Large‐scale connectivity, cryptic population structure, and relatedness in Eastern Pacific Olive ridley sea turtles (Lepidochelys olivacea)

Abstract Endangered species are grouped into genetically discrete populations to direct conservation efforts. Mitochondrial control region (mtCR) haplotypes are used to elucidate deep divergences between populations, as compared to nuclear microsatellites that can detect recent structuring. When prior populations are unknown, it is useful to subject microsatellite data to clustering and/or ordination population inference. Olive ridley sea turtles (Lepidochelys olivacea) are the most abundant sea turtle, yet few studies have characterized olive ridley population structure. Recently, clustering results of olive ridleys in the Eastern Tropical Pacific Ocean suggested weak structuring (F ST = 0.02) between Mexico and Central America. We analyzed mtCR haplotypes, new microsatellite genotypes from Costa Rica, and preexisting microsatellite genotypes from olive ridleys across the Eastern Tropical Pacific, to further explore population structuring in this region. We subjected inferred populations to multiple analyses to explore the mechanisms behind their structuring. We found 10 mtCR haplotypes from 60 turtles nesting at three sites in Costa Rica, but did not detect divergence between Costa Rican sites, or between Central America and Mexico. In Costa Rica, clustering suggested one population with no structuring, but ordination suggested four cryptic clusters with moderate structuring (F ST = 0.08, p < .001). Across the Eastern Tropical Pacific, ordination suggested nine cryptic clusters with moderate structuring (F ST = 0.103, p < .001) that largely corresponded to Mexican and Central American populations. All ordination clusters displayed significant internal relatedness relative to global relatedness (p < .001) and contained numerous sibling pairs. This suggests that broadly dispersed family lineages have proliferated in Eastern Tropical Pacific olive ridleys and corroborates previous work showing basin‐wide connectivity and shallow population structure in this region. The existence of broadly dispersed kin in Eastern Tropical Pacific olive ridleys has implications for management of olive ridleys in this region, and adds to our understanding of sea turtle ecology and life history, particularly in light of the natal‐homing paradigm.


