High levels of population genetic differentiation in the American crocodile (Crocodylus acutus)

The American crocodile (Crocodylus acutus) is a widely distributed species across coastal and brackish areas of the Neotropical region of the Americas and the Greater Antilles. Available information on patterns of genetic differentiation in C. acutus shows a complex structuring influenced by interspecific interactions (mainly hybridization) and anthropogenic actions (mostly historical hunting, recent poaching, habitat loss and fragmentation, and unintentional translocation of individuals). In this study, we used data on mitochondrial DNA control region and 11 nuclear polymorphic microsatellite loci to assess the degree of population structure of C. acutus in South America, North America, Central America and the Greater Antilles. We used traditional genetic differentiation indices, Bayesian clustering and multivariate methods to create a more comprehensive picture of the genetic relationships within the species across its range. Analyses of mtDNA and microsatellite loci show evidence of a strong population genetic structure in the American crocodile, with unique populations in each sampling locality. Our results support previous findings showing large degrees of genetic differentiation between the continental and the Greater Antillean C. acutus. We report three new haplotypes unique to Venezuela, which are considerably less distant from the Central and North American haplotypes than to the Greater Antillean ones. Our findings reveal genetic population differentiation between Cuban and Jamaican C. acutus and offer the first evidence of strong genetic differentiation among the populations of Greater Antillean C. acutus.


Background
The American crocodile is widely distributed along coastal swamps, estuarine rivers, and lakes of the Neotropical region of the Americas and the Insular Caribbean.It is primarily a coastal crocodilian inhabiting brackish waters and the saltwater sections of rivers, coastal lagoons, and mangrove swamps (1,2).The species also occurs in freshwater areas located inland (3)(4)(5).This wide-ranging species overlaps with three other new world crocodilians: the Cuban crocodile (Crocodylus rhombifer) in Zapata, Cuba (6), Morelet's crocodile (Crocodylus moreletii) in the Yucatan Peninsula, Mexico (7,8) and Belize (9), and the Orinoco crocodile (Crocodylus intermedius) in Venezuela (10).It hybridizes in the wild with C. moreletii (8,11,12)and C. rhombifer (6,12).The ability to tolerate and thrive in highly saline environments (13,14) allowed C. acutus to occupy a unique ecological niche and extend geographically more than any other new world crocodilian.Like many crocodilians, populations of American crocodiles have declined due of intense hunting pressures since the early 1800s.The species was listed in the Convention on International Trade in Endangered Species of Flora and Fauna (CITES) at the Convention's inception in 1975 (15).Since then, a combination of sustainable farming (15), habitat protection (16) and conservation and monitoring programs (17) have aided the recovery of the species.
However, this recovery has not been even throughout its range, and new conservation challenges have arisen.New threats have been driven by changes in water regimes and water quality of coastal wetlands, illegal hunting (18), human-mediated movements of animals across the range (19), and habitat loss and fragmentation due to coastal development (16).The species is categorized as vulnerable by the International Union for Conservation of Nature (IUCN) Red List of Endangered Species (20) and listed in CITES Appendix I (16), which prohibits trade, except for the Cuban population which was transferred to Appendix II in 2007 (1).
An important challenge in the conservation of any widespread species is to identify unique clusters that need to be considered as independent management units (21).These clusters are usually de ned considering habitat types within the species range, local conservation status, threats, enforcement and legislation (16,22), as well as patterns of genetic variation within and among subpopulations.Up until recently, most of the work on C. acutus genetic variation has focused on zones where it can hybridize with C. rhombifer and C. moreletti (6,23).Though information about genetic differentiation has been used as a tool for delineating local management units on some species of Crocodylus (24), thereis still a need to incorporate population-level genetic information into conservation planning for many species of crocodilians, including the American crocodile.
Current information on patterns of genetic differentiation of C. acutus show complex population structuring in uenced by interspeci c interactions (mainly hybridization) and anthropogenic actions (mainly poaching and unintentional translocation of individuals).For example, an analysis of variation in mtDNA control region revealed a minimum of 11 haplotypes in the American crocodile (19,23).Nine of these haplotypes are also present in Central and North American populations (19), and only two are exclusive to the Greater Antilles (23).C. acutus populations from South America, however, have not been incorporated in these analyses, and thus, the relationship of Antillean crocodiles with South American lineages is unclear.Moreover, hybridization with other crocodilians and humanmediated migration across the species range may affect the distribution and frequency of the reported haplotypes (19).A study based on mitochondrial (mtDNA) and nuclear DNA markers revealed interspeci c hybridization, admixture, and signi cant patterns of population substructure of crocodiles in Southern Florida as a result of humanmediated migrations from Latin America and the Greater Antilles (19).More recently, a study on populations of C. acutus in the Greater Antilles and the Americas recognized distinguishable phylogenetic relationships and high genetic divergence between these two groups (25).C. acutus populations along the Paci c coast of Costa Rica exhibited small to moderate levels of inbreeding and signi cant levels of population differentiation potentially attributed to the lack of connectivity between some localities and the occurrence of population bottlenecks in the past (26).In addition, Pacheco et al (2018)'s extensive sampling throughout Mexico revealed several unique lineages as a result of historical and present hybridization between C. acutus and C moreletti.In parallel to C. acutus' species designation debates, it is imperative to strengthen the comparative research of extant populations to inform local and regional management, and incorporate lacking information into the broader species assessment.
De ning conservation management units based on molecular data has proven more effective when combining a wide range of genetic markers (27,28).The hypervariable control region of the mitochondrial DNA has been one of the most informative and widely used markers to uncover population subdivision (27,29).In addition, nuclear polymorphic microsatellite markers have proven useful to detect subtle patterns of population structure (30,31), and to accurately assign population origin (24).Information on population differentiation of American crocodiles derived from mtDNA markers has been useful to detect hierarchies of structuring but has not resolved subtle patterns of population subdivision (19).
To better understand patterns of genetic differentiation in the American crocodile we used data on mtDNA and nuclear DNA to assess the degree of population structuring between and among localities in South America, North America, Central America and the Greater Antilles.We incorporated previously unsampled areas (two sites in Jamaica, and one site in Venezuela) and populations previously studied to create a more comprehensive picture of the genetic relationships within the species across its range.Our work aims at understanding the population genetic structuring of C. acutus, and to provide relevant information for the conservation and recovery of the species.

