香京julia种子在线播放

To further assess the connectivity and genetic diversity of manatees in the southern extent of the range on the eastern Atlantic coast of South America, this study included samples from the Antillean subspecies of West Indian manatees in Brazil and a few from Venezuela/Guyana, as well as Amazonian species samples from portions of their range in Brazil. Samples from the mouth of the Amazon River and further upstream were included to investigate the degree of hybridization found between the Amazonian and eastern Antillean manatees in our samples using nuclear and mtDNA. Nuclear DNA accounts for genetic material from both parents and can provide increased resolution to address the level of introgression and temporal hybridization patterns (Vaha and Primmer, 2006). The three sampled regions included: (i) the coast of Brazil across the fragmented distribution found in the states of PA, Maranhão (MA), Piauí (PI), Ceará (CE), Rio Grande do Norte (RN), Paraíba (PB), Pernambuco (PE), and AL, n = 82; (ii) a limited assessment of the state of AP (Brazil), Venezuela (VE) and Guyana (GY), n = 12; and (iii) the Amazon River, including the river mouth and upstream in the Brazilian states of AP and PA (n = 18; Figure 1). Nine individuals were sampled from captivity at Centro Nacional de Pesquisa e Conservação de Mamíferos Aquáticos but were assigned to region (i) based on their original rescue location (ICMBio/CMA database). During collection of samples in the field, individuals were identified as T. manatus (regions i and ii) or T. inunguis (region iii) based on collection location, and these assignments were utilized until microsatellite data could provide genetic assignments. The genetic data (haplotypes and nuclear DNA) of the nine manatees released into the SE and BA states were analyzed in conjunction with the data from their original stranding locations. One individual, named “Poque,” genetically characterized in previous studies (Vianna J. A. et al., 2006; Vianna J. D. A. et al., 2006), was included and identified from region (ii) based on previous findings (Luna, 2001; Luna and Passavante, 2010).

Blood or skin tissue from wild carcasses and stranded calves was collected by manatee researchers from 2009 to 2011 for DNA analysis. Blood from wild-born captive manatees with known rescue locations was also used for this study. Blood and tissue samples were preserved in lysis or tissue buffer, respectively (Amos and Hoelzel, 1991; White and Densmore, 1992; Proebstel et al., 1993). While additional samples would increase the statistical power to detect substructure and relatedness, sample collection from endangered and elusive manatees is difficult, costly, and requires extensive efforts that would delay timely information necessary for implementation of conservation measures.

Genomic DNA extraction, amplification and fragment analysis were performed at the U.S. Geological Survey, Conservation Genetics Laboratory in Gainesville, FL, United States. DNA extractions were carried out using DNeasy Blood and Tissue Extraction Kits (QIAGEN, Valencia, CA, United States). Due to the state of decomposition, or the small physical sample size, three samples were isolated using phenol-chloroform extraction (Hillis et al., 1996). Polymerase chain reaction (PCR) amplifications were completed in a Biometra UNOII, T-Gradient thermocycler (Biometra, Gottengen, Germany) or on a PTC-100 or PTC-200 (MJ Research, Waltham, MA, United States) thermocycler.

Nuclear microsatellite DNA amplifications (n = 16 loci) were conducted using the following conditions: initial denaturation at 95°C for 5 min; 35 cycles consisting of a denaturation step at 95°C for 30 s, specific annealing temperature for each primer for 60 s, and extension at 72°C for 30 s; a final extension at 72°C for 10 min (Supplementary Table 1). Amplifications were performed in a total volume of 12.7 μL, with 10 ng target DNA, 1× Sigma PCR Buffer, 2.5 mM of MgCl₂, 0.2 mM (each) dNTP, 0.01 U μL^–1 of Sigma JumpStart Taq polymerase (Sigma–Aldrich, St. Louis, MO, United States), specific quantities of each primer (Supplementary Table 1), and 1 mg mL^–1 bovine serum albumin (BSA) when indicated by García-Rodríguez et al. (2000) and Pause et al. (2007).

For fragment analysis, the forward primers were labeled with the fluorescent dyes VIC, HEX or 6-FAM for processing and visualization on an Applied Biosystems ABI 3130xl Automated DNA Analyzer (ThermoFisher Scientific Inc., Waltham, MA, United States). Fragments from the PCR products were analyzed using GENEMARKER, v. 1.5 (SoftGenetics, LLC, State College, PA, United States) to determine allele sizes. Allele sizes were standardized using previously analyzed Florida manatee samples for comparison and data binning.

