Global Phylogeography of the Ridley Sea Turtles (Lepidochelys Spp.) As Inferred from Mitochondrial DNA Sequences

Genetica 101: 179–189, 1998. 179

c 1998 Kluwer Academic Publishers. Printed in the Netherlands.

Global phylogeography of the ridley sea turtles (Lepidochelys spp.) as inferred from mitochondrial DNA sequences

B.W. Bowen1;2,A.M.Clark3, F.A. Abreu-Grobois4, A. Chaves5, H.A. Reichart6 & R.J. Ferl7 1 Dept. of Fisheries and Aquatic Sciences, University of Florida, 7922 NW 71st Street, Gainesville, FL 32653-3071, USA; (E-mail: [email protected]fl.edu); 2 Archie Carr Center for Sea Turtle Research, 223 Bartram Hall, University of Florida, Gainesville FL 32611, USA; 3 Molecular Services Core, University of Florida, P.O. Box 110700, Gainesville FL 32611, USA; 4 Estacion´ Mazatlan,´ Instituto de Ciencias del Mar y Limnología, Apdo. Postal 811, Mazatlan,´ Sinaloa 82000, Mexico;´ 5 Programa de Tortugas Marinas, Escuela de Biología, Universidad de Costa Rica, Apdo. 18-3019, San Pablo de Heredia, Costa Rica; 6 Surinam Forest Service, P.O. Box 436, Paramaribo, Surinam; 7 Program in Plant Molecular and Cellular Biology, Horticultural Sciences Department, University of Florida, Gainesville, FL 32611, USA

Accepted 5 December 1997

Key words: biogeography, conservation, control region, Kemp’s ridley, olive ridley

Abstract

The Kemp’s ridley sea turtle (Lepidochelys kempi) is restricted to the warm temperate zone of the North Atlantic Ocean, whereas the olive ridley turtle (L. olivacea) is globally distributed in warm-temperate and tropical seas, including nesting colonies in the North Atlantic that nearly overlap the range of L. kempi. To explain this lopsided distribution, Pritchard (1969) proposed a scenario in which an ancestral taxon was divided into Atlantic and Pacific forms (L. kempi and L. olivacea, respectively) by the Central American land bridge. According to this model, the olive ridley subsequently occupied the Pacific and Indian Oceans and recently colonized the Atlantic Ocean via southern Africa. To assess this biogeographic model, a 470 bp sequence of the mtDNA control region was compared among 89 ridley turtles, including the sole L. kempi nesting population and 7 nesting locations across the range of L. olivacea. These data confirm a fundamental partition between L. olivacea and L. kempi (p=0.052- 0.069), shallow separations within L. olivacea (p=0.002-0.031), and strong geographic partitioning of mtDNA lineages. The most divergent L. olivacea haplotype is observed in the Indo-West Pacific region, as are the central haplotypes in a parsimony network, implicating this region as the source of the most recent radiation of olive ridley lineages. The most common olive ridley haplotype in Atlantic samples is distinguished from an Indo-West Pacific haplotype by a single nucleotide substitution, and East Pacific samples are distingushed from the same haplotype by two nucleotide substitutions. These shallow separations are consistent with the recent invasion of the Atlantic postulated by Pritchard (1969), and indicate that the East Pacific nesting colonies were also recently colonized from the Indo-West Pacific region. Molecular clock estimates place these invasions within the last 300,000 years.

Introduction cumglobal distribution. Both species occupy coastal habitats as adults, but the olive ridley moves offshore The genus Lepidochelys contains two species; the in some areas to feed on pelagic prey (Plotkin et al., Kemp’s ridley turtle (L. kempi) is found primarily in 1995). Although several species of sea turtle nest on warm-temperate waters of the North Atlantic and Gulf oceanic islands, ridleys are almost exclusively main- of Mexico, and the olive ridley (L. olivacea) has a cir- land nesters (Pritchard & Trebbau, 1984). The olive ridley typically occupies a more tropical habitat than

Representative sequences from this report have been deposit- Kemp’s ridley, but cold-temperate conditions probably ed in the GenBank database as accession numbers AF051773- limit the distribution of both species. AF051777 180

Figure 1. Arrows indicate the biogeographic model proposed by Pritchard (1969), in which the L. olivacea may have spread westward from the eastern Paciﬁc through the Indian-Paciﬁc region and recently colonized the Atlantic Ocean via southern Africa. Circles indicate sample sites for olive ridley, and the square represents the sole nesting population of the Kemp’s ridley.