Mitochondrial sequence data
We identi ed a total of ten mtDNA haplotypes across populations of C. acutus in the Americas and Greater Antilles (Table 2).Four haplotypes had not been previously described, while the other six matched previously identi ed ones (11,19).Haplotype β, described by Cedeño-Vazquez et al. (2008) was observed in the Greater Antillean population.The most common haplotype in the Everglades was Ca1; noteworthy, we also found two individuals with haplotype β and one with a newly reported one (Ca10).. Crocodiles from Belize exhibited the two more common haplotypes for the species in Central and North America (Ca1 and Ca2).. We found three haplotypes for South American crocodiles (Ca11, Ca12, and Ca13)..
Crocodiles from Jamaica and one locality in Cuba (WRMC) showed the lowest values of mtDNA genetic diversity (Table 2).Maternal lineages sampled from these populations seemed to be identical in terms of molecular distances, with the same mtDNA haplotype (β) and values of zero for pairwise differences among sequences and nucleotide diversity.Likewise, samples from Costa Rica had values of zero in all genetic diversity estimators and only one haplotype.However, as GeneBank available sequences encompassed only ve samples, this may not represent the true number of haplotypes present in Costa Rica.Samples from Mexico exhibited the highest values of haplotypic diversity and was followed by samples from Venezuela, Cuba (ZAP_CU) and the Everglades in the United States.Molecular distances, however, were relatively similar among Mexican samples and were higher for Zapata, the Everglades, and Venezuela.
We detected strong patterns of haplotype distribution across geographic locations.The median-joining haplotype network (Figure 1B) reveals differences in haplotype identity between the Americas and the Greater Antilles, in accordance with recent ndings (25).Cuba and Jamaica comprise a cluster separated from all other populations.
Haplotypes present in C. acutus from the Everglades, Central America and Venezuela are closely related than those in the Antilles.

Microsatellite data
The analysis of microsatellite data did not nd evidence of LD and the null hypothesis of HW equilibrium could not be rejected (p > 0.05).We found no signi cant differences between the expected heterozygosity under HW and the observed heterozygosity in the data for any of the putative populations (Table 3).Loci Cj109 and Cj127 had the highest allele counts, while CUJ131 and CUD68 showed the lowest.
The results from Micro-checker showed no evidence of scoring error due to stuttering or large allele dropout and crocodile populations from the four countries sampled (United States, Cuba, Jamaica and Belize).There was evidence of shared null alleles (Cj20, Cj35, Cj109, Cj131) in all localities.Microsatellite null alleles are commonly encountered in population genetics studies (32) but appear to have little effect on the outcome in Bayesian assignment analyses (33).

