicates of a heuristic search strategy, with an as-is addition sequence and TBR branch swapping.

PAUP* 4.0 (Swofford 2003) was also used to perform NJ analyses. Distances based on the TIM+G model were used in the analysis, and confidence in the tree was assessed using 1,000 bootstrap replicates.

Genetic diversity

Average p distances (total number of differences between two sequences divided by total length of sequence compared) within and between sites were calculated using PAUP* 4.0 (Swofford 2003). In addition, distances were estimated based on the TIM+G model. The number of haplotypes and haplotype diversity for each site were determined using DnaSP 4.10.4 (Rozas and Rozas 1999).

Population structure

We characterized population structure within Ciona species and sites using an analysis of molecular variance (AMOVA) in Arlequin 3.0 (Schneider et al. 2000). Pairwise ΦST values between sites were calculated by AMOVA and P-values were obtained by a 110-replicate permutation test.

Results

With the exception of Mediterranean Type A, for which all individuals had the same haplotype, haplotype diversities within collection sites of C. intestinalis were high, ranging from 0.916 to 0.956 (Table 2). Haplotype diversity was similarly high in C. roulei (0.933), and only slightly lower in C. savignyi (0.752). Analyses based on AMOVA suggest that most variation was within sites, and that there was relatively little differentiation among sites within a region (Table 3; Northeast Pacific Type A C. intestinalis ΦST = −0.073 (P > 0.05), Northwest Atlantic Type B C. intestinalis ΦST = 0.103 (P > 0.05), C. savignyi ΦST = −0.099 (P > 0.05)). But within both Type A and Type B C. intestinalis, differentiation existed between regions [Table 3; Type B ΦST = 0.226 (P < 0.0001), Type A ΦST = 0.476, (P < 0.002)].

Table 2 Ciona spp. haplotype diversity [calculated in DnaSP 4.10.4 (Rozas and Rozas 1999)] and number of haplotypes in each species or region within species
Full size table
Table 3 Distribution of variation within and among sites and among regions resulting from AMOVA tests performed in Arlequin 3.0 (Schneider et al. 2000)
Full size table

In contrast to the limited variation among sites of the same species/type, there was substantial differentiation between Type A and B C. intestinalis and between C. intestinalis and C. savignyi. Within C. intestinalis, Type A and B differed in mtDNA sequence by ∼12%. Ciona savignyi exhibited 13.5–16.2% mtDNA sequence divergence from either of the other Ciona species (Table 4).

Table 4 Sequence divergence within and between Ciona species/regions within species
Full size table

Maximum Parsimony and Maximum Likelihood methods produced bootstrap majority-rule consensus trees with identical topologies. Both the 50% majority rule bootstrap MP tree (1,000 replicates; Fig. 1) and the 50% majority rule bootstrap ML tree (250 replicates; not shown) support the monophyly of C. savignyi and place it sister to a C. intestinalis/C. roulei clade. The C. intestinalis/C. roulei clade is divided into two well-supported sub-clades in both trees (Fig. 1). One sub-clade, with 100% bootstrap support, defines Type A C. intestinalis and includes haplotypes from both the Northeast Pacific and the Mediterranean. A second sub-clade, also with 100% bootstrap support, includes Type B C. intestinalis from the Northeast and Northwest Atlantic, together with C. roulei. Thus Type B C. intestinalis is paraphyletic with respect to C. roulei. The NJ tree defines the same major clades revealed by the MP and ML analyses, and clearly shows that interspecific (or inter-type) differences dwarf intraspecific differences (Fig. 2).

Fig. 1
figure 1

Bootstrap majority rule Maximum Parsimony tree generated from the 56 unique haplotypes found in 107 individuals (106 ingroup sequences and 1 outgroup sequence, listed in Materials and methods). Numbers along branches indicate bootstrap percentages. Nodes with bootstrap percentages <50% have been collapsed. Numbers in parentheses beside each site refer to the number of individuals sequenced from each site

Full size image
Fig. 2
figure 2

Neighbor Joining phylogram generated from the 56 unique haplotypes found in 107 individuals (106 ingroup sequences and 1 outgroup sequence, listed in Materials and methods). Numbers along each branch indicate the number of changes along that branch and (in parentheses) the bootstrap percentages. Bootstrap percentages <50% not shown. Numbers in parentheses beside each site refer to the number of individuals sequenced from each site

Full size image

Ciona sp.

