In marine environments, biological introductions exert important pressure on natural ecosystems. Introduction of non-indigenous species (NIS) has occurred at an increasing rate since the 20th century, in pace with the increasing range and intensity of vectors (Nunes et al., 2014). Because successful eradications of invasive species in marine systems are very rare (Sambrook et al., 2014; Ojaveer et al., 2015), both pre-border prevention measures (e.g., limiting the introduction of NIS by ballast waters) and early detection focused on introduction hotspots (e.g., marinas and harbors) have been recommended (Ojaveer et al., 2014). Accurate identification is a pre-requisite for the implementation of policies and regulations such as the Marine Framework Strategy Directive (MSFD), in which Descriptor 2 is dedicated to NIS, with one criterion involving the detection and identification of new NIS, and another concerning the changing spatial extent of already-reported NIS.

However, identifying new or recently introduced NIS can be challenging when based only on traditional methods (Darling et al., 2017), such as rapid assessment surveys (RAS), which rely heavily on field recognition of species and have been commonly used to survey NIS in marinas and harbors (e.g., Cohen et al., 2005; Arenas et al., 2006; Campbell et al., 2007; Bishop et al., 2015a; Lehtiniemi et al., 2015). Many marine animal NIS in introduction hotspots belong to taxonomic groups, such as bryozoans, hydrozoans and tunicates, that require substantial taxonomic expertise. In addition, the development of molecular studies in the last three decades has revealed a large number of cryptic species (i.e., species that are not distinguishable based on morphological traits) in these taxonomic groups (Knowlton, 2000; Appeltans et al., 2012). Consequently, an increasing number of studies have recommended the use of molecular tools to complement the traditional methods, and thus achieve reliable taxonomic identification of marine NIS (Comtet et al., 2015; Darling et al., 2017; Dias et al., 2017).

The usefulness of the molecular barcoding approach is well established. The benefits of the approach have been demonstrated by numerous studies (for a review, see Comtet et al., 2015). Such an approach can be used to ascertain the presence of a new NIS (e.g., Asterocarpa humilis: Bishop et al., 2013), to reveal false morphology-based NIS identification (e.g., Crepidula spp.: McGlashan et al., 2008) or to determine the taxonomic status of previously unrecognized NIS (Ordóñez et al., 2016). However, the robustness of the molecular barcoding approach relies on some important pre-requisites, notably the availability of reliable reference sequence data for the targeted species. Thus, the prompt updating of databases is essential following taxonomic revision and discovery of unrecognized cryptic species. The sequences delivered in public databases such as GenBank (Benson et al., 2013) or BOLD [Barcode of Life Data System; (Ratnasingham and Hebert, 2007)] should thus be error-free and validated by taxonomic experts. BOLD is expected to provide such a framework. However, the public portal also retrieves non-curated sequences from Genbank that can propagate errors. There is so far a limited number of initiatives specifically aimed at delivering reference data for marine NIS. Although rare, such initiatives should be strongly encouraged, notably to support international regulations and policies such as the MSFD in Europe (Darling et al., 2017). Such an initiative has recently been undertaken in Western Australia, where a reference collection (vouchers, DNA and sequences for one marker (COI)) was assembled for 75 species (out of 79) of the “Western Australian Prevention List for Introduced Marine Pests” (Dias et al., 2017). Similarly, facing the issue of cryptic diversity, integrative taxonomy coupling molecular phylogenetics with morphology and other species attributes (e.g., ecology, life-history traits) is needed (Pante et al., 2015).