Mitochondrial sequence data
The among-groups component of the AMOVA analysis was signi cant when both the haplotype frequencies and the molecular distances were considered (F ST : 0.96, p<0.0001;Φ st = 0.75, p<0.001).Similarly, the χ 2 showed signi cant differentiation among populations (χ 2 = 659.03,p<0.001, df = 72).We found signi cant structure between the same pairs of populations when performing the exact test of population differentiation (Table 4).Pairwise comparisons showed signi cant structuring for all pairs except for Mexico -Belize, Mexico -Costa Rica, and Cuba -Jamaica.Greater Antillean C. acutus populations exhibited strong differentiation when compared with all other populations with xation indices ranging from 0.45 to 1 (F ST ).

Microsatellite data
The among-groups component of the AMOVA analysis was signi cant (F ST : 0.20, p<0.0001).In contrast to the data from the mtDNA, pairwise comparisons between Greater Antillean populations from Cuba and Jamaica did reveal population structure; congruent to the mtDNA analysis, all remaining comparisons between population pairs were also signi cant (Table 5).
Spatially explicit Bayesian clustering analysis implemented in Geneland yielded a modal number of four populations.All ten independent runs converged in four populations.The comparison of the posterior probability of assignment of individuals to populations allocated each individual to the population from which it originated.Non-spatial Bayesian clustering implemented in STRUCTURE with no prior distribution revealed subdivision of the samples into three populations (K = 3) with Delta K = 950.33(Figure 1D); clustering individuals from the United States and Belize and separating Cuba from Jamaica.The second-best con guration assigned individuals into four populations: Belize, Cuba, Jamaica, and the United States with Delta K = 147.37(Figure 1D).

Discussion
We present new and relevant information to understanding C. acutus evolutionary history and reveal patterns of genetic partitioning relevant to its conservation.As a widespread species living across a diversity of fresh and brackish water habitats, able to move long distances across saltwater, the American crocodile has been often considered "homogenous" across its range.With most research focusing on hybrid zones (6,8,11,16,34), our study contributes to current debates on the taxonomic complexity of C. acutus (25,35) Overall, our ndings improve our understanding of populations of C. acutus across seven countries and provide sound evidence of genetic structuring with direct consequences into management of local, unique populations.
Analyses for both mtDNA and nuclear markers show evidence of population genetic structure in the American crocodile with unique populations in North America, Central America, South America, and the Greater Antilles.In accord with other studies (6,19,25) our results show the greatest degree of genetic differentiation between the continental and Greater Antillean C. acutus.The suite of genetic differentiation estimators used in our mtDNA analysis indicated that this differentiation is strong, reaching pairwise comparisons estimator values (F ST and Φ ST ) as high as one between Greater Antillean and continental populations.The same estimators found a lack of differentiation between populations in Mexico and Costa Rica, suggesting deep matrilineal phylogenetic divergences between Central America populations and their conspeci cs in the Insular Caribbean.
We report four newly discovered haplotypes: three in Venezuela (Ca11, Ca12, and Ca13) and one in the Everglades (Ca10) (Accession number pending).Our results show that Venezuelan haplotypes are considerably less distant from Central and North American haplotypes than Greater Antillean haplotypes (Figure 1B).Notably, the addition of South American sampling locations into haplotype designation was useful to improve the understanding of C. acutus intraspeci c population genealogies.
A study by Bloor et al. (2015), using mtDNA cytochrome b and cytochrome oxidase I gene sequences in captive C. acutus from Colombia, revealed two distinct lineages: one closely related to Central American haplotypes and a second one unique to Colombian C. acutus.Our ndings strengthen such evidence and highlight the importance of future research to better understand haplotype relationships within South American C. acutus.
In addition, all but one sample from the Greater Antilles share the same haplotype (β).. Haplotype β is closer to Cuban crocodile haplotype α than to any other haplotypes from the Americas (6,23); mitochondrial capture may have occurred during an ancient hybridization event between Greater Antillean C. acutus and C. rhombifer (19).An alternative scenario postulates multiple colonization events to Cuba by C. rhombifer and later by C. acutus, with sustained periods of hybridization and dispersion across the island (23).In fact, hybridization between Cuban and American crocodiles in Cuba seems to have taken place both historically and in recent times (6).Similarly, ancestral and present hybridization has occurred between the American and Morelet's crocodiles in the Yucatan Peninsula (8,35).Regardless of potential evolutionary explanations for haplotype origins, the geographic distribution of C. acutus haplotypes suggests that strong genetic structuring shaped haplotype identities, followed by differences in haplotype frequencies.
The new haplotype (Ca10) found in the Everglades seems to be closer to South and Central American haplotypes.In addition, we found an individual in the Everglades with haplotype β.Rodriguez et al. (2011) reported the presence of several haplotypes from Latin America and the Caribbean within Southern Florida and attributed genetic structuring as a result of the admixture of local haplotypes with those of foreign and captive American crocodiles.In addition, the authors reported only one haplotype (Ca1) in core nesting areas in the Everglades.In this context, the two additional haplotypes present in our samples from the Everglades (β and Ca10) are most likely introduced or admixed individuals, suggesting that individuals with foreign haplotypes are being released and/or moved from areas nearby, where non-local haplotypes have been previously reported.
Our analysis of microsatellite data further uncovered patterns of genetic subdivision between populations of the Greater Antilles, Belize and the United States with at least three populations.It is more likely that four populations occur within these sampling locations as found by Geneland and the DAPC, as the Evanno method may underestimate K when a hierarchical structure is present.Spatial explicit models applied to other studies have detected biologically meaningful clusters where STRUCTURE failed to detect any population subdivision (37,38) and might be more accurate for populations exhibiting some degree of isolation by distance (39).Overall, we found sound evidence supporting that Cuban and Jamaican C. acutus represent two distinct populations.Presence of admixture in Jamaica and Cuba could be explained by natural or human-mediated migration of individuals between localities.

