Genetic connectivity of lionfish (Pterois volitans) in marine protected areas of the Gulf of Mexico and Caribbean Sea

Abstract Lionfish (Pterois volitans) have rapidly invaded the tropical Atlantic and spread across the wider Caribbean in a relatively short period of time. Because of its high invasion capacity, we used it as a model to identify the connectivity among nine marine protected areas (MPAs) situated in four countries in the Gulf of Mexico and the Caribbean Sea. This study provides evidence of local genetic differentiation of P. volitans in the Gulf of Mexico and the Caribbean Sea. A total of 475 lionfish samples were characterized with 12 microsatellites, with 6–20 alleles per locus. Departures from Hardy–Weinberg equilibrium (HWE) were found in 10 of the 12 loci, all caused by heterozygous excess. Moderate genetic differentiation was observed between Chiriviche, Venezuela and Xcalak, México localities (F ST = 0.012), and between the Los Roques and the Veracruz (F ST = 0.074) sites. STRUCTURE analysis found that four genetic entities best fit our data. A unique genetic group in the Gulf of Mexico may imply that the lionfish invasion unfolded both in a counterclockwise manner in the Gulf of Mexico. In spite of the notable dispersion of P. volitans, our results show some genetic structure, as do other noninvasive Caribbean fish species, suggesting that the connectivity in some MPAs analyzed in the Caribbean is limited and caused by only a few source individuals with subsequent genetic drift leading to local genetic differentiation. This indicates that P. volitans dispersion could be caused by mesoscale phenomena, which produce stochastic connectivity pulses. Due to the isolation of some MPAs from others, these findings may hold a promise for local short‐term control of by means of intensive fishing, even in MPAs, and may have regional long‐term effects.


| INTRODUC TI ON
The lionfish species complex is composed of two congeneric species (Pterois volitans and Pterois miles) that were initially introduced on the coast of Florida. The first sighting occurred in 1985 (Schofield, 2009). By 1992, the population had increased and dispersed rapidly to the north, following the coast of the United States toward New York, and later progressing south to the Bahamas (Betancur-R et al., 2011). Currently, lionfish have spread throughout the Caribbean and have recently reached Brazil (Ferreira et al., 2015). Several features have made lionfish extremely good invaders. First, females can release 10,000-40,000 eggs at a spawning event (Morris & Akins, 2009) and show asynchronous releases (Morris, Sullivan, & Govoni, 2011;Murua & Saborido-Rey, 2003). Second, lionfish are opportunistic feeders, and feed widely on fish and crustaceans, with a list of more than 70 different species included in their diet (Valdez-Moreno, Quintal-Lizama, Gómez-Lozano, & García-Rivas, 2012). Third, lionfish possess venomous spines that prevent most native predators from eating them (Morris & Akins, 2009). Finally, lionfish can occupy a range of habitats from shallow waters to waters 300 m deep (Albins & Hixon, 2013) and tolerate a broad range of salinity (Jud, Nichols, & Layman, 2014). Little is known about the behavior of P. volitans/miles larvae; it is only known that they are dispersed by ocean currents and they have then 35-40 days of the pelagic stage (Morris, Akins, Barse, Cerino, & Freshwater, 2008;Freshwater et al., 2009). No real coordinated controls are in place for the larger region, but local control strategies involve intensive fishing of lionfish included the Marine Protected Areas (MPAs).
Marine protected areas are tools used for the administration and management of marine resources (Pomeroy, Watson, Parks, & Cid, 2005), and their success depends on connectivity among other MPAs to promote population persistence in marine ecosystems (Planes, Jones, & Thorrold, 2009). Connectivity has always been a key point in the success of MPAs strategies. The connectivity process allows transportation of organisms between populations, contributing to increased genetic diversity and colonization from distant populations (Palumbi, 2003). Connectivity occurs through larval dispersal and adult migration, the former being the primary mechanism of connectivity between reefs (Sale et al., 2010). Currently, in ecology, different definitions are used for connectivity, depending on the context, and the type of evaluation method (Lowe & Allendorf, 2010). Genetic connectivity is the amount of gene flow occurring among populations over a timescale of several generations. It determines the extent of genetic differences among populations (Sale et al., 2010). Even when the connectivity assessment is specific and unique to one species, biological models are frequently used to predict the behavior of other species with similar characteristics. Several connectivity studies have been carried out in the Caribbean, using predictive models (Abesamis, Stockwell, Bernardo, Villanoy, & Russ, 2016;Cowen, Paris, & Srinivasan, 2006) to direct genetic studies of reef species (Bakker et al., 2016;Rippe et al., 2017;Villegas-Sánchez, Pérez-España, Rivera-Madrid, Salas-Monreal, & Arias-González, 2014;Villegas-Sánchez, Rivera-Madrid, & Arias-González, 2010). However, to date, regional connectivity among MPAs in the Caribbean and the Gulf of Mexico along the Mexican coasts has not been estimated through direct methods. Studies of lionfish in the Caribbean have explored the use of molecular technologies for the analysis of the populations of P. volitans/miles from different approaches (Betancur-R et al., 2011;Bors, Herrera, Morris, & Shank, 2019;Butterfield et al., 2015;Johnson, Bird, Johnston, Fogg, & Hogan, 2016;Pérez-Portela et al., 2018;Toledo-Hernández et al., 2014;Freshwater et al., 2009). However, the current study is the first to use microsatellites. The main objective of this study was to explore the level of genetic connectivity among P. volitans populations in Marine Protected Areas in the Gulf of Mexico and Caribbean Sea, using specific microsatellite markers for P. volitans/ miles (Schultz, Fitzpatrick, Wilson Freshwater, & Morris, 2013). The rapid dispersion of this specie and environmental geographical barriers could influence locally differentiated populations, making it possible to evaluate connectivity matrices. Investigating this hypothesis may help us understand the dispersal behavior of this invasive species on a local scale and help develop guidelines for effective local and regional management. genetic structure, as do other noninvasive Caribbean fish species, suggesting that the connectivity in some MPAs analyzed in the Caribbean is limited and caused by only a few source individuals with subsequent genetic drift leading to local genetic differentiation. This indicates that P. volitans dispersion could be caused by mesoscale phenomena, which produce stochastic connectivity pulses. Due to the isolation of some MPAs from others, these findings may hold a promise for local short-term control of by means of intensive fishing, even in MPAs, and may have regional long-term effects.

