This is an Open Access article Special issue “Proceedings of the 2nd International Invasive Sea Squirt Conference”

A colonial tunicate belonging to the genus Didemnum has recently been found in many temperate coastal regions throughout the world, as well as large areas of Georges Bank in the NW Atlantic. It continues to spread rapidly and compete aggressively with native, hard substrate species (e.g., mussels, barnacles, bryozoans, other ascidians). In addition, it can form dense mats on deepwater cobble-gravel substrates and influence the abundance and species composition of benthic epifauna and infauna. Thus, its ever-increasing presence is creating potentially severe detrimental economic and ecological impacts. This invasive species, referred to in recent publications as Didemnum sp. A, has been misidentified as five previously described species native to the regions where Didemnum sp. A has been discovered and has been described as two new species based solely on morphological characteristics. There are relatively few diagnostic characters and a great deal of variability in the relevant characters, making the task of identification very difficult. Adding to the confusion has been the widespread and apparently disjunct distribution of the species. Here, we present molecular data on both mitochondrial and nuclear genes from colonies sampled from Europe, east and west coasts of North America, Japan, and New Zealand. These data strongly indicate that Didemnum sp. A is a single species, possibly native to the northwestern Pacific Ocean, that has become established globally. Considering genetic and morphological evidence, the most appropriate name for this species is Didemnum vexillum Kott, 2002.


Introduction
The identification and control of invasive species, both terrestrial and marine, are of increasing concern around the globe because of the economic and ecological toll exotic species can exact on their new habitats (e.g., Naeem et al. 1994;Pimentel et al. 2000).Invasive tunicates are often important spatial competitors and can alter community structure and local biodiversity (e.g., Stachowicz et al. 2002;Dijkstra et al. 2007;Osman and Whitlatch 2007).They can also cause economic damage by overgrowing aquaculture equipment (USGS 2008).
Over the last three decades, previously unrecorded populations of superficially similar colonial tunicates of the genus Didemnum have been discovered in many temperate coastal regions throughout the world (Minchin and Sides 2006;Bullard et al. 2007;Coutts and Forrest 2007;Gittenberger 2007;Lambert 2009).The invaders grow rapidly and are aggressive competitors with other hard substrate species both native and exotic (e.g., mussels, barnacles, bryozoans) and on both artificial and natural substrates (Figure 1a, Minchin 2007;Osman and Whitlatch 2007;USGS 2008).These tunicates are also of particular concern because of their unique ability to overgrow and cement cobble/gravel substrates together (Figure 1b).Currently, over 230 km 2 of Georges Bank are covered with mats of Didemnum sp.A, causing concerns about potential impacts on food availability for economically important juvenile fishes (Valentine et al. 2007), and on other organisms residing in these habitats (Mercer et al. 2009).
Identification of this invader, or even the determination of how many species are involved, has been problematic.Various populations have been identified under as many as five different, previously described Didemnum species (D. lahillei Hartmeyer, 1909 -France;D. helgolandicum Michaelsen, 1921 -France and northern Europe;D. lutarium Van Name, 1910 -New England;D. carnulentum Ritter and Forsyth, 1917 -California and Washington;and D. pardum Tokioka, 1962 -Japan) as well as two new species described from New Zealand and New Hampshire (Didemnum vexillum Kott, 2002 and Didemnum vestum Kott, 2004, respectively) (Lambert 2009).However, morphological re-examination of material collected around the world and comparison with the type specimens of several of the abovementioned species suggests that all of the recently discovered populations are in fact the same species (Lambert 2009), nominally designated Didemnum sp.A, and are not any of the previously described species suggested.Part of the confusion surrounding the identity of this species is due to the difficulty in identifying any species belonging to the genus Didemnum, since species of the genus show high morphological similarities.Many morphological characters are difficult to study due to the small size of zooids, larvae, and spicules.Poor preservation techniques, inadequate sampling, intra-and intercolony variability, and even timing of collection during non-reproductive periods can obscure or eliminate many important characters, making identification difficult.
Molecular markers are therefore useful where there has been confusion using standard taxonomy because they are independent of the morphological characters in question (e.g.Tarjuelo et al. 2001;Nydam and Harrison 2007).Thus, samples that cannot be identified from morphological characters, such as lack of larvae or reproductive structures, can still be usable for molecular studies.Furthermore, molecular data from any one sample can be precisely recorded and stored for subsequent comparison with other samples at any other time.Finally, a great number of DNA sequences are publicly available (http://www.ncbi.nlm.nih.gov),allowing selection of markers with a level of interspecific variation that suit the required taxonomic resolution (e.g., rapidly evolving genes for intra-specific studies, slowly evolving genes for comparisons at the species-level or higher).Due to the difficulty involved with identification of didemnids in general and the confusion surrounding the identity of Didemnum sp.A in particular, our objective was to develop and apply molecular markers for use as an additional character in order to identify the species Didemnum sp. A. In this study, we used a nuclearencoded gene (tho2) and a mitochondrialencoded gene (coI) as independent characters to test the morphologically-derived hypothesis of a single, broadly distributed species and determine what, if any, described species is genetically consistent with Didemnum sp. A.