Conclusion
Although sampling was limited and extremely challenging as a result of the current political and administrative environment where C. acutus is distributed, our study provides a rst comparison among populations across the species range and robust evidence of genetic differentiation among populations of the Greater Antilles.Information at the regional-scale is crucial for regional planning and conservation of the species.With this evidence and that of previous studies (25), we propose the incorporation of independent conservation management units for Cuban and Jamaican C. acutus.The di culty to access Black River Lower Morass restricted our sampling in Jamaica and biased our results to samples from Portland Bright Protected Area, were we allocated signi cantly more sampling effort.We strongly suggest that future studies expand the sampling area to further clarify potential patterns of subtle structuring between American crocodiles at these two locations in Jamaica.
Future research should also incorporate comparative analysis for other sampling localities in South America and Central America.A broader assessment of C. acutus will further aid conservation efforts and population management decisions to achieve an effective range-wide conservation strategy for the species.Additionally, future studies should aim at using whole-genome sequencing to resolve population structuring further and to understand the phylogenetic evolution of the American crocodile and its relationships with other crocodilian species in the Americas and the Insular Caribbean.Finally, as research looks deeper into C. acutus lineages across its range and suggests the occurrence of potential cryptic species as a result of ancient hybridization and other evolutionary processes (25,35) our study aims at providing population-level data to inform current management and conservation as we continue our debate into C. acutus species designations.

Sample collection
As part of crocodilian conservation and monitoring programs, we collected tissue samples of C. acutus from ve countries (United States, Belize, Venezuela, Southeastern Cuba, and Jamaica).The dataset was supplemented with mtDNA sequences obtained from Genbank from Mexico, Costa Rica, and Southwestern Cuba (Figure 1A; Table 1).Tissue samples were removed from the dorsal section of the base of the tail and stored at -20 °C in 95 % ethanol until DNA extraction.Sexes of the individuals were unknown.Protocols for animal handling and collecting biological samples were approved by the Field Veterinary Program of the Wildlife Conservation Society and conducted in accordance with guidelines of the IUCN Crocodile Specialist Group (40).Samples from all sites except Venezuela were transported to the Sackler Institute for Comparative Genomics at the American Museum of Natural History, New York, USA (CITES export permits C0001733 and JM2320, and USFWS permit 1018-0093) for laboratory analyses including DNA extraction, ampli cation and sequencing of mtDNA control region, and ampli cation and sequencing of 11 microsatellite loci.Samples from Venezuela were transported to the Museo Nacional de Ciencias Naturales of Madrid, Spain (CITES permit 1395/VE9120190) where laboratory analysis, including mtDNA extraction, ampli cation and sequencing were conducted.Nuclear DNA was not recovered from samples originating from Venezuela, and thus, microsatellite loci could not be included in our analyses.