| INTRODUC TI ON
Understanding the population genetics of endangered species is critical to identifying where and how many distinguishable populations there may be in a region, thus aiding in developing conservation plans for those populations. For sea turtle conservation, this is often done by designating management units (MUs), which are genetically discrete groupings of nesting assemblages (Komoroske, Jensen, Stewart, Shamblin, & Dutton, 2017). Nesting assemblages are obvious choices for defining turtle populations, as females are easily accessible for sampling as they come ashore to nest, and typically display natal homing (Lohmann, Putman, & Lohmann, 2008;Lohmann, Witherington, Lohmann, & Salmon, 2017). Defining MUs is important for developing effective conservation plans and is continually highlighted as a priority for global sea turtle research (Hamann et al., 2010;Rees et al., 2016).
Genetic analyses of mitochondrial DNA and nuclear microsatellites have allowed researchers to designate more informative MUs that capture much of the genetic variation within a species regionally and globally (Bowen & Karl, 2007;Komoroske et al., 2017).
Mitochondrial control region (mtCR) sequences (haplotypes) are maternally inherited, and sea turtles typically share haplotypes within regions (such as isolated islands or the northern and southern areas of an ocean basin; Bowen & Karl, 2007), due to maternal natal homing for reproduction. Microsatellite loci are highly variable repeating units found throughout the nuclear genome that may provide novel insights into population structure relative to mtCR haplotypes due to their high mutation rates and biparental heredity. Microsatellites reflect more contemporary gene flow and demographic changes than mtCR haplotypes due to their high mutation rates, and may further be used to conduct analyses to better understand the mechanisms behind MU population structuring (see Blouin, 2003;Putman & Carbone, 2014 for reviews of these analyses).
When populations are unknown, exploratory methods may be used to identify genetically discrete populations. Software that implements clustering population inference (i.e., STRUCTURE; Pritchard, Stephens,& Donnelly, 2000) and ordination population inference (i.e., DAPC; Jombart, Devillard, & Balloux, 2010) is commonly used toward this end. Software that implements clustering population inference typically uses a Bayesian or maximum-likelihood framework to cluster individuals into arbitrary populations, and then subsequently assesses the likelihoods of those populations and their genetic signatures. Software that implements ordination population inference plots individuals as points on a coordinate plane, and then uses variance between points or groups of points to identify putative populations. These methods have known shortcomings, differ in the assumptions they make of sample data and of the best inferred populations, and may suggest different population structuring when analyzing identical data (see Jombart et al., 2010;Kalinowski, 2011, andCarbone, 2014). It is therefore critical to use multiple analytical methods when making inferences about population structure in a data set. When studying threatened species such as sea turtles, such thorough analysis will enable researchers and managers to identify population structuring at multiple scales and determine the most suitable MUs for effective conservation and management.
MUs are not well defined for olive ridley sea turtles (Lepidochelys olivacea; Figure 1). Olive ridleys are the most abundant sea turtle globally (Abreu-Grobois & Plotkin2008), and entire ocean basins constitute the few existing MUs (see Komoroske et al., 2017 and references therein). Olive ridleys display unique reproductive traits relative to other sea turtle species (excluding the congeneric Kemp's ridley, Lepidochelys kempii) that influence their conservation status and population structure (Bernardo & Plotkin, 2007). Olive ridleys often nest en masse during "arribada" events, which may comprise tens of thousands of individual turtles (Bernardo & Plotkin, 2007).
The size of arribada events lends support to olive ridleys' "threatened" (rather than endangered) status on the IUCN Red List (Abreu- Grobois & Plotkin 2008). Olive ridleys also likely display limited natal fidelity to nesting beaches relative to all other sea turtle species (Dornfeld, Robinson, Tomillo, & Paladino, 2015;Kalb, 1999). Sea turtles typically home to the region from which they hatched to reproduce, which engenders population structure around nesting beaches (Bowen & Karl, 2007). Mating offshore of arribada events may facilitate admixture if thousands of olive ridleys from distant beaches are involved (Jensen, Abreu-grobois, Frydenberg, & Loeschcke, 2006).
Both of these unique behaviors could weaken signatures of population structure that might be inferred from mtCR and microsatellite genotype data. However, adequate region-wide genetic studies of olive ridleys have not been undertaken globally to assess population structure or determine MUs below the scale of entire ocean basins.
The Eastern Tropical Pacific olive ridley population is robust, with multiple arribada nesting sites and high-density solitary nesting sites (López-Castro & Rocha-Olivares, 2005;Valverde et al., 2012). Despite their contemporary abundance, adult olive ridleys and eggs were extensively harvested in the Eastern Tropical Pacific in the mid-late 20th century (Márquez, Peñaflores, & Vasconcelos, 1996;Spotila, 2004) and adult olive ridleys constitute a large proportion of contemporary fisheries bycatch (Moore et al., 2009). The number of individuals participating in arribadas has also exhibited decades-long declines (from millions to tens of thousands; Fonseca, Murillo, Guadamúz, Spínola, & Valverde, 2009). Costa Rica hosts two of the most prominent arribada beaches and index sites in the Eastern Tropical Pacific (Playa Nancite and Ostional) and is a global focal point for research on olive ridley biology and conservation. Playa Nancite historically hosted arribada assemblages as large as 115,000 individuals, which decreased into the early 21st century and has stabilized at ~8,000 individuals per arribada (Fonseca & Valverde, 2010). Playa Ostional has exhibited a recent increase in olive ridley abundance (Eguchi, Gerrodette, Pitman, Seminoff, bottleneck, conservation genetics, haplotypes, kin, marine connectivity, ordination, sea turtle & Dutton, 2007) and is estimated to have hosted assemblages as large as 476,550 individuals (Valverde et al., 2012). Since the late 1980s, locals have legally harvested ~20% of the eggs from arribada events, although local consumption was likely ongoing prior to this (Valverde et al., 2012).
There appears to be minimal population structuring within the Eastern Tropical Pacific Basin, and phylogeographic studies of olive ridleys have suggested that the Eastern Tropical Pacific population may be no more than 250,000-300,000 years old (Bowen et al., 1997;Jensen et al., 2013). Earlier genetic population assessments based on mtCR data indicated that olive ridleys nesting across the Baja California Peninsula (Mexico) may comprise a discrete MU (López-Castro & Rocha-Olivares, 2005). This finding was later supported by microsatellite analyses in a basin-wide study that additionally proposed a weak but significant partition (F ST = 0.028, p = .000) between Mexican and Central American populations across the entire Eastern Tropical Pacific (Rodríguez-Zárate et al., 2018). Though comprehensive, the study did not survey genetic variability of key population index sites in Costa Rica (but did survey index sites in Mexico and Nicaragua,  Bowen et al. (1997) to report mtCR haplotype data from olive ridleys south of Mexico. We compared mtCR haplotype data from Costa Rican olive ridleys to previously published data for Mexican olive ridleys to look for structuring across the Eastern Tropical Pacific. We subjected both microsatellite data sets to clustering and ordination population inference analyses, and investigated the mechanisms behind inferred structuring using metrics of population differentiation, bottleneck analysis, and relatedness analysis.
This study furthers our understanding of Eastern Tropical Pacific olive ridley sea turtle biology, and provides valuable information to assist management conservation of the species in this region.