Kemp’s ridley is the most endangered marine turtle. onization may have proceeded from the Indian Ocean In the latter half of the twentieth century, the number to the Atlantic coast of Africa, followed by transplan- of nesting females dropped from tens of thousands tation across the Atlantic Ocean to South America. to a few hundred (Woody, 1985). This trend towards Hughes (1972) noted that the distribution of olive rid- extinction is especially well documented because L. leys around Africa was consistent with this biogeo- kempi nests primarily at a single location in the west- graphic model (see also Hendrickson, 1980). ern Gulf of Mexico (Hildebrand, 1963; Carr, 1963). Molecular genetic studies have proven useful for In contrast, the olive ridley is the most abundant sea examining hypotheses about the historical distribution turtle (Reichart, 1993). Nesting aggregates in excess of marine taxa (Avise, 1994; Palumbi, 1994; Bowen of 100,000 females have been reported from Pacific & Grant, 1997). Previous mitochondrial (mt) DNA Mexico, Pacific Costa Rica, and eastern India (Anony- surveys have resolved aspects of global biogeogra- mous, 1976; Marquez,´ Villanueva & Penaflores,˜ 1976; phy in two other Cheloniid turtles: the green sea turtle Pritchard & Trebbau, 1984). Nesting aggregates have (Chelonia mydas; Bowen et al., 1992) and the logger- also been reported from the Indo-West Pacific, both head sea turtle (Caretta caretta; Bowen et al., 1994). sides of tropical Africa, and from the Atlantic coast of Here the global phylogeography of ridley species is South America. The presence of nesting populations inferred from mtDNA sequence comparisons, to assess in Guyana, French Guiana, and Surinam indicates that the model proposed by Pritchard (1969) and elaborat- the ranges of the olive ridley and Kemp’s ridley are ed by Hughes (1972) and Hendrickson (1980). The separated only by the Caribbean Sea (Reichart, 1993). sequence chosen for this survey is a 470 bp segment of Although the olive ridley is not considered globally the mtDNA control region (or d-loop). In turtles and endangered, many areas of former abundance have other vertebrates, this noncoding segment has a high- been greatly reduced by human activities (Anonymous, er mutation rate than that observed in protein-coding 1976; Ross, 1982). sequences or restriction fragment comparisons (Allard In the first thorough examination of ridley mor- et al., 1994; Norman, Moritz & Limpus, 1994; Lamb phology and systematics, Pritchard (1969) suggested et al., 1994; Walker et al., 1995; Dutton et al., 1996). that an ancestral Lepidochelys form was isolated into This higher mutation rate yields demonstrably greater Atlantic (L. kempi) and Pacific (L. olivacea) cohorts resolution of population separations and biogeographic by the formation of the Central American land bridge patterns in sea turtles (Lahanas et al., 1994; Encalada et (now known to be 3–4 MY BP). Under this model, al., 1996; 1998). Furthermore, the relatively large body the olive ridley spread through the Pacific and Indian of control region sequences for sea turtles (see Abreu- Oceans during the late Pliocene and Pleistocene, and Grobois et al., 1996) provides a rich background for recently colonized the Atlantic Ocean (Figure 1). Col- the interpretation of results from ridley turtles. 181

Materials and methods primers in a robotic work station (Applied Biosystems model 800), and the labelled extension products were Seven L. olivacea nesting locations were sampled analyzed with an automated DNA sequencer (Applied between 1988 and 1994, including Nancite Beach, Biosystems model 373A) in the DNA Sequencing Core Guanacaste Province, Paciﬁc Costa Rica (n=18); at University of Florida. Raw data from the sequencer McLure Island Group, Northern Arnhem Land, North- was edited and aligned using Sequencher version 3.0 ern Territory, Australia (n=8); Kijal, Malaysia (n=4) (Gene Codes Corp.). All samples were sequenced in and Paka, Malaysia (n=1); southwestern coast of Sri the forward direction. Those mtDNA sequences that Lanka (n=17); Orango National Park, Guinea Bis- matched known haplotypes were collated for analysis, sau (Atlantic coast of Africa, n=4); Sergipe, Brazil whereas new haplotypes were resequenced to assure (n=15); and Eilanti Beach, Surinam (n=13). Addition- the accuracy of nucleotide sequence designations. al samples were collected at major nesting locations Sequence divergences between haplotypes were in India and Colombia, but permit requests to export estimated with the Kimura two-parameter model these specimens were denied (see Bowen & Witzell, (Kimura, 1980) with a 4:1 transition/ transversion ratio, 1996). Samples were collected as fresh or frozen tis- using the algorithm in PHYLIP version 3.57 (Felsen-

sues (embryos and hatchlings) or blood from nesting stein, 1993). Nucleotide diversity ( ; equation 10.19 females. In the latter approach, a small aliquot of blood in Nei, 1987) and haplotype diversity (h; equation 8.5 (˜1 ml) was removed from the cervical sinus of nest- in Nei, 1987) within each region and overall were cal- ing turtles after egg deposition, using the technique culated with REAP version 4.0 (McElroy et al., 1992). described by Owens and Ruiz (1980). The proportion of genetic diversity distributed with- Blood was stored in a lysis buffer (100 mM Tris- in and among nesting populations was estimated with HCl, pH 8; 100 mM EDTA, pH 8; 10 mM NaCl; an analysis of molecular variance (AMOVA version 1.0% sodium dodecyl sulfate) in approximately a 1:10 1.5; Excofﬁer, Smouse & Quattro, 1992), and a chi- ratio of blood to buffer as recommended by White and squared test of haplotype frequencies among nesting Densmore (1992). Our experience indicates that blood populations was performed with the program CHIRXC collected for genetic analyses can be preserved in lysis (Zaykin & Pudovkin, 1993) using 1000 randomiza-

buffer for extended periods (> 1 year) at room temper- tions of the original data matrix to estimate a proba- ature without substantial DNA degradation. Females bility distribution for each test (see Roff & Bentzen, renest on cycles of approximately 13-17 days, so spec- 1989). An estimate of average gene ﬂow among nest- imens at each location were collected within a 10-day ing locations (Nm: number of migrants per generation) time frame to avoid resampling the same female. Sam- was calculated with the private allele method (Slatkin ples of L. kempi were obtained from strandings on the & Barton, 1989). Maternal gene ﬂow between pairs of Atlantic coast of the United States and Gulf of Mexi- nesting colonies was estimated using the relationship

co, and presumably these are afﬁliated with the single Nm = 1/2 (1/Gst -1) (Takahata& Palumbi, 1985), using

st nesting population in the western Gulf of Mexico. the Gst analog ( )inAMOVA. Total genomic DNA was isolated with a phenol- Locations with sample sizes below n=8 (Malaysia, chloroformprocedure and stored in a Tris/EDTA buffer n=4; Guinea Bissau, n=5) were excluded from chi- (Hillis, Moritz & Mable, 1996). Amplifications of con- squared tests of independence, estimates of migration trol region sequences were accomplished with stan- between rookery pairs, and rookery-specific estimates dard PCR conditions (Innes, Gelfand & Sninsky, 1995) of genetic diversity. However, these samples were using primers described by Allard et al. (1994). An 18- included in species-wide estimates of genetic variance, base ‘universal’ M13 sequence was added to the 5 0 end haplotype diversity, and nucleotide diversity. of primers to facilitate automated sequencing.Standard To elucidate relationships among haplotypes, evo- precautions, including negative controls (template-free lutionary trees were generated with the parsimony PCR reactions) were used to detect contamination and approach of PAUP version 3.1.1 (Swofford, 1993) and related problems. Sample sets from each location were the neighbor-joining algorithm (Saitou & Nei, 1987) in processed separately to safeguard against accidental MEGA (version 1.02; Kumar, Tamura & Nei, 1994). mixing. Support for nodes in both neighbor-joining and parsi- PCR products were purified with Millipore 30,000 mony approaches were assessed with a bootstrap con- MW filters. Double-stranded sequencing reactions fidence level using 500 replicates. An unrooted parsi- were conducted with fluorescently labelled M13 182 mony network was constructed to elucidate phylogeo- graphic patterns.