DNA extraction, sequencing and genotyping
We extracted genomic DNA from fresh scale tissue using the DNeasy Blood and Tissue kit (Qiagen, Valencia, CA) following manufacturer recommendations.The nal DNA concentration of each sample was measured using a NanoDrop 2000c.A fragment of 525 base pairs (bp) of the mitochondrial control region was ampli ed with primers drL15459 and CR2HA (2).Polymerase chain reaction (PCR) was performed at a nal volume of 25μl with 1.25 unit of AmpliTaq Gold (Applied Biosystems, Foster City, California, USA), 1x PCR Buffer, 0.2mM each dNTP, 0.2 μM each primer, 1mM MgCl 2 and 50ng/μl DNA.An initial 2-minute denaturation cycle at 94°C was followed by 33 cycles of denaturation at 94°C for 30 seconds, annealing at 56°C for 1min, and elongation at 72°C for 7 minutes.Amplicons were puri ed and concentrated via ethanol precipitation.Double-stranded DNA was sequenced with the dideoxy chain termination method using an ABI 3730 automated sequencer.
We used Arlequin to calculate the number of alleles, observed (H O ) and expected heterozygosities (H E ) in our microsatellite marker set.We used GENEPOP 4.3 (53) to test Hardy-Weinberg (HW) expectations per locus as well as genotypic linkage disequilibrium (LD) between loci.Sequential Bonferroni corrections (54) were used to adjust departures for HW and LD (p≤0.05).We took into account that genotyping errors could occur as a result of primersite mutations or contamination and could produce null alleles and/or allele drop out.Accordingly, genotyping inconsistencies were assessed using Micro-checker 2.2.3 software (55).
The partitioning of genetic variation among putative populations was assessed through a nested analysis of molecular variance (AMOVA) implemented in Arlequin.We computed genetic distances in Arlequin, including pairwise F ST (haplotype frequencies only) and Φ ST statistics (pairwise differences between haplotypes).The signi cance of values (Fst and ΦST) was tested using 10,000 nonparametric random permutations.Chi-square analysis was conducted in DnaSP to test for signi cant differences in haplotype frequency distributions among sampling localities.

Microsatellite data
We analyzed microsatellite data from 225 individuals from four countries: 1) Turneffe Atoll, Belize, 2) Wildlife Refuge Monte Cabaniguán, Cuba, 3) Black River Lower Morass, and Portland Bight Protected Area, Jamaica, and 4) Everglades National Park, United States (Figure 1, Table 1).We assessed spatial structure through AMOVA analysis, and estimated pairwise F ST statistics in Arlequin, using Weir and Cockerham's estimator (56), which assumes an in nite allele model of mutation.We evaluated the degree of partitioning among putative populations using spatially explicit Bayesian clustering methods.
The package Geneland 3.1.4(57,58) implemented in R 3.2.3(59) uses an algorithm based on a spatial model which assigns individuals into a number of genetic clusters (K) making use of genotypes as well as spatial coordinates of sampled individuals.As is common to all explicit clustering methods, Geneland weights information of an individual's location in the search for the most likely K instead of assuming that all clustering solutions are equally possible.Whereas, STRUCTURE 2.3.4 (60-62) uses the number of populations (K) as a xed parameter to estimate the log-likelihood of the data for the pre-de ned K values, and assigns memberships for all individuals in the total sample set.Both programs infer unknown parameters through Markov Chain Monte-Carlo (MCMC) computations, assume Hardy-Weinberg equilibrium with linkage equilibrium between loci, and do not require an a priori de nition of putative populations.STRUCTURE, too, allows for the incorporation of sampling location priors.Location priors allow STRUCTURE to assign individuals into genetically similar clusters considering a priori the geographic origin of each individual (63).
For Geneland, we determined K across 10 independent runs with 1,000,000 MCMC iterations and allowing K to vary from 1 to 8. We used a correlated allele model and set the maximum rate of the Poisson process at 225 (the number of individuals) and the maximum number of nuclei in the Poisson-Voronoi tessellation at 675 (three times the number of individuals) (58).The uncertainty of spatial coordinates of the collection was set at 20 meters.In our study, the uncertainty of spatial coordinates accounted for the recording error.We inferred the most likely number of clusters as the modal K with the highest posterior probability.
We ran STRUCTURE without sampling location priors, through 20 independent runs for K = 1-8.We set a burn-in period of 100,000 and 1,000,000 MCMC iterations to identify the genetic clusters.We assumed an admixture model with correlated allele frequencies, which allows individuals to have mixed ancestry.We determined the optimal value of K according to the ΔK method (64) using STRUCTURE HARVESTER 0.6.94online application (65).We performed consensus analyses for the average scores for the inferred K value in CLUMPP 1.1.2software (65).
Finally, we performed a Discriminant Analysis of Principal Components (DAPC) using the Adegenet 2.0.1 package (66) to further explore the structure of the population without making assumptions of panmixia and to complement Bayesian inference analyses.This method is useful to identify population clusters without prior knowledge and provides a robust alternative to traditional Bayesian approaches (67).

