Molecular Population Genetics of the Southern Elephant Seal Mirounga leonina

Corresponding author: Robert W. Slade, Queensland Institute of Medical Research, Post Office, Royal Brisbane Hospital, Queensland 4029, Australia., roberts{at}qimr.edu.au (E-mail).

Communicating editor: R. R. HUDSON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Southern elephant seals breed on sub-Antarctic islands and have a circumpolar distribution. We assayed mitochondrial DNA (mtDNA) and nuclear DNA (nDNA) variation in the three main populations in the south Atlantic, south Indian, and south Pacific oceans, and a smaller continental population in South America. Population structure of mtDNA was strong and not consistent with isolation by distance. The nDNA loci, although less informative, were consistent with the mtDNA results. Geographic structure appears to be dominated by historical processes, not contemporary gene flow. Uncorrected levels of nucleotide diversity for mtDNA control region I (2.86%) and nDNA (0.09%) were similar to those in humans and mice. Mutation rates for control region I (75 x 10^-9 substitutions per site per year) and nDNA (1.23 x 10^-9) were similar to those in other mammals. Female effective population size and total effective population size were roughly equal at ~4 x 10⁴, indicating a twofold greater rate of drift for mtDNA. Effective breeding sex ratio of four to five females per male was estimated from nucleotide diversity and mutation rates for mtDNA and nDNA, and was much less than behavioral observations would suggest. There was no evidence for selection at any of the assayed loci.

THE southern elephant seal Mirounga leonina has a circumpolar distribution, and breeding colonies are concentrated on sub-Antarctic islands near the Antarctic convergence (Figure 1). The three main populations are centered on South Georgia (SG) in the south Atlantic Ocean, the geographically close Heard (HD) and Kerguelen Islands in the south Indian Ocean, and Macquarie Island (MQ) in the south Pacific Ocean (LING and BRYDEN 1992 ). There are several small breeding populations on other sub-Antarctic islands, and there is a continental breeding population in the south Atlantic ocean at Península Valdés (PV) in Argentina (CAMPAGNA and LEWIS 1992 ). On the basis of skull characters, LYDEKKER 1909 proposed that three subspecies be recognized, falclandicus, crosetensis, and macquariensis, corresponding to type localities in the south Atlantic, south Indian, and south Pacific oceanic regions, respectively. There are different growth patterns between populations, and these probably underlie the skull character differences. Studies during the 1950s and 1960s showed that individuals from MQ and HD grew at a slower rate and had a smaller ultimate size than individuals from SG (CARRICK et al. 1962A ; BRYDEN 1968 ). BRYDEN 1968 suggested that the growth pattern differences were environmentally determined. Alternatively, the growth pattern differences could be genetically determined, in which case we might expect a closer genetic relationship between the MQ and HD populations than either is to SG.

View larger version (52K): In this window In a new window Download PPT slide   Figure 1. Past and present distribution of the southern elephant seal (taken from LING and BRYDEN 1992 with permission).

Recent studies on population size change have also indicated similarities between HD and MQ compared with SG. Population size was estimated as 350,000 for SG, 80,000 for HD, 157,000 for Kerguelen, and 136,000 for MQ, and the three main populations account for 96% of total population size (MCCANN 1985 ). These figures represent population size during the early 1950s for SG, HD, and MQ, and during the 1970s for Kerguelen. Although the SG population has been stable since that time (MCCANN and ROTHERY 1988 ), populations in the south Indian and south Pacific oceans have declined markedly (HINDELL and BURTON 1987 ). The MQ population, for example, has declined by 50% in 30 yr at a rate of -2% per year. One possible explanation for these changes is that they are due to large-scale movement of individuals between populations (HINDELL and BURTON 1987 ).

One of the aims of this study was to assess the extent of geographic structure in the southern elephant seal and to determine if the genetic relationship between populations corresponded to the relationship indicated by the morphometric and demographic studies. A previous study analyzed the genetic relationship between only the MQ and HD populations, between which there were significant differences in allozyme frequency (GALES et al. 1989 ). A second aim of this study was to obtain more detailed information on other population genetic parameters for both mitochondrial DNA (mtDNA) and nuclear DNA (nDNA). These other parameters include mutation rate, effective population size, and selection for both mtDNA and nDNA. The framework for estimating these parameters relies upon sequence data information on the pattern of variation within and between species. Our previous studies have described the general approach for detecting variation in orthologous nDNA genes within the southern elephant seal and other organisms (SLADE et al. 1993 ), and between the southern elephant seal and other pinniped species (SLADE et al. 1994 ), and have also compared levels of mtDNA variation in southern and northern elephant seals (HOELZEL et al. 1993 ). In this article, we investigate several areas of the molecular population and evolutionary genetics of the southern elephant seal, including level of genetic diversity, rate of molecular evolution, effective population size and effective breeding ratio, geographic distribution of genetic variation, rate of gene flow, and population divergence time.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Samples and molecular methods:
The southern elephant seal blood tissue samples from Macquarie Island (MQ) in the south Pacific Ocean and Heard Island (HD) in the south Indian Ocean were collected by the Australian Antarctic Division. Samples from South Georgia (SG) in the south Atlantic Ocean were collected by the British Antarctic Survey using ronguers to clip ~1–2 g of skin from the webbing in the rear flippers. The MQ and HD samples were from beaches representing the geographic extremes of each island. The samples from other pinniped species were described in SLADE et al. 1994 . For initial sequencing, to determine levels of variation and to search for diagnostic markers, one individual was sequenced from each of MQ, HD, and SG. Several genes were then selected for further sequencing of a total of five individuals from each population (Table 1). For large-scale screening of diagnostic markers, ~30 individuals from each of MQ, HD, and SG were selected. The gene fragments amplified and sequenced are shown in Table 1. Here we detail only those protocols and primer sequences that have not been described previously (SLADE et al. 1993 , SLADE et al. 1994 ).