Nuclear microsatellite loci were tested for the presence of null alleles with MICRO-CHECKER (Van Oosterhout et al., 2004). Probability of identity (P_ID) and P_ID among siblings (P_(ID)sib) was calculated in GENECAP to remove bias and GENALEX to calculate observed values following Hunter et al. (2010) methods for each species. A range of samples sizes to adequately identify individuals in a population were estimated using the unbiased 1/P_ID and 1/P_(ID)sib for the liberal and conservative estimates, respectively. Departures from the expected genotypic frequencies in Hardy-Weinberg Equilibrium (HWE) (Jorde et al., 2007) were calculated using the Markov chain Monte Carlo (MCMC) method (dememorization 10,000, batches 1,000, and iterations per batch 10,000) in GENEPOP (Raymond and Rousset, 1995). GENEPOP was also used to test for linkage disequilibrium (LD) among loci (Raymond and Rousset, 1995). Alpha values were adjusted with Bonferroni correction for pairwise comparisons in HWE and LD tests (Rice, 1989).

Population structure was evaluated in various Bayesian and multivariate analyses. The program STRUCTURE v. 2.3.4 was used to infer populations (Falush et al., 2007). A first approach using all samples aimed to test species identification of T. manatus, T. inunguis, or potential hybrids {K = 1–8}. Hierarchical analyses were then conducted on the species individually {K = 1–5} to identify substructure in the Amazonian manatees and to assess whether the gaps at the distribution of Antillean manatees (Figure 1) act as barriers and influence structure of the population. Hierarchical analyses were conducted in which the identified sub-clusters from the initial run were each analyzed independently. This procedure was repeated consecutively until K = 1 was achieved for all sub-clusters. We used the following parameters for all simulations of the initial and consecutive hierarchical analyses: the admixture model without a priori assignment, a run-length of 100,000 MCMC repetitions following the burn-in period of 50,000 iterations, and 20 independent iterations simulated for each value of K. The most probable number of clusters was determined by calculating ΔK and the mean estimation of ln probability in STRUCTURE HARVESTER (Earl and VonHoldt, 2009). Individuals were assigned to clusters with a threshold of q ≥ 0.80, which indicates little likelihood of belonging to a different cluster. Using the GENECLASS2 software, Bayesian assignment tests were run using Rannala and Mountain (1997) methods to identify F1 and F2 hybrid individuals with a threshold of 0.05. Samples with q ≥ 0.90 in STRUCTURE were used as a reference population for each geographic grouping to evaluate the assignment of admixed individuals (q < 0.80) identified in STRUCTURE (Piry et al., 2004). With the aim to compare between two methods, we also ran BAYESASS in order to infer the proportion of recent migrants among populations (Wilson and Rannala, 2003). BAYESASS was run for 10⁷ iterations, 10⁶ burn-in period, and a sampling frequency of 1,000. We ran several iterations with varying delta parameters to maximize the log likelihood and to ensure the acceptance rate of each was between 40 and 60%. However, in all assessments of the various groups, the pairwise comparison of migration rates showed that the derived means and 95% confidence intervals (CIs) did not differ from the expected reference CIs. Since the two derived and reference CIs did not differ, the dataset did not have sufficient information to calculate migration rates using this method; thus, we did not include this analysis in our results.

A pairwise genetic distance matrix among all individuals was created in GENALEX v. 6.501 (Peakall and Smouse, 2006) and then used to plot a Principle Component Analysis (PCA) (adegenet, Jombart, 2008; RCoreTeam, 2019). A PCA was performed as a multivariate approach by creating synthetic variables of genetic distance to determine population structure. To assess overall genetic differentiation at the population level, GENALEX was used to calculate F_ST using the infinite alleles model and R_ST using the stepwise mutation model through an analysis of molecular variance (AMOVA). Distance values represent 10 loci since data were not complete for the eastern Antillean manatees of Venezuela and Guyana and “Poque” at loci TmaE26, TmaA02, and TmaM79.