Results 0 The 50 and 3 ends of ampliﬁed PCR product were not resolvable in all samples, so this analysis refers to a

470 bp fragment that begins 29 bp from the 50 primer. In a comparison of 80 olive ridleys and 9 Kemp’s ridleys, we observed 51 variable sites, which included 36 transitions, 6 transversions, 4 single-nucleotide indels (insertions or deletions), and one 7 bp indel (Table 1). These polymorphisms resolved 16 haplotypes among the 8 sampled locations (Table 2). Within L. olivacea, sequence comparisons revealed 26 variable sites, including 16 transitions, 2 transversions, 3 single-nucleotide indels, and one 7 bp indel. Eleven of the 12 olive ridley haplotypes were separated by no more than 6–7 sequence changes (p=0.002-0.015 sequence divergence). The exception was a distinctive haplotype observed in specimens from Sri Lanka (‘K’ in Table 1) that differs from other olive ridley haplotypes by 11–15 sequence changes (p=0.022-0.031). Haplotype diversity within locations and overall was moderate or lower than values obtained from other sea turtle species (Table 3). Nucleotide diversity within regions and overall was very low, reﬂecting the shal-

low separations among haplotypes (Table 3). spp. Within L. kempi, sequence comparisons revealed four haplotypes characterized by four transitions, a single transversion, and no indels. Sequence divergence among these four haplotypes ranged from p=0.004- Lepidochelys 0.009. Between the most closely-related L. kempi and L. olivacea haplotypes, we observed 20 transitions, 3 transversions, and 2 indels. Sequence divergences between the two species ranged from p=0.052-0.069. Notably, the 7 bp deletion described above, and 4 transversion substitutions, were shared between L. kempi and the divergent L. olivacea haplotype (‘K’ in Table 1). The low divergence among L. olivacea haplotypes is reﬂected in a neighbor-joining tree (Figure 2). The topology of this tree indicates that the distinctive haplotype ‘K’ observed in the Indian Ocean is closest to the Kemp’s ridley haplotypes in the neighbor-joining tree, and this orientation is supported at a bootstrap . Variable sites observed in control region sequences of value of 98% in the parsimony tree and 97% in the neighbor-joining tree. Table 1 183

Table 2. Distribution of haplotypes among Kemp’s ridley (presumed to be a single nesting population) and seven olive ridley nesting beaches. Abbreviations: LK=L. kempi, SU=Surinam, BR=Brazil, GB=Guinea Bissau, SL=Sri Lanka, MA=Malaysia, AU=Northern Australia, CR=Paciﬁc Costa Rica

Location Haplotype LK SU BR GB SL MA AU CR

A2 B1 C1 D5 E2 F11154 G3 H2 I4 J355 K8 L4Figure 2. Neighbor-joining tree describing the relationships among M1four Kemp’s ridley haplotypes (A-D) and eight olive ridley hap- N7lotypes (E-P). Bootstrap values for critical nodes are derived from O5neighbor-joining analysis (above the branch) and a parsimony analysis (below the branch). The branch order observed in parsimony P1 analysis (PAUP, branch and bound, 50% majority rule) is identical to the neighbor-joining tree except for an unresolved polycotomy among the four shallow lineages in the olive ridley (clusters E-F, Table 3. Haplotype diversity (h) and nucleotide diver- G-I, J, and L-P).

sity ( ) estimates for Kemp’s ridley, for olive ridley (overall), and for ﬁve olive ridley nesting populations. Sample sites in Malaysia (n=5) and Guinea Bissau (n=4) were excluded from population-level estimates of diversity due to low sample size

Haplotype Nucleotide significant test was a comparison of Atlantic nesting diversity diversity colonies (sites in Brazil and Surinam, separated by about 2500 kms), and this pair of rookeries had a corre- Kemp’s ridley 0.69 0.0033 spondingly high estimate of interpopulation migration Olive ridley (overall) 0.81 0.0108 (Nm=4.5; Table 4). In general, Nm estimates greater Pacific Costa Rica 0.76 0.0029 than 1–4 indicate that gene flow is sufficient to maintain Australia 0.54 0.0034 Sri Lanka 0.72 0.0207 a relatively homogeneous gene pool (Birky, Maruya-

Brazil 0.0 0.0 ma & Fuerst, 1983; Slatkin, 1987). Most estimates Surinam 0.28 0.0006 based on st were below this range, and the average number of migrants per generation based on endemic

alleles (Slatkin & Barton, 1989) was Nm 0.6. These estimates should not be overinterpreted, as recent his- Analysis of molecular variance within L. olivacea torical events (such as colonizations) cannot be distin- revealed that approximately 59% of observed haplo- guished from contemporary movement between nest-

type diversity is partitioned among nesting colonies ing colonies. In particular, the low mtDNA diversity

( st =0.586), a ﬁnding that indicates strong popula- observed in Atlantic nesting sites (with one haplotype tion structuring among assayed nesting colonies. In observed in 94% of samples) limits conclusions from haplotype frequency comparisons, nine out of ten chi- a Brazil-Surinam comparison. Nonetheless, the segre- square tests were highly signiﬁcant (Table 4), reinforc- gation of mtDNA haplotype among assayed rookeries ing the conclusion of substantial population structure indicates strong population separations on a global across the global range of L. olivacea. The single non- scale. 184

Table 4. Measures of population genetic separations among ﬁve olive ridley nesting colonies. Sample collections from