For mtDNA, amplification with the THR and TDKD primers from 15 individuals (5 from each major population) resulted in 444 bp of sequence data, and analysis focused on a highly variable 299-bp subset comprising all 264 bp of control region I (CRI) and 35 bp of flanking sequence. This CRI subset corresponds to sites 67–365 in the GenBank sequence. (All site numbers refer to the sequences in GenBank; accession numbers in Table 1.) This enabled us to make direct use of published mtDNA sequence data (HOELZEL et al. 1993 ) for the southern elephant seal populations at SG and PV. There were a total of 60 CRI individual sequences available for analysis comprising 28 from SG, 21 from PV, 6 from HD (one individual was heteroplasmic; see RESULTS), and 5 from MQ.

For nDNA, an ~800-bp gene fragment of the ß-globin gene, spanning exons 2–3, was amplified with primers BETA3/BETA4 using cycling parameters of 35 x 94° for 1 min, 65° for 1 min, 72° for 3 min; 168 bp at the 5' end was sequenced with BETA3, and 351 bp at the 3' end was sequenced with BETA4. Sequencing with the BETA4 primer revealed a pentameric microsatellite (GGAAA)_n, the 3' end of which was located 297 bp upstream from the beginning of exon 3. Primers BETA6 and BETA5 were designed to regions flanking the microsatellite. The BETA6/BETA5 amplified product consisted of 149 bp of flanking sequence and from 6 to 20 repeats. The cycling parameters for BETA6/BETA5 were 35 x 94° for 1 min, 65° for 1 min, 72° for 1 min. An ~260-bp fragment of the Corticotropin-releasing factor gene was amplified and sequenced with primers OSCRFA1/OSCRFA2 using cycling parameters of 35 x 94° for 1 min, 50° for 1 min, and 72° for 1 min. An ~1-kb fragment of the lysozyme gene was amplified with primers LYSO1/LYSO2 using cycling parameters of 35 x 94° for 1 min, 55° for 1 min, 72° for 3 min, and 303 bp at the 5' end was sequenced using LYSO1, and 311 bp at the 3' end was sequenced with LYSO2. The PCR primers not described previously are (5' 3') LYSO1 (AGGTCTTTGRACGDTGTGA), LYSO2 (GGGGTTTTGCCATCATTACA), DQA4 (ACACATACCATTGGTAG), BETA6 (AATTGGGCATGTGATGTATGAG), and BETA5 (AATTAGTATGATGCTGGGCTGTC).

For microsatellite PCR, the BETA5 primer was end-labeled with [-³³P]dATP according to a standard protocol (SAMBROOK et al. 1989 ). This was used in a regular PCR reaction with the BETA6 primer except that the cycling parameters involved denaturation at 94° for 30 sec, variable annealing temperatures (see below), and extension at 72° for 2 min. The annealing temperatures followed a "touchdown" procedure to reduce nonspecific amplification (DON et al. 1991 ) and were 67°, 30 sec for the first 2 cycles, then 66°, 30 sec for 2 cycles, and then 65°, 30 sec for 25 cycles. Formamide stop solution was added to the product, which was heated to 95° for 2 min, and 4 µl was loaded onto a 4.5% sequencing gel. The bands were visualized after autoradiography. Electrophoresis revealed up to four bands per individual, indicating that at least two microsatellite loci were being amplified. Direct sequencing of the BETA6/BETA5 product from several individuals revealed a clean sequence with one variant site, located within the repeat unit. It is likely, therefore, that there are at least two ß-globin genes in the southern elephant seal and that these are the result of a very recent duplication or possibly a very recent homogenizing event between two preexisting ß-globin genes. The presence of two ß-globin genes in the southern elephant seal was previously suggested by isozyme studies (SEAL et al. 1971 ). Although it was not possible to allocate alleles to separate microsatellite loci, the clear and reproducible dosage differences did allow scoring of the combined loci allele frequencies. For example, for individuals with three bands there was always one band with increased intensity, indicating the presence of two alleles for that number of repeats.

For each sequence with a variant nucleotide site, a computer search was made using the "MacVector" program (IBI-Kodak) for corresponding variant restriction sites. For mtDNA CRI there were several variant restriction sites, and three enzymes (RsaI, FokI, BslI) were chosen, on the basis of their geographic distribution, to use in a large-scale survey (Figure 2A). For ALD-A (Figure 2B), there was a variant site at position 202 that was part of a NspI restriction site (SLADE et al. 1993 ). The presence/absence of the 6-bp indel in intron 2 of Mhc-DQA was assayed by double digestion of the PCR fragment with Sau96I and NlaIV and subsequent PAGE. The restriction digests were conducted as in SLADE et al. 1993 .

View larger version (45K): In this window In a new window Download PPT slide   Figure 2. (A) The 16 different sequence haplotypes of CRI from 15 individuals. (B) The 5 different sequence haplotypes of ALD-A from 15 individuals. The site numbers correspond to the sequences in GenBank, and the reference sequences in GenBank are SG1 and haplotype a for CRI and ALD-A, respectively. Diagnostic RFLP sites are in bold, and for CRI these are FokI @ 196 and 326, RsaI @ 169, and BslI @ 305. For CRI, the heteroplasmic individual contained haplotypes HD3 and HD4. SEE HOELZEL et al. 1993 for CRI haplotypes of the PV and other SG sequences.

Data analysis:
For those nDNA sequences with multiple variant sites (i.e., ALD-A and Lysozyme), the haplotypes were determined. For ALD-A, the method of phase determination was described in SLADE et al. 1993 . Briefly, the haplotypes of all double heterozygotes were separated by digestion with NspI, and the separate fragments sequenced. For Lysozyme, phase was determined by reference to the unambiguous haplotypes observed after sequencing homozygotes. There was also one individual that was heteroplasmic for mtDNA, and the two haplotypes were determined by reference to the unambiguous haplotypes. Phase can be inferred from the unambiguous haplotypes assuming that there has been no recombination that is reasonable for short PCR fragments (CLARK 1990 ).