Sample collection and DNA isolation
Colonies of Didemnum sp.A were sampled from around the world (Annex 1), including specimens of D. vexillum from New Zealand and two specimens from the type location of D. vestum (Kott 2002(Kott , 2004)).Additionally, samples of three congeneric species were collected: Didemnum albidum Verrill, 1871 from the Gulf of Maine; Didemnum psammathodes Sluiter, 1895 from Panama; and a species designated Didemnum sp.B from New Zealand (see Annex 1).Colonies were collected subtidally from both natural and artificial substrates either via SCUBA or by pulling up ropes and other artificial structures hanging off of docks.Samples from Georges Bank were collected with dredges.Samples were preserved either directly in 95% ethanol or first in 70% isopropanol and then transferred to 95% ethanol.Fifty to one hundred thoraces were dissected from each colony and stored in 95% ethanol at -20°C until DNA isolation.Only the thoraces were used for three reasons: 1) to avoid contamination from including gut contents, 2) to remove most of the tunic which is primarily an acellular matrix of cellulose, and 3) to avoid larvae which could cause confusion with nuclear genes.Because Didemnum sp.A is capable of fusion (pers.obs.), thoraces were taken from a small area of each colony (<1 cm 2 ) to avoid sampling two genomes from a single colony.Forty-eight of the seventyone invasive didemnid samples analyzed in this study were identified as Didemnum sp.A by G. Lambert before any molecular work was undertaken; the remainder of the samples were identified by L. Stefaniak.Voucher specimens for all samples sequenced have been deposited with the National Museum of Natural History Naturalis, Leiden, The Netherlands.
Ethanol supernatant was removed carefully from the tunicate thorax samples using a pipette, and the remaining ethanol was evaporated by heating at 55°C for 5-10 minutes.Thoraces were then resuspended in 200μl DNA extraction buffer (0.1 M EDTA, 1% sodium dodecyl sulfate, and 40 ng proteinase K [Invitrogen, Carlsbad, CA]) and incubated at 55°C for 72 hours.Total DNA was extracted using a CTAB method, following Zhang and Lin (2005), with slight modifications.DNA was isolated by adding 33 μl each of 5 M NaCl and 10% cetyltrimethylammonium bromide (CTAB, Sigma, St. Louis, MO) in 0.7 M NaCl and incubating at 55°C for 10 minutes, followed by one chloroform extraction.The DNA was purified from the aqueous layer using DNA Clean and Concentrator-25 columns (Zymo Research, Orange, CA), eluting twice each with 30 μl 10mM Tris, pH 8.0.DNA quality and quantity were measured using a Nanodrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE).

RNA isolation, cDNA library construction, and gene screening
Portions of live Didemnum sp.A colonies (<1 cm 2 ) were obtained from Avery Point, CT, USA and washed three times with 300 ml of 0.22-μmfiltered, autoclaved sea water and dried in a clean paper towel.The entire sample, zooids and tunic, was cut into small pieces and immersed into 1 ml of Trizol solution (Invitrogen, Carlsbad, CA), and total RNA was isolated according to Zhang et al. (2006).The cDNA library was constructed following the manufacturer's instructions of the SMART cDNA Library Construction Kit (Clontech, Mountain View, CA), with some modification.The Polymerase Chain Reaction (PCR)-amplified cDNA library was purified using a DNA Clean and Concentrator column (Zymo Research, Orange, CA), directly ligated into a T-vector, and transformed into bacterial competent cells as in Zhang and Lin (2002).Forty-eight of the resultant colonies were randomly picked up and their cDNA plasmids were isolated.Eight clones with insert DNA size of >500 bp were sequenced according to Zhang and Lin (2002).Sequences obtained were used to query GenBank databases, using BLAST to identify genes.