Malaysia (n=5) and Guinea Bissau (n=4) were excluded due to low sample size. Above the diagonal: Estimates of

st migration (Nm) among nesting colonies based on G analog ( st ). Below the diagonal: results of chi-squared tests

of independence among nesting colonies. The probabilities for all tests were P < 0.001 except for the comparison of Brazil and Surinam (in bold), which was not signiﬁcant (P=0.206)

Paciﬁc Costa Rica Australia Sri Lanka Brazil Surinam

Paciﬁc Costa Rica – 0.2 0.4 0.1 0.1 Australia X2=26 – 1.1 0.1 0.2 Sri Lanka X2=35 X2=25 – 0.5 0.1 Brazil X2=33 X2=23 X2=32 – 4.5 Surinam X2=31 X2=21 X2=30 X2=2.49 –

of Panama (now known to be 3-4 MY BP; Donnel- ly, 1989), and is consistent with the suggestion by Pritchard (1969) that ancestral populations of Kemp’s and olive ridleys were initially isolated by the closure of a marine corridor between West Atlantic and East Pacific. The olive ridley is characterized by low haplotype diversity, low nucleotide diversity, and a shallow mtDNA phylogeny relative to other Cheloniid sea turtles (see Bowen & Karl, 1996). However, a phylogeo- graphic pattern is apparent in the neighbor-joining tree and the parsimony network (Figure 3). The deepest lineage within L. olivacea (haplotype K) and the haplotype central to the parsimony network (haplotype J) are located in the Indian and West Pacific region. An Figure 3. Parsimony network based on the nucleotide sequence Atlantic cluster (haplotypes E-F) and an East Pacific changes observed in Lepidochelys spp., including 36 transitions, 6 cluster (haplotypes L-P) radiate from the Indian-Pacific transversions, and 5 indels. All sequence changes are denoted as cluster by one and two site substitutions, respectively. dashes on the bars between haplotypes. Asterisk indicates one case While bootstrap support for the monophyly of Atlantic of suspected homoplasy at site 400. and East Pacific lineages was low (Figure 2), samples from these regions always clustered together in the phylogenetic analyses. Discussion How does this pattern of haplotype relationships compare to the biogeographic model based on distrib- Historical biogeography ution and morphology? Pritchard and Trebbau (1984, p. 343) noted a lack of morphological differentiation The primary feature of the control region comparisons across the range of the olive ridley, concluding that is a relatively deep separation between Kemp’s rid- ‘populations of Lepidochelys olivacea even in the East ley and olive ridley sequences (p=0.052-0.069). This Pacific (where the most distinctive populations of other observation is consistent with prior comparisons of cheloniid species are located) are very similar to those ridley mtDNA sequences in a phylogenetic context in the Atlantic and Indian Oceans’. A virtually identi- (Bowen, Meylan & Avise, 1991; Bowen, Nelson & cal statement can be made about mtDNA lineages. The Avise, 1993; Dutton et al., 1996). Under a molecu- most common haplotypes in each ocean basin (F, J, and lar clock proposed for marine turtle control regions N corresponding to Atlantic, Indian-Pacific, and East (approximately 2%/MY; Encalada et al., 1996), the Pacific Oceans respectively) are separated by only three separation of Kemp’s and olive ridleys would date to nucleotide substitutions (p=0.004-0.006). Under the about 2.5 to 3.5 MY BP. This time frame is in good aforementioned molecular clock for marine turtle con- general agreement with the closure of the Isthmus 185