The pairwise genetic distances between sequences were calculated with the "MEGA" program (KUMAR et al. 1993 ) using pairwise-deletion to account for gaps. The genetic distances were the uncorrected number or proportion of differences, and for some analyses of the CRI data, distances were corrected using the Tamura-Nei gamma model (TAMURA and NEI 1993 ). For the latter, the scale parameter for the negative-binomial distribution was estimated by maximum-likelihood using the macro "kfit" (CRAWLEY 1993 ) for the program "GLIM" (Release 3.772; Royal Statistical Society, London). The input data were the number of substitutions that occurred at each site, and this was estimated from a parsimony tree (WAKELEY 1993 ) constructed in "PHYLIP" (FELSENSTEIN 1993 ). The frequency distribution of the inferred number of substitutions per site was compared with expected poisson (SOKAL and ROHLF 1969 , p. 86) and negative binomial (BLISS and FISHER 1953 ) distributions.

Nucleotide diversity within populations was calculated as in equation 10.6 in NEI 1987 . For nDNA data, this was based on the uncorrected proportional distance, and for CRI data it was based on both the uncorrected proportional distance and the Tamura-Nei gamma corrected distance. The variance of nucleotide diversity was that because of estimation errors of nucleotide substitutions. For the uncorrected proportional distance, the variances were calculated as in NEI and JIN 1989 using the program "SEND." For the Tamura-Nei corrected distance, the maximum variance was calculated as in TAKAHATA and TAJIMA 1991 using a BASIC program modified by R. W. SLADE from one kindly provided by Y. SATTA.

The distribution of genetic variation between populations was estimated in several ways. For CRI data only, we utilized a procedure analogous to analysis of variance with the "AMOVA" program (EXCOFFIER et al. 1992 ). As recommended, the distance matrix used was the number of nucleotide differences between pairs of sequences, and the significance levels were determined after 1000 permutations. Also, for CRI data only, we calculated gross and net nucleotide divergence between populations as in equations 10.20 and 10.21, respectively, in NEI 1987 . For CRI haplotype and nDNA allele data, ² tests of independence were conducted using "Monte" (ROFF and BENTZEN 1989 ) in the "REAP" suite of programs (MCELROY et al. 1992 ), and significance levels were determined after 1000 permutations.

To estimate levels of long-term gene flow, we calculated analogs of G_ST between pairs of populations from the CRI sequence data and from the nDNA allele frequency data. For the CRI sequence data, estimates of G_ST were calculated between pairs of populations as K_ST = (HUDSON et al. 1992 ), where K_S is the average nucleotide diversity within the two populations, and K_T is the total nucleotide diversity among all sequences from the populations. The degree of maternal gene flow (N_fm_f) between each pair of populations was then estimated from K_ST = , where = ()² and n is the number of populations exchanging migrants (CROW and AOKI 1984 ), in this case four. For the nDNA data, estimates of G_ST between pairs of populations were obtained from allele frequency data from several loci using the program "BIOSYS-1^1.7" (SWOFFORD and SELANDER 1989 ), and total gene flow (Nm) between pairs of populations was estimated from K_ST = (CROW and AOKI 1984 ).

The CRI sequences were analyzed for any pattern of selection among haplotypes using the methods of TAJIMA 1989 and FU and LI 1993 . For the latter, a northern elephant seal sequence was used as the outgroup. All sequences were analyzed for selection using 2 x 2 tables of within-species polymorphism and between-species fixed differences for two loci as in KREITMAN and AGUADE 1986 . For this test, a significant departure from a neutral model could be because of linkage and stochastic factors rather than selection (HUDSON et al. 1987 ). However, the ease of the ² test means that it is an appropriate first choice and that, subsequently, any significant associations can be revisited with the more conservative HKA test. For comparisons between mtDNA and nDNA sequences, the difference in effective population size needs to be accounted for (NACHMAN et al. 1994 ). All nDNA was assumed to be autosomal, and because N_f was found to be equal to N_e, the polymorphism within the mtDNA was corrected down by one-half.

The total effective population size (N) and the female effective population size (N_f) were estimated from the formulae = 4Nu_g and = 2N_fu_g, respectively. The neutral parameter was taken to be equivalent to nucleotide diversity (). The mutation rate per site per generation (u_g) was calculated as the mutation rate per year multiplied by the generation time. The generation time for the southern elephant seal of 8 yr was calculated from female life tables of the stable population at South Georgia (MCCANN 1985 ) using the formula , where x is age in years, l_x is the proportion of females surviving to age x, and m_x is the number of females produced per female at age x (BEGON et al. 1986 ). The effective breeding ratio was estimated from data on mutation rates (µ_n and µ_mt) and levels of diversity (_n and _mt) for nDNA and mtDNA, respectively. The mutation rates do not depend on effective population size, although diversity does, and so for neutrally evolving sequences, and assuming mutation-drift equilibrium, the following should be true:

(1)

N is composed of male and female components such that N = (WRIGHT 1931 ). Substitution of the right-hand side into Equation 1 above results in the following:

(2)

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

From mtDNA, we analyzed 299 bp of CRI from 60 individuals and assayed three diagnostic restriction sites from 115 individuals. From nDNA we sequenced 3594 bp of mostly noncoding regions from seven single-copy loci from 6 haplotypes (three individuals) from the geographic extremes of the range, and 877 bp of this was sequenced from 30 haplotypes. Three nDNA diagnostic markers (restriction site, indel, microsatellite) were assayed from ~186 haplotypes.

Sequence variation:
nDNA: The pilot analysis of one individual per main population of SG, HD, and MQ (i.e., a total of 6 haplotypes) detected variation in 7 of the 15 nDNA gene segments (Table 1). From this initial screening, several segments were selected to increase the sample size from one individual per population to five individuals per population (i.e., a total of 30 haplotypes). The segments chosen for further analysis were ALD-A, exons 2 and 3 of Mhc-DQA, and two segments of ß-globin. Subsequently, it was discovered that two ß-globin loci were being amplified simultaneously and that it was not possible to distinguish alleles from loci (see MATERIALS AND METHODS). Therefore, the 734 bp of ß-globin sequence was not included in further analyses of sequence variation. The 3594 bp of nDNA analyzed (excluding the 734 bp of ß-globin) comprised 877 bp sequenced from 30 haplotypes and 2717 bp sequenced from 6 haplotypes. Of the total 3594 bp, there were 2838 bp of silent sites comprised of 467 bp sequenced from 30 haplotypes and 2371 bp sequenced from 6 haplotypes. The silent sites were assumed to be the synonymous sites in exons, all intron sites except for the AG/GT splice sites, and all of the 5' UTR and pseudogene sites.