Primer design for coI and tho2
Previous attempts to amplify coI from Didemnum sp.A using the universal primers (Folmer et al. 1994) were unsuccessful (B.Swalla and J. Collie, pers. comm.).In this study, we designed a set of tunicate-specific primers.Tunicate sequences were obtained from GenBank and aligned using the ClustalW Multiple alignment function in BioEdit.Primer sites were selected by hand and tested for self-dimmers, pair-dimmers, and primer hairpins in the program PrimerSelect 4.03 (DNASTAR, Madison, WI).Specificity of the primers for tunicates was determined by BLAST analysis of the primers against GenBank to make sure that they did not match the coI sequences of other taxa.One pair of primers was selected for further analysis (Table 1).A tho2 gene homolog was obtained by cDNA library screening Several sets of primers were designed using Beacon Designer 3.0 (PREMIER Biosoft, Palo Alto, CA) (Table 1) and tested for specificity to Didemnum against genomic DNA from Didemnum sp.A and some phytoplankton.The primer set DidTho2 F3-R3 proved to amplify a 1-kb PCR fragment specifically yet efficiently from Didemnum genomic DNA and was selected for further analysis.

PCR and Sequencing
PCR amplifications were carried out in 25-μl reactions containing 0.2 mM each dNTP, 0.2 μM each primer, 1x buffer, 2.0 mM MgCl 2 , 0.5 units of ExTaq polymerase (Takara Bio USA, Madison, WI), and 23-185 ng total DNA.Cycling conditions for coI consisted of an initial denaturing step of one minute at 94°C, followed by 60 cycles of 10-second denaturing at 94°C, 30-second annealing at 50°C, and 50-second elongation at 72°C, with a final elongation of 10 minutes at 72°C.To amplify tho2, the initial denaturing step was followed by 35 cycles of 20-second denaturing at 94°C, 30-second annealing at 56°C, and 1-minute elongation at 72°C.PCR products for coI were purified using DNA Clean and Concetrator-5 columns (Zymo Research, Orange, CA) with a single 8-μl elution by 10 mM Tris pH 8.0 and were then directly sequenced in both directions using the PCR primers with Big Dye v 3.1 Terminator chemistry and analyzed on an ABI 3130 automated sequencer (Applied Biosystems, Foster City, CA) at the University of Connecticut COMMA sequencing facility, Groton, CT.Because tho2 is a nuclear gene with potentially multiple alleles in a single sample, PCR products for the tho2 amplifications were cloned into a T-vector and transformed into competent cells -1 blue (Stratagene, La Jolla, CA).Eight of the resultant colonies were randomly selected and their plasmids were isolated (Zhang and Lin 2002).Insert length was analyzed by digesting plasmids with EcoRI and HindIII and running electrophoresis on a 1% agarose gel.Four plasmids from each PCR product were sequenced using the T3 universal primer with Big Dye v3.1 Terminator chemistry on an ABI 3730 automated sequencer at the DNA Analysis Facility on Science Hill at Yale University, New Haven, CT.Because tho2 PCR amplifications for D. psammathodes, D. albidum, and Didemnum sp.B produced a number of different sized products, up to 24 clones were screened and sequenced for each species.Clones of tho2 were identified via BLAST against GenBank.Base calls for all sequences were checked by hand and sequences deposited in GenBank (coI Accession numbers: EU419401-EU419459, EU742661-EU742677; tho2 accession numbers: EU419460-EU419569).Single nucleotide polymorphisms found in only one of all tho2 clones sequenced (>200 clones) were considered to be artifacts derived from PCR error.