Genetic diversity metrics were used to investigate the clusters identified in STRUCTURE and the PCA including, number of alleles (N_A), effective number of alleles (E_A), number of private alleles (N_P), heterozygosity observed (H_O), and expected (H_E) for each subgroup of samples calculated in GENALEX. Kruskal–Wallis (KW) tests were performed to determine significant differences in genetic diversity among populations (stats, RCoreTeam, 2019). If significant differences were found, a Dunn’s Test was performed on pairwise comparisons to determine which pair of groupings was driving the significant difference (Dinno, 2017; RCoreTeam, 2019). The inbreeding coefficient, F_IS, was calculated in FSTAT v. 2.9.3 with significance evaluated by a 95% (CI) confidence interval (Goudet, 2002). The presence of a potential bottleneck was estimated using BOTTLENECK with 13 loci and sample groups (Cornuet and Luikart, 1996). Parameters were modified based on the use of imperfect microsatellite repeat loci as suggested, with 95% of the Infinite Allele Model (IAM) in the two-phase mutation model (TPM) with a variance of 12 applied and 10,000 iterations performed (Piry et al., 1999; Cristescu et al., 2010). Since most of the 13 microsatellites were imperfect dinucleotide repeats, bottlenecks were considered to have occurred if excessive heterozygotes were found significant for the IAM and TPM sign tests (Piry et al., 1999; Cristescu et al., 2010). Effective population sizes (N_E), for each genetic grouping identified, were calculated in NEESTIMATOR v. 1.3 using the single sample linkage disequilibrium method (Peel et al., 2004).

The mtDNA control region displacement loop was amplified with primers CR-4 and CR-5 (Southern et al., 1988; Palumbi et al., 1991), following the PCR techniques of Hunter et al. (2010). Briefly, the PCR reaction conditions were as follows: 10 ng DNA, 1× Sigma PCR buffer (10 mM L^–1 Tris–HCl, pH 8.3, 50 mM L^–1 KCl, 0.001% gelatin; Sigma-Aldrich, Inc., St. Louis MO, United States), 3 mM L^–1 MgCl₂, 0.8 mM L^–1 dNTP, 0.24 mM L^–1 of each primer, 0.04 U μL^–1 of Sigma Jump Start TaqDNA polymerase. The PCR cycling profile was: initial denaturation at 94°C for 5 min; 35 cycles consisting of a denaturation step at 94°C for 1 min, annealing temperature of 55°C for 1 min, and extension at 72°C for 1 min; a final extension at 72°C for 10 min. Amplified products were purified using the QIAquick PCR purification kit (QIAGEN, Valencia, CA, United States). DNA sequencing was accomplished in the DNA Sequencing Core at the University of Florida, Gainesville, FL with the BigDye terminator protocol developed by Applied Biosystems Inc. (ThermoFisher Scientific Inc., Waltham, MA, United States) using fluorescently labeled dideoxynucleotides (ddNTPs).

Control region mtDNA sequences were cleaned and aligned with GenBank sequences using GENEIOUS v. 10.0.9 (https://www.geneious.com) and MESQUITE v. 3.51 software (Maddison and Maddison, 2018). Sequences published by Vianna J. A. et al. (2006) were obtained from GenBank (Accession numbers: AY963852-56 and AY963859-61) to compile haplotypes for the Venezuela/Guyana populations. Haplotypes of each individual were identified using reference sequences in GenBank and visualized as a Median-Joining Network (MJN) using POPART v. 1.7 software (Bandelt et al., 1999). The MJN was run for simplicity and with complexity (epsilon = 0 and 1, respectively) to visualize relationships among the three groupings but only the simplified network was reported. Variable characters within haplotypes were identified using Mesquite software (Maddison and Maddison, 2018). Samples grouped by localities were evaluated for genetic diversity through the number of haplotypes (HT), polymorphic sites (S), nucleotide diversity (π), Tajima’s D of site neutrality, and haplotype diversity (h) using DNAsp v. 4 (Rozas et al., 2003). Pairwise Φ_ST distance values among groupings were calculated using ARLEQUIN v. 3.5.2.2 (Excoffier and Lischer, 2010).