The variants detected included 9-point substitutions, a 6-bp indel, and a pentameric microsatellite. Of the nine variant sites, five were found in introns, two were silent polymorphisms in exons, and there were two nonsynonymous polymorphisms, glycine/arginine in exon 2 of Mhc-DQA and glycine/arginine in exon 2 of Lysozyme. There were five transitions and four transversions resulting in a roughly twofold bias of transitions over transversions. The total diversity (uncorrected) for nDNA of 0.09 ± 0.03% was calculated from 877 bp from 30 haplotypes and 2717 bp from 6 haplotypes. The level of silent site diversity of 0.08 ± 0.03% was calculated from 467 bp from 30 haplotypes and 2371 bp from 6 haplotypes. The diversity calculated from the samples from the initial screening was the same, and therefore we assume that no systematic bias was introduced by selecting some gene fragments for further sampling. Phase was determined for those nDNA genes with more than one polymorphic site, ALD-A and Lysozyme. For ALD-A, the phase of the variant sites is shown in Figure 2B. for Lysozyme, the T/C polymorphism at site 239 was in phase with the A/G polymorphism at site 273.

mtDNA: The level of polymorphism in the 299 bp of mtDNA CRI sequence was considerably higher with 16 different haplotypes observed from 15 individuals, 5 from each of the three main populations (Table 1, Figure 2A). One HD individual was heteroplasmic. All 26 variant sites were transitions. All further analyses of CRI sequence variation combined this data set of 16 sequences with the data set of HOELZEL et al. 1993 , giving a total of 60 individual sequences with 38 variant sites (five transversions) and 38 different haplotypes. The uncorrected nucleotide diversity in these 60 CRI sequences was 2.86 ± 0.49%. Several studies have shown substitution rate variation among sites in the mammalian control region as evidenced by departure from a Poisson distribution of the number of substitutions per site (e.g., KOCHER and WILSON 1991 ). The mean of the inferred number of substitutions per site from the 60 CRI sequences was less than the variance (0.2843 and 0.8820, respectively), indicating a non-Poisson distribution. The frequency distribution of the inferred number of substitutions per site did not fit a Poisson distribution (² = 161.3, P < 0.001) but did fit a negative-binomial distribution (² = 0.98, P > 0.25). The scale parameter for the negative-binomial distribution was estimated by maximum-likelihood as 0.1042, and therefore we supplied this to the Tamura-Nei gamma model of pairwise distances giving a nucleotide diversity of 4.75 ± 2.67%.

Selection: Using the method of TAJIMA 1989 on the 60 CRI sequences, the test statistic D was equal to -0.427, which is not significant (P > 0.1). Using the method of FU and LI 1993 on the 60 CRI sequences, with the northern elephant seal as the outgroup, the test statistic D was equal to 0.55, which is not significant (P > 0.05). The Kreitman and Aguadé ² test of independence on all the variant nDNA sequences and the 60 CRI sequences showed no significant deviations from the neutral expectation of levels of polymorphism within species, given the number of fixed differences between species.

Rate of molecular evolution, N_e and N_f:N_m: The rate of neutral evolution for ~1200 bp of silent sites in pinniped nDNA (see Table 1 for genes sequenced) was previously estimated with a Kimura two-parameter model (SLADE et al. 1994 ) from three fossil-record calibrations (divergence times of 4.5, 14.5, and 23 mya) as being 1.23 ± 0.24 substitutions per nucleotide site per 10⁹ yr (i.e., 1.23 ± 0.24 x 10^-9). For the 299 bp of CRI, and using the Tamura-Nei gamma model with one fossil-record calibration (the 4.5 mya divergence between Mirounga spp. in one clade and Hydrurga leptonyx and Leptonychotes weddelli in the other clade; RAY 1976 ; SLADE et al. 1994 ), the rate of CRI evolution was estimated as 75 ± 46 x 10^-9.

The total effective population size of roughly N_e = 4 x 10⁴ (Table 2) was calculated from the silent site nucleotide diversity given above. We considered that there was not enough nDNA variation to properly estimate N_e within populations. The female effective population size calculated from the 60 CRI sequences was roughly N_f = 4 x 10⁴, and each of the three main populations has an N_f of roughly 2–3 x 10⁴ (Table 2). The PV population is an order of magnitude smaller than the other southern elephant seal populations, and the effective number of northern elephant seals is an order of magnitude smaller than that for the southern elephant seal. The effective breeding ratio was estimated from Equation 2 to be four to five females to one male.

Geographic structure:
Distribution of CRI sequence variation: The levels of CRI nucleotide diversity and divergence within and between populations are shown in Table 3. Diversity was highest within the SG and HD populations and very low by comparison in the PV population. The greatest divergence was between the PV and MQ populations, and the least was between the SG and HD populations. Although PV and SG are in the same oceanic region, there is almost an order of magnitude of greater divergence between those two populations than between the separate oceanic populations of SG and HD. The AMOVA results showed that 57% of CRI variation was distributed between populations. This distribution of mtDNA variation was significantly different from random (P < 0.001).

For a hierarchical AMOVA, we compared oceanic regions, with SG and PV representing the south Atlantic ocean, HD the south Indian ocean, and MQ the south Pacific ocean. This resulted in a decomposition of variation of -5% among the three oceanic regions, 61% among SG and PV within the south Atlantic oceanic region, and 44% within populations. The breakdown into oceanic regions thus explained none of the variation. This is because of heterogeneity within the south Atlantic ocean between SG and PV and the close relationship between haplotypes from SG and HD representing different regions.