Figure 4. Parsimony network illustrating the relationship among Atlantic, Indian-West Pacific, and East Pacific haplotypes for L. olivacea. These findings indicate a radiation of lineages from the Indian-West Pacific westward into the Atlantic and eastward into the East Pacific. trol region sequences (2%/MY), these lineages share a 2) a recent dispersal event from the Indian Ocean to common ancestor at approximately 200,000-300,000 the Atlantic Ocean. However, one expectation of the years BP. The most divergent haplotypes observed in model may not be supported by the mtDNA data. If L. olivacea would coalesce at approximately 1.5 MY the ancestors of Kemp’s and olive ridley were isolated BP. Such clock estimates must be considered with cau- by the Central American land bridge, then East Pacific tion because calibrations are imprecise,but the mtDNA populations of L. olivacea may be expected to con- data are concordant with morphological comparisons tain the oldest mtDNA lineages and those closest to L. (Pritchard, 1969) in indicating that widely separated kempi. In contrast, the closest olive ridley haplotypes olive ridley populations shared a common ancestor in to L. kempi and the most divergent lineages observed recent evolutionary history. within L. olivacea are in the Indian Ocean and West Based on a synthesis of geographic distribution and Pacific regions. The radiation of mtDNA lineages (Fig- comparative morphology, Pritchard (1969) suggested ure 4) appears to indicate colonization outward from that the Atlantic Ocean was recently colonized by olive the Indo-West Pacific rather than the East Pacific. How ridleys from the Indian Ocean via the Cape of Good may this discrepancy be explained? It is possible that Hope (Figure 1). Contemporary oceanic current pat- sampling is a factor; several major nesting colonies terns and the presence of olive ridleys in waters around were unavailable for this study, and the East Pacif- southern Africa are consistent with this interpretation ic is represented by a single nesting colony. Surveys (Hughes, 1972), and this route has been widely sited of additional nesting sites could alter the structure of as a path of colonization for Indo-Pacific fauna invad- the mtDNA genealogy and corresponding conclusions. ing the Atlantic Ocean (Briggs, 1974). The mtDNA However, a detailed analysis of ridley populations in data support this model by demonstrating a recent link the East Pacific in progress has not uncovered addition- between ridleys in the Atlantic and Indian Oceans; al deep mtDNA lineages (Abreu-Grobois et al., 1996 haplotype ‘J’ observed in the Indian Ocean and West and unpublished data). Pacific samples is distinguished from the most com- The putative discrepancy between genetic data and mon Atlantic haplotype by a single site substitution. the biogeographic model may merely reflect the limits The relationships between Atlantic and Indian-Pacific of inference available from the mtDNA genealogy. The haplotypes (Figure 3), and the low mtDNA diversity deepest node in Figure 2 reflects the original divergence among Atlantic haplotypes (Table 3), are both consis- of Kemp’s and olive ridleys (p=0.052-0.069). The next tent with a recent colonization of the Atlantic. deepest node, characterizing the outermost differenti- The mtDNA data are in good general agreement ation within L. olivacea (p=0.022-0.031), is approxi- with the biogeographic model proposed by Pritchard mately half of the depth of the kempi-olivacea split. In (1969) in supporting: 1) a vicariant separation of L. considering the structure of this tree, it is apparent that kempi and L. olivacea by the Isthmus of Panama, and the mtDNA gene genealogy within L. olivacea traces 186 only the most recent half of the history of this species. Comparative phylogeography Whatever clues existed about the early history of L. olivacea were erased during the mtDNA lineage sort- How does the phylogeography of ridleys compare ing process that led to the most recent radiation in the to patterns in other globally-distributed marine tur- Indian-West Pacific region. Additional gene genealo- tles? A range-wide survey of the tropical green turtle gies, including those from the nuclear genome, would reveals two primary mtDNA lineages corresponding doubtless prove informative (Karl, 1996). to Atlantic-Mediterranean and Indian-Pacific cohorts The central position of Indian-West Pacific haplo- (Bowen et al., 1992), and these lineages are distin- types in the mtDNA genealogy does not necessarily guished at approximately p=0.044 in control region contradict the biogeographic model, because conclu- comparisons (Encalada et al., 1996) as contrasted sions about the early history of olive ridleys based to p=0.052-0.069 between Kemp’s and olive ridleys. on the genetic data are limited. However, this finding These findings reinforce the premise that land masses invokes a point about the transient nature of sea tur- that separate Atlantic and Indian-Pacific basins con- tle nesting habitats, particularly in the East Pacific. A stitute long-standing barriers to dispersal of tropical growing body of evidence indicates that the climate species. in the tropical eastern Pacific has not been stable dur- The mtDNA data for the loggerhead turtle, which ing recent evolutionary time (Guilderson, Fairbanks & has a more temperate distribution, reveals evidence Rubenstone, 1994; Kotilainen & Shackleton, 1995). of a long separation between Atlantic-Mediterranean Cold water extentions across the equator have proba- and Indian-Pacific populations followed by recent bly eliminated tropical faunas repeatedly over the last exchange via southern Africa (Bowen et al., 1994). few million years. In contrast, the Indo-Pacific region This interoceanic exchange may be facilitated by has been one of the warmest bodies of water on earth a rookery in Natal, South Africa, located within for at least 20 million years, providing relatively stable 1000 kms of the Atlantic Ocean. These findings illus- habitat for tropical species. Perhaps ancestral popula- trate that contemporary continental barriers are less tions of L. olivacea in the East Pacific were extirpated restrictive to temperate-adapted marine species. by cold water conditions, and subsequently recolo- Like the tropical green turtle and the temperate log- nized from the more stable Indo-West Pacific region gerhead turtle, the genus Lepidochelys contains two on a time frame under 300,000 years BP (based on the primary mtDNA lineages. In all three cases this bifur- aforementioned clock estimate). It is notable that East cation is attributable to long-standing continental bar- Pacific populations of the green sea turtle have low riers. However, the olive ridley and loggerhead tur- mtDNA nucleotide diversity as well, possibly indicat- tle have recently traversed these continental barriers, ing shallow histories (Bowen et al., 1992). invoking a point about the age of intraspecific separa- One final