Our final dataset was composed of 95 samples, containing 67 from Brazil, 11 from Venezuela and Guyana, and 17 Amazonian manatees. We originally assessed 16 loci but evidence of null alleles and HWE disequilibria across population groupings resulted in the removal of three loci: TmaH13, TmaE01, and TmaE11. Manatees from Brazil and the Amazon amplified at 13 nuclear microsatellite loci, while Venezuela and Guyana (n = 11) only amplified at 10 out of 13 loci (Table 1). Observed estimated P_(ID) and P_(ID)sibs were greater, but the same order of magnitude, than the unbiased estimates for both eastern Antillean groupings (Table 2). For the Amazonian species, unbiased P_(ID) and P_(ID)sibs estimates were at least one order of magnitude greater than observed estimates (Table 2). Individual identity values allowed us to confidently distinguish individuals separately with ten loci. MICROCHECKER found evidence of null alleles in Brazilian manatees for loci TmaSC5, TmaSC13, TmaE08, and TmaA02, but found no evidence of stutter alleles and since Hardy Weinberg disequilibrium was not found at these loci, they are not likely to influence the results. Of the 13 loci, deviations from HWE were detected for the Amazonian manatees at locus TmaF14, and Brazilian manatees at loci TmaK01 and TmaA02 after alpha was adjusted using the Bonferroni method (α = 0.0004, Table 1). Some loci for all groupings were not able to be evaluated by MICROCHECKER or for HWE due to monomorphic loci. No evidence of LD was found after alpha was adjusted using Bonferroni method (α = 0.0004).

Nuclear Bayesian methods in the program STRUCTURE were used to assess the divergence of Amazonian and eastern Antillean manatees using log-likelihood and ΔK analyses (Figure 2A). In the first analysis, the two species were analyzed together, and Amazonian and West Indian samples formed K = 2 clusters, supported by the highest ΔK and lowest log-likelihood values. There, “Poque” was assigned to each cluster equally (q = 0.5; Figure 2A). Although the K = 2 hypothesis had the highest support, when the K = 3 hypothesis is considered, Amazonian, Brazil and Venezuela/Guyana populations separated into three distinct clusters. The second hierarchical analysis assessed the Amazonian sample cluster alone and identified K = 1, with no substructure in those samples (Supplementary Figure 1). There, “Poque” strongly assigned to the second cluster supported in the K = 2 hypothesis (q = 0.95; Supplementary Figure 1).

The third hierarchical STRUCTURE analysis assessed connectivity among only the Antillean manatees and identified Brazil and Venezuela/Guyana as separate clusters with support from the ΔK and log-likelihood values (ΔK = 2; Figure 2B).

There, “Poque” grouped with Venezuela/Guyana (q = 0.98). Although the Antillean Brazil and Venezuela/Guyana samples separated into two clusters, admixture (q ≤ 0.80) was identified in 17 of the 78 Antillean samples (Figure 2B). The admixed samples identified in STRUCTURE all originated from Brazil (17/67; 25%) and were found as far south as the state of PB. GENECLASS2 assigned two samples to Venezuela/Guyana (85 and 95%) and one to Brazil at 66%, but 14 samples to Brazil at 90 and 100%. No prominent geographic patterns were observed in Brazil, or in relation to admixture between the Venezuela/Guyana and Brazilian samples. The limited samples collected in the southernmost states (PE and AL) were not admixed with the Venezuela/Guyana cluster, however, RN to the north did have admixed samples.

The Brazil and Venezuela/Guyana samples were analyzed separately in the fourth and fifth hierarchical STRUCTURE analyses. No substructure was identified in either group with K = 1 being the highest supported hypothesis. When “Poque” was assessed with the three independent STRUCTURE clusters in GENECLASS2, he was assigned to Venezuela/Guyana at 100 and 0% with the Amazon and Brazil clusters.

The PCA revealed three distinct groups: Brazilian, Venezuela/Guyana, and Amazonian (Figure 3), further supporting the K = 3 clusters identified in STRUCTURE (Figure 2A). The first two eigenvalues include 91.4% of the variation in the dataset. Overall significant genetic differentiation was found in the F_ST and R_ST distance values for all pairwise comparisons of populations after Bonferroni adjustment (α = 0.017, p = 0.001) except between the Amazonian species and Venezuela/Guyana grouping (p = 0.22; Table 3). When calculating F_ST distances, AMOVA results showed the highest estimated variance within individuals (61%), then among populations (35%), and least among individuals (5%). When calculating R_ST distances, AMOVA results showed the highest estimated variance among individuals (79%), then among populations (18%), and least within individuals (2%). Pairwise comparison of Amazonian and Venezuela/Guyana manatees was the only insignificant differentiation found in R_ST values (R_ST = 0.01, p = 0.224).