Distribution of nDNA allele and CRI haplotype frequencies: Diagnostic markers were screened in ~30 individuals from each of the three main populations. For CRI, the combined RFLP haplotypes and their population frequencies are shown in Figure 3. Most haplotypes were found in more than one population; however, none of the MQ haplotypes were shared with HD, SG, or PV. Haplotype frequency differences were analyzed by ², and significant heterogeneity was found among all populations (² = 224.76, P < 0.001). For each of the pairwise comparisons, there was significant heterogeneity (P < 0.001 for each comparison).

View larger version (21K): In this window In a new window Download PPT slide   Figure 3. The CRI RFLP haplotypes and their frequencies in four populations. Only the polymorphic sites are shown. Presence of a restriction site = 1 and absence = 0. For each of the 21 individuals from PV (HOELZEL et al. 1993 ), the 299 bp of CRI sequence data was converted to a single RFLP haplotype. An extra polymorphic restriction site was observed for the RsaI restriction site @ 251–254 in the large-scale screening that was not observed in the smaller sequence sample shown in Figure 2.

For the nDNA loci, the markers analyzed were the presence/absence of the Mhc-DQA 6-bp deletion, the ALD-A NspI restriction site, and the two ß-globin pentameric microsatellites (Table 4). There were no departures from the Hardy-Weinberg equilibrium (HWE) for genotypes at the Mhc-DQA or ALD-A loci. The ß-globin microsatellite data were not tested for HWE because single-locus genotypes could not be determined. There was no significant heterogeneity among the three populations for alleles at the Mhc-DQA and ALD-A loci. The ² analysis of the two microsatellite loci treated them as a single multiallelic locus in a tetraploid. This approach (UTTER et al. 1992 ), or a similar one (GHARRETT et al. 1987 ; BARTLEY and GALL 1990 ; GALL et al. 1992 ), has been used previously for duplicated loci in salmonid fish (see WAPLES 1988 ; WAPLES and AEBERSOLD 1990 for discussion). This is a conservative approach for detecting population structure because for any one microsatellite allele, the frequency differences between populations could only be the same or greater by partitioning into two loci. The genotype of individuals with fewer than four bands was deduced from the relative band intensitites. Of the 93 individuals assayed, there were 0 single-band, 7 double-band, 31 triple-band, and 55 quadruplex-band genotypes, indicating a high degree of heterozygosity at both loci. That is, 59% (55/93) of individuals were heterozygous at both loci, and a minimum of 92% (55+31/93) were heterozygous for at least one locus. There was significant heterogeneity of allele frequencies for the microsatellite loci among the three populations. Pairwise comparisons showed significant differences in microsatellite allele frequency between MQ and the other two populations, but not between HD and SG.

Gene flow: For the 60 CRI sequences, the spatial variance of gene frequency (K_ST), the number of female migrants exchanged between populations per generation (N_fm_f), and the proportion of female migrants exchanged per generation (m_f) were estimated from the uncorrected and Tamura-Nei corrected distances (Table 5). To estimate m_f from N_fm_f, N_f was assumed to be the arithmetic mean of the N_f estimates from the two populations (CHAKRABORTY and NEI 1974 ). There was little difference in the estimates between the uncorrected and Tamura-Nei corrected distances, because the relative levels of diversity and divergence are similar despite a large difference in the absolute amounts of estimated variation. There was limited maternal gene flow between all populations, except between South Georgia and Heard Island. For the latter, maternal gene flow was roughly an order of magnitude greater than among other populations. For the nDNA data, the level of sequence variation and the number of loci assayed were too low to use the sequence-based approach. However, it was possible to obtain an estimate of total gene flow between Heard and Macquarie Islands using allele frequency data from three polymorphic allozyme loci, Acp-1, Ada-1, and Pgm-1 (GALES et al. 1989 ), and the two polymorphic nDNA loci from this study (Mhc-DQA, ALD-A). This resulted in an estimate of Nm between Heard and Macquarie Islands of 4.11.

Time since divergence: An alternative approach is to assume that there has been no gene flow between populations and that the level of nucleotide divergence between populations reflects only time since population divergence. The time of divergence between southern elephant seal populations and between the southern and northern elephant seal populations was estimated from CRI data using the Tamura-Nei net divergences and the estimated mutation rate of 75 x 10^-9 (Table 6). In this historical association model, MQ and PV last shared a common ancestral population some 600,000 yr ago, SG and HD separated as recently as 20,000 yr ago, whereas all other population divergences occurred some 200,000–300,000 yr ago. The northern and southern elephant seals were estimated to have diverged 800,000 yr ago.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In this study, we assessed the contribution of mutation rate, effective population size, gene flow, and selection to the amount and distribution of genetic variation in some regions of mtDNA and nDNA of the southern elephant seal.

Selection:
Two approaches were used to infer that selection had not played a significant role in the level or distribution of genetic variation in the gene regions analyzed. First, selection can be detected from patterns of DNA sequence variation using several recently developed methods (KREITMAN and AGUADE 1986 ; TAJIMA 1989 ; FU and LI 1993 ), although the power of at least some of these tests is limited, given the sample size in this study (BRAVERMAN et al. 1995 ; SIMONSEN et al. 1995 ). For all these methods, the distributions of the mtDNA and nDNA variants in the southern elephant seal (with the exception of the ß-globin microsatellites that were not tested) were compatible with neutrality. Second, it is reasonable to assume that selection is not a factor if there are concordant geographic patterns among several unlinked loci. Although not as informative, the available data from nDNA loci are concordant with the mtDNA data. Three allozyme loci (Ada, Pgam, Pgm) showed significant differences in allele frequency between HD and MQ (GALES et al. 1989 ) as did the ß-globin microsatellites, and there are nonsignificant indications from the latter of differences between HD and SG.

