Connecting paths between juvenile and adult habitats in the Atlantic green turtle using genetics and satellite tracking

Abstract Although it is commonly assumed that female sea turtles always return to the beach they hatched, the pathways they use during the years preceding their first reproduction and their natal origins are most often unknown, as it is the case for juvenile green turtles found in Martinique waters in the Caribbean. Given the oceanic circulation of the Guiana current flowing toward Martinique and the presence of important nesting sites for this species in Suriname and French Guiana, we may assume that a large proportion of the juvenile green turtles found in Martinique are originating from the Suriname–French Guiana beaches. To confirm this hypothesis, we performed mixed stock analysis (MSA) on 40 green turtles sampled in Martinique Island and satellite tracked 31 juvenile green turtles tagged in Martinique to (a) assess their natal origin and (b) identify their destination. Our results from MSA confirm that these juveniles are descendant from females laying on several Caribbean and Atlantic beaches, mostly from Suriname and French Guiana, but also from more southern Brazilian beaches. These results were confirmed by the tracking data as the 10 turtles leaving Martinique headed across the Caribbean–Atlantic region in six different directions and 50% of these turtles reached the Brazilian foraging grounds used by the adult green turtles coming from French Guiana. One turtle left the French Guianan coast to perform the first transatlantic migration ever recorded in juvenile green turtles, swimming toward Guinea‐Bissau, which is the most important nesting site for green turtles along the African coast. The extensive movements of the migrant turtles evidenced the crossing of international waters and more than 25 exclusive economic zones, reinforcing the need for an international cooperative network to ensure the conservation of future breeders in this endangered species.

The migration of sea turtles is directly related to their complex life cycle which encompasses neritic, marine and/or oceanic habitats. Except the flatback and the leatherback turtles that adopt a particular life history pattern (exclusively neritic and exclusively oceanic, respectively), the five remaining species alternate between neritic and oceanic habitats during their life cycle (Bolten, 2003). After passively drifting for several years in oceanic environments, that is, the "lost years" (Putman & Naro-Maciel, 2013;Reich, Bjorndal, & Bolten, 2007), juvenile turtles then recruit to shallow neritic habitats (Musick & Limpus, 1997). They thereafter spend many years foraging in these developmental habitats (Tomaszewicz et al., 2015) until they reach sexual maturity before migrating back to their adult habitats, either nesting site or foraging grounds (Bjorndal, Bolten, & Chaloupka, 2005;Patrício, Velez-Zuazo, Diez, Van Dam, & Sabat, 2011).
Based on the assumption that female sea turtles always return on the beach they hatched (Meylan, 1999), genetic studies of juveniles may contribute to decipher their future adult habitats. In the Western Atlantic, several studies have been conducted to assess the natal origin of juvenile green turtles and they have shown a strong genetic diversity, especially within the Caribbean (Bass & Witzell, 2000;Bass, Epperly, & Braun-McNeill, 2006;Lahanas et al., 1998;Luke, Horrocks, LeRoux, & Dutton, 2003). Yet, the pathways juvenile green turtles use during the years preceding their first reproduction and their natal origins are unknown, as it is the case for the immature green turtles found in Martinique waters in the Caribbean. In Martinique Island, the green turtle was overexploited for its flesh and its eggs for 500 years (Chevalier, 2006), leading to the disappearance of nesting females on this island. However, many juvenile green turtles are found year-round in this habitat (Chevalier, 2006), likely due to the high abundance of seagrass (in comparison with the lack of seagrass in French Guianese waters) and a protection from predators . But to date, no information on their natal origin and developmental migrations is available.
Given the oceanic circulation of the Guiana current flowing toward Martinique (Borstad, 1982) and the presence of important nesting sites for green turtles in Suriname and French Guiana (Chambault, de Thoisy, Kelle, et al., 2016), we assume that a large

| Animal capture and tagging
The fieldwork was conducted randomly in the Anses d'Arlet, in the (from June to October of each year), 425 juvenile green turtles were captured by teams composed of three free divers. After spotting a static turtle feeding or resting at the bottom, a free diver would dive close to the head of the turtle, as discreetly as possible in order to avoid detection. Once close enough, the free diver would catch the turtle by the pygales plates of the shell (located behind the nuchal), and bring it to the surface. A second diver would then hold the foreflippers and help lift the individual into the boat for measurements and tagging. Each turtle was placed in a pen, and standard morphometric data recorded, that is, curved carapace length (CCL) and body mass using an electronic dynamometer.

| Genetic analysis
The use of the genetic resources was declared to the French Ministry of Environment under the reference TSP 79.585, in compliance with the Access and Benefit Sharing procedure implemented by the Loi pour la Reconquête de la Biodiversité.