We here illustrate the issues described above using the colonial tunicate genus Botrylloides as a particularly relevant case study. At least two Botrylloides species, B. violaceus Oka, 1927 and B. diegensis Ritter and Forsyth, 1927, with the recent addition of B. giganteus (Pérés, 1949) (Rocha et al., 2019) are globally invasive. B. violaceus and B. diegensis are both native to the N Pacific. While the former is native to the NW Pacific, there is more uncertainty regarding the native range of the latter: although B. diegensis was originally described from the NE Pacific (southern California), it is likely that the species had been introduced from the western or southern Pacific (Carlton, 2009). The two species are important members of the fouling community colonizing artificial substrates on the Pacific coast of the United States, for instance in harbors and marinas (e.g., Nydam and Stachowicz, 2007). The same two species have also been introduced in Europe, notably in the English Channel (EC), where they are well established in marinas of the United Kingdom and N France (Bishop et al., 2015a, b). In Europe, one putatively native species, B. leachii (Savigny, 1816), is also recognized, often showing coloration somewhat similar to the two-tone color pattern seen in B. diegensis (Supplementary Figure S3, panel 6). The original description (Savigny, 1816) of B. leachii was based on material “Communiqué par M[onsieur]. Leach, directeur du Musée britannique” (i.e., William Elford Leach, then working in the Department of Natural History of the British Museum) and the specimen was thought by Savigny to have originated probably from the English coast (“Habite les côtes de l’Angleterre?”). Conversely to the introduced species, the native taxon seems uncommon in artificial habitats in some parts of the EC (J. Bishop, L. Lévêque, and F. Viard, pers. obs.).

The first reports of non-indigenous Botrylloides spp. in the English Channel were made after RAS in 2004 (Arenas et al., 2006). A single species, B. violaceus, was reported, recognized by its single-colored colonies and large and morphologically distinctive brooded larvae. Colonies of B. violaceus could occur in a range of single colors, including violet, cream, yellow (Supplementary Figure S3, panel 4), brick-red and (commonly) orange. A second species was encountered during the 2004 surveys, but was not reported by Arenas et al. (2006) since its identity was uncertain (Bishop et al., 2015a, b). This second species showed a strong and distinctive two-colored pattern featuring a broad ring of solid pigmentation (commonly orange or cream) surrounding each buccal orifice, against a contrasting darker background (Supplementary Figure S3, panel 2), and was subsequently identified as B. diegensis. This distinction, based on color morphs, was used by Lambert and Lambert (2003) to distinguish B. violaceus and B. diegensis in southern California, United States. In 2011, the present authors and colleagues collected extensive samples on both sides of the English Channel for an intended population-genetic study of B. violaceus; single-color colonies were thus collected, with or without confirmation of the distinctive but seasonally brooded larvae of that species, under the assumption that these would all be the target species. Preliminary molecular analyses revealed that these single-color colonies included specimens attributable to B. diegensis, so that B. diegensis had been misidentified in the field as B. violaceus on the criterion of possessing single-colored, rather than two-colored, colonies. B. diegensis in the English Channel could thus occur in a single-color morphotype (e.g., Supplementary Figure S3, panel 1) easily confused with B. violaceus, even by those supposedly familiar with the taxa. In addition, during our surveys some color morphs believed to belong to B. leachii were encountered that had a two-color pattern similar to the two-color pattern of B. diegensis, although differing in detail (Supplementary Figure S3, panel 6).

Following from these preliminary observations and molecular results, in this study we examined a large number of Botrylloides colonies displaying a single-color pattern, using sequencing data to make an in-depth assessment of the association between color-morph and species. We chose to analyze a fragment of the mitochondrial gene region cytochrome oxidase I (COI) which has been previously investigated to study B. violaceus in its N American introduced range (Bock et al., 2011; Lejeusne et al., 2011), besides being a marker commonly used in barcoding studies of marine invertebrates (Comtet et al., 2015) and a standard for BOLD (Hebert et al., 2016). This dataset was completed with data from specimens kindly provided by taxonomic experts (see acknowledgment section); this allowed us to get reliable reference sequences for the three Botrylloides species, B. violaceus, B. diegensis and B. leachii, putatively present in the EC. We further analyzed if similar confusion existed in the Mediterranean, where six Botrylloides species have been reported: B. leachii (native), B. violaceus (introduced), B. pizoni Brunetti and Mastrototaro, 2012 (likely a synonym of the invasive B. giganteus; Reem et al., 2018), B. niger Herdman, 1886 (likely a Lessepsian migrant), B. anceps (Herdman, 1891) (likely a Lessepsian migrant), and B. israeliense Brunetti (2009) (uncertain status). To this end, we used all existing Mediterranean sequences of the genus in GenBank, complemented with new samples, for comparison with our reference sequences. The objectives of our work were four-fold: (i) ascertaining the distribution of the two NIS in the study EC area, (ii) providing further molecular reference data, specific to European seas, (iii) assessing if field identification, based on the typical single-color pattern of B. violaceus, is truly reliable, and (iv) updating Mediterranean records of Botrylloides spp. backed by sequence data.