| Site and sampling description
Blood samples and skin samples from 118 olive ridley turtles collected in 1999 (Playa Nancite, n = 7; Figure 2; Clusella-Trullas, Spotila, & Paladino, 2006), 2011; Playa Grande, n = 33; Figure 2) were processed in 2014. All three sites host solitary nesting, but only Playa Grande does not host arribada nesting. Playa Grande is situated between Playas Nancite and Ostional and has been a national park since the early 1990s.
Olive ridley nesting at Playas Nancite and Ostional is as described  Dutton, personal communication). "*" indicates one haplotype that is in the NMFS nomenclature but could not be reliably named. "T" indicates haplotypes first reported in this study frozen at −20ºC. Samples were diluted to 25 ng/μl before PCR.
PCR products were purified using ExoSAP-IT (Thermo) and sent to Genewiz (New Jersey, USA) for sequencing. Forward and reverse sequences were trimmed to approximately 800 bps, assembled, and aligned in Geneious v.11 (Kearse et al., 2012) using the CLUSTALW algorithm (Thompson, Gibson, & Higgins, 2003
Loci were tested for deviations from Hardy-Weinberg equilibrium and linkage disequilibrium using GENEPOP v.4.0.10 (Rousset, 2008;Rousset & Raymond, 1995). All samples were tested for heterozygote deficiency (evidence of null alleles) or heterozygote excess. A sequential Bonferroni correction was applied to account for multiple pairwise comparisons (Rice, 1989)

| Population inference
Clustering ( STRUCTURE output files were analyzed in StructureSelector (Li & Liu, 2018) to determine the best estimate of k using multiple metrics. The estimated log probability of the data given a particular value of K (pr(X|Z,P), where X is the data and Z [K] is a grouping of individuals with P allele frequencies, allows the estimation of the most likely number of clusters (Pritchard et al., 2000). The ad hoc delta-K method (Evanno, Regnaut, & Goudet, 2005) reports the second-order rate of change of the log probability of each K, which typically peaks at the appropriate value of K. The admixture model calculates the fractional probability (Q) of individuals belonging to each population. Puechmaille's (2016) four estimators (included in StructureSelector) base the likelihood of K on whether or not subpopulations (i.e., sampling sites) have at least 50% assignment to clusters for each K. Clusters that do not receive at least 50% assignment within subpopulations are defined as spurious, and lower the likelihood of that clustering configuration. These estimators are found to perform better than the log probability method in STRUCTURE (Pritchard et al., 2000) and the delta-K method (Evanno et al., 2005).
A discriminant analysis of principal components (DAPC; Jombart et al., 2010) was run for both data sets using adegenet (Jombart, 2008) as implemented in R. DAPC was run with sampling sites as groups, and inferred clusters as groups. Genotype data were transformed into a coordinate format for principal component analysis (PCA) using read.genepop(). The most likely number of clusters was determined using find.clusters(), which employs k-means clustering and a Bayesian information criterion to identify clusters. DAPC was run for each configuration (Costa Rican sites and inferred clusters, and Eastern Tropical Pacific sites and inferred clusters) using dapc().
As suggested by Jombart et al. (2010), 100% of the initial PCs were retained when identifying K, PCs accounting for ~ 80% of variance were retained during DAPC, and all axes of the DA were retained.