Three individuals from Banyuls-sur-Mer Harbor, France, had haplotypes that were distinctive from those seen in all other individuals (Table 4, Figs. 1, 2). Sequence divergences between these haplotypes and haplotypes found in C. intestinalis and C. savignyi ranged from 10.8 to 14.0% (Table 4). These individuals (hereafter referred to as Ciona sp.) were also morphologically distinct from C. intestinalis and C. roulei found in Banyuls-sur-mer. Furthermore, they did not match morphological or ecological descriptions of C. edwardsi (the fourth shallow-water species) from the region (Fiala-Medioni 1974; Copello 1981). The tunics of these individuals were reddish-orange and opaque, and the body clear. In contrast, the C. roulei individuals had clear tunics and deep red bodies, while C. intestinalis individuals (from this region) had clear tunics and milky-white, opaque bodies. The tunics of Ciona sp. were very thin, well-attached, and difficult to remove, in contrast to the tunics of C. intestinalis and C. roulei. The three individuals assigned to Ciona sp. were 3–4 cm long fully extended with tunic intact, which was considerably smaller than C. intestinalis and C. roulei from this region. The buccal and atrial siphons were proportionally longer compared to those of C. intestinalis and C. roulei, and extended half the length of the body. The atrial siphon had six red pigment dots, and the buccal siphon had at least six. Photos of Ciona sp. have been placed on the Dutch Ascidians website: http://www.ascidians.com.

Discussion

Mitochondrial DNA sequence data confirm that C. intestinalis and C. savignyi exhibit substantial genetic divergence. The mtDNA phylogeny also shows C. roulei haplotypes to be embedded within and not clearly distinct from haplotypes found in Type B C. intestinalis. Furthermore, Type A and Type B C. intestinalis are exclusive groups (Figs. 1, 2), with sequence divergence that suggests a most recent common ancestor in the Pliocene, assuming a molecular clock rate of about 3%/my [estimated from mtDNA clocks in three echinoderm genera: Lessios et al. (1999), McCartney et al. (2000), Lessios et al. (2001)].

Suzuki et al. 2005 compared 120 kb of cosmid sequence from British Type B C. intestinalis with the whole genome sequence from a Northeast Pacific Type A C. intestinalis. This revealed a broad range of sequence divergence, with predicted coding regions showing the least divergence. Twenty-four randomly chosen ESTs from British Type B C. intestinalis showed 95.2% average sequence identity to orthologues found in the Japanese Type A C. intestinalis cDNA database (Suzuki et al. 2005). Unfortunately, the selection of genomic regions in these analyses does not allow comparisons with patterns of divergence in other marine invertebrate taxa, because of the difficulty in identifying homologous gene regions. However, mitochondrial COI data allow us to compare distances between Type A and B C. intestinalis to distances between other closely related marine taxa.

We searched the literature for studies that estimated amounts of interspecific and intraspecific mtDNA COI sequence divergence between congeneric marine invertebrate species. These comparisons used a variety of distance corrections, depending on what authors judged to be the most appropriate model of molecular evolution. Only a very few studies provided uncorrected sequence divergence. We calculated an average distance, regardless of correction method. Based on 464 comparisons of 101 nominal species of cnidarians, annelids, arthropods, bivalves, and echinoderms, the average interspecific divergence is 0.159 ± 0.002 (data from (Arndt et al. 1996; Hart et al. 1997; O’Foighil et al. 1998; Hellberg 1998; Metz 1998; Bucklin et al. 1999; Lessios et al. 1999; McCartney et al. 2000; Dawson and Jacobs 2001; Hill et al. 2001; Hurtado et al. 2002; Biermann et al. 2003; Landry et al. 2003; Zigler et al. 2003; Holland et al., 2004; Zigler and Lessios 2004). The average intraspecific difference is 0.109 ± 0.01, based on 80 comparisons of 45 nominal species of cnidarians, arthropods, annelids, bivalves, echinoderms and urochordates (van Syoc 1994; Burton 1995; Edmands et al. 1996; Hoeh et al. 1996; Palumbi et al. 1997; Hart et al. 1997; O’Foighil et al. 1998; King et al. 1999; Medina et al. 1999; Lessios et al. 1999; Lee 2000; Tarjuelo et al. 2001; Hill et al. 2001; Holland et al., 2004; Hurtado et al. 2004; Tarjuelo et al. 2004; Glover et al. 2005; Jolly et al. 2005). The average TIM+G corrected distance between Type A and B C. intestinalis (0.2, Table 4) is twice as high as the average corrected intraspecific distance for marine invertebrates, and is even higher than the average corrected interspecific distance. Although the intraspecific comparisons from the literature may be biased towards species with substantial differentiation between sites, correcting such a bias would only increase the disparity between C. intestinalis and other marine invertebrates. These data highlight the substantial differentiation between Type A and B C. intestinalis compared with intraspecific differentiation in other marine invertebrates.