limitation on the genetic data bears tions in marine turtles. The seven extant marine turtle consideration here: when applying an mtDNA gene species have existed for millions of years (the ridleys genealogy to reconstruct historical biogeography, being the youngest species in terms of genetic diver- researchers cannot readily distinguish between true gences; Bowen & Karl, 1996), and some lineages such introductions (taxon range extensions) and replace- as the green turtle may have evolved independently ments (introgression of migrant lineages into preex- for tens of millions of years (see Bowen, Nelson & isting populations). Although the genetic evidence for Avise, 1993; Dutton et al., 1996). Yet the separations a recent invasion of the Atlantic by L. olivacea is observed between Atlantic-Mediterranean and Indian- unequivocal, the mtDNA data cannot eliminate the Pacific populations of Ch. mydas, Ca. caretta,andL. possibility that olive ridleys existed in the Atlantic pri- olivacea probably date from a few thousand to a few or to this interval, but were extirpated or replaced by million years. Over moderate evolutionary timescales lineages invading from the Indian Ocean. In this case, (104 to 106 years) the barriers between ocean basins are the hypothesis of a range extension into the Atlantic breached by sea turtles and other warm water species (rather than a replacement) is supported by analyses with strong dispersal capabilities. of climate, distribution, and morphology, as well as mtDNA data. Additional fossil records and nuclear Population structure gene genealogies would be valuable to test the possibility that olive ridleys existed in the Atlantic prior to Like the other marine turtles, ridleys are known to the most recent invasion. migrate between feeding habitats and nesting beaches, 187 and at least some females have demonstrated fidelity Acknowledgements to a particular nesting beach (Reichart, 1993; Mar- quez, 1994; Plotkin et al., 1995). Earlier researchers This project was made possible by the outstanding con- proposed that natal homing was the cause of this site tributions of A.L. ‘Sam’ Bass, John Avise, Damien fidelity, an expectation that was recently tested and Broderick, Suhashini Hewavisenthi, Steve Johnson, confirmed in three Cheloniid sea turtles (but not in rid- RuthEllen Klinger, Colin Limpus, Maria ‘Neca’ Mar- leys) with mtDNA data (Bowen et al., 1992; 1994; covaldi, and Bruno Paris. For field and lab assistance Broderick et al., 1994; Bass et al., 1996; FitzSim- we gratefully recognize D. Amorocho, C. Abrew, R.M. mons et al., 1997). Do olive ridley populations have Ball, R. Briseno-Due˜ nas,˜ V. Burke, C. Campbell, S. the genetic signatures of natal homing behavior? Nest- Cornelius, S. Epperly, J. Frazier, N. FitzSimmons, ing aggregates from widely separated locations are sig- K. Ibrahham, H. Kalb, G. Marcovaldi, A. Meylan, nificantly different in haplotype frequencies (Table 4), C. Moritz, S. Morreale, J. Parker, R. Pitman, P.T. but a rigorous test of the natal homing model requires Plotkin, S. Sadove, C. Thome, J. Thome, and M. comparison of nesting populations which overlap on Tiwari. We thank K. Bjorndal, A. Bolten, J. Fra- feeding grounds (Meylan, Bowen & Avise, 1990). The zier, W.S. Grant, G.R. Hughes, P.T. Plotkin and P.C.H. oceanic feeding grounds of L. olivacea are known to Pritchard for comments which influenced the project span thousands of kms (Plotkin et al., 1995), but have design and interpretation. Special recognition is due not been demonstrated to span entire ocean basins. It to the staff of Projeto Tartarugas Marinhas (TAMAR; is possible that olive ridley nesting colonies in Brazil Brazil) for steadfast support of this research over the and Surinam overlap on feeding grounds and there- last decade. For logistic assistance we thank F. Baal, G. by provide a test of natal homing, but results of this Cruz, L. Mendez, J.P. Ross, Okeanos Ocean Research comparison are inconclusive due to the low level of Foundation, Orango National Park (Guinea Bissau), observed mtDNA diversity. Indian-Pacific populations Queensland Dept. of Environment and Heritage (Aus- in Sri Lanka and northern Australia are significant- tralia), Dept. of Fisheries (Malaysia), Servicio de Par- ly different in haplotype frequencies, but these nesting ques Nacionales de Costa Rica, Surinam Forest Ser- locations are separated by about 8000 kilometers,prob- vice (STINASU), World Conservation Union (IUCN), ably too distant to provide a robust assessment of natal the Victor Hasselblad Hatchery (Kosgoda, Sri Lan- homing behavior. Comparisons of mtDNA sequences ka) and the Interdisciplinary Center for Biotechnology among rookeries in the eastern Pacific may resolve Research at University of Florida. The DNA Sequenc- this issue (Abreu-Groboisand colleagues, in progress), ing Core at U.F. generated the sequences in this report, and hypervariable nuclear DNA assays (FitzSimmons, and we are endebted to E. Almira, S. Shanker, S. Moritz & Moore, 1995; Peare & Parker, 1996) will Encalada-Boomershine, and A. Garcia-Rodriguez for doubtless provide additional insights into the popula- their conscientious assistance. Funding was provided tion structure and reproductive biology of ridley turtles. by the National Science Foundation Conservation and Finally, rookery-specific mtDNA markers (and cor- Restoration Program, and by CONABIO grant G007 responding haplotype frequency shifts) have recently to F.A.A.-G. been used to assign foraging populations of marine turtles to nesting locations of origin (Broderick et al., 1994; Bowen et al., 1995, 1996; Bolten et al., 1998; References Lahanas et al., 1998). This approach can be very effective in indicating which nesting populations are Abreu-Grobois, F.A., A.L. Bass, R. Briseno-Due˜ nas,˜ P.H. Dutton, impacted by human activities, a necessary prerequi- S.E. Encalada & N.N. FitzSimmons, 1996. Mitochondrial DNA D-loop sequences of marine turtles, pp. 147–162 in Proceed- site to effective management and recovery plans for ings of the International Symposium on Sea Turtle Conservation depleted populations. The olive ridley survey present- Genetics, edited by B.W. Bowen and W.N. Witzell. NOAA Tech. ed here does not provide sufficient coverage of nesting Memo. NMFS-SEFSC-396 (National Technical Information Ser- vice, Springfield, VA). colonies to allow such analyses, but the initial find- Allard, M.W., M.M. Miyamoto, K.A. Bjorndal, A.B. Bolten & B.W. ings indicate that sufficient mtDNA diversity exists in Bowen, 1994. Support for natal homing in green turtles from the Indian and Pacific Oceans to allow this possibility. mitochondrial DNA sequences. Copeia 1994: 34–41. However, the cooperation of all nations that host nest- Anonymous, 1976. Huge turtle slaughter in India. Oryx 13: 325. Avise, J.C., 1994. Molecular Markers, Natural History, and Evolu- ing habitats will be necessary to conduct these feeding tion. Chapman and Hall, New York, NY. ground assessments. 188