Rodríguez-Zárate et al. (2018) report pairwise F ST and D for
Eastern Tropical Pacific sampling sites in their original publication.
Alpha levels for pairwise F ST and D were adjusted using a sequential Bonferroni correction for multiple testing.

| Bottleneck analysis
A rapid decrease in population size leads to a reduction of the number of alleles present in the population, and therefore creates a heterozygosity deficiency (more heterozygotes are expected than actually exist). Rapid expansion after a bottleneck event leads to an increase in the number of alleles in the population, and therefore creates a heterozygosity excess (fewer heterozygotes are expected than actually exist). Recent extractive take from Eastern Tropical Pacific olive ridleys may have left such bottleneck signatures in genetic data from inferred populations, which should necessitate conservation actions. We therefore looked for bottleneck signatures in inferred structuring using BOTTLENECK (Cornuet & Luikart, 1996;Piry, Luikart, & Cornuet, 1999). BOTTLENECK can detect moderate bottlenecks with confidence for as many as 250 generations after they have occurred (Cornuet & Luikart, 1996) far longer than the approximately five to six generations of olive ridleys that have come into existence since the peak of take in the Eastern Tropical Pacific (Spotila, 2004 . The TPM is thought to be more representative of actual processes of mutation and evolution than other mutation models (Di Rienzo et al. 1994, Piry et al., 1999. The TPM is also more conservative than other models in inferring bottleneck events, as alleles that differ by more than one repeat still have a probability of coming from one mutational event, rather than multiple mutational events (Sainudiin, Durrett, Aquadro, & Nielsen, 2004). All BOTTLENECK runs were accompanied by tests for L-shaped distributions of allele frequencies to determine whether any inferred structuring exhibited evidence of a bottleneck-induced mode-shift.

| Relatedness analysis
Relatedness (r) is measured for two individuals on a 0 to 1 scale (0 being unrelated, 1 being identical) based on how much of the genome these two individuals are estimated to share (see Blouin, 2003, for a review of relatedness theory, methods, and studies). As sea turtles typically display natal breeding fidelity (with a margin of error but typically within MUs; Lohmann et al., 2008), we expect a higher probability of relatedness between individuals from within the same inferred cluster than between individuals from different inferred clusters. We therefore examined average pairwise relatedness within inferred structuring using two different algorithms (LRM: Lynch & Ritland, 1999;and QGM: Queller & Goodnight, 1989) over 10,000 iterations in GenAlEx (Peakall & Smouse, 2006). Queller and Goodnight's (QGM; 1989) estimator is a coefficient based only on the estimated identity by descent (IBD; Grafen, 1985). Lynch and Ritland (1999) estimator uses a regression calculation to determine relatedness coefficients for any pair of individuals based on shared IBD alleles, but can perform poorly if few related individuals are sampled, or if loci are too highly polymorphic (Blouin, 2003). Both estimators may also have high variances when few loci (n < 20) are used, but can provide a good estimation of relatedness between groups of individuals (Blouin, 2003;Queller & Goodnight, 1989).
GenAlEx tests specifically for significantly high relatedness within groups (i.e., inferred structuring) relative to global relatedness, and we therefore focused on relative values of relatedness and the significance of inferred structuring relatedness rather than on specific thresholds of relatedness.
First, to verify that both microsatellite data sets were powerful enough to exclude incorrect parent pairs (in light of the absence of known parents), we calculated P3Exc (the probability of excluding incorrect parent pairs when both parent genotypes are unknown) in GenAlEx for increasing combinations of loci for inferred structuring in Costa Rican and Eastern Tropical Pacific olive ridleys. We ran COLONY with all loci for each data set (n = 6 for Costa Rican data, n = 10 for Eastern Tropical Pacific data) for five "very long" runs of the full-likelihood method of determining parentage and sibships with "very high" likelihood precision. We allowed for both male and female polygamy, as multiple paternity is commonly documented in sea turtles (see Lee, Schofield, Haughey, Mazaris, & Hays, (2018) for a review), and also allowed for inbreeding. We did not include a sibship prior. We examined the number of parent-offspring, full-sibling, and half-sibling pairs within and between inferred structuring groups to determine whether inferred structuring groups contained more parent-offspring or sibling pairs (i.e., family lineages) than pairs with members in two different groups.

