The Miseno Lake (~40.793111N - 14.072287E, central-western Mediterranean Sea) is a brackish-water body of around 40.25 hectares and a total perimeter of 2800 meters, located in the town of Bacoli (Naples, Italy) (Leccese and Speziale, 1967). It is included in the list of Sites of Community Importance (SCI no. IT8030017; European Commission Habitats Directive 92/43/EEC) as a Special Area of Conservation (SAC), and it also lays within the Campi Flegrei Regional Park (https://www.parcodeicampiflegrei.it/). As an area subject to protection, access and recreational activities are only allowed to authorized personnel. The lake has a regular outline, somehow resembling a right-angled triangle with the right angle cut away by a rectangular portion, and in its southern part is separated from the Tyrrhenian Sea by a 200 m-wide stretch of coast (Leccese and Speziale, 1967; ENEA, 2002; Figure 1). The entire lake is bordered by an artificial dock wall (~50–60 cm in height) composed of tuff, a friable volcanic rock. Water exchange with the sea is allowed by two channels, the Miliscola channel (MC), located on the western side, and the Casevecchie channel (CC), located on the eastern side. The MC is 250 m long, has a sandy bottom with scattered rocks and an average depth of ~1.5 m, and faces a stretch of open sea, the Procida Strait. The outer part is characterized by rock boulders laying on a sandy bottom, covered by littoral assemblages dominated by photophilic algae. The CC is 100 m long, has a rocky bottom and an average depth of ~1 m, and opens directly into a sheltered bay that hosts two marinas, namely the Bacoli marina and Marina Piccola, and a maritime police facility operative until 2020. Outside the marinas, two mussel farms are still active nowadays. Little is known on the lake’s bathymetry and main substrates, except for preliminary characterizations of the area made about 50 and 20 years ago (Leccese and Speziale, 1967; Rigillo Troncone, 1975; ENEA, 2002; La Magna et al., 2002).

Ascidian samples collected were brought to the Laboratory of Benthos of Stazione Zoologica Anton Dohrn (SZN, Naples, Italy) for proper taxonomic identification. When necessary, specimens were analyzed and dissected under a Zeiss Axio Zoom.V16 (Carl Zeiss, Oberkochen, Germany) stereomicroscope. Then, single specimens (in case of solitary ascidians) or a sub-sample of each colony (in case of colonial ascidians) were isolated for molecular analyses. As a general rule (whenever possible), conspecificity between samples was tested by sequencing 5–6 samples per putative species, collected from different places of the lake.

To do so, total genomic DNA was extracted from muscular tissue of single zooids using the DNeasyBlood & Tissue^® kit (Qiagen, Hilden, Germany), following the producer’s protocol. Amplifications of a partial region of the cytochrome c oxidase subunit I (COI) gene were obtained through polymerase chain reactions (PCRs) with the following primer sets by Brunetti et al. (2017): dinF (5’-CGTTGRTTTATRTCTACWAATCATAARGA-3’) and Nux1r (5’-GCAGTAAAATAWGCTCGRGARTC-3’). Polymerase chain reactions (PCRs) were carried as follows: first denaturation at 95°C for 5 min, followed by 35 cycles of denaturation at 95°C for 1 min, annealing at 48–50°C for 1 min, extension at 72°C for 1 min, with a final elongation at 72°C for 5 min. The PCR products were purified and sequenced at the SZN Molecular Biology and Sequencing Service through an Automated Capillary Electrophoresis Sequencer 3730 DNA Analyzer (Applied Biosystems, CA, USA), using the BigDye^® Terminator v3.1 Cycle Sequencing Kit (Life Technologies, Renfrew, UK).

Sequences obtained were quality checked, assembled, and edited using Unipro UGene v.39 (Okonechnikov et al., 2012). Basic Local Alignment Search Tool (BLAST; https://blast.ncbi.nlm.nih.gov/Blast.cgi) (Morgulis et al., 2008) was then used to assess the identity of each sequence against the GenBank database (last research: 30.XII.2021).

Sequenced and other collected specimens were finally fixed in 99.9% ethanol and deposited in the collection of the Laboratory of Benthos (SZN), under different accession codes (Supplementary Table 1).

Since the sequences obtained for Distaplia bermudensis Van Name, 1902 gave unclear BLAST results (see below), we also carried out phylogenetic analyses in order to assess the phylogenetic position of our specimens within other D. bermudensis collected worldwide.

A total of thirty-two COI partial sequences were used in the final alignment, including the five sequences of D. bermudensis amplified in this study and twenty-one sequences of D. bermudensis retrieved from GenBank (accession numbers reported in Supplementary Tables 1, 2). Six additional representatives of the order Aplousobranchia Lahille, 1886, collected in this study, were used as outgroup: Clavelina lepadiformis (Müller, 1776) (SZN_B_1879ASC15G), Clavelina sabbadini Brunetti, 1987 (SZN_B_2707ASC36A), Didemnum pseudovexillum Turon & Viard, 2020 (SZN_B_1902ASC22A), Lissoclinum weigelei Lafargue, 1968 (SZN_B_3172ASC49A), Aplidium accarense (Millar, 1953) (SZN_B_1822ASC18B), and Polyclinum constellatum Savigny, 1816 (SZN_B_2884ASC43A). The selected sequences were aligned and trimmed with Unipro UGene v.39 (Okonechnikov et al., 2012), and the best-fit evolutionary model (GTR + I + G) was found with JModelTest 2 v.0.1.10 (Darriba et al., 2012), selected with the Akaike information criterion (AIC) method. The maximum likelihood (ML) analyses were run with RAxML v.8.2 (Stamatakis, 2014) using 1000 rapid bootstrap pseudo-replicates. The Bayesian Inference (BI) was performed with MrBayes v.3.2.5 (Ronquist et al., 2012) run for 10 million generations, sampled every 1000 generations, with a standard 25% burn-in. Convergence of the MCMC runs was checked with Tracer v1.7.1 (Rambaut et al., 2018).