K E Y W O R D S
Caribbean Sea, founder event, genetic structure, invasive species, lionfish, marine protected areas, microsatellites 2 | ME THODS
As part of the Pterois species local control strategies, personnel in charge of each MPA were involved in collecting the individuals. A fragment of the caudal peduncle muscle was extracted from each sample, preserved in 70% alcohol and stored at 4°C until DNA extraction and analysis were performed at the laboratory. The current study concentrates on the most abundant species, P. volitans.

| DNA extraction and identification of P. volitans
Genomic DNA was extracted from all samples using the DNeasy Blood and Tissue Kit (Qiagen) following manufacturer's instructions.
We added eight cycles of 15 s at 94°C, 15 s at 53°C, and 30 s at 72°C with a final extension at 72°C for 5 min to incorporate the M13 primer.
PCR reactions were analyzed using LI-COR Model 4300 DNA and microsatellite sizes were determined with Saga 3.1.2 software. values between all sites also to validate the structure and we made a sequential Bonferroni adjustment for multiple tests (Sokal & Rohlf, 1995). Additionally, we calculate a pairwise Hedrick's standardized F ′ ST index, which is a standardized measure of population differentiation. This distance is allowing comparison between studies with different types of markers and different effective population sizes (Hedrick, 2005;Meirmans, 2006;Meirmans & Hedrick, 2011). All calculations were made with GenAlEx 6.5 (Peakall & Smouse, 2006) and with Genetix (Belkhir, Borsa, Chikhi, Raufaste, & Bonhomme, 2001).

| Data analysis
Lastly, we investigated the genetic structure using a Bayesian approach considering all individuals within the same genetic entity. STRUCTURE 2.3.4 was used to optimize the variance in each genetic population. The analysis was conducted with 100,000 generations for burnin and 500,000 recorded generations for each K, repeated, for K = 1 to K = 9, 10 times for each K. Then, we used sampling locations as prior (model LOC-PRIOR, Hubisz, Falush, Stephens, & Pritchard, 2009) for three regions as locality priorities. The number of the most likely clusters was determined using the ∆K approach (Evanno, Regnaut, & Goudet, 2005), using Structure Harvester (Earl, 2012).

| RE SULTS
A total of 143 alleles were found in the 12 microsatellite loci at nine sites of 475 individuals. Observed allele size range (bp) and number of alleles (Na) in each locus were slightly wider than published by Schultz et al. (2013), with a range of 20 to 6 alleles per locus for an average of 468 amplified samples (Table 1). Departures from Hardy-Weinberg equilibrium (HWE) were found in 10 of the 12 loci (Table 1)  frequency.
The first axes of the PCoA explained 43.7% and 18.49% of the variation of the allelic configuration. The first axis of the PCoA shows two clear clusters ( Figure 2); one from the Venezuela, Xcalak, Utila, Chinchorro, and Cozumel MPAs; and a second from the Tuxpan and Veracruz MPAs. In addition, the Guatemala and Los Roques sites' data were found isolated from the two main clusters described.
We found small to moderate genetic difference across sites (Table 4). The lowest genetic differentiation occurred between Venezuela and Xcalak (F ST = 0,012). The highest differentiation values were found between Los Roques and Veracruz MPAs (F ST = 0,074). All genetic differences between sites were significant, even after a sequential Bonferroni adjustment for multiple tests.
Additionaly we obtained pairwise Hedrick's standardized F ′ ST index (Table 5), for all interested in will make comparison between studies with different types of markers and different population sizes. For more information about of the distribution of F ST data, please go to supporting information.
The number of genetic groups in the Bayesian analyses computed with all samples suggests that the best subdivision is found with ∆K = 4, followed by ∆K = 2 (Figure 3a   Our results offer a view of the same patterns with a finer resolution. The interesting thing is the structure of the populations was preserved in some nearby areas, such as Cozumel (Figure 3, K = 4).
The distribution of P. volitans genetic frequencies of our study, suggests strong self-recruitment and that the main dispersal factor, may be the influence of mesoscale phenomena, transporting larvae from one site to another (Johnston & Purkis, 2015).
When considering this hypothesis, the question is then how a species that is able to colonize rapidly the entire American Atlantic coasts will remain isolated after an initial colonization and not receive new migrants from elsewhere? At this stage, we can only formulate hypotheses and consider just the initial colonization. If we consider that the offspring from the initial colonization will rapidly occupy the suitable habitat, then new migrants will face intraspecific competition on a species that shows strong cannibalism (Dahl et (Carrillo, Johns, Smith, Lamkin, & Largier, 2015;Huang, Walker, Hsueh, Chao, & Leben, 2013). These currents could be trans- and has followed a path from east to west, traveling countercurrent.