List Of Abbreviations
bp: base pairs BRLM_JM: Black River Lower Morass, Jamaica BZ: Turneffe Atoll, Belize CITES: Convention on International Trade in Endangered Species of Flora and Fauna IUCN: International Union for Conservation of Nature CR: Costa Rica DAPC: Discriminant Analysis of Principal Components EVNP: Everglades National Park, Florida, United Sates H(n): number of haplotypes Hd: haplotypic diversity HE: expected heterozygosity HO: observed heterozygosity K: mean number of pairwise differences among sequences MCMC: Markov Chain Monte-Carlo mtDNA: mitochondrial DNA MX: Quintana Roo, Mexico n: sample size, PBPA_JM: Portland Bight Protected Area, Jamaica VE: Lago Maracaibo, Cuare, and Rio Chico in Higuerote, Venezuela WRMC_CU: Wildlife Refuge Monte Cabaniguán, Cuba ZAP_CU: Zapata Swamp National Park, Cuba π: nucleotide diversity Declarations critical review of the manuscript.BW contributed with the collection of C. acutus samples in Jamaica, interpretation of data and substantial review of the manuscript.KM, JW, and JB contributed with the collection of C. acutus samples in Jamaica and critical editing and review of the manuscript.FM contributed with the overall research design, collection of C. acutus samples in the Everglades, interpretation of data and substantial review of the manuscript.The late RC contributed with the overall research design, collection of C. acutus samples in the Everglades, and interpretation of data.EH contributed with the collection of C. acutus samples in Belize, laboratory and data analyses for Belizean C. acutus samples, interpretation of data and substantial review of the manuscript.MAT, YAG, and ML contributed with the collection of C. acutus samples in Cuba, data analyses for Cuban C. acutus samples, interpretation of data and substantial review of the manuscript.GEL and GA contributed in the conception and design of the research, collection of C. acutus samples in Cuba, interpretation of data, and substantial review of the manuscript.JRG and KAV contributed in the conception and design of the research, interpretation of data, and substantial review of the manuscript.The late BT contributed with the conception and design of the research and collection of samples from Belizean C. acutus.

Figures
Figures

Table 1
Geographic distribution and description of the American crocodile samples used in this study.

Table 2
Genetic diversity indices for mtDNA data.
n: sample size, H(n): number of haplotypes, Hd: haplotypic diversity, π: nucleotide diversity, and K: mean number of pairwise differences among sequences.Haplotype counts are presented in brackets and standard errors in parentheses.Total number of samples and average Hd, π and K are highlighted in bold.

Table 3
Genetic diversity indices of the microsatellite data.HO: observed heterozygosity, HE: expected heterozygosity.Standard deviations are in parentheses.Total number of samples and loci per locality are highlighted in bold.

Table 4
Pairwise estimates of FST and ΦST between sampling localities based on mtDNA.Pairwise Fst and Φst values above and below the diagonal, respectively.Statistically signi cant values after sequential Bonferroni correction for pairwise comparisons with p<0.05 are highlighted in bold.Asterisks (*) below the diagonal indicate signi cant values (p<0.05) for the exact test of population differentiation.

Table 5
Pairwise estimates of FST between sampling localities based on mtDNA.