Three species delimitation methods were then used on the same alignment to infer Operational Taxonomic Units (OTUs) boundaries within the Distaplia clade: Automated Barcode Gap Discovering (ABGD, Puillandre et al., 2012), Poisson Tree Processes (PTP, Zhang et al., 2013), and Generalized Mixed Yule Coalescent (GMYC, Fujisawa and Barraclough, 2013). The ABGD analysis was performed on the ABGD web platform (https://bioinfo.mnhn.fr/abi/public/abgd/abgdweb.html) with Jukes-Cantor (JC69) model and standard parameters. The PTP and GMYC analyses were run on the Exelixis Lab web server (The Exelixis Lab, https://species.h-its.org). The PTP was run for 500000 generations, 25% burn-in, and 100 thinning based on the rooted ML tree obtained in the aforementioned analysis. The GMYC was performed in single threshold (s-GMYC) method. As input, an ultrametric tree was obtained with BEAST v.1.10.4 (Suchard et al., 2018) based on the same alignment as above.

Once a solid identification of the specimens was achieved, species were classified into the following categories based on the most up to date literature (see Essl et al., 2018): i) native for species with a putative Atlantic-Mediterranean origin; ii) cryptogenic for species with uncertain native range; iii) non-indigenous for species native in other biogeographic sectors and locally introduced through human activities (European Commission, 2014).

Finally, the digital photographic archives (2014–2021) of one of the authors (G.V.), who regularly surveyed and mostly photographed the lagoon biota in the past decade, was screened in order to backdate as much as possible the most probable first introduction event of selected NIS in the Miseno Lake.

Images obtained (photoquadrats) were subsequently analyzed in Photoquad v1.4 (Trygonis and Sini, 2012). Each picture was calibrated according to the known frame measures. After picture calibration, abundance and cover of each species were calculated counting single specimens (for solitary ascidians) and colonies (single separated clusters) for each quadrat. Data obtained from transects and photoquadrats were then collated in a matrix in Excel (Supplementary Data Sheet 1). Data are reported as total number of individuals counted on the seabed and on the dock wall, as well as mean count per transect on the seabed and mean percentage cover on the dock wall ± standard error (SE).

The abundance (number of individuals per transect) of the most commonly encountered NIS ascidians, namely Botrylloides niger Herdman, 1886, D. bermudensis, P. constellatum, and A. accarense, as well as of all NIS and cryptogenic species combined together, was modeled with Generalized Additive Models (GAMs; Hastie and Tibshirani, 1990). Due to severe overdispersion (residual deviance of the fitted models much higher than the residual degrees of freedom), and thus violation of the assumption of equal mean and variance in the response variables, the negative binomial distribution was used instead of the Poisson distribution (Lindén and Mäntyniemi, 2011). The expected values of the number of individuals counted in each transect were related to spatial and environmental covariates, according to the general formulation f(E[counts]) = LP_i = c + Σ_ms_m (z_mi ) + Σ_rF_ri , where f is the link function (in this case link=log), LP is the linear predictor, s_m (·) is the one-dimensional smooth function for covariate m, F_r are categorical predictors, and z_mi is the value of spatial covariate m for the ith transect. The smooth function s_m (·) was represented using penalized regression splines (cubic splines with basis dimension q=10), estimated by penalized iterative least squares (Wood, 2006). The optimum degree of smoothing was defined by generalized cross validation (GCV), where the amount that the effective degree of freedom of each model counts in the GCV score was increased by a factor c=1.4 according to Kim and Gu (2004) to avoid occasional overfitting. The mgcv package implemented in R v4.1.0 was used for model fitting (Wood, 2006; R Core Team, 2021).

Four predictor variables were used for modelling the abundance of NIS on the seafloor, namely: i) depth; ii) hard, the amount of hard substrate items counted in each transect — i.e. the sum of rocks and litter items; iii) section, a categorical variable with three levels representing the distinct seafloor areas investigated in the lake (channel, inner, and peripheral); iv) Caulerpa, a categorical variable with two values, 0 for absence and 1 for presence, indicating the presence/absence of C. prolifera within each transect, as the seagrass can act as substrate for epiphytic ascidians. The degree of collinearity among these variables was measured by calculating the variance inflation factor (VIF) for each covariate based on the negative binomial models. Since all VIF values were <3, there was no substantial multicolinearity among the independent variables and thus were all kept in the analyses. Overall, eight candidate models g_i were used in the analyses linked to the following hypotheses: h1 - abundance is affected by depth; h2 - abundance is restricted by the availability of proper substrate (hard+Caulerpa); h3 - the three different sections have varying conditions (other than depth and substrate) — e.g. hydrodynamism, that may affect the abundance of the species (Supplementary Table 3). Furthermore, the abundance of native species on the seafloor was modelled. For this, in addition to the same three hypotheses, a fourth one was investigated: h4 - abundance of native species is restricted by the abundance of NIS and cryptogenic species. An additional predictor variable was included — i.e. the abundance of NIS and cryptogenic species (AlCr), and the set of candidate models for native species included overall sixteen models (Supplementary Table 4). This hypothesis was also tested using the estimated cover of native vs NIS/cryptogenic species along the dock wall perimeter (photoquadrat sampling), based on only four models, the null model, one including the cover of NIS and cryptogenic species as a predictor variable, one including section (peripheral vs channels) as a predictor variable, and one including both predictors.

Akaike’s information criterion was used for model selection (Akaike, 1973; Burnham and Anderson, 2002), and the AIC differences Δ_i=AIC_i - AIC_min were computed over all candidate models g_i, where AIC_min is the minimum AIC corresponding to the best model. To quantify the plausibility of each model, given the data and set of candidate models, the ‘Akaike weight’ w_i of each model was calculated, where w_i = exp(-0.5∆_i )/Σ_j exp(-0.5∆_j ) (Burnham and Anderson, 2002). The relative importance of each of the hypotheses made was estimated by summing the Akaike weights across all the models in the set that were linked with each hypothesis (Burnham and Anderson, 2002).