| Fine scale genetic differentiation
The genetic structure we found shows a particular pattern regard-

| Could differences in the type of reefs cause settlement differentiation?
It is worth mentioning that the Marine Protected Areas where the collections were made differ in the type of reefs they contain, and those that have some similar physical characteristics also show a similarity in the genetic group identified in Figure 3 (K = 4). The Gulf of Mexico reefs present characteristics totally different from those of the Caribbean, due to a greater introduction of fresh water, sediments from rivers, estuaries, and lagoons that flow into its waters.
It is predominantly rocky with low coral cover, poor visibility, and relatively shallow waters (González-Gándara, 2015 (Weil, 2003). Los Roques is a reef type atoll, which is located far from the coast of Venezuela (Weil, 2003). Chiriviche in Venezuela is dominated by mangrove forests that line the coast and includes reefs around small calcareous islands patches (Weil, 2003). This is interesting because it suggests that the genetic differentiation of this invasive species could be influenced by the natural selection. This should be further analyzed with more specific studies that could lead to a complementary interpretation illuminating more thoroughly the distribution of this species.

| The probable number of the founding population
The first hypothesis about the spread of P. volitans suggests that the invasion began with six to eight individuals that escaped from an aquarium in Miami, Florida, USA (Courtenay, 1995;Morris & Akins, 2009;Whitfield et al., 2002). Betancur-R et al. (2011) mentioned the minimum number of founding individuals required to explain the observed genetic diversity in WA lionfish it was between eight and 12.
Our work revealed that the number of alleles at Pmv21 Locus was 20 (Table 1), suggesting that the number of original invasive organisms must have been at least more than 10, based on the conservative assumption that each organism had a unique pair of alleles, and without taking into account the probability of allele mutations. A recent estimate suggests a minimal source population of 118, and when accounting for allele effects, the number of colonizers increased to 272 (Selwyn et al., 2017). The conclusion of Selwyn and collaborators is more consistent with the probability of occurrence of the results that we obtained in our study.
Present microsatellites data show HW disequilibrium for excess of heterozygotes. Taking into account the study of Selwyn et al. (2017) and comparing the first studies that mentioned that the number of mitochondrial DNA Haplotypes of P. volitans is nine (Freshwater et al., 2009), it could be considered that some founding organisms were probably related or belonged to the same populations.
The excess of heterozygotes found in the study could be attributed to biological realities of this P. volitans population, because it violates the criteria of an ideal population (Selkoe & Toonen, 2006).
Although it is known that populations that present a recent bottleneck show apparent excess of heterozygotes (Cornuet & Luikart, 1996;Sakai et al., 2001), our analyses found heterozygosis in all loci; therefore, it is likely to be a reflection of the founder effect noted in other studies (Freshwater et al., 2009). Even so, this information differs with the studies of Pérez-Portela et al. (2018) and Bors et al. (2019). However, the differences may be inherent to the different molecular markers used.
Overall, our data clearly show that the early invasive history of the lionfish in the Atlantic is not as simple as a few individuals released around the 1980s. Also, this species is extremely faithful to its territory and that its movement as an adult is relatively limited to a few hundred meters (Jud & Layman, 2012). In this context, it is more likely this invasion was caused by the introduction of more than a hundred organisms of P. volitans from the same geographical area, and that were released in a single exhibition, that diverse introductions of organisms that probably took place at different chronologic moments within the last 30 years. Which makes us think about the probability that the black market of ornamental marine fish was behind of the cause of the massive liberation in an entrance port.
Which of these hypotheses is close to reality? It is material for another study. But the history of the stepwise invasion in the Atlantic should be revisited in light of this approach.

| CON CLUS ION
The origin of the Atlantic Lionfish invasion, the number of individuals that initiated it, and the connectivity patterns in the Caribbean and the Gulf of Mexico is more complex than has been previously suspected. Our findings demonstrate that there is genetic differentiation on both regional, local, and fines scales. We suggest that the colonization of each geographic area has been in temporary invasive species, acting at local scales that may have real short-term effects at regional scale.