Bass, A.L., D.A. Good, K.A. Bjorndal, J.I. Richardson, Z.-M. Hillis, leucine tRNA and control regions of mitochondrial DNA. Mol. J.A. Horrocks & B.W. Bowen, 1996. Testing models of female Phylogen. Evol. 5: 511–521. migratory behavior and population structure in the Caribbean Encalada, S.E., P.N. Lahanas, K.A. Bjorndal, A.B. Bolten, M.M. hawksbill turtle, Eretmochelys imbricata, with mtDNA control Miyamoto & B.W. Bowen, 1996. Phylogeography and population region sequences. Mol. Ecol. 5: 321–328. structure of the green turtle (Chelonia mydas) in the Atlantic Birky, C. W. Jr., T. Maruyama & P.Fuerst, 1983. An approach to pop- Ocean and Mediterranean Sea: a mitochondrial DNA control ulation and evolutionary genetic theory for genes in mitochondria region sequence assessment. Mol. Ecol. 5: 473–484. and chloroplasts, and some results. Genetics 103: 513–527. Encalada, S.E., K.A. Bjorndal, A.B. Bolten, J.C. Zurita, B. Schroed- Bolten, A.B., K.A. Bjorndal, H.R. Martins, T. Dellinger, M.J. Bis- er, E. Possardt, C.J. Sears & B.W. Bowen, 1998. Population coito, S.E. Encalada & B.W. Bowen, 1998. Trans-Atlantic devel- structure of loggerhead turtle (Caretta caretta) nesting colonies opmental migrations of loggerhead sea turtles demonstrated by in the Atlantic and Mediterranean regions as inferred from mtDNA sequence analysis. Ecol. Appl. 8: 1–7. mtDNA control region sequences. Mar. Biol. In press. Bowen, B.W. & S.A. Karl, 1996. Population structure, phylogeog- Excoffier, L., P.E. Smouse & J.M. Quattro, 1992. Analysis of mole- raphy, and molecular evolution, pp. 29–50 in The Biology of Sea cular variance inferred from metric distances among DNA haplo- Turtles, edited by P.L. Lutz and J.A. Musick. CRC Press, Boca types: application to human mitochondrial restriction data. Genet- Raton, FL. ics 131: 479–491. Bowen, B.W. & W.N. Witzell, 1996. Introduction to the international Felsenstein, J., 1993. PHYLIP (Phylogeny Inference Package). Ver- symposium on sea turtle conservation genetics, pp. 1–7 in Pro- sion 3.5. Computer program and manual distributed by the author. ceedings of the International Symposium on Sea Turtle Conser- FitzSimmons, N.N., C. Moritz & S.S. Moore, 1995. Conservation vation Genetics edited by B.W. Bowen and W.N. Witzell. NOAA and dynamics of microsatellite loci over 300 million years of Tech. Memo. NMFS-SEFSC-396, National Technical Informa- marine turtle evolution. Mol. Biol. Evol. 12: 432–440. tion Service, Springfield, VA. FitzSimmons, N.N., C.J. Limpus, J.A. Norman, A.R. Goldizen, Bowen, B.W. & W.S. Grant, 1997. Phylogeography of the sardines J.D. Miller & C. Moritz, 1997. Philopatry of male marine turtles (Sardinops spp.): assessing biogeographic models and population inferred from mitochondrial DNA markers. Proc. Natl. Acad. Sci. histories in temperate upwelling zones. Evolution 51: 1601–1610. USA 94: 8912–8917. Bowen, B.W., A.B. Meylan & J.C. Avise, 1991. Evolutionary dis- Guilderson, T. P., R. G. Fairbanks & J. L. Rubenstone, 1994. Trop- tinctiveness of the endangered Kemp’s ridley sea turtle. Nature ical temperature variation since 20,000 years ago: modulating 352: 709–711. interhemispheric climate change. Science 263: 663–665. Bowen, B.W., A.B. Meylan, J. P. Ross, C.J. Limpus, G.H. Balazs Hendrickson, J.R., 1980. The ecological strategies of sea turtles. & J.C. Avise, 1992. Global population structure and natural his- Amer. Zool. 20: 597–608. tory of the green turtle (Chelonia mydas) in terms of matriarchal Hildebrand, H.H., 1963. Hallazgo del area de anidacion de la tor- phylogeny. Evolution 46: 865–881. tuga marina ‘lora’, Lepidochelys kempi (Garman), en la costa Bowen, B.W., W.S. Nelson & J.C. Avise, 1993. A molecular phy- occidental del Golfo de Mexico. Ciencia, Mexico 22: 105–112. logeny for marine turtles: trait mapping, rate assessment, and Hillis, D. M., C. Moritz & B. K. Mable, eds., 1996. Molecular conservation relevance. Proc. Natl. Acad. Sci. USA 90: 5574– Systematics, Second edition, Sinauer, Sunderland, MA. 5577. Hughes, G.R., 1972. The olive ridley sea-turtle (Lepidochelys oli- Bowen, B.W., N. Kamezaki, C.J. Limpus, G.H. Hughes, A.B. Mey- vacea) in South-east Africa. Biol. Cons. 4: 128–134. lan & J.C. Avise, 1994. Global phylogeography of the loggerhead Innes, M. A., D. H. Gelfand & J. J. Sninsky, 1995. PCR Strategies. turtle (Caretta caretta) as indicated by mitochondrial DNA hap- Academic Press, New York, NY. lotypes. Evolution 48: 1820–1828. Karl, S.A., 1996. Hybridization and taxonomy of marine turtles: Bowen, B.W., F.A. Abreu-Grobois, G.H. Balazs, N. Kamezaki, C.J. anonymous nuclear DNA sequence analyses, pp. 99–108 in Pro- Limpus & R.J. Ferl, 1995. Trans-Pacific migrations of the logger- ceedings of the International Symposium on Sea Turtle Conser- head sea turtle demonstrated with mitochondrial DNA markers. vation Genetics, edited by B.W. Bowen and W. Witzell. NOAA Proc. Natl. Acad. Sci. USA 92: 3731–3734. Tech. Memo. NMFS-SEFSC-396, National Technical Informa- Bowen, B.W., A.L. Bass, A. Garcia, C. E. Diez, R. van Dam, A. tion Service, Springfield, VA. Bolten, K.A. Bjorndal, M.M. Miyamoto & R.J. Ferl, 1996. The Kimura, M., 1980. A simple method for estimating evolutionary rate origin of hawksbill turtles in a Caribbean feeding area as indicated of base substitutions through comparative studies of nucleotide by genetic markers. Ecol. Appl. 6: 566–572. sequences. J. Mol. Evol. 16: 111–120. Briggs, J.C., 1974. Marine Zoogeography. McGraw-Hill Book Co. Kotilainen, A. T. & N. J. Shackleton, 1995. Rapid climate variability New York, NY. in the North Pacific Ocean during the past 95,000 years. Nature Broderick, D., C. Moritz, J.D. Miller, M. Guinea, R.J. Prince & 377: 323–326. C.J. Limpus, 1994. Genetic studies of the hawksbill turtle: evi- Kumar, S., K. Tamura & M. Nei, 1994. MEGA: Molecular evolu- dence for multiple stocks and mixed feeding grounds in Aus- tionary genetic analysis softwear for microcomputers. Cabios 10: tralian waters. Pacific Conserv. Biol. 1: 123–131. 189–191. Carr, A.F., 1963. Panspecific reproductive convergence in Lepi- Lahanas, P.N., M.M. Miyamoto, K.A. Bjorndal & A.B. Bolten, dochelys kempi. Ergebn. Biol. 26:298–303. 1994. Molecular evolution and population genetics of Greater Donnelly, T.W., 1989. History of marine barriers and terrestrial Caribbean green turtles (Chelonia mydas) as inferred from mito- connections: Caribbean paleogeographic inference from pelagic chondrial DNA control region sequences. Genetica 94: 57–67. sediment analysis, pp. 103–118 in Biogeography of the West Lahanas, P.N., K.A. Bjorndal, A.B. Bolten, S. Encalada, M.M. Indies: Past, Present, and Future, edited by C.A. Woods. Sandhill Miyamoto, R. Valverde & B.W. Bowen, 1998. Genetic compo- Crane Press, Gainesville, FL. sition of a green turtle feeding ground population: evidence for Dutton, P.H., S.K. Davis, T. Guerra & D. Owens, 1996. Molecular multiple origins. Mar. Biol. In press phylogeny for marine turtles based on sequences of the ND4- Lamb, T., C. Lydeard, R. B. Walker & J. W. Gibbons, 1994. Mole- cular systematics of map turtles (Graptemys): a comparison of 189