| Costa Rican mtDNA
We observed ten ~ 800 bp haplotypes from 60 turtles nest- Note: Also shown are the number of individuals from each site (n), mean (±SD) haplotype diversity (H), and mean (±SD) nucleotide diversity (π). "T" indicates haplotypes first reported at nesting beaches from this study. "*" indicates one haplotype that is known to NMFS but could not be named with certainty. Jensen et al., 2013;López-Castro & Rocha-Olivares, 2005) and did not vary between sites (as suggested by overlapping standard deviations; Table 2).
There was no evidence that any of the three sites were ge- comprised 68% of the haplotypes we found, which is consistent with (albeit lower than) López-Castro and Rocha-Olivares (2005) findings from Mexican olive ridleys (~90%) and suggests a lack of mitochondrial differentiation between Mexican and Costa Rican nesting assemblages.

| Costa Rican microsatellite analysis
All eight loci amplified successfully, albeit not in every individual (97.1%±3 SD amplification success; Table S1, also see supplementary data). We did not detect any deviations from HWE, but three loci (OR7, OR16, and OR22) were found to be in linkage disequilibrium (p < .000001). OR7 and OR22

| Population inference
In Costa Rican olive ridleys, analysis of clustering population inference results initially suggested K = 2 as the most likely population structure, but assignment plots for K = 2 suggested admixture.

Ordination population inference by nesting site for Eastern
Tropical Pacific olive ridleys confirmed clustering population inference results: Mexican and Central American nesting beaches split along the first axis, PAR separated from other Mexican nesting beaches, and the Central American nesting beaches displayed admixture ( Figure 3b). However, assignment proportions were low (mean = 65.3 ± 0.002 SE).
Ordination population inference elucidated 9 discrete clusters with high assignment proportions (mean = 0.99 ± 0.003 SE;  Table 1). In F, sites are in North to South order. Note that PAR is the southernmost site in the Mexican population and GH the northernmost site in the Central American population.  Central American cluster 3 displayed elevated F IS (F IS = 0.075-0.109, p < .0001; Table 4). All clusters displayed significant heterozygote deficiencies in global HWE exact tests (p < .0001; Table 4).

| Bottleneck analysis
BOTTLENECK results varied depending on the proportion of SMM in the TPM, and on the test used to validate the significance of results. In general, TPM with 95% SMM inferred more population expansion after bottleneck events than TPM with 0% SMM, which only inferred one instance of heterozygosity excess ( With no SMM in the TPM, there were no inferred bottleneck events in either data set. This may be due to the constraints and limitations of the mutation models used in BOTTLENECK (Luikart, Allendorf, Cornuet, & Sherwin, 1998, Piry et al., 1999Putman & Carbone, 2014). With 95% SMM in the TPM, there was still no evidence of bottleneck events in Costa Rican olive ridleys (Table 4).

However, BOTTLENECK found that both Mexican and Central
American populations had significant heterozygote deficiency (p = .00098 and .014 respectively; Table 4). The Wilcoxon test detected bottlenecks in three out of four Mexican ordination population inference clusters, and three out of five Central American ordination population inference clusters (p < .05; Table 4). However, the sign test only detected a bottleneck in cluster #2 (p < .05; Table 4).
Despite this, none of the L-shaped distribution tests suggested mode-shifts in allele frequencies.

| Relatedness analysis
LRM and QGM showed agreement in general patterns of relatedness, but differed in exact values of relatedness within nesting sites and putative populations. In general, LRM was more conservative than QGM. In Costa Rica, relatedness was negligible overall (Table 5).
Relatedness was significantly high (p < .001) within ordination population inference clusters, and ranged from 0.310 to 0.570 (LRM) and 0.053 to 0.235 (QGM). Relatedness was higher in Mexican and Central American populations overall than in Costa Rica (Table 5). COLONY results for individual ordination population inference clusters are found in Table 5. COLONY did not identify parent-offspring relationships in either data set. In Costa Rican olive ridleys (

| D ISCUSS I ON
Olive ridleys were thought to display minimal population structuring within ocean basins (Bowen et al., 1997;Bowen & Karl, 2007) due in part to their low nesting site fidelity (Kalb, 1999) and broad foraging ranges (Plotkin, 2010). While we show that mtCR haplotypes are un-