Several other lines of evidence also point to substantial differentiation between Type A and B C. intestinalis. Hoshino and Nishikawa (1985) recognized differences between C. intestinalis from Naples and Japan (Type A) and the Northeast Atlantic (Type B) in the rate of change with growth of the number of inner longitudinal vessels. In addition, British Type B C. intestinalis and Japanese Type A C. intestinalis are partially reproductively isolated. Although crosses between British Type B C. intestinalis eggs and Japanese Type A C. intestinalis showed normal fertilization rates (as estimated by the frequency of “normal two-cell stage embryos”), the reciprocal cross yielded many fewer fertilized eggs (Suzuki et al. 2005).

Given this genetic, morphological and reproductive evidence, we tentatively advocate the separation of C. intestinalis into two species according to the phylogenetic species concept. This cryptic speciation in C. intestinalis is consistent with recent studies of other marine taxa in which molecular data uncover cryptic species previously hidden by anthropogenic transport or morphological similarity (Dawson and Jacobs 2001; Holland et al. 2004).

Ciona roulei is not a monophyletic group with respect to mtDNA haplotypes; some C. roulei haplotypes are more closely related to Type B C. intestinalis haplotypes than to other C. roulei haplotypes. Crosses between C. roulei and Mediterranean Type A C. intestinalis are successful in one direction (Lambert et al. 1990), leading the authors to suggest that C. roulei and C. intestinalis diverged more recently than C. intestinalis and C. savignyi (between which hybrids have not been produced). This distinctness of C. savignyi is strongly supported by the mitochondrial COI data. The position of C. roulei nested within the larger Type B C. intestinalis clade suggests that C. roulei was recently derived from Type B C. intestinalis. Phylogenetic analyses lend support to a hypothesis that Northeast Atlantic Type B C. intestinalis individuals moved into the Mediterranean Sea, where they diverged from the ancestral Type B C. intestinalis, giving rise to C. roulei. Reproductive compatibility between C. roulei and Type B C. intestinalis has not been investigated; successful crosses between C. roulei and Type B C. intestinalis would reinforce the Northeast Atlantic origin of C. roulei’s ancestors.

The substantial intraspecific divergence in C. intestinalis, the placement of C. roulei within the Type B C. intestinalis clade, and the phylogenetically distinct Ciona sp. highlight the evolutionary and taxonomic ambiguities within the genus Ciona. Type A and B C. intestinalis are 12% divergent at the mitochondrial COI locus, and yet they exhibit partial reproductive compatibility. These two types are sympatric in southern England, and exhibit pre-zygotic reproductive isolation at this site (P. Sordino, personal communication). Similarly, mtDNA COI sequences differ by 12.5% between the sympatric C. roulei and Mediterranean Type A C. intestinalis, but viable F1 hybrids are produced from one of the two reciprocal crosses. Although they occur along the same coastline, this pair of species may be ecologically isolated. Nothing is known about reproductive isolation between C. roulei and Type B C. intestinalis. Differences in compatibility between individuals from allopatric and sympatric sites of these two species will help clarify the evolutionary processes generating reproductive isolation in Ciona.

While successful crosses between other pairs of ascidian species have been well documented in the laboratory (Lambert et al. 1981; Jeffery and Swalla 1990), hybrids between species have never been identified in nature. This lack of natural hybridization suggests that pre-zygotic isolating mechanisms may have a large role in the maintenance of reproductive isolation in ascidians. Even so, genetically distinct ascidian species such as C. intestinalis and C. savignyi have been confused historically (Hoshino and Nishikawa 1985), raising the possibility that hybrids could be morphologically cryptic as well. These uncertainties necessitate further study of species level relationships within Ciona in the context of improved understanding of intrinsic barriers to gene exchange.