We used COI sequencing to examine >750 colonies of the genus Botrylloides and Botryllus with a special interest in three species, two introduced species, namely B. violaceus and B. diegensis, and one species putatively native to Europe, namely B. leachii. Combining phylogenetic analyses and barcoding, we showed that COI is a robust barcode for distinguishing the three species. We also confirmed the presence of the two NIS in the EC. In addition, this study went well beyond our expectations: it pointed out (1) the risk of mis-identification between the two NIS when rapidly identified in the field based on simple external criteria (colony color patterns), (2) the presence of errors in databases, with reference data available for B. leachii (native) at the time of the analysis (last checked in May 2019) actually being sequences corresponding to B. diegensis (introduced in the study regions), (3) the presence of B. diegensis, probably previously mistaken as B. leachii, in the Mediterranean Sea, based on our custom reference database, thus pointing to the need to revise all previous reports of B. leachii, as it has likely been confused with B. diegensis and (4) the existence of a cryptic divergent lineage, displaying a single-color pattern similar to B. violaceus.

Thus the molecular findings indicate that single-color colonies lacking larvae cannot be identified to species level with confidence from external appearance alone (see brief discussion in Bishop et al., 2015a), at least until additional distinguishing characters are recognized. Preliminary morphological observations based on external features of the species investigated are presented in Supplementary Figure S3. Identification by internal morphology, i.e., microscopical examination of the zooids and brooded larvae, requires laboratory facilities and training, and would be very time-consuming if large numbers of specimens were to be processed, for instance to estimate the relative abundance of the different species in multiple samples. In addition, some key morphological features, such as gonads and larvae, may be present only seasonally. Recourse to molecular means of identification appears a fruitful option, particularly for colorless preserved specimens that may not be reproductive. Furthermore, clear distinctions based on internal anatomy are not yet very consistently reported in the literature, given the current level of taxonomic confusion. Morphological characterization of specimens that have been pre-sorted into putative taxonomic categories based on DNA sequence information may help to achieve the full clarification of any differences in internal anatomy. However, discovery of reliable differences in zooidal morphology can prove challenging in this genus. Saito et al. (1981) distinguished B. simodensis Saito and Watanabe, 1981 from B. violaceus on the basis of reproductive characteristics and the color of live colonies, while preserved, non-reproductive colonies were very hard to separate on the basis of zooidal morphology. Atsumi and Saito (2011) were subsequently able to distinguish two additional species from B. simodensis based on differences in reproductive seasonality and the color of living colonies, backed up by mtDNA sequence information, despite them sharing very similar zooidal features.

As shown by the phylogenetic tree (Figure 2), COI is reliable to distinguish the three accepted species with a clear barcoding gap among them (Supplementary Figure S1). This is in line with previous studies showing that COI can be a relevant barcoding marker for metazoans (Bucklin et al., 2011), notably considering its properties as a rapidly evolving marker (as compared to the nuclear genome). This study also highlights a clear benefit of the barcoding approach, i.e., to enable species identification in groups that lack easy morphological traits to use in taxonomic determination. We carried out extensive sampling of Botrylloides spp. colonies that externally looked like B. violaceus, based on one simple field criterion, namely the single-color patterns of the colonies. The results of COI sequencing and comparison with a newly generated reference dataset clearly indicated that we made a large number of errors in assigning species name in the genus Botrylloides. Error rates varied 0 to 100% (Figure 1) according to the population. We could compare the results of the molecular identifications of the supposedly ‘B. violaceus’ population samples with results of RAS carried by one of us (JB) while sampling for the intended molecular study in the nine populations from the United Kingdom. During the RAS, a semi-quantitative abundance of the distinctive two-color morph B. diegensis was recorded on a four-point scale of abundance. Plotting the percentage of mis-identification as a function of the abundance of two-color B. diegensis shows that the highest mis-identification rates were observed where two-color B. diegensis colonies were relatively common (Figure 4). From this plot, it is clear that the typical two-color morph is easily identified as compared to B. violaceus. It also confirms that the distinctive ‘B. diegensis’ two-color morph is associated with a conspecific uniformly colored morph. Although the number of sites where the two morphs were both identified is low, the correlation further suggests that these two morphs within B. diegensis occur in relatively constant proportions from place to place. Interestingly, the single-color morph of B. diegensis shares haplotypes with the two-color morph (Supplementary Table S1 compared to Supplementary Table S3). In addition, colonies of the same color can be found in both single-color B. diegensis and in B. violaceus.

Even before becoming the barcoding marker of choice for many groups, COI had been extensively used in phylogeographic studies, thus to investigate intra-specific diversity and evolutionary history (Avise, 2000; Beheregaray, 2008; Hickerson et al., 2010). Such population-based investigations often revealed divergent lineages, which, through an integrative taxonomy approach, were sometimes assigned to new species (Pante et al., 2015). Botryllus and Botrylloides are complex taxa in which species status has already been debated (Brunetti, 2009; Reem et al., 2018). For instance, Bock et al. (2012) showed that at least three divergent clades that might be cryptic species characterize B. schlosseri. Griggio et al. (2014) further indicated that the clade A (as defined in Bock et al., 2012) may in turn be undergoing speciation. Our eight reference sequences for this accepted species were obtained from specimens collected in Brittany and south-western United Kingdom. They are also indicative of the presence of two divergent clades in the W English Channel (Figure 2), which are corresponding to clade A and E described by Bock et al. (2012). These authors also showed they co-occur in three populations (namely Brest, Falmouth and Brixham) from EC. Similarly, we found one haplotype, which we called BvX-H6, from colonies externally somewhat similar to B. violaceus, that showed a very large sequence divergence from the other studied Botrylloides species, with a divergence similar (ca. 20%) to the one measured between reference sequences of B. violaceus and B. diegensis (Table 2 and Supplementary Figure S1).

Two populations from United Kingdom, namely Southsea and Gosport, where the first dataset revealed the presence of BvX-H6 were sampled a second time in June 2012. Out of the 13 and 35 specimens examined, 77 and 24% specimens showed this particular haplotype in Southsea and Gosport, respectively. In Gosport, the other specimens were all B. diegensis and in Southsea they were all B. violaceus (Supplementary Figure S2). As part of another population-genetics study, we used the microsatellite markers specifically developed to target B. violaceus (Molecular Ecology Resources Primer Development Constorium et al., 2010) on the colonies assigned here to B. violaceus (with haplotypes Bv-H1 to Bv-H5) and to this new clade. Interestingly, whereas all the colonies with COI haplotypes Bv-H1 to Bv-H5 were easily genotyped with these microsatellites, no PCR product could be obtained for any of the BvX-H6 colonies (data not shown, FV, CR, JB unpublished data). This amplification failure strongly suggests that the colonies with haplotype BvX-H6 represent a distinct cryptic species in the EC. The clustering of this form with the Botrylloides colony from Miseno Lagoon (Italy) and with Botrylloides israeliense (Figure 3) indicates that the same divergent clade is represented in the Mediterranean Sea, but the reported divergence between the two geographical regions (ca. 6%), based on comparison of just 395 base pairs, leaves it unclear whether the EC colonies with haplotype BvX-H6 should be regarded as a somewhat divergent lineage of B. israeliense or as a sister species which is possibly undescribed. Discrepancies between the outcome of the GYMC and mPTP species delimitation methods left us with uncertainties regarding the status of this new lineage. The external appearance of B. israeliense shown by Reem et al. (2018: Figure 1b) does not suggest close kinship with the entity represented by BvX-H6 (Supplementary Figure S3, panels 7 and 8), but further molecular studies, including new species delimitation analyses with other markers, ideally coupled with detailed morphological comparison, will be necessary to resolve the status of this lineage and the exact relationship between the EC and Mediterranean populations.