| Limitations
The analyses and results presented here were not without limitations. We were unable to directly compare the microsatellite data generated here and those generated by Rodríguez-Zárate et al.
(2018). This would require costly and time-intensive calibrations, and we believe our side-by-side analyses are still of value toward understanding olive ridley population structure. We used relatively few microsatellite loci (n = 6) to study population structure among Costa Rican olive ridleys. Six loci have previously been used to study olive ridley population structure (Jensen et al., 2006), but Jensen et al. (2006) did not examine relatedness or population bottlenecks. Relatedness estimates improve in accuracy with increasing loci (Blouin, 2003), and BOTTLENECK may require more than six loci to adequately identify population bottlenecks (Peery et al., 2012;Williamson-Natesan, 2005). Further, bottleneck signatures may be confounded by relatedness among individuals, which in some cases may lead to similar heterozygote deficiencies (as seen in some ordination population inference clusters here; Table 4). We therefore cautiously interpret relatedness and BOTTLENECK results from Costa Rican olive ridleys.

| Connectivity
Analyses of mtCR haplotypes did not support or refine population structure in Costa Rican or Eastern Tropical Pacific olive ridleys. This may be due in part to relatively recent colonization of the Eastern Tropical Pacific by olive ridleys. Chelonian mitochondrial DNA has been shown to accrue mutations on a scale of tens of thousands to hundreds of thousands of years (Avise, Bowen, Lamb, Meylan, & Bermingham, 1992). Past phylogeographic studies of olive ridleys have suggested that the Eastern Tropical Pacific population may be no more than 250,000-300,000 years old (Bowen et al., 1997;Jensen et al., 2013). Thus, while at least 14 mtCR haplotypes are reliably documented from Eastern Tropical Pacific olive ridleys (Bowen et al., 1997;Jensen et al., 2013;López-Castro & Rocha-Olivares, 2005 (Bowen et al., 1997; but see Hahn, 2013), and a new effort to study ridley phylogeography will necessitate better organization of mtCR haplotypes.
Failure of mtCR haplotypes to refine population structure is also likely due to broad connectivity in Eastern Tropical Pacific olive ridleys. Previous studies have suggested that Eastern Tropical Pacific olive ridleys display little population structuring (Bowen et al., 1997;Jensen et al., 2013;Rodríguez-Zárate et al., 2018) and site fidelity (Dornfeld et al., 2015;Kalb, 1999), and studies have found only weak structuring at broad scales (i.e., turtle species (i.e., Roberts, Schwartz, & Karl, 2004, Carreras et al. 2007). Male-mediated gene flow likely contributes to observed connectivity in Eastern Tropical Pacific olive ridleys. Finally, arribada events likely facilitate connectivity in Eastern Tropical Pacific olive ridleys (Jensen et al., 2006). There were historically six arribada beaches in the Eastern Tropical Pacific (Montero, Rincon, Heppell, & Hall, 2016), although some assemblages were extirpated in the 20th century due to extensive anthropogenic take of turtles (Spotila, 2004). Arribada beaches were and are foci for hundreds of thousands of breeding olive ridleys, and likely have fostered and continue to foster broad genetic connectivity in Eastern Tropical Pacific olive ridleys.