Mutation rate, genetic diversity, and effective population size:
The rate of silent substitution in pinniped nDNA was previously estimated as 1.23 ± 0.24 substitutions per site per 10⁹ yr (SLADE et al. 1994 ). This was based on three different divergences and independent fossil-record dates of 4.5, 14.5, and 23 mya, each of which gave very similar rate estimates of 1.37, 0.93, and 1.40 x 10^-9, respectively. As discussed previously (SLADE et al. 1994 ), this is at the low end of rates among mammals, that is, 6.5 x 10^-9 in rodents, 3 x 10^-9 in primates and artiodactyls, and 1 x 10^-9 in humans (LI et al. 1987 ). A slower rate of nDNA evolution in carnivores compared with other mammals has also been indicated by immunological distances (SARICH 1985 ). However, these estimates should be treated with caution because they are based on the fossil record, and it is clear that reinterpretation of parts of the mammalian fossil record is necessary (WILSON et al. 1987 ; EASTEAL 1990 ).

The rate of mtDNA evolution was estimated for 299 bp of CRI and was calibrated against only the 4.5 mya fossil-record date. Other calibrations were not possible because of alignment difficulties between more divergent sequences. The rate of evolution in CRI was estimated as 75 ± 46 x 10^-9. It is clear that the rate is at least an order of magnitude higher than that for nDNA, and it may be close to two orders of magnitude higher. In terms of percent change, the nDNA is evolving at roughly 0.1%/million years (myr) along a lineage, and the mtDNA CRI is evolving at a rate of 5–10%/myr. The rate of substitution averaged over the entire mtDNA molecule in other animals was estimated as roughly 1%/myr along a lineage (WILSON et al. 1985 ), and the silent rate for the ND4 and ND5 mtDNA genes in primates was estimated to be ~5%/myr (BROWN et al. 1982 ). Estimates of the rate of control region evolution include 103 x 10^-9 and 74 x 10^-9 for CRI and CRII, respectively, in great apes (HORAI et al. 1995 ), 75 x 10^-9 for the entire control region in great apes (TAMURA and NEI 1993 ), 114 x 10^-9 for the entire control region in humans (STONEKING et al. 1992 ), and 20–40 x 10^-9 for CRI in the horse (ISHIDA et al. 1995 ). Therefore, the estimate presented here of 75 x 10^-9 for CRI in the southern elephant seal is consistent with estimates for other mammals.

The uncorrected level of nucleotide diversity in the 299 bp of CRI was 2.86 ± 0.49%. This is among the higher levels so far observed among vertebrate species (see Table 2 in SLADE 1998 ) and is similar to that observed in control regions I and II in African humans of 2.08% (VIGILANT et al. 1991 ). Consistent with its lower mutation rate, nucleotide diversity in nDNA was much lower at 0.09%. Less comparative data are available for nDNA compared to mtDNA. Diversity in southern elephant seal nDNA is similar to the 0.11% found for humans (LI and SADLER 1991 ) and to the 0.08% found for mice (NACHMAN 1997 ) but lower than that in Drosophila pseudoobscura, D. simulans, D. ananassae, or D. melanogaster (2.2%, 1.4%, 1.0%, and 0.4%, respectively; summarized in AQUADRO 1992 ). The heterozygosity of 3.2% in the protein products of 35 nDNA genes (GALES et al. 1989 ) is similar to the mammalian average of 4.1% (NEVO et al. 1984 ).

Given that the neutral mutation rate and nucleotide diversity in southern elephant seal nDNA is similar to that found in humans, it is not surprising, then, that the total effective population size (N_e) of 4 x 10⁴ is also similiar to that in humans, which is ~10⁴ to 10⁵ depending on the time scale over which the estimates are calculated (TAKAHATA 1993 ). If the worldwide population size of southern elephant seals is 750,000 (MCCANN 1985 ), then the ratio of effective to current population size (N_e/N) would be 0.05. N_e/N ratios are highly variable among species; in a summary of ratios, those of nine species of mammal ranged from 0.08 to 0.76 (FRANKHAM 1994 ). The low ratio for the southern elephant seal would reflect, in part, its highly polygynous breeding system. Such a breeding system would also explain the similarity between N_f and N_e (BIRKY et al. 1983 ).

We estimated the effective breeding ratio as four to five females to one male. This ratio is perhaps lower than expected considering that, in one study, the average number of cows per bull in a harem varied between 28 and 53 (CARRICK et al. 1962A ). However, not only are estimates of copulatory success often overestimated by behavioral observations (e.g., PEMBERTON et al. 1992 ; AMOS et al. 1993 ; LAMBERT et al. 1994 ) but also, although males may sire a large number of offspring in the one or two years they are a beachmaster, their lifetime reproductive success would be much lower than observation during beachmaster years would indicate. A female at South Georgia that survives to breeding age produces, on average, five offspring in her lifetime. [This figure was derived from the female life tables in MCCANN 1985 and from the estimated fecundity rate of 0.391 females born per female per year (HINDELL 1991 ).] Considering this, the above data suggest that the average breeding male would sire 20 offspring in his lifetime, although there would likely be a much larger variance associated with this figure than that for the average female. We should be cautious when interpreting the estimated effective breeding ratio for two reasons. One, there is a large variance associated with each of the mutation rates and diversities that would probably be amplified in the estimated ratio. Second, the ratio may reflect the presence of population structure and sex-biased gene flow, rather than a sex-biased breeding ratio.

Geographic structure:
This is the first global survey of genetic variation in the southern elephant seal. For mtDNA, there was significant geographic structure among the four populations of the southern elephant seal for distribution of sequence variation and distribution of haplotype frequencies. A previous phylogeographic analysis had also shown geographic structure for distribution of lineages (SLADE 1998 ). Most of the geographic structure was because of the divergent MQ and PV populations, the lineages of which formed discrete monophyletic groups. The SG and HD populations shared haplotype lineages, but nonetheless there was significant geographic structure for the distribution of those lineages as well as for the distribution of genetic variation and haplotype frequencies. The limited variation at the nDNA loci allowed only an analysis of allele frequency differences, which, among the populations of SG, HD, and MQ, did not exhibit as strong a geographic structure as the mtDNA data. There were no significant differences in allele frequency markers at the ALD-A and Mhc-DQA loci, whereas for the ß-globin microsatellites a significant difference was found between MQ and the other two populations. The difference in the microsatellite allele frequencies between SG and HD was significant at the 10%, but not the 5%, level and suggests that further analysis of the distribution of mDNA variation between these two populations is warranted, preferably with highly variable markers such as microsatellites. The genetic discreteness of the HD and MQ populations has also been shown in a previous study of allozyme variation among those two populations (GALES et al. 1989 ). It is clear, therefore, that there is strong geographic structure for the female (i.e., mtDNA) component among the four populations, but for the biparental (i.e., nDNA) component this is clear only between MQ and both HD or SG. The genetic relationship between populations does not correlate with the relationship determined by growth patterns, thus supporting the hypothesis that the growth pattern differences are environmentally determined (BRYDEN 1968 ). The degree of genetic structure also did not support the hypothesis that the demographic changes within populations were because of differences in movement of individuals between populations (SLADE 1998 ).