Based on these results, three groups were identified comprising the Brazilian, Venezuela/Guyana, and Amazonian clusters separately. For nuclear DNA, the KW test only showed a significant difference in number of alleles (N_A) among the three clusters. Inbreeding levels were not significant based on the 95% CI (−0.126, 0.213), but were similar for Venezuela/Guyana (F_IS = 0.14) and Amazonian manatees (F_IS = 0.13) and lower for Brazil manatees (F_IS = 0.03). The Brazilian and the Amazonian manatees had a significant difference in N_A (p = 0.04), but when the pairwise comparison results were adjusted using Bonferroni correction, no significant differences were found (α = 0.017). The number of private alleles was highest for the Amazonian species (N_P = 21), then Venezuela/Guyana (N_P = 11), and lowest for the Brazilian grouping (N_P = 5). No evidence of a bottleneck was found for the Brazilian or Amazonian groupings. Venezuela/Guyana manatees showed evidence of a bottleneck under the IAM model (p = 0.027). Effective population size estimates were low for all three groupings. Venezuelan and Guyanese manatees had the lowest estimation (N_E = 14.5, 95% CI = 8.7–32.9), then the Amazonian species (N_E = 26.5, 95% CI = 17.5–48.7), and Brazilian manatees had the highest estimation (N_E = 49.9, 95% CI = 35.8–74.9).

We successfully amplified mtDNA sequences from 93 individuals from the three geographic areas of this study. Sequenced samples corresponded to the three nuclear clusters: Atlantic coast of Brazil including eight captive individuals (n = 72), Venezuela and Guyana, including “Poque” (n = 5), and the Amazonian species within the Amazon River mouth (AP and PA states of Brazil) (n = 16). Control region mtDNA was trimmed to 410 bp sequences for each individual and aligned to evaluate haplotype distribution among regions. A total of eight new haplotypes was identified in Amazonian (n = 7) and West Indian species (n = 1) and were added to GenBank (Accession numbers: MW091459-1466; Table 4 and Supplementary Table 2).

The two mtDNA MJNs supported the distinction between southeastern Antillean and Amazonian manatees (Figure 4A). The complex MJN showed a more direct relationship between the Amazonian species and Venezuelan/Guyanese Antillean manatees but still aligned with the shape of the simplified network (Supplementary Figure 2). Haplotypes P01 and T01 from eastern Antillean individuals grouped in the Amazonian species complex (Figure 4A), as reported previously (Vianna J. A. et al., 2006). The haplotype network was then broken down to include only samples grouping as Brazil (Figure 4B). One other T01 haplotype was found in the state of PA matched to the hybrid “Poque” (Figure 4B). Haplotype M01 was the most common haplotype in Brazil, with all the sampled areas having at least one M01 individual, except for the state of MA. Less common haplotypes (M03 and M07), were found in MA and PI states (Figure 4B). The cluster on the left-hand side in Figure 4A, contained Antillean samples with haplotypes found in Amazonian manatees, possibly indicating past hybridization.

The genetic diversity and differentiation values were calculated for the three nuclear clusters (Tables 1, 4). The Brazilian manatees had the highest sample size in this study but the lowest values for every mitochondrial diversity metric calculated (Table 4). Tajima’s site neutrality D values were not significant. Distance values were highest between Brazilian and Amazonian groupings (Φ_ST = 0.9), lower between the two eastern Antillean groupings (Φ_ST = 0.7), and lowest between Amazonian and Venezuelan/Guyanese groupings (Φ_ST = 0.4); all statistically significant (p < 0.001).