| Lineages
Despite broad connectivity and weak large-scale structuring, ordination population inference produced cryptic clusters in both Costa Rican and Eastern Tropical Pacific microsatellite data. These clusters presented with moderate F ST and D (Table 3A and B), but also had notable internal relatedness (Table 5) (Tables 4 and 5). Eastern Tropical Pacific clusters in particular contain more full-sibling pairs within than between clusters, save for cluster 3 (Table 5). These results suggest that ordination population inference did not identify subpopulations, but instead identified family lineages in Eastern Tropical Pacific olive ridleys. Members of these family lineages primarily correspond to Mexican and Central American subpopulations, but these lineages also corroborate broad connectivity in Eastern Tropical Pacific olive ridleys as evidenced by their basin-wide geographic distribution, and the ubiquitous presence of full-and half-sibling pairs between clusters.
In Mexico, these lineages persisted despite intensive take, which may have engendered genetic bottleneck signatures (Márquez et al., 1996;Spotila, 2004; Table 4). Reduced genetic diversity due to bottlenecking may have accentuated differential survival and genetic signatures of family lineages in Mexico, hence positive F IS and heterozygote deficiency (Table 4), and relatively high proportions of intra-versus intercluster full-and half-sibling pairs in all primarily Mexican ordination population inference clusters (Table 5). These clusters even exhibited some spatial heterogeneity (Figure 3f), which may only be apparent due to reduced genetic diversity in Mexican olive ridleys overall, and further highlight relatively reduced connectivity between Mexican lineages and other lineages.
In Central America, take of eggs and adult turtles was not as severe as in Mexico (Spotila, 2004). While significant heterozygote deficiency and internal sibship pairs (particularly full siblings; Table 5) still suggest primarily Central American ordination population inference clusters represent family lineages, these lineages do not exhibit the degree of positive F IS as do all Mexican lineages (save for cluster 3; Table 4). Central American lineages also display consistently lower proportions of intra-versus intercluster half-sibling pairs ( Table 5).
The spatial homogeneity of Central American ordination population inference clusters (Figure 3f) Rica (Montero et al., 2016;Spotila, 2004). The two arribada sites sampled in the present study are located within 100 km of each other ( Figure 2) and may have a regional draw for breeding olive ridleys, who then mate with individuals from Mexican and Central American populations and other family lineages. However, this enhanced connectivity may be an artifact of using fewer loci (n = 6 in Costa Rica, n = 10 in Mexico and Central America

| Implications
The existence of broadly distributed family lineages is a first and pre- and offspring, and in this case, siblings (Ronce, Gandon, & Rousset, 2000). Eastern Tropical Pacific olive ridleys are less likely to compete for nesting sites with kin if they disperse vast distances from natal beaches to reproduce. Dispersal also reduces the probability of inbreeding, which may ultimately reduce individual fitness (Perrin & Goudet, 2001) and may ensure that lineages persist despite stochastic, but spatially isolated, disturbances. For instance, olive ridleys from a primarily Central American lineage that were sampled while nesting in Mexico may survive disturbances that impact Central America, and continue contributing offspring to that lineage undisturbed. There may be some individual olive ridleys who never display natal fidelity as a consequence of this pressure to disperse from kin. These individuals would constitute fascinating exceptions to the natal-homing paradigm known for sea turtles (Lohmann et al., 2008), if genetic capture-mark-recapture (i.e., Dutton & Stewart, 2013)  Pacific olive ridleys given increasing pressure by unregulated fisheries and other threats in this region (Dapp, Arauz, Spotila, & O'Connor, 2013;Hope, 2002;Moore et al., 2009). Further knowledge of migratory routes used by olive ridleys, gained via satellite telemetry, will also be necessary to better understand how to protect olive ridleys at sea and to maintain genetic diversity, population structure, and family lineages in this region over time. Eastern Tropical Pacific olive ridley lineages merit investigation, and studies that examine individual fitness in light of kin dispersal in multiple lineages over multiple generations would provide insight into the evolutionary drivers behind lineage persistence and dispersal.
As mentioned before, genetic capture-mark-recapture, and neonate and adult satellite telemetry of males and females, throughout entire nesting seasons, will aid in elucidating the extent of natal homing and breeding fidelity in Eastern Tropical Pacific olive ridleys. It is as of yet unclear to what extent female olive ridleys display limited site fidelity between nesting events (perhaps traveling thousands of kilometers), or whether to some extent olive ridleys (including breeding males) do not exhibit natal homing for reproduction. Family lineages may underlie known population structure in other sea turtle species. Future studies that integrate molecular, spatial, and behavioral ecology techniques to investigate this phenomenon will provide novel insight for management of populations in these highly mobile species and will have an impact on our current understanding of sea turtle behavioral and spatial ecology.

ACK N OWLED G EM ENTS
We thank two anonymous reviewers for invaluable feedback and suggestions. We thank Susanna Clusella-Trullas and Leatherback Trust biologists at Playa Grande for collecting and processing tissue samples. We thank the Jack W. Schrey Distinguished Professorship at Purdue University Fort Wayne for providing funding. Work was conducted in accordance with Purdue University ACUC, SINAC, and CONAGEBIO guidelines.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data available in this study are obtained from DNA sequences (GenBank Accession nos. MK749418-MK749421) and microsatellite genotypes deposited on Dryad repository (https://doi.org/10.5061/ dryad.c866t 1g4f).