Difference between mtDNA and nDNA: Differences between mtDNA and nDNA in the level of observed geographic structure may reflect either differences in rates of gene flow between males and females, or differences in characteristics of the genes, such as mutation rate and/or rate of genetic drift. If males and females have an equal probability of migrating, then Nm is expected to be twice that of N_fm_f. The observed result showing Nm between HD and MQ to be five to eight times greater than N_fm_f suggests that the rate of male gene flow may be up to four times greater than female gene flow. The result must be viewed cautiously, because the estimate of N_fm_f was derived from one locus, and because the estimate of Nm was derived from only one pairwise population comparison. Nonetheless, male-biased gene flow is consistent with ecological studies on dispersal of male and female southern elephant seals (SLADE 1998 ) and with sex-biased dispersal during the pelagic feeding phase of the annual cycle in which males generally travel farther than females to reach their respective feeding grounds (HINDELL et al. 1991 ).

However, we suggest that most of the difference between mtDNA and nDNA in the observed level of geographic structure is accounted for by differences in the characteristics of the genes. The greater sensitivity of mtDNA compared with nDNA is usually explained by two factors: its smaller effective population size and its higher mutation rate (BIRKY et al. 1983 ; WILSON et al. 1985 ). For the southern elephant seal, N_f is roughly equal to N_e, and so the rate of drift of mtDNA is expected to be roughly twice that of the nDNA, thereby contributing to the increased resolution of the mtDNA variation. However, most of the increased sensitivity for detecting subdivision from mtDNA point mutations compared with nDNA point mutations would come from the 50–100 times greater mutation rate. The higher mutation rate would decrease the frequency of shared alleles when one population separates into two and would increase the chance that isolated populations contain unique alleles that arose within each population after separation (SLADE et al. 1994 ). That differences in mutation rate account for a significant part of the difference in detecting subdivision is supported by the results from the highly variable microsatellites, which alone among the nDNA loci in this study showed evidence of geographic structure. Microsatellite loci have mutation rates of the order of 10^-2–10^-4 per locus per generation (WEBER and WONG 1993 ). This is similar to the mutation rate per locus per generation for CRI, which is 1.8 x 10^-4 (i.e., 75 x 10^-9 x 299 bp x 8 yr). Conversely, the per locus mutation rate of the other 250–450-bp nDNA gene fragments is 2.5–4.5 x 10^-6, some two orders of magnitude lower.

Gene flow vs. historical association: Geographic structure reflects some combination of contemporary levels of gene flow and recent historical association. Considering the former, the dispersal capacity of adult males and females is such that movement between oceanic regions is certainly possible. The longest interisland movement recorded was some 3000 km between Heard and Marion Islands in the south Indian ocean (CARRICK et al. 1962B ), and males and females regularly make long-distance round trips of up to 6000 km from breeding sites to feeding areas (HINDELL et al. 1991 ; MCCONNELL and FEDAL 1996). However, because only very small amounts of gene flow are required to generate panmixia (WRIGHT 1931 ), it is reasonable to suggest that geographic structure in the southern elephant seal, especially with respect to females, is not dominated by any substantial amounts of contemporary gene flow. This suggestion is supported by the observation that mtDNA gene flow (or genetic distance) between populations does not correspond very well with geographic distance (Table 5), particularly when comparing SG and PV (N_fm_f = 0.53, distance = 2400 km) with SG and HD (N_fm_f = 7.74, distance = 8000 km), and that resight records of ~20,000 individuals tagged mostly in the south Indian and south Pacific ocean regions between 1950 and 1980 showed no movement between regions but some movement, mostly by immatures, within regions (SLADE 1998 ).

If historical association, rather than contemporary gene flow, dominates the observed geographic structure of the southern elephant seal, then how do we explain the close genetic relationship between HD and SG, despite their being as geographically separate as HD and MQ? In the historical model, SG and HD separated as recently as 20,000 yr ago, and all other population divergences occurred over 200,000 yr ago. We suggest that the estimated relatively recent separation time between SG and HD is linked with the last ice age 18,000 yr ago (CLIMAP 1976). It is likely that the distribution of southern elephant seals during that period would have been different from the current distribution, because of the unsuitability of current breeding beaches that were iced over at that time (R. W. SLADE, unpublished results). Rather than being distributed on sub-Antarctic islands, the southern elephant seal was probably distributed primarily on the southern edge of the continental land masses of Australia, Africa, and South America, and we suggest that at that time HD and SG were a single breeding population on the coast of South Africa. Records of southern elephant seals in South Africa indicate that it is currently used as a haul-out for immature seals to moult, although several births have occurred there also (OOSTHUIZEN et al. 1988 ). It is known that Australia was a former breeding colony (PEMBERTON and SKIRA 1989 ), and PV may be the remnant of the population breeding on the coast of South America.

	FOOTNOTES

¹ Present address: Queensland Institute of Medical Research, Australia.
² Present address: Biological Sciences, University of Durham, England.

	ACKNOWLEDGMENTS

Thanks to HAMISH MCCALLUM, YUKI TAKAHATA, and MONTY SLATKIN for advice; TOM ARNBOM and IAN BOYD for sample collection at South Georgia; CHRIS SCHNEIDER, DICK HUDSON, and two reviewers for comments on the manuscript; and especially ANITA HEIDEMAN for assistance in the lab. This work was funded principally by grant no. 66 from The Antarctic Science Advisory Committee of Australia. Logistic and financial support was also provided by The Australian Antarctic Division.