Data analysis
Sequences for coI and tho2 were aligned separately using ClustalX.Both analyses included sequences from Didemnum sp.A, Didemnum sp.B, and D. psammathodes.Sequences from D. albidum were also included in analyses of coI.For use as an out group in constructing phylogenetic trees, coI sequence for Diplosoma spongiforme was obtained from GenBank (Accession number AY600972).Both introns and exons of tho2 aligned among Didemnum sp.A samples.However, while the locations of the introns in the amplified fragment of tho2 were conserved among all Didemnum species in this study, the sequences and the lengths of the introns varied dramatically among species, making alignment difficult.Therefore, only the exon regions (227 bp) were used in this analysis.Maximum likelihood trees were generated with PhyML (Guindon and Gascuel 2003) using 1000 bootstrap replicates.Two nucleotide substitution models were used based on the results of modeltest (Posada and Crandall 1998): HKY85 (Hasegawa et al. 1985) and GTR (e.g.Lanave et al. 1984).For both models, the proportion of invariable sites was estimated, and for the HKY85 model, the transition:transversion ratio was also estimated.
Sequences of coI for Didemnum sp.A were collapsed in haplotypes using DNASP v4.0 (Rozas et al. 2003) The first subgroup containing samples from only a single sampling location is more divergent from the other Didemnum sp.A samples, however, the second subgroup does not reveal any exclusive geographic pattern at the present time, as samples from all but one of eight locations were also found in the third subgroup.
tho2 sequences and phylogenetic trees Each colony of Didemnum sp.A produced one, two, or three alleles from the exons of tho2.Three alleles from a single colony are probably the result of accidentally including some abdomen tissue, fertilized eggs, or larvae in the DNA extraction.While there were four sequences from each sample, only one copy of each unique allele per sample was included in the analysis, resulting in 108 alleles of tho2 from 57 colonies of Didemnum sp. A. Additionally, one D. psammathodes and one Didemnum sp.B allele were amplified and sequenced.No sequences of tho2 were found from the total of 48 D. albidum clones screened.
All of the tho2 sequences from samples identified as Didemnum sp.A grouped together in a single clade with 68% bootstrap support (Figure 4).Within the Didemnum sp.A clade, sequences from different geographic locations, including those identified as D. vexillum, formed several small (2 -10 sequences) subgroups; however, these came with short branches and no bootstrap support, in general.In contrast, the clade was well-separated from Didemnum sp.B and D. psammathodes.

Discussion
The strong grouping of Didemnum sp.A and D. vexillum for both coI and tho2 suggests that they are most likely the same species and that they are distinct from the other superficially similar and sympatric species included in the analysis.Two samples from the type locality of D. vestum (collected the same year as the D. vestum holotype), which were morphologically identified as conspecific to Didemnum sp.A, also fall into the same clade as Didemnum sp.A and D. vexillum for both genes.This suggests that all three species are, in fact, a single species, particularly given that the published descriptions of D. vexillum and D. vestum overlap to a large extent and that D. vestum was described from poorly preserved colonies that contained no larvae (Kott 2002(Kott , 2004)).Morphological evidence based on examination of numerous samples including zooids and larvae also supports this grouping (Lambert 2009, this issue).Morphological and molecular data do not justify the specific separation between Didemnum sp.A and D. vexillum Kott 2002.Therefore, Didemnum sp.A, to date, must be designated as D. vexillum Kott 2002(Lambert 2009, this issue).
Additionally, a comparison of 464 congeneric pairs of marine invertebrates produced an average interspecific divergence in coI of 15.9±0.2%(Nydam and Harrison 2007).This estimate may be low because of the inclusion of cnidarians, which generally have a lower interspecific divergence than other animals (Hebert et al. 2003); however, the amount of congeneric interspecific divergence between D. vexillum and the other species analyzed in this study (with the exception of D. psammathodes) is well above the average (Table 2) and D. psammathodes is highly distinct in both morphology and ecology to D. vexillum (Monniot 1983).Because of the minimal sampling in this study of non-D.vexillum species and D. vexillum from each of the study localities, the comparison of sequence variation within D. vexillum to that between D. vexillum and other Didemnum species alone may not be adequate to confirm that D. vexillum is a single and distinct species.However, the phylogenetic evidence here presented and morphological evidence (Lambert 2009) strongly supports D. vexillum as a single species.The recent discovery (<20 years ago) of D. vexillum in many parts of the world and the highly   identical gene sequences suggests that this species has become globally distributed recently, probably as a result of a human-mediated invasion (Lambert 2009).
The initial populations of an invasion can undergo a phenomenon known as a founder effect: a genetic bottleneck caused by only a small number of individuals from the native range taking part in the invasion (Sakai et al. 2001;Wares et al. 2002;Lindholm et al. 2005).Thus, invasive populations should have relatively little genetic diversity compared to the source population (Garnatje et al. 2002;Dawson et al. 2005; but see Roman 2006 for a recent example of the effects of multiple invasions on signals of founder effects).Intraspecific variation in regions of coI congruent to the region amplified in D. vexillum in this study ranges from approximately 4-26% in other species of ascidians (Table 3).At 5.1% intraspecific variation, D. vexillum falls within the lower limit of this range.However, in the study that produced the lowest intraspecific variation (3.8%), samples of Pseudodistoma crucigaster were taken from a relatively small geographic distribution (<100 km on the Mediterranean coast of Spain and two offshore islands, Tarjuelo et al. 2004).In studies with larger range sampling, intraspecific variation is approximately 26% (Lopez-Legentil and Turon 2005;Lopez-Legentil et al 2006).A comparison of the intraspecific variation in coI of 45 species of marine invertebrates produced an average intraspecific variation of 10.9±1% (Nydam and Harrison 2007), which, again, is probably low because of the inclusion of cnidarians (Hebert et al. 2003).Finding low intraspecific variation in groups that tend to have a large amount of intraspecific variation could indicate that some of the populations are invasive.For example, Turon et al. (2003) found only 8.7% variation in Clavelina lepadiformis (Ascidiacea, Aplousobranchia) populations on both the Atlantic and Mediterranean coasts of Spain but determined that several harbor populations on the Mediterranean side were, in fact, introduced from the Atlantic.For a sessile organism such as D. vexillum, with short-lived, non-feeding larvae that could limit gene flow, to have a low genetic variation (based on coI) on a global scale, the most likely explanation is recent invasion in multiple regions, an explanation that is supported by the timeline of discovery of D. vexillum populations world wide (Lambert 2009).
It must be noted that the amount of interspecific variation between D. vexillum and D. albidum (Table 2) is only marginally greater than the intraspecific variation found in the colonial ascidian Botyrllus schlosseri (Lopez-Legentil et al. 2006) and the interspecific variation between D. vexillum and Didemnum sp.B and between D. vexillum and D. psammathodes are much less.This molecular evidence alone could suggest that all four Didemnum spp.are conspecific.However, morphological evidence shows them to be highly distinct.Additionally, the amount of congeneric interspecific divergence can vary greatly within groups (Hebert et al. 2003).Botryllus schlosseri belongs to a different suborder than Didemnum and can be expected to have a somewhat different rate of mutation.Given that the interspecific variation  Turon (2008) in the genus Didemnum produced here is well above the average produced by Nydam and Harrison (2007), it is safe to conclude that the three superficially similar species are distinct; the interspecific variation between D. vexillum and D. psammathodes is less than the average produced by Nydam and Harrison (2007), however, the large morphological and ecological differences between these two species suggest that they are separate species.Finally, the large (16.6%) intraspecific variation found in D. albidum compared to Didemnum sp.B and D. vexillum can be explained in two ways.The difference between D. albidum (16.6%) and Didemnum sp.B (0%) may be due to a sampling artifact.Both samples of Didemnum sp.B were collected at the same site, increasing the possibility that the two samples are closely related, while the two D. albidum samples were collected at sites over 100 kilometers apart.The distance between the D. albidum sites is relatively large compared to larval lifespans in colonial ascidians (e.g.Jackson 1986, Svane andYoung 1989).As a result it is possible that considerable genetic differentiation has built up between these two populations of the native D. albidum.As a recently introduced species, D. vexillum is less likely to have accumulated as many mutations among populations.Also, while many colonies of D. vexillum were sampled globally, relatively few samples were collected within each region; it is possible that not all of the available genetic diversity in D. vexillum has been uncovered (e.g., Muirhead et al. 2008).
Identification of native or source ranges of species has important ecological implications.Understanding the ecology of an invasive species in its native range can help predict interactions with native biota and the environment in introduced ranges and determine effective modes of biological control (Wares et al. 2005).Information on the native range of an invasive species or even the definitive identification of a species as invasive can also influence management decisions.Incomplete characterization of the fauna of a region and/or records of baseline biotas before and during historical invasions can lead to species being erroneously defined as either native or invasive (Carlton 1996;Kott 2004).Improper designation of an invasive species as native or cryptogenic can delay or reduce nationally-coordinated efforts to eradicate or manage species that then become costly pests (e.g., D. vexillum in New Zealand, Coutts and Forrest 2007).
The globally low coI intraspecific variation found in the present study could suggest that the native range of D. vexillum has yet to be sampled.However, our sampling scheme was aimed at comprehensive global coverage with less than twenty samples per region.Except in areas that underwent a severe bottleneck event during the invasion process, it is unlikely that we have sampled the full range of genetic diversity in any given region.The earliest records of a didemnid tunicate exhibiting a combination of the lobe/tendril-forming growth form and other morphological characters peculiar to D. vexillum are from 1926 in Japan (Nishikawa 1990).That sample was unfortunately only identified to genus at the time and later identified as Didemnum pardum by Nishikawa (1990), but it is now considered to be conspecific to D. vexillum (Nishikawa, pers. comm., Lambert 2009).The greatest regional number of coI haplotypes was found in Japan (eight haplotypes from thirteen samples); however, many more samples will need to be analyzed from the entire current range of D. vexillum before the question of where is the native range of D. vexillum can be adequately addressed.Highly variable molecular markers such as microsatellites or introns may also be useful in that attempt.