mitochondrial DNA restriction site versus sequence data. Syst. Pritchard, P.C.H. & P. Trebbau, 1984. The Turtles of Venezuela. Biol. 43: 543–559. Contrib. Herpetol. 2, Society for the Study of Amphibians and Marquez,´ R., 1994. Synopsis of Biological Data on the Kemp’s Rid- Reptiles, Fundacion de Internados Rurales, Caracas, Venezuela. ley Turtle, Lepidochelys kempi (Garman, 1880). NOAA Tech. Reichart, H.A., 1993. Synopsis of biological data on the olive ridley Memo. NMFS-SEFSC-343, National Technical Information Ser- sea turtle Lepidochelys olivacea (Eschscholtz, 1829) in the west- vice, Springfield, VA. ern Atlantic. NOAA Tech. Memo. NMFS-SEFSC-336, National Marquez,´ R., O. Villanueva & C. Penaflores.˜ 1976. Sinopsis de Technical Information Service, Springfield, VA. datos biologicos´ sobre la tortuga golfina Lepidochelys olivacea Roff, D. A. & P. Bentzen, 1989. The statistical analysis of mitochon- (Eschscholtz, 1829). INP Sinopsis Pesca No. 2 SIC-SP-INP drial DNA polymorphisms: chi-square and the problem of small (Mexico). samples. Mol. Biol. Evol. 6: 539–545. McElroy, D., P. Moran, E. Bermingham & I. Kornfield, 1992. REAP: Ross, J.P., 1982. Historical decline of the loggerhead, ridley, and an integrated environment for the manipulation and phylogenetic leatherback sea turtles, pp. 189–195 in Biology and Conservation analysis of restriction data. J. Heredity 83: 157–158. of Sea Turtles, edited by K.A. Bjorndal. Smithsonian Institution Meylan, A.B., B.W. Bowen & J.C. Avise, 1990. A genetic test of Press, Washington D.C., pp. 189–195. the natal homing versus social facilitation models for green turtle Saitou, N. & M. Nei, 1987. The neighbor-joining method: a new migration. Science 248: 724–727. method for reconstructing phylogenetic trees. Mol. Biol. Evol. 4: Nei, M., 1987. Molecular Evolutionary Genetics. Columbia Univ. 406–425. Press, New York, NY. Slatkin, M., 1987. Gene flow and the geographic structure of natural Norman, J.A., C. Moritz & C.J. Limpus, 1994. Mitochondrial DNA populations. Science 236: 787–792. control region polymorphisms: Genetic markers for ecological Slatkin, M. & N.H. Barton, 1989. A comparison of three indirect studies of marine turtles. Mol. Ecol. 3: 363–373. methods for estimating average levels of gene flow. Evolution 43: Owens, D.W. & G.W. Ruiz, 1980. New methods of obtaining blood 1349–1368. and cerebrospinal fluid from marine turtles. Herpetologica 36: Swofford, D.L., 1993. PAUP: Phylogeny analysis using parsimony, 17–20. version 3.1.1, Illinois Natural History Survey, Champaign, IL. Palumbi, S.R., 1994. Genetic divergence, reproductive isolation, and Takahata, N. & S.R. Palumbi, 1985. Extranuclear differentiation and marine speciation. Ann. Rev. Ecol. Syst. 25: 547–572. gene flow in the finite island model. Genetics 109: 441–457. Peare, T. & P.G. Parker, 1996. The use of multilocus minisatel- Walker, D., V.J. Burke, I. Barak & J.C. Avise, 1995. A comparison lite DNA fingerprinting to examine local genetic structure within of mtDNA restriction site vs. control region sequences in phylo- green turtle rookeries, pp. 87–94 in Proceedings of the Interna- geographic assessment of the musk turtle (Sternotherus minor). tional Symposium on Sea Turtle Conservation Genetics, edited Mol. Ecol. 4: 365–373. by B.W. Bowen and W.N. Witzell. NOAA Tech. Memo. NMFS- White, P.S. & L.D. Densmore, 1992. Mitochondrial DNA isolation, SEFSC-396, National Technical Information Service, Spring- pp. 29–58 in Molecular Genetic Analysis of Populations: a prac- field, VA. tical approach, edited by A.R. Hoezel. IRL Press at Oxford Univ. Plotkin, P.T., R.A. Byles, D.C. Rostal & D.W. Owens, 1995. Inde- Press, New York, NY. pendent versus socially facilitated oceanic migrations of the olive Woody, J.B., 1985. Kemp’s ridley continues decline. Mar. Turtle ridley, Lepidochelys olivacea. Mar. Biol. 122: 137–143. Newsl. 35: 4–5. Pritchard, P.C.H., 1969. Studies of the Systematics and Reproductive Zaykin, D.V. & A.I. Pudovkin, 1993. Two programs to estimate sig- Cycles of the Genus Lepidochelys. Ph.D. Thesis, University of nificance of X2 values using pseudo-probability tests. J. Heredity Florida, Gainesville, FL. 84: 152.