The bathymetric survey recorded a total of 228000 reliable depth measurements, interpolated to produce a detailed bathymetric map of the lake (Supplementary Figure 1). The average depth was ~2 m, whereas the maximum depth was ~3.5 m. The habitat characterization revealed the presence of muddy bottoms almost dominated by a dense meadow of the native alga C. prolifera in several grid sectors but mostly in the inner parts of the lake (A1, B2–B5, C3–C7, D2–D7, E2–E6, F2–F6, G3–G7). The meadow only becomes sparse in the majority of the peripheral parts of the lake and underneath the dock wall, nearby freshwater inputs, and in the proximity of the two channels (A2–A4, B1, B6, C1, C2, C8, D1, D8, E1, E7–E9, F1, F7, G1, G2), where the substrate is again unconsolidated but with a wider granulometric range, from mud to gravel. Worth a mention, the sector C1 was mostly characterized by a sand bank with a very low depth, and thus it was excluded from the analysis of the transects-based data, although retained for the dock wall analysis through photoquadrats. The area was also characterized by the presence of sparse rocks (n=455) and widely impacted by litter items (n=1801), mostly originating from commercial or recreational activities in the nearby town of Bacoli. Their presence and distribution was generally influenced by geographic position and conformation of the nearby substrate. Rocks were rare to absent in the inner parts of the lake and mostly found in grid sectors close to the dock wall — C2 (43), D1 (59), D8 (68), E1 (47), F1 (53), G1 (54), as they constitute waste material from the previous levee. Litter items were found in all transects, with the sectors close to the CC being the ones most impacted — D8 (296), E8 (91), and E9 (80).

A total of 24 taxa were observed during our survey, belonging to the three orders of the class Ascidiacea Blainville, 1824 (nine Aplousobranchia, six Phlebobranchia Lahille, 1886, and nine Stolidobranchia Lahille, 1886) (Table 1, Figures 2, 3 and Supplementary Table 1). Of these, ten are native, nine are NIS, and five are cryptogenic species. Distribution in taxonomic orders varied based on the species’ status in the Mediterranean Sea. Among native species, Aplousobranchia and Phlebobranchia accounted for four taxa each, whereas Stolidobranchia only accounted for two. On the contrary, NIS accounted for five Stolidobranchia, followed by three Aplousobranchia, and one Phlebobranchia. Finally, cryptogenic species accounted for two species of Aplousobranchia and Stolidobranchia each, and only one Phlebobranchia (Table 1, Figures 2, 3 and Supplementary Table 1).

The amplification of the COI partial region gave a total of 114 sequences, ranging from 504 to 828 base pairs (bp), with a number of specimens sequenced per species varying from one to six. The morphological identification was supported by BLASTn hits (~95–100% similarity) for 22 species (Supplementary Table 1). However, several ambiguities in the GenBank database were found, as follows: i) single to a few mismatches in four species — Lissoclinum perforatum (Giard, 1872), Phallusia mammillata (Cuvier, 1815), Pyura dura (Heller, 1877), and B. niger — were presumably due to misidentifications or incorrect uploads; ii) mismatches in three species — L. weigelei, A. accarense, and Ciona robusta Hoshino & Tokioka, 1967 — were due to identification not provided at a species level by some authors when uploading sequences in GenBank or lack of GenBank updates after the species complexes were resolved (e.g. Caputi et al., 2007); iii) mismatches in four additional taxa — C. lepadiformis, D. bermudensis, Botryllus schlosseri (Pallas, 1766), and Styela canopus (Savigny, 1816) — were due to the fact that they belong to species complexes (Turon et al., 2003; Nydam et al., 2017; Barros and Rocha, 2021; Salonna et al., 2021; Nuzzo et al., 2022), and thus the sequencing of local specimens allowed clarification of their exact taxonomic allocation and also identification of a new clade in D. bermudensis (see below); for the sake of text clarity, these species are still listed in the subsequent chapters as monotypic taxa; iv) mismatches in two species — P. constellatum and Symplegma brakenhielmi (Michaelsen, 1904) — have already been reported and presumably require further taxonomic work to solve the related taxonomic issues, including sequencing of specimens from the type localities of the taxa involved (Mastrototaro et al., 2019; Montesanto et al., 2022); these species are listed in the subsequent chapters following the most recently updated literature on the Mediterranean biota.

Finally, with regards the two remaining species, mismatches in Ascidia malaca (Traustedt, 1883) may be due to misidentifications or even uncertainties resulting from the use of the COI marker within the genus Ascidia Linnaeus, 1767 (see Streit et al., 2021), whereas Ascidia colleta Monniot & Monniot, 1970 accounted for a new GenBank entry, as it only showed 80.4–81.2% similarity with Ascidia conchilega Müller, 1776 (MN064596–7: Couton et al., 2019). Given taxonomic issues regarding worldwide species of this genus, these taxa are also further illustrated and discussed in Supplementary Data Sheet 2.

BLAST results of D. bermudensis sequences from the Miseno Lake (SZN_B_1743ASC10H–1747ASC10N) gave high similarity (99.71%) to another sequence of a specimen identified as D. bermudensis (MT637947) from Puerto Rico marinas by Streit et al. (2021), but also significant differences (80.33–86.5%) from other D. bermudensis worldwide (Supplementary Table 1).

After sequences-editing steps, we obtained a trimmed alignment (575 bp) of COI partial sequences, with 250 parsimony-informative sites. Both the ML and BI analyses produced congruent tree topologies, hence we showed the best ML tree with bootstrap (bs) and posterior probability (pp) values reported at nodes (Figure 4). Both tree topologies clustered the newly collected D. bermudensis from Miseno with the specimen from Puerto Rico marinas by Streit et al. (2021), forming a different clade with respect to the other samples. This clade, that we first name here as clade C, was well supported (bs=99; pp=1) and the sister (bs=82; pp=0.99) of clade B sensu Evans et al. (2021). Results of the species delimitation analyses also well supported the three-clades grouping well (Figure 4 and Supplementary Data Sheet 3). The ABGD analysis inferred three OTUs within the D. bermudensis sequences in the primary partition. The PTP and GMYC analyses were also consistent, reporting the three-clade grouping in all the analyses carried out, with PTP Bayesian supporting values ranging from 0.18 to 0.88 (Supplementary Data Sheet 3).

The digital photographic archives of the Miseno Lake biota were composed of ~400 pictures, ranging from May 2014 to October 2021. By subsequently excluding widespread NIS ascidians introduced in the entire Mediterranean Sea in historical times (C. robusta and S. plicata), and notwithstanding limitations in identifying ascidians on the basis of underwater photographs, unequivocal evidence was found regarding the presence of at least five NIS before our survey, namely D. bermudensis, A. accarense, P. constellatum, B. niger, and P. zorritensis. In particular, four of them were recorded since 2014, and one (D. bermudensis) since 2019, with three sightings that predate first-record dates and localities in Italy, one of which also accounts for the first sighting in the Mediterranean Sea (Table 2 and Figure 5). It is noteworthy that four of these species were the most common NIS encountered during our survey (see below). Finally, photographic evidence of the impact of D. bermudensis on dock wall communities were also found, and illustrated in Figures 5F, G.

Overall, a total of 20197 ascidian individuals/colonies were counted in the Miseno Lake. Of these, 4537 were censused during transects on the seabed of the lake, 11570 were found in the dock wall, and the remaining 4090 in transects and dock walls in the Casevecchie and Miliscola channels (CC=3482; MC=608) (Table 1 and Supplementary Data Sheet 1).

Fifteen species were found in the seabed transects, among which B. niger (n=2696, n_m=29 ± 62.3) and D. bermudensis (n=796, n_m=8.3 ± 32.8) were the most common (Table 1 and Supplementary Data Sheet 1). The distribution of the four most abundant NIS is depicted in Figure 6. However, the peripheral parts of the lake hosted all species mentioned above, whereas the ascidian diversity was highly reduced in the inner parts of the lake, with only seven species found, among which only B. schlosseri and B. niger ranked high in numbers, which was also reflected in their presence in the transects (33/48 and 28/48, respectively) (Supplementary Data Sheet 1). With regard to substrates: 13 species were found on/below rocks, with B. niger and A. accarense being among the most commonly found, 12 species were found associated with litter, with B. niger and D. bermudensis almost completely dominating the group, five species were found associated with other biogenic substrates, among which D. bermudensis and B. niger were again the most commonly encountered, and two species only were found as epiphytes on C. prolifera leaves, namely B. niger and B. schlosseri (Supplementary Data Sheet 1).

Thirteen species were found in the photoquadrats along the dock wall perimeter, among which the NIS B. niger (n=5770, c_m= 1.4% ± 0.2%), P. constellatum (n=2115, c_m=1.1% ± 0.2%), A. accarense (n=1327, c_m=0.3%), and D. bermudensis (n=975, c_m=2% ± 0.6%) were the most widespread (Table 1 and Supplementary Data Sheet 1). The distribution of these four species is depicted in Figure 7. Such abundances were also somewhat reflected in the overall presence of the species in the photoquadrats, with B. niger appearing in all photoquadrats (72/72), followed by P. constellatum (68/72), and A. accarense (64/72). On the other hand, D. bermudensis was locally abundant in only few sectors (25/72) (Table 1, Figure 7 and Supplementary Data Sheet 1). The dock wall sectors with the highest mean total cover were E8 (c_m=17.0% ± 7.2%), mostly dominated by D. bermudensis (n=214, c_m=10.4% ± 3.8%), B. niger (n=855; c_m=1.1% ± 0.1%), and P. constellatum (n=132, c_m=0.7% ± 0.2%), followed by C8 (c_m=14.7% ± 9.2%) and A3 (c_m=8.7% ± 4.4%) (Supplementary Data Sheet 1).

Fourteen and thirteen species were respectively found in CC and MC, with assemblages differing between them and both channels hosting exclusive taxa. With regards CC, eight species were found in the transect, with D. bermudensis (n=1542) and B. niger (n=794) being the most common, whereas 12 species were found in the photoquadrats, with D. bermudensis (n=680, c_m=13.1% ± 2.3%) dominating the assemblages (Table 1 and Supplementary Data Sheet 1). The CC also accounted for the highest cover in a quadrat (c=44.3%) and for the highest mean cover per sector (c_m=22.3% ± 4.6%) (Supplementary Data Sheet 1). With regards MC, seven species were found in the transect, whereas 13 species were found in the photoquadrats; B. niger was the most common species in both of them (n=272; n=113, c_m=0.6% ± 0.3%) (Table 1 and Supplementary Data Sheet 1). Microcosmus polymorphus Heller, 1877 ranked low in number of specimens found (n=5), but high in cover (c_m=0.2% ± 0.2%) due to the big size of the specimens found. The overall mean cover (c_m=2.6% ± 0.8%) was much lower than in CC (Supplementary Data Sheet 1).

Through the GAM analyses it was found that, when NIS and cryptogenic species were combined, the best model for their overall abundance on the seafloor was the full model (g₇), including all predictor variables (Figure 8 and Supplementary Table 3). Their abundance declined with depth, increased with the availability of hard substrates, and was the highest in the channels, particularly in the CC, and the lowest in the inner zone (Figure 8). The best model for the abundance on the seafloor of all native species combined together was g₁ (Supplementary Table 4), with only section as a predictor variable, having substantially higher support than the null model. For native species only h3 had high support, with no strong apparent effect of NIS or cryptogenic species (low support for h4, although it cannot be totally discarded) (Table 3 and Supplementary Figure 2). With regards the dock wall perimeter, the model using the abundance of NIS and cryptogenic species as predictor variable was the best, although its Akaike difference with the null model was only 0.6, and thus h4 does not have substantially higher support than the null hypothesis; hypothesis h3 had even less support (Supplementary Table 4). Hence, although the cover of the native ascidians was negatively correlated to the cover of NIS and cryptogenic species (Supplementary Figure 3), no solid conclusion can be made at this stage.

With regards the four dominant NIS encountered in the Miseno Lake, namely D. bermudensis, A. accarense, P. constellatum, and B. niger, when analyzing the seabed, the best models had significantly higher support by the data in comparison to the null model (with Akaike differences >20), indicating that the assumption of completely random distribution of the species in Miseno Lake has no support (Supplementary Table 3). In all four cases, section was included in the best model, which indicates that the conditions among the two channels and the peripheral and the inner zones of the lake were sufficiently different to affect the abundance of NIS ascidians; hypothesis h3 had high support (Table 3). Three of the four species (D. bermudensis, P. constellatum, and B. niger) had the lowest abundance in the inner part of the lake (zero in the case of D. bermudensis), and the highest in the channels and in particular in the CC (Supplementary Figures 4, 6, 7), whereas A. accarense was absent from the channels but still had higher abundance in the peripheral than in the inner zone (Supplementary Figures 5). Depth (hypothesis h1) was negatively correlated with abundance in the cases of B. niger and P. constellatum (Supplementary Figures 6, 7 and Supplementary Table 3); in the case of B. niger, h1 had high support (Table 3). In particular, B. niger and A. accarense seemed to be favored by the existence of appropriate substrates, with higher local abundances with increased availability of rocks and litter items; however, the hypothesis h2 had only medium support for these species (Table 3).

Marine reserves are protected areas aiming to achieve conservation of nature with its associated ecosystem services and cultural values. Notwithstanding legal limitations that are in place in these areas, a growing amount of literature concerns their vulnerability to biological invasions, in particular when they host, coincide with, or are close to potential hotspots of NIS arrival or spreading (Otero et al., 2013; Giakoumi et al., 2019; Bilecenoğlu and Çınar, 2021). Both qualitative and quantitative results of the present study match this statement. The benthic survey carried out here revealed that the Miseno Lake hosts a previously undetected thriving community of ascidians, with native taxa that still rank high in species richness (10 taxa: ~41.5%), but that are fewer than NIS (9 taxa: 37.5%) plus cryptogenic species (5 taxa: ~21%) combined.

Moreover, among NIS, only two (C. robusta and S. plicata) are relatively old introductions and thus almost ubiquitous in the Mediterranean Sea (Caputi et al., 2007; Pineda et al., 2016a), whereas the majority of them are only known from a few records at a regional level, with some taxa that were never found in the study area (P. constellatum, P. zorritensis, and S. brakenhielmi), others that were only known from single opportunistic sightings (D. bermudensis, A. accarense, and B. niger: Montesanto and Mastrototaro in Orfanidis et al., 2021; Montesanto et al., 2021; Della Sala et al., 2022), and Microcosmus squamiger Michaelsen, 1927 which was the only species widely recorded in the Gulf of Naples, but has not been reported again in grey or published literature for more than 20 years (since 1996: Turon et al., 2007). However, the absence of records also holds true at Mediterranean level and is particularly surprising for the four NIS dominating the assemblage. Indeed, the most abundant NIS in the Miseno Lake was B. niger, only known from a few specimens sequenced from Israel (Rubinstein et al., 2013; Griggio et al., 2014) and Italy (Taranto: Salonna et al., 2021; Fusaro Lake, Naples: Della Sala et al., 2022), followed closely by P. constellatum, only known by few localities in the eastern Mediterranean (Halim and Abdel Messeih, 2016; Aydin-Onen, 2018; Montesanto et al., 2022), and A. accarense, the only species apparently more widespread, but only in some areas of the western Mediterranean (López-Legentil et al., 2015; Montesanto et al., 2021). Finally, D. bermudensis is again only known from a few records in Spain (Balearic Islands: Pérès, 1957) and Italy (Taranto: Mastrototaro and Brunetti, 2006; Miseno Lake: Montesanto and Mastrototaro in Orfanidis et al., 2021). The present paper highlights a high local invasiveness for all these species, report P. constellatum for the first time in the western Mediterranean, document the presence of another species (hereby named clade C) in the complex of D. bermudensis, and finally backdate and georeference to the Miseno Lake the first finding of A. accarense and B. niger in Italy and of P. constellatum in the Mediterranean Sea.

Cryptogenic species also accounted for a significant part of the ascidian biota of the lake. In particular, S. canopus and B. schlosseri were widespread, while C. lepadiformis, D. pseudovexillum, and Perophora viridis Verrill, 1871 were rare. The first two taxa are common in shallow sheltered waters and account for at least five different clades each in the Mediterranean (Barros and Rocha, 2021; Salonna et al., 2021). Specimens of S. canopus from the Miseno Lake well match the lineage G10 sensu Barros and Rocha (2021), already known from India, Angola, and Mediterranean Spain, and first recorded here from Italy, whereas specimens of B. schlosseri belong to the subclade A1 sensu Salonna et al. (2021), widely recorded worldwide including in Italy. Clavelina lepadiformis is represented by two clades in the Mediterranean (Turon et al., 2003; Nuzzo et al., 2022), and specimens from the Miseno Lake belong to C. lepadiformis sp. B sensu Nuzzo et al. (2022), commonly found in semi-enclosed basins and only known with certainty in Italy from the nearby Fusaro Lake (Nuzzo et al., 2022). Finally, with regards the two remaining species, P. viridis was described from the northwestern Atlantic (Buzzard’s Bay, Massachusetts, USA) (Verrill, 1871) but is very similar to several congeneric species, thus suggesting that at least part of the Mediterranean historical records may be based on misidentifications and making it difficult to trace a putative colonization history of the basin (Zenetos et al., 2017). This species was already recorded from the area by Neppi (1921), Salfi (1931), and Chimenz et al. (1985), and the Gulf of Naples is also the locality of description of its junior synonym Perophora dellavallei Neppi, 1921. Sequences of specimens from the Miseno Lake are the first based on Mediterranean specimens and proved conspecific with those from the USA deposited in GenBank, thus strengthening the synonymy between P. viridis and P. dellavallei and suggesting that historical records from the Gulf of Naples may be true. Finally, D. pseudovexillum was only known so far from Roscoff (Atlantic France) and Catalonia (Mediterranean Spain) and the present record widens its distribution to Italy. Prior to its formal description, this species was often mixed up with other didemnids, such as Didemnum vexillum Kott, 2002 (see discussions in Turon et al., 2020), and even the present finding was incidental, as the only colony found in the Miseno Lake was initially misidentified in the field as the native didemnid Trididemnum cereum (Giard, 1872).

Finally, some native species found here deserve attention. Among them, four — Phallusia fumigata (Grube, 1864), P. mammillata, M. polymorphus, and P. dura — are almost ubiquitous in Mediterranean coastal areas (e.g. Brunetti and Mastrototaro, 2017), whereas those of the family Didemnidae Giard, 1872 (L. perforatum, L. weigelei, and T. cereum) usually live in fouling communities of harbors and semi-enclosed basins (e.g. Chimenz et al., 1985; Casso et al., 2018), and even their presence in the Miseno Lake may be due to intra-Mediterranean transfers, as already speculated for the occurrence of both species of the genus Lissoclinum Verrill, 1871 on ship hulls in the Gulf of Taranto (Lafargue and Tursi, 1975). Integrative approaches, which include literature reviews and population genetic studies, may shed further light on that. It is noteworthy that the three remaining species (C. sabbadini, A. colleta, and A. malaca) are apparently quite rare in the Mediterranean Sea. The two former species were never recorded in Italy outside their area of description, namely the Adriatic Sea (Brunetti and Mastrototaro, 2017; Mastrototaro and Montesanto, 2021): for A. colleta this may be due to misidentifications with the very similar Ascidia muricata Heller, 1874, while C. sabbadini is almost unmistakable even in the field, suggesting that it may have ecological requirements that limit its distribution. On the other hand, A. malaca was originally described by Traustedt (1883) based on specimens collected in the Gulf of Naples, and considered conspecific with Ascidia depressa sensu Heller (1874) from the Lesina Lagoon (Adriatic Sea), an environment similar to the one investigated here. We figure topotypical specimens for the first time in more than a century. Although taxonomic uncertainties in the genus Ascidia prevented us from reaching any sort of taxonomic conclusion, the implementation of the genetic information on this genus, with two species newly sequenced here, may represent an advance toward resolving its taxonomy.

The present revealing results from an investigated area that is less than 1 km² in extent are somewhat unexpected, but are likely due to two main previous limitations, namely taxonomic impediments (or general absence of zoologists and taxonomic specialists in the area) and lack of specific monitoring programs. Although the ascidian fauna of the central-western Mediterranean Sea is considered as the most studied in the entire basin (Koukouras et al., 1995), the rich literature published on the Gulf of Naples and nearby areas mostly dates back to a century ago, when early pioneers focusing on Mediterranean biodiversity investigated the local biota (e.g. Della Valle, 1877; Della Valle, 1881; Traustedt, 1883; Salfi, 1929; Salfi, 1931; Parenzan, 1959). Then, in the subsequent historical periods, faunistic and taxonomic work has been often considered old-fashioned, resulting in a lack of proper studies but for a single one assessing seasonal diversity and shifts of ascidian species in Ischia harbor (Chimenz et al., 1985) and few NIS records generated opportunistically by specialists (López-Legentil et al., 2015; Montesanto and Mastrototaro in Orfanidis et al., 2021; Montesanto et al., 2021). All this somehow makes it difficult to discuss the ascidian assemblage found in the Miseno Lake in a local perspective. However, Chimenz et al. (1985) reported 12 species in the investigated area, seven of which — C. lepadiformis, T. cereum, C. intestinalis (now C. robusta), P. viridis, B. schlosseri, S. partita (synonym of S. canopus), and S. plicata — were shared with the Miseno Lake. The absence of the NIS now found dominating the Miseno Lake suggests that they may have arrived only recently in the Gulf of Naples, although no genetic support was available for the 1985 study, and the knowledge of the Mediterranean ascidian biota was more limited, and thus we cannot exclude that putative discrepancies in taxonomically difficult families (e.g. Didemnidae, Ascidiidae Müller, 1776, Styelidae Herdman, 1881, and Pyuridae Hartmeyer, 1908) may be the result of misidentifications. On the other hand, studies at the Mediterranean level are indeed more widespread. Knowledge of ascidian biota in selected areas was gained either through targeted studies (e.g. Gewing et al., 2017; Casso et al., 2018; Arroyo et al., 2021), data gathering for the creation of checklists (e.g. Lafargue et al., 1986; Koukouras et al., 1995; Moreno et al., 2014), or scattered records in surveys of fouling communities (e.g. Leclerc and Viard, 2018; Lezzi et al., 2018; Giangrande et al., 2021). Studies investigating port environments also often focused on NIS, including ascidians (e.g. López-Legentil et al., 2015; Marchini et al., 2015; Tempesti et al., 2020), whereas specific surveys in enclosed and semi-enclosed basins were carried out in several Mediterranean localities (e.g. Pérez-Ruzafa, 1989; Mastrototaro et al., 2008; Chebbi et al., 2010; Davis and Davis, 2010). In general, the native ascidian assemblage found in the Miseno Lake is substantially coherent with the ones reported in the aforementioned studies, whereas no studies reported such a high and contemporary presence of NIS and cryptogenic species in a single and somewhat restricted study site. All this identifies the Miseno Lake as a major hotspot of NIS and cryptogenic ascidian species in the Mediterranean basin, highlights that the spreading of at least some of these taxa went partially overlooked for about a decade, and suggests that the Miseno Lake protected area deserves the establishment of appropriate monitoring programs for new NIS detections at early stages of arrival.

The spatial analyses of the ascidian biota of the Miseno Lake revealed a general homogeneity of assemblages in terms of species composition, with dominant species occurring in almost all investigated transects and areas of the lake, although their cover and abundance often varied on the basis of the topography of the lake as well as of the main benthic substrate and presence/absence of additional substrates available for settlement (whether natural or anthropogenic). Finally, both channels also hosted exclusive taxa, notably different between them.

The dock wall represented the main pool of ascidians in the basin, although showing species-specific distributional patterns and abundances. As an example, A. accarense and B. niger were somewhat widespread in terms of mean cover and number of colonies in the entire lake, whereas P. constellatum was mostly found in the western side of the lake, where colonies also seemed to be bigger in average size, and D. bermudensis displayed a marked segregation, being mainly restricted to sectors close to the CC except for a cluster of colonies found in the southern side of the lake. Other species did not show a clear distribution pattern but were widely scattered, namely P. dura, P. zorritensis, and S. canopus, whereas the natives C. sabbadini and M. polymorphus and the NIS M. squamiger were rarely encountered.

The inner sectors hosting the dense C. prolifera meadow were not a preferred habitat for almost any of the ascidian species found in the Miseno Lake. In particular, although 15 species were censused in the transects and seven in the Caulerpa meadows, only B. niger and B. schlosseri were found as epiphytes on the seaweed, with B. schlosseri being widespread in small scattered colonies evenly distributed within the meadow, while B. niger was limited to the outermost parts of the meadow, thus close to the dock wall where it is dominant. However, where present, B. niger was apparently more aggressive in its colonization, fouling a large part of the leaves. On the other hand, in the peripheral parts of the lake, where the seaweed meadow becomes sparse or ends abruptly, the substrate was mainly sandy/muddy and ascidians were mostly found on/under scattered rocks, litter, or other biogenic substrates, with a community similar to the one living in the dock wall, but with the addition of A. colleta and P. fumigata. As revealed by GAMs, the presence of hard substrates such as rocks and litter also influenced ascidians distribution in the lake. Indeed, although litter was found in all the lake, the sectors near the dock wall were mostly characterized by heavy-weight sunken objects (e.g. tyres, cans, tarpaulins, and hard plastic), whereas the inner part was mostly littered by lower-density drifting objects (e.g. tissues and soft plastic), still often detached from the substrate. Therefore, litter colonization by ascidians mostly occurred in the peripheral parts of the lake, where litter objects soon turn into additional anthropogenic substrates available for the resident fauna. Finally, when present, other kinds of natural substrates were also often overgrown by tunicates, such as sunken woods or polychaete tubes.

Although we did not investigate here chemico-physical parameters such as water salinity or turbidity, the observed differences in ascidians’ distribution are presumably related to the species-specific traits of the species found. Unfortunately, little is known of species’ autoecology for the majority of the taxa found in the Miseno Lake, although several of them, namely A. accarense, P. constellatum, B. niger, and P. zorritensis, are well known to adapt to different environmental conditions (Van Name, 1945; Mastrototaro and Brunetti, 2006; Mastrototaro et al., 2008) and that is presumably why they thrived in almost all the lake from the calm inner side, under rocks or inside crevices, to the unsheltered areas of the channels. Other species, such as P. mammillata, C. robusta, and S. plicata, are known to tolerate calm waters and poor environmental conditions (Chimenz et al., 1985; Mastrototaro et al., 2008; Caputi et al., 2013), and thus their distribution in the lake is likely to be influenced by additional factors. However, the renowned invasiveness of several NIS indeed implies wide environmental tolerances, with species able to adapt to various environments and being able to resist temperature shifts (Shenkar and Loya, 2008; Grey, 2011) and sunlight exposure (Forward et al., 2000), or even to cope with the local environmental metabolome (Palanisamy et al., 2018). With regards to this, antifouling properties due to secondary metabolites are widely reported from the alga Caulerpa prolifera and congeneric species (Smyrniotopoulos et al., 2003; Dobretsov et al., 2006), and thus they may also act locally as inhibitors for larval settlement of ascidians, being thus at the basis of the absence of the majority of the species in the inner parts of the lake. Interestingly, the two only species fouling the leaves (B. niger and B. schlosseri) are phylogenetically closely related, which may led to the speculation that these species possess antagonist strategies against the antifouling molecules produced by the alga. In addition, litter material has already been shown to be a collector for sessile species on unconsolidated bottoms, enhancing the exploitable substrate (Katsanevakis et al., 2007; Crocetta et al., 2020) and influencing the overall biomass of benthic assemblages (Ramirez-Llodra et al., 2013). In our case, litter was mostly fouled by NIS (e.g. D. bermudensis, A. accarense, P. constellatum, and B. niger), thus possibly enhancing their invasiveness in the lake, as also supported by the GAM analyses.

Finally, the two lake channels differed markedly in terms of species diversity. The CC showed an ascidian diversity which is comparable to the rest of the lake, although with a significantly higher NIS cover and dominance. It also hosted four exclusive taxa, namely C. lepadiformis, D. pseudovexillum, T. cereum, and P. viridis. In contrast, the MC showed an overall lower ascidian cover, as well as a significantly different community, with L. perforatum, L. weigelei, and S. brakenhielmi only found there. The importance of the channels as species and biomass enhancers in lagoon habitats was already acknowledged by several studies (e.g. Sordino et al., 1989; Procaccini and Scipione, 1992; Macali et al., 2013), and this is mostly the result of the peculiar hydro-dynamic conditions occurring in these areas and of the dynamic balance between the massive arrival of larvae and the interference that currents may cause to larval settlement (Bingham and Young, 1991; Valentine et al., 2009). The higher ascidian cover, and particularly of NIS, found in the CC dock wall is presumably influenced by the general similarity of conditions between the eastern side of the Miseno Lake, the CC, and the nearby Bacoli Bay, an additional enclosed bay connected to the Miseno Lake through the CC itself. On the other hand, the MC is characterized in its outer part by an open marine environment dominated by photophilic algae, thus suggesting that only this channel has a true role as an ecotone, with a somewhat impoverished presence of brackish-water ascidians colonizing the dock wall. The marked difference observed in transects is, instead, likely due to the different substrate composition (rocky in the CC and sandy with scattered rocks in the MC).