Samples from the primary sympatric zone at the mouth of the Amazon River in Brazil, were used to comprehensively assess the degree of contemporary hybridization between the Amazonian and Antillean manatees (Luna et al., 2008b, 2018). No evidence for contemporary hybridization was detected using nuclear multi-locus genotypes in the Antillean and Amazonian manatee samples, beyond the single previously identified captive hybrid. Nuclear microsatellite markers with samples from local Amazonian and Antillean reference populations are well suited to detect hybridization considering manatees have a long generation time of 16–23 years (Vaha and Primmer, 2006). However, more in-depth genomic investigations and future genetic monitoring with additional samples would help to investigate hybridization events in the evolutionary past and those occurring more recently. In the 95 assessed samples, the single hybrid known as “Poque” was identified by a nuclear genotype equally divided between the two species (q = 0.5 probability) and also contained an Amazonian manatee haplotype of T01 (Vianna J. A. et al., 2006). “Poque” was rescued from the north coast of Brazil, but his capture location, natal origins, and prior history were not well documented – making it difficult to draw conclusions about hybridization in wild populations. Amazonian mitochondrial haplotypes were shared with two Antillean samples, likely indicating historic hybridization or incomplete lineage sorting after a speciation event. More data are needed to assess this relationship (see Scornavacca and Galtier, 2017; Wang et al., 2018).

While we did not find evidence of hybridization in our samples from the two manatee species or at the mouth of the Amazon River, previous studies recently identified a “hybrid swarm” of multigenerational hybridization in six manatees from French Guiana and Guyana (Vianna J. A. et al., 2006; Vilaça et al., 2016; Lima et al., 2019). Lima et al. (2019) detected hybrids using one nuclear locus and one mtDNA locus which, due to incomplete lineage sorting, is not a strongly reliable approach to identify hybrids in contrast to multi-locus markers (Wang et al., 2018). Vilaça et al. (2016) assessed ∼2,000 SNPs to detect hybrids but only used a single West Indian genotype obtained from the Florida manatee genome for the Antillean genotype. Florida manatees are a separate subspecies at the northern edge of the range, which could introduce challenges during population assignment in Bayesian clustering and maximum-likelihood estimators. These estimators use the proportion of alleles inherited from the parental species, which in this case would be the single Florida sample. Therefore, additional loci and additional “parental” reference samples from local populations could allow for comparisons with local genotypes. Lastly, manatees have been translocated to closed systems as weed control agents in French Guiana and Guyana, and to captive settings which could have led to admixture outside of the species standard population range (Allsopp, 1960, 1969; Haigh, 1991; de Thoisy et al., 2003; Hollowell and Reynolds, 2005).

The Amazonian manatees exhibited evidence of a genetically diverse, panmictic population from the mouth of the Amazon River to at least 1,000 km westward into the interior basin of the Santarém region, which supported findings from a previous mtDNA study (Cantanhede et al., 2005). Conversely, the nuclear genetic diversity of the Brazilian Antillean manatee was low, containing the lowest number of private alleles among the three clusters (N_P = 5) and the largest sample sizes. Furthermore, the number of alleles in Brazilian manatees (N_A = 3) was similar to those in Belize (N_A = 3.4), but lower than studies of manatee populations in Puerto Rico (N_A = 3.9), Florida (N_A = 4.8), and Colombia (N_A = 3.5 to 7) (Hunter M. E. et al., 2010; Hunter et al., 2012; Satizábal et al., 2012; Tucker et al., 2012). In fact, the Brazilian diversity values were lower than other mammalian populations which were demographically challenged through pollution, harvesting, or habitat fragmentation (Garner et al., 2005; DiBattista, 2007; Torres-Florez et al., 2014). Low genetic diversity in marine mammal populations can be common and is often related to glacial events, founder effects, bottlenecks, anthropogenic mortality, and habitat degradation (Waldick et al., 2002; Roman and Palumbi, 2003; Paster et al., 2004). The nuclear and mtDNA genetic diversity found for the Venezuelan/Guyanese samples should be used with caution as the sample size was low (N = 10 and 13, respectively).

The low genetic diversity in the Brazilian Antillean manatee may be a combination of a founder effect and the decrease in abundance of manatees due to habitat loss and intensive historical hunting. Despite the decreased abundance due to hunting pressure, no evidence of a bottleneck within the last ∼0.2–4 N_E generations (∼32–92 years) was detected using the generation time of 16–23 years (Whitehead, 1978; Vaha and Primmer, 2006). Only approximately five manatee generations (∼100 years) have occurred since indiscriminate hunting was reduced in Brazil which is likely not sufficient to produce a detectable bottleneck in our analyses (Cornuet and Luikart, 1996; Luna et al., 2008b, 2018; Luna and Passavante, 2010). Additionally, the southern and westward colonization of manatees along the coastline of Brazil may have proceeded through a series of bottlenecks, according to population size estimates and records (Whitehead, 1978). The colonization may be too distant in time to be detected with these analytical tools. The eastern Antillean manatees of Venezuela and Guyana showed evidence of a bottleneck; however, it should be carefully considered as the sample size was very low. The evidence of gene flow between the Venezuela/Guyana and Brazil Antillean clusters may also reduce the genetic effects of bottlenecks in Brazil.