| DNA extraction and D-loop control region genotyping
Biopsy punches were prelysed within the EasyMAG lysis buffer (BioMérieux, Marcy l'Etoile, France) at 4°C overnight, then in Tris-SDS buffer and proteinase K at 56°C before grinding. DNA was isolated using NucliSENS EasyMAG® bio-robot (BioMérieux, Marcy l'Etoile, The cycling conditions included an initial denaturation step at 95°C for 5 min followed by 36 denaturation cycles at 94°C for 30 s, annealing for 30 s at 50°C, and elongation at 72°C for 30 s. A final 10 min extension step at 72°C followed the last cycle. Beckman Coulter Genomics carried out sequencing (Beckman Coulter Genomics, Takeley, UK). estimating the contribution of different source populations (nesting sites) to a mixed population (Bolker, Okuyama, Bjorndal, & Bolten, 2007). Haplotypes found at feeding grounds, which cannot be tracked back to the source population ("orphan" haplotypes), were excluded from the analyses. In the same way, haplotypes found at nesting sites but not at feeding ground of Martinique were excluded from the analyses.
The many-to-one Bayesian approach of mixed stock analyses has used and evaluated the contribution of each nesting population to Martinique population by considering that each nesting population analyzed would have the same probability to contribute to the mixed stock. The Gelman and Rubin (1992) diagnostic has been used to test the convergence strength of chains. Estimated shrink factors close to one indicate convergence, and acceptable values are less than 1.2 (Kass, Carlin, Gelman, & Neal, 1998).

| Tracking data analysis
Data were downloaded daily via Argos Message Retriever (WC-DAP, Wildlife Computers-Data Analysis Programs). In order to provide optimum location accuracy and increase the number of positions available to counterbalance errors caused by the proximity to the shore, the GPS-SRDL tags were programmed to simultaneously record Argos and GPS locations. The GPS sampling interval was set to 15 min. Tag position estimates (Argos data) were enhanced by applying a Kalman-filtering algorithm to account for Argos error (CLS, Collecte Localisation Satellites, Toulouse, France).
The proximity to the shore and the possible Argos error resulted in 51% of the positions being found on land after applying the Kalman filter. We used the altimetry product provided by the Hydrographic and Oceanographic Service of the French Navy (SHOM) at a 25 m 2 resolution to identify these erroneous locations and discard them. Positions associated with a travel speed of over 10 km/hr were discarded (8%), and also those associated with location class Z (0.1%), that were considered insufficiently accurate.

| Genetic diversity, population structure, and Mixed Stock Analyses
A total of 40 green turtles of the mtDNA control region in the feeding ground of Martinique (hereafter MT) and from 15 nesting populations have been used (Table 1). The results are presented in Figure 1.  The tag instruments transmitted on average (mean ± SD) 515 ± 349 locations (range: 142 to 886, #149691 vs. #164548, respectively).
The tracking duration was on average 166 ± 79 days, and the total distance travelled ranged from 1,370 km (#149697) to 7,821 km (#149696)-see Table 2.

| Migratory routes
The 10 individuals headed in six different directions, with 50% to the North Brazilian coast (n = 5), 30% to the Caribbean (n = 3), 10% to TA B L E 1 Haplotypes and haplotype frequency of the Martinique feeding ground (MT, shaded line) investigated and nesting sites considered in this study

| D ISCUSS I ON
This study made it possible to assess the natal origins of juvenile green turtles inhabiting Martinique waters, and to identify their connecting paths between the developmental habitat of Martinique and already known habitats as inhabited by adult green turtles across the Caribbean-Atlantic region. Prior to our study, there was very little information on the movements of juvenile green turtles in the Caribbean.
Our results show that the genetic diversity of the Martinique juvenile turtles is high. Mixed stock analysis enables to show the contribution of different nesting sites to a feeding ground (Jordao et al., 2017;Okuyama & Bolker, 2005). The mitochondrial CMA-5 haplotype that dominates in the Caribbean is the most prevalent (Bass et al., 2006;Bjorndal et al., 2005;Lahanas et al., 1998). Other haplotypes from other regions were found, such as CMA-8 which is widely reported in southern Atlantic rookeries (Jordao et al., 2017) and in feeding grounds in the Caribbean and Atlantic (Lahanas et al., 1998;Luke et al., 2003  , and Guinea-Bissau (CCL range: 92-103 cm, Godley, Almeida, et al., 2003;Godley, Lima, et al., 2003), this suggests that our migrant individuals are close to reach sexual maturity. These turtles might therefore initiate their migration, either to discover new developmental habitats (Carr, Carr, & Meylan, 1978) or to directly reach adult habitats (foraging grounds or breeding sites).
Among the 31 green turtles tracked from Martinique, 21 remained in the shallow waters of Martinique. The smaller size of these 21 resident turtles (mean CCL: 75.5 ± 10.5 cm) compared to the size of the migrants (85.9 ± 3.8 cm) suggested a threshold beyond which individuals start migrating, therefore reinforcing the assumption that the migrant turtles may be close to sexual maturity.
The multiple origins of these juvenile turtles in the Caribbean-Atlantic region have been confirmed by the tracks of migrant turtles that headed toward adult habitats (natal beaches and foraging ground) (see Figure 3). Similar behavior was observed for some juvenile green turtles in the Indian Ocean which had shorter CCL than our migrant individuals (73.5 ± 4.9 cm vs. 85.9 ± 3.8 cm), but successfully reached adult foraging areas (Pelletier, Roos, & Ciccione, 2003 G G G G G G G G G G G G G G G G G G G G G  G G G G G G G G G G G G G G G G Days Atlantic Ocean BRAZIL Thalassia testudinum, the same species that is found in Martinique (Ogden, Robinson, Whitlock, Daganhardt, & Cebula, 1983). Despite the availability of different foraging grounds and nesting sites along her route, turtle #149694 targeted one site located in Florida rather than exploring new developmental habitats: after crossing the many nesting sites located north of West Palm Beach (Roberts, Collins, Paxton, Hardy, & Downs, 2014;Shamblin et al., 2014), and the juvenile reef habitats (Stadler, Salmon, & Roberts, 2014), this turtle ended its migration in Biscayne Bay which hosts different communities of seagrasses and macroalgae, as well as many megafauna groups, including sea turtles (Lirman et al., 2014).  (Catry et al., 2002), and is also a major contributor to the genetic diversity of the West Atlantic foraging grounds (Jordao et al., 2015). The probability that this turtle ended its journey either in the nesting site of Guinea-Bissau or Cape Verde is reinforced by the simulations performed by Putman and Naro-Maciel (2013), who demonstrated that the dispersal east of 14°W accounted for 5%-65% of the simulated turtles found in the Caribbean and North American foraging grounds after 5 years of simulations (Putman & Naro-Maciel, 2013).  (Baudouin et al., 2015; and juvenile green turtles (Godley, Almeida, et al., 2003;Godley, Lima, et al., 2003;Lima, Lagueux, Castro, & Marcovaldi, 1999;Lum, Lima, & Santos, 1998). These results confirm the significant contribution of the Suriname-French Guianan and the southern Brazilian nesting sites to genetic diversity found in the Brazilian feeding grounds (Jordao et al., 2015(Jordao et al., , 2017. For turtles originating from the Suriname-French Guiana nesting beaches, there will be a need for an active swimming behavior toward specific destinations when comparing the tracks of these five turtles with the oceanic circulation. Similarly to the behavior observed in adult green turtles originating from French Guiana , these juvenile individual swam against both the north-westward flowing Guiana current and North Brazilian current (Baklouti et al., 2007;Fratantoni & Glickson, 2002).
For adequate geographic and ecological scaling, species conservation plans should take into account the genetic structure and demographic history of populations (Lande, 1988). Genetic clades may vary according to gender, age, and bioecological function (e.g., feeding vs. breeding) taken into account (Jordao et al., 2017). The genetic structure of Chelonia mydas in the West Atlantic revealed movements of individuals between northern (this study) and southern feeding areas (Jordao et al., 2017), and northern nesting rookeries, and reinforces the conservation importance of regional corridors.
The extensive movements of the 10 migrant turtles evidenced the crossing of international waters and more than 25 exclusive economic zones. The multidirectional migrations performed by these 10 individuals, reinforced therefore the need for an international cooperative network to ensure the conservation of future breeders in this endangered species. Our result is also in accordance with the genetically mixed stock of green turtles inhabiting common developmental habitats but originating from diverse natal beaches dispersed across the Caribbean (Bass & Witzell, 2000;Jordao et al., 2015;Luke et al., 2003). Using the same genetic analyses as those already carried out for loggerhead (Engstrom, Meylan, & Meylan, 2002), hawksbill (Meylan, 1990), and green turtles (Bass et al., 2006;Luke et al., 2003), we will be able to find the natal origins of these juvenile turtles and confirm whether they are heading for adult habitats or rather exploring new developmental habitat until they migrate toward adult habitats. Since this population also benefits of a long-term follow up via capture and tagging (PIT) of individuals in Martinique foraging sites, the recapture of tagged turtles in the multiple nesting sites spread across the Atlantic will soon make it possible to link the developmental habitat of Martinique to natal nesting beaches (i.e., French Guiana and Suriname, and likely others). In order to support the conservation of this endangered species at multiple scales and stages, that is, developmental habitat, foraging areas, and reproduction sites, it is crucial to identify the key areas used by juvenile green turtles to ensure the protection of these future breeders, via the track of additional individuals in the different Caribbean islands. Data from these studies will consolidate decision making for the application of urgent conservation measures in this extensive area composed of multiple jurisdictional waters. This study provides up-to-date and useful genetic and tracking data that could more accurately support the revision of the geographic limits of Regional Management Units (Wallace et al., 2010).

ACK N OWLED G M ENTS
This study was carried out within the framework of the Plan

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

AUTH O R CO NTR I B UTI O N S
PC performed the data analysis and wrote the manuscript. DC designed the experiment, collected the data, and supervised the analy-

DATA ACCE SS I B I LIT Y
DNA sequences were deposited to GenBank, accession numbers range from MK060180 to MK060219.