Manuscript received August 19, 1997; Accepted for publication May 11, 1998.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

AMOS, W., S. TWISS, P. P. POMEROY, and S. S. ANDERSON, 1993  Male mating success and paternity in the grey seal, Halichoerus grypus: a study using DNA fingerprinting. Proc. R. Soc. Lond. Ser B 252:199-207[Medline].

AQUADRO, C. R., 1992  Why is the genome variable? Insights from Drosophila. Trends Genet. 8:355-362[Medline].

BARTLEY, D. M. and G. A. E. GALL, 1990  Genetic structure and gene flow in chinook salmon populations of California. Trans. Am. Fish. Soc. 119:55-71.

BEGON, M., J. L. HARPER and C. R. TOWNSEND, 1986 Ecology. Blackwell Scientific Publications, Oxford.

BIRKY, C. W., JR., T. MARUYAMA, and P. FUERST, 1983  An approach to population and evolutionary genetic theory for genes in mitochondria and chloroplasts, and some results. Genetics 103:513-527[Abstract/Free Full Text].

BLISS, C. I. and R. A. FISHER, 1953  Fitting the negative binomial distribution to biological data and note on the efficient fitting of the negative binomial. Biometrics 9:176-200.

BRAVERMAN, J. M., R. R. HUDSON, N. L. KAPLAN, C. H. LANGLEY, and W. STEPHAN, 1995  The hitchhiking effect on the site frequency spectrum of DNA polymorphisms. Genetics 140:783-796[Abstract].

BROWN, W. M., E. M. PRAGER, A. WANG, and A. C. WILSON, 1982  Mitochondrial DNA sequences in primates: tempo and mode of evolution. J. Mol. Evol. 18:225-239[Medline].

BRYDEN, M. M., 1968  Control of growth in two populations of elephant seals. Nature 17:1106-1108.

CAMPAGNA, C. and M. LEWIS, 1992  Growth and distribution of a southern elephant seal colony. Marine Mam. Sci. 8:387-396.

CARRICK, R., S. E. CSORDAS, and S. E. INGHAM, 1962a  Studies on the southern elephant seal, Mirounga leonina (L.). IV. Breeding and development. C.S.I.R.O. Wildlife Res. 7:161-197.

CARRICK, R., S. E. CSORDAS, S. E. INGHAM, and K. KEITH, 1962b  Studies on the southern elephant seal, Mirounga leonina (L.). III. The annual cycle in relation to age and sex. C.S.I.R.O. Wildlife Res. 7:119-160.

CHAKRABORTY, R. and M. NEI, 1974  Dynamics of gene differentiation between incompletely isolated populations of unequal sizes. Theor. Pop. Biol. 5:460-469[Medline].

CLARK, A. G., 1990  Inferences of haplotypes from PCR-amplified samples of diploid populations. Mol. Biol. Evol. 7:111-122[Abstract].

CLIMAP PROJECT MEMBERS,, 1976  The surface of the ice-age earth. Science 191:1131-1137[Abstract/Free Full Text].

CRAWLEY, M. J., 1993 Glim for Ecologists. Blackwell, Oxford.

CROW, J. F. and K. AOKI, 1984  Group selection for a polygenic behavioral trait: estimating the degree of population subdivision. Proc. Natl. Acad. Sci. USA 81:6073-6077[Abstract/Free Full Text].

DON, R. H., P. T. COX, B. J. WAINWRIGHT, K. BAKER, and J. S. MATTICK, 1991  Touchdown' PCR to circumvent spurious priming during gene amplification. Nucleic Acids Res. 19:4008[Free Full Text].

EASTEAL, S., 1990  The pattern of mammalian evolution and the relative rate of molecular evolution. Genetics 124:165-173[Abstract].

EXCOFFIER, L., P. E. SMOUSE, and J. M. QUATTRO, 1992  Analysis of molecular variance inferred from metric distances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131:479-491[Abstract].

FELSENSTEIN, J., 1993 PHYLIP (Phylogeny Inference Package) version 3.5c. Distributed by the author. Department of Genetics, University of Washington, Seattle.

FRANKHAM, R., 1994  Conservation of genetic diversity for animal improvement. Proc. 5th World Congr. Genetics Appl. Livestk. Prod. 21:385-392.

FU, Y. X. and W. H. LI, 1993  Statistical tests of neutrality of mutations. Genetics 133:693-709[Abstract].

GALES, N. J., M. ADAMS, and H. R. BURTON, 1989  Genetic relatedness of two populations of the southern elephant seal, Mirounga leonina.. Mar. Mamm. Sci. 5:57-67.

GALL, G. A. E., D. BARTLEY, B. BENTLEY, J. BRODZIAK, and R. GOMULKIEWICZ et al., 1992  Geographic variation in population genetic structure of chinook salmon from California and Oregon. Fish. Bull. (U.S.) 90:77-100.

GHARRETT, A. J., S. M. SHIRLEY, and G. R. TROMBLE, 1987  Genetic relationships among populations of Alaskan chinook salmon (Oncorhynchus tshawytscha). Can. J. Fish. Aquat. Sci. 44:765-774.

HINDELL, M. A., 1991  Some life-history parameters of a declining population of southern elephant seals, Mirounga leonina.. J. Anim. Ecol. 60:119-134.

HINDELL, M. A. and H. R. BURTON, 1987  Past and present status of the southern elephant seal (Mirounga leonina) at Macquarie Island. J. Zool. Lond. 213:365-380.

HINDELL, M. A., H. R. BURTON, and D. J. SLIP, 1991  Foraging areas of southern elephant seals, Mirounga leonina, as inferred from water temperature data. Austral. J. Mar. Freshwater Res. 42:115-128.

HOELZEL, A. R., J. HALLEY, S. J. O'BRIEN, C. CAMPAGNA, and T. ARNