Genetics reveal the identity and origin of the lionfish invasion in the Mediterranean Sea

Following aquarium releases, invasive lionfishes have colonized large areas of the Caribbean and western Atlantic, resulting in an immense ecological damage. The early stages of that invasion are poorly known. Indeed, a lag of time between the introduction and detection often preclude genetic characterization of that crucial phase. With elevated awareness, the recent invasion of Pterois miles was quickly detected in the Mediterranean Sea. We hereby show that the very first individuals establishing populations in the Mediterranean Sea display haplotypes that nest within the large genetic diversity of Red Sea individuals, thus indicating an invasion via the Suez Canal. We also show that only two haplotypes are detected in the Mediterranean Sea, suggesting that few individuals may have been involved in the invasion. Thus, we conclude that the Mediterranean invasion is the result of a movement of individuals from the Red Sea, rather than from other means, and that low genetic diversity does not seem to have a negative effect on the success and spread of lionfish into the Mediterranean Sea.


Results
DNA sequencing. Fourteen P. miles mitochondrial control region sequences were obtained from Lebanon (6), Cyprus (7) and Greece (1). Alignment of control region sequences resulted in 352 base pairs (bp) with no gaps being necessary for perfect alignment. Out of those 352 bp, 10 bp were variable, resulting in two haplotypes. Thirteen individuals were identical and differed from the last individual by 10 nucleotide positions.

Phylogenetic reconstructions and molecular identity of Mediterranean samples.
To present our results in a broader context, we included previously published lionfish sequences in our analyses. Neighbour-Joining and Maximum Likelihood methods resulted in identical tree topologies (Fig. 1). For reasons of graphical simplification, only eight out of the 13 previously published haplotypes were used (Fig. 1). The inclusion of the remaining five haplotypes, and all sequences available in Genbank (a total of 105 individuals) did not change the tree topology (Supplementary Figure S1).
Sequences from Mediterranean individuals clustered with published sequences assigned to Pterois miles with very strong bootstrap support (100% of replicates). This result was due to the presence of only a single nucleotide difference between the GenBank control region (P. miles, Red Sea, accession number LK022697) and the sampled haplotype MA, as opposed to 35 changes with the P. volitans sequence, a congeneric species to P. miles 26,27 . Our results were further illustrated by the phylogenetic analysis revealing a close relationship of our Mediterranean individuals to those collected in the Red Sea (Fig. 1).  Fig. 1). These haplotypes differed at 10 nucleotide positions, revealing a difference well within the observed range of divergence among P. miles individuals collected in the Red Sea 28 . Both haplotypes were very closely related to those previously described as Red Sea haplotypes ( Fig. 1; haplotypes 1 and 35 28 ).

Discussion
Origin of the Mediterranean population. Four main hypotheses may explain the establishment of the lionfish, Pterois miles, in the Mediterranean Sea: Figure 1. Neighbour-Joining (NJ) tree-based K2P distances of mitochondrial control regions of the common lionfish, Pterois miles, from the Red Sea (red, from GenBank) and the Mediterranean Sea (blue, sampled) (identical topology obtained using Maximum Likelihood, ML). Bootstrap support values higher than 50% are shown next to the corresponding nodes where the first value represents the NJ distance and the second value represents the ML value. PMI: Pterois miles; CYP, Cyprus; RHO, Rhodes in Greece; LEB, Lebanon. GenBank accession and haplotype numbers are indicated 28 . In black is the control region extracted from a complete mitochondrial genome of P. miles from the Gulf of Aqaba and as an out-group from its congeneric P. volitans.
1. an accidental introduction by an aquarium release event, similar to the situation observed in the western Atlantic, 2. a larval transport via ballast water through the Suez Canal 25, 29, 30 , 3. a long-distance larval dispersal event from an Atlantic source, or 4. an immigration of individuals through the Red Sea via the Suez Canal (Lessepsian immigrants), as previously described for several other NIS 22,31 .
While a definite exclusion of the former three hypotheses is not possible, our results strongly favour an arrival through the Suez Canal. All samples analysed nested within samples collected from the Red Sea (where very limited exports for the pet trade occur). The possibility of ship mediated ballast transport was considered as less plausible since only large adults have been reported and not necessarily close to major ports of the Mediterranean Sea [14][15][16][17][18][19][20] . Moreover, the fact that lionfish sightings followed the common invasion pattern of Lessepsian immigrants -first established in the Levant, and then gradually spreading into the eastern Mediterranean (stepping stones), is consistent with an invasion of the Mediterranean Sea through the Suez Canal, which also exclude the Atlantic origin possibility. To this end, the most likely explanation for the Mediterranean invasion is therefore an active movement of individuals from the Red Sea into the Mediterranean through the Suez Canal. Our results revealed the presence of two haplotypes that may also indicate the entry of two individuals. Furthermore, the recent sightings in several places in the Mediterranean Sea (Lebanon, Cyprus, Turkey, Greece, Tunisia) in relatively short intervals, also point to a rapid spread in a specific direction rather than multiple isolated introductions. Such a rapid spread is consistent with the presence of the same, most frequent haplotype, in all three sampled localities (Lebanon, Cyprus and Rhodes). That common haplotype was also found in our earliest collection, 2013.
Haplotype diversity and frequency. While the presence of a population bottleneck in the early phase of invasion, due to founder effects seems intuitive, little evidence in support of this has been found for Lessepsian immigrants 32,33 . Even for studies based on very few samples and early detection, such as for the parrotfish Scarus ghobban, there was no evidence of population bottlenecks as shown by unique haplotypes for each of the five sampled individuals 34 . For the Mediterranean lionfish, only two haplotypes out of 14 samples were found, one being prominent (13 individuals, MA) while the other one (MB) only seen in a single individual. Furthermore, no spatial or temporal partitions were found. The single haplotype MB was found in Cyprus, where the other haplotype MA is also present, and both MA and MB haplotypes were collected in 2016, while the common haplotype MA was collected in all sampling years (2013, 2015 and 2016). However, in a previous study on P. miles performed on 88 individuals collected in the Gulf of Aqaba and the northern Red Sea, 13 haplotypes were uncovered thus setting up an expectation of 2.07 haplotypes out of 14 samples 26 . This expectation matches our observed haplotype frequency (2 haplotypes out of 14). Since the standard expectation that a NIS should experience a reduction in genetic diversity has rarely been the case with Lessepsian immigration, the cause could be the multiple entry possibility via the Suez Canal, which acts as a permanent open gateway.
Significance of repeated introductions. The single individual of P. miles reported in 1991 from the Levant 21 was most possibly a non-successful invasion attempt since no other individual was reported thereafter. The re-appearance of P. miles in the Mediterranean Sea, almost two decades later, is almost certainly due to the arrival of new propagules that resulted in a successful establishment. True introduction rates have changed over  35 . A similar pattern of repeated introductions was observed for another highly invasive species in the Mediterranean Sea, the bluespotted cornetfish Fistularia commersonii. The first specimen was captured in 1975 but the real establishment and spread started in 2000 and covered the entire Mediterranean Sea 36-38 . The conspicuous morphology of both the lionfish and the bluespotted cornetfish makes their presence in the interim extremely unlikely. Both cases may suggest that the susceptibility of the Mediterranean to biological invasions has been changing as a result of various biotic and abiotic events and processes 39 and led to the establishment of new propagules from adapted lineages. At a larger scale, the high number of recent records of newly introduced NIS with historically low introduction rates, such as the lionfish, shows a shift in the nature of Lessepsian immigration 35 .
The future of the lionfish invasion. The lionfish invasion in the western Atlantic has revealed huge detrimental ecological effects on the local ecosystems. While a model predicted that a natural invasion of the Mediterranean Sea by lionfish was unlikely to be successful 24 , and suggested that the Mediterranean lionfish were probably released through means of aquarium trade, our data suggest that this is unfortunately not the case. The recent and rapid expansion of the lionfish in the Mediterranean is therefore alarming and requires the immediate action of all concerned stakeholders in the area.

Materials and Methods
Sampling. Samples were collected for this study by spear or trammel nets from several localities in Lebanon, Cyprus and Greece (Rhodes), between 2013 (one year after its first record in the Mediterranean Sea) and 2016 (Table 1, Fig. 2). They correspond to the very first individuals to be collected from each region. Prior to analysis, all species used in this study were identified following 40 . Fin tissue samples were preserved in 95% ethanol at room temperature. All animal procedures were performed in accordance with UCSC Institutional Animal Care and Use Committee (IACUC -BERNG1601).
DNA extractions and PCR amplifications. DNA was extracted using standard extraction kits following the manufacturer's protocol (Blood and Tissue DNA extraction, Qiagen). In order to compare our data with previously published sequences 28 , we amplified and sequenced the mitochondrial control region. Amplification of the mitochondrial control region (also called D-loop) was accomplished with universal primers CR-A and CR-E 41  After purification following the manufacturer's protocol (ABI, Perkin-Elmer), sequencing was performed with the primers used in the PCR amplification on an ABI 373 automated sequencer (Applied Biosystems, Foster City, CA). To supplement our data, we used GenBank data of sequences for P. miles and its congeneric, P. volitans, where 13 out of 88 individuals were Red Sea individuals 28 . All sequences from this study were deposited in Genbank under the accession numbers KY713609-KY713622. Phylogenetic analyses. Amplification products were sequenced and the computer program Clustal W implemented by Geneious was used to align the mitochondrial control region sequences. Phylogenetic reconstructions were performed based on the Neighbour-Joining method generated in R 42 with the use of the ape package 43 . Genetic distances were calculated using a Kimura 2 parameter method. We also used the maximum likelihood (ML) method as a second phylogenetic reconstruction approach, as implemented in GARLI 44 . To estimate support for the nodes, 1000 bootstrap replicates were performed and we retained only the values supporting the nodes accounting for more than 50% of the bootstrap replicates. Data Availability. 14 sequences were deposited at Genbank under the accession numbers KY713609-KY713622.
Ethical Statement. All animal procedures were carried according to protocols approved by the IACUC at the University of California -Santa Cruz, and all methods were carried out in accordance with relevant guidelines and regulations (IACUC -BERNG1601).