Another facet of this work has been the documentation of some errors in public databases. Such errors have been pointed out previously (e.g., Harris, 2003; Vilgalys, 2003, and references herein). In our case, this issue was assessed by using simple identification searches in the BOLD system and in the NCBI portal. The outcome is illustrated in the Mediterranean Sea with a phylogenetic tree in Figure 3, where it is clear that all our B. diegensis references are mixed with sequences named B. leachii in Genbank. Further, new sequences obtained confirmed that B. diegensis in the Mediterranean can also have a two-color pattern (Venice or Sète specimens) or a single reddish coloration (Ebro Delta). We failed to find any sequence, either published or new, of Mediterranean B. leachii. Considering that this is the Botrylloides species most commonly mentioned in this sea, our results suggest that the presence of B. diegensis might have been strongly underestimated in the Mediterranean region. We also confirm the presence of B. violaceus in the Venice Lagoon, with external aspect identical to single-color colonies of B. diegensis.

The case of B. diegensis and B. violaceus is a particularly acute issue as these are two NIS, widespread in the northern hemisphere. The lack of reliability of public databases for notorious NIS is an important shortcoming for their early detection and thus effective management (Darling et al., 2017). In this context and for the specific case examined here, it is noteworthy that this mis-identification of B. diegensis occurred during a survey of Mediterranean harbors with the specific aim of examining the distribution of NIS in these well-known introduction hotspots (López-Legentil et al., 2015), and during collections by JB and FV in the EC, specifically targeting B. violaceus in marinas for population-genetic studies. We now have evidence for the presence of B. diegensis in Venice (Italy), Delta of Ebro (Spain), along the Catalan coast (Spain) and in Sète (France). Nevertheless, B. diegensis has not been formally reported in the Mediterranean, although Brunetti and Mastrototaro (2017) mention personal communication by A. and W. Bay-Nouailhat indicating its presence in the W Mediterranean. B. diegensis might be not only present but common and widespread along the Mediterranean coast, although commonly mis-identified as B. leachii. Confusion between B. diegensis and B. leachii might have contributed to statements that these species share very similar morphology (e.g., Van Name, 1945), and Brunetti (2009) synonymized them, although Brunetti and Mastrototaro (2017) treat them as separate. There is little doubt that the distinctive two-color morph now attributed to B. diegensis is a relatively recent arrival in NW Europe, rather than a long-established, potentially native, species such as B. leachii. The very striking two-color morph is not illustrated or mentioned in accounts of Botrylloides in less recent identification guides to the region (e.g., Eales, 1939; Berrill, 1950; Barrett and Yonge, 1958; Millar, 1970; Picton, 1985; Hayward and Ryland, 1990; Gibson et al., 2001). Furthermore, the continuing spread of the distinctive two-color morph to new regions of the English coast from an apparently recent origin on the south coast has been documented in repeated field surveys since 2004 (e.g., Wood et al., 2016). RAS rely substantially on clearly defined field characteristics, which may be elusive in some groups; detailed molecular and morphological studies are called for before field identification can be deemed reliable for the important ascidian genus Botrylloides.

The haplotypes generated by Sanger sequencing have been deposited in GenBank (accession numbers MK978800–MK978824 for the EC sampling and reference sequences and MN076465–MN076483 for the Mediterranean sampling).

FV and JB contributed to the conception and design of the study. FV, JB, SB, and XT provided the samples and/or carried out the sampling. CR, SB, and XT performed the molecular work or provided the sequence data. FV and CR made the analyses. FV wrote the first draft of the manuscript with substantial contributions of JB for some sections and XT for the Mediterranean part of the manuscript. All authors contributed to the manuscript revision, and read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.