Figure 3 .
Figure3.Maximum likelihood tree of Didemnum spp.coI (586bp).One thousand bootstrap replicates were run using the GTR nucleotide substitution model.All nodes with greater than 50 percent bootstrap support are labeled.For Didemnum vexillum samples, sequences are labeled as 'Sampling location sample number.'For all other species, species name is also included.

Figure 4 .
Figure 4. Maximum likelihood tree of Didemnum spp.tho2 exons (227bp).One thousand bootstrap replicates were run using the GTR nucleotide substitution model.All nodes with greater than 50 percent bootstrap support are labeled.For Didemnum vexillum samples, sequences are labeled as 'Sampling location, sample number -clone number.'For all other species, species name is also included.

Table 1 .
Primer sequences for coI and tho2 designed and used in this study.

Table 2 .
. Haplotypes from Didemnum sp.A were subsequently compared to coI sequences from D. albidum, D. psammathodes, and Didemnum sp.B to determine the percentage of variable sites in co1 between Didemnum sp.A and other species in the genus.Seventy-one coI sequences were generated from Didemnum sp.A colonies, one from D. psammathodes, and two each from D. albidum and Didemnum sp.B colonies.The 71 Didemnum sp.A samples analyzed in this study produced 10 unique coI haplotypes.Thirty variable sites were found in the 586 bp coI fragment from the 18 Didemnum sp.A haplotypes (Figure 2).Haplotypes of coI from Didemnum sp.A averaged 26.8% variable sites in the 586 bp fragment compared to D. albidum, 12.3% variable sites compared to D. psammathodes, and 17.7% variable sites compared to Didemnum sp.B (Table 2).Similar tree topologies resulted from the two substitution models.Branches with bootstrap support over 50% were the same between models, though the location of branches within a clade differed somewhat.Therefore, only the GTR model trees are presented here.All Didemnum spp.and D. psammathodes sequences formed a monophyletic clade to the exclusion of D. albidum sequences.The D. albidum clade branched off earlier than the Didemnum spp./D.psammathodes clade and had strong support (100%).The Didemnum sp.A Percentage of variable sites among Didemnum spp.for the 586 bp fragment of coI amplified in this study.

Table 3 .
Intraspecific variation of coI in various colonial ascidian species.
Summary of sampling locations, approximate geographic coordinates, sampling date, museum reference numbers, and sequence accession numbers.Haplotype number is also listed for Didemnum sp.A and D. vexillum samples.Didemnum sp.A from D. vestum type location