The Mediterranean Sea is widely acknowledged as hotspot of native biodiversity, but at the same time it also hosts a high number of NIS (Coll et al., 2010; Zenetos et al., 2017), arriving in the basin through different pathways and with varying patterns by ecoregion (Galil, 2009; Zenetos et al., 2012; Katsanevakis et al., 2013; Katsanevakis et al., 2014; Convention on Biological Diversity, 2014). However, once NIS arrive in a biogeographic sector, the high functional connectivity in the Mediterranean Sea favors their secondary spread, often aided by small-scale (e.g. fishing and recreational boats) to medium- and large-scale (e.g. shipping) human transport (Katsanevakis et al., 2014; Ferrario et al., 2017), thus obscuring their routes and complicating the tracing of putative introduction pathways unless at early stages of arrival.

The conspicuous community of NIS ascidians found here in syntopy in the Miseno Lake immediately raises the question about their origin. The assemblage was composed of species native to different localities worldwide (from Western Atlantic to Indo-Pacific), thus suggesting that such co-occurrences are the result of different introduction events. However, the majority of them are renowned worldwide invaders that also already colonized other Mediterranean areas, and also sometimes co-occur there (e.g. Mastrototaro et al., 2008; López-Legentil et al., 2015; Casso et al., 2018). In addition, they are widely known to colonize new areas through shipping, whether in ballast water or as fouling communities of ship hulls (Gewing and Shenkar, 2017; Lambert, 2019), or through aquaculture (Mastrototaro et al., 2019; Ramos-Esplá et al., 2020), with the sole exception of P. constellatum, commonly considered to be a Lessepsian migrant (Halim and Abdel Messeih, 2016; Aydin-Onen, 2018; Montesanto et al., 2022). However, even for the latter species, the present record from the Miseno Lake, as well its presence in the area since 2014, would dismiss such hypothesis and rather suggest that transport-stowaway may have played a primary role also here, or that at least a combination of both pathways may have occurred in invading the Mediterranean Sea. Unfortunately, with few exceptions (e.g. Pineda et al., 2011; Pineda et al., 2016a; Pineda et al., 2016b for S. plicata and Rius et al., 2008 for M. squamiger), no detailed genetic information is known at a population level for all these species in the Mediterranean Sea and even worldwide, which makes it difficult to trace putative sources and pathways. This is also particularly relevant in the case of D. bermudensis, whose clade C is also present at least in Puerto Rico (see Streit et al., 2021), but for which no additional information is available from other Mediterranean localities, thus making it unknown whether other lineages of the complex are invading the basin yet. Finally, the general absence of genetic data in support of faunal studies worldwide is also why at least two of the species found here are considered as cryptogenic in the Mediterranean Sea, namely P. viridis and D. pseudovexillum. The former species was already discussed above, whereas, in the case of the latter species, further integrative taxonomic approaches focused on “whitish/creamish didemnids” could easily reveal that its distribution is not only overlooked in the Mediterranean, but even worldwide. Notwithstanding these premises, the known auto-ecology of these species and the presence of mussel farms and two active marinas in the vicinity of the Miseno Lake point towards the speculation that these areas may be the first sites of impact and thus the true reservoir of these species. Further field studies to be carried out in these localities will confirm or dismiss such a hypothesis.

Moreover, although this benthic survey did not focus on species-specific impacts of NIS, and no certain evidence of competitive restriction of native ascidians was found, several additional considerations can be made. As already mentioned above, lack of knowledge and mostly of field studies presumably hindered so far the recognition of true invasiveness of these species in the Mediterranean Sea. However, among these, B. niger is widely known to dominate fouling communities in native and invaded areas (Sheets et al., 2016; Nydam et al., 2021; Ramalhosa et al., 2021), and this also holds true for A. accarense (López-Legentil et al., 2015; Montesanto et al., 2021), M. squamiger (Turon et al., 2007; Rius et al., 2008), P. zorritensis (López-Legentil et al., 2015; Casso et al., 2018; Ramalhosa et al., 2021), and S. brakenhielmi (Mastrototaro et al., 2019; Ramos-Esplá et al., 2020), which are worryingly expanding in the Mediterranean and becoming stable presences in littoral environments. In addition, the eminent invader P. constellatum is also considered a serious potential threat to native communities (Tovar-Hernández et al., 2010; Govindharaj et al., 2022). All these statements well match the high abundances of NIS observed here. Two of these species, namely D. bermudensis and B. niger, were also found to aggressively overgrowing tubes of the polychaete Chaetopterus variopedatus in the sectors close to CC (C8, D8, E7, E8, E9), sometimes totally covering the aperture of the tubes and thus providing evident stress to the native polychaete (see Figure 2G). Such interaction was never reported before in the Mediterranean Sea, but aggressive overgrow on benthic organisms by other members of the genus Distaplia Della Valle, 1881 was already reported for Distaplia viridis Kott, 1957 and Distaplia cf. stylifera (Kowalevsky, 1874), and may possibly lead to mass mortality events of the local biota (Russ, 1982; Moreno-Dávila et al., 2021). Although true impacts of these NIS on C. variopedatus have still to be properly evaluated, the observed occurrence was worrying since most of the tubes observed were affected, and D. bermudensis already showed aggressive behaviour, as also highlighted by comparison of CC dock wall communities before and after its arrival. Finally, the finding of S. brakenhielmi in a typically marine environment in the outermost part of the MC, and its absence in other areas of the lake widely colonized by NIS, somewhat disagrees with previous Mediterranean records of this species, generally confined to mussel farms and port environments (Mastrototaro et al., 2019; Ramos-Esplá et al., 2020), and suggests that this species may easily thrive even outside the areas usually targeted for NIS detection.