Since manatees have special habitat requirements and are vulnerable to stochastic events, it is imperative to continue to monitor and reduce anthropogenic threats (Marmontel, 1995; Rathbun et al., 1995; Reynolds et al., 1995) for both species in Brazil (Luna et al., 2018). The high level of past and current manatee mortality in Brazil has resulted in a reduced population size and potentially limited the population’s genetic diversity. It is imperative that species identification – and contemporary hybridization – be tested to allow for more accurate estimates of census and effective population sizes and any subsequent changes over time. Hybridization needs to be identified through nuclear genetic data derived from appropriate reference populations and sample sizes, as opposed to potentially ambiguous morphology traits and mtDNA (Allendorf et al., 2001; Vianna J. A. et al., 2006; Bohling and Waits, 2015; Laikre et al., 2016; Lima et al., 2019).

Maintenance of population connectivity is necessary to allow for genetic diversity growth (Bouzat, 2010; Whiteley et al., 2015; Laikre et al., 2016). Based on genetic distance values, connectivity has likely been occurring within the last 40–100 years between the two Antillean clusters identified here. Of note, mtDNA sequences from these three South American countries were assigned to all three of the mtDNA clusters delimited by Vianna J. D. A. et al. (2006), potentially supporting historical geneflow. Continued gene flow could help support sharing of advantageous alleles and prevent the Brazilian population from experiencing negative genetic effects (e.g., inbreeding, the extinction vortex etc.; Frankham et al., 2002; Tigano and Friesen, 2016), assuming interspecific hybridization is low. Gene flow between small and endangered populations is important to counteract the effects of random genetic drift and inbreeding, and can be achieved primarily by increasing effective population size and minimizing potential stochastic effects (Frankham, 1995; Westemeier et al., 1998; Madsen et al., 1999; Storfer, 1999). Our study indicates that although the population has low genetic diversity, manatees are moving along the coast, connecting areas that were previously separated by human impacts (Normande et al., 2016). Reintroduction of manatees into distribution gaps and the protection of conservation units would help to continue to encourage gene flow along the species distribution in Brazil. Protecting corridors and habitat types that include mangroves, seagrass beds and available fresh water may improve broader stepping-stone movement patterns. The management of wide-range travel corridors between South American countries could facilitate gene flow and allow for augmentation of the species without direct human intervention.

Genetic mutations that become fixed over many generations can increase genetic diversity, which is more likely to occur in larger populations. Although no physiological or genetic implications of inbreeding depression have been identified to date, small manatee populations could be susceptible to detrimental genetic effects, such as limited adaptation potential during climate change (Couvet, 2002). Population sizes of N_E > 50 effective breeders are recommended for short-term sustainability, while N_E > 500 are needed for long-term survival and the prevention of excessive inbreeding (Wright, 1951; Jamieson and Allendorf, 2012). This implies that census population levels in the upper thousands are necessary to maintain evolutionary stability (Franklin, 1980; Lande, 1995). All three groups of manatees in this study had estimates of N_E < 75, well below the 500-minimum threshold.

The rapid growth of anthropogenic activities has reduced the availability of habitat that is required for breeding and parental care (Luna et al., 2011; ICMBio, 2018b, 2019). As numerous stranded calves are rehabilitated, release locations should target natal habitats to avoid genetic swamping of locally adapted alleles (Luna et al., 2011, 2012; ICMBio, 2018a). To achieve near-term stability of the Brazilian Antillean manatee population, habitat restoration and protections, and direct anthropogenic threats (boats, oil extraction, etc.) are needed to encourage gene flow, bring N_E to >50 effective breeders, and improve connectivity through the recolonization of fragmented manatee habitats (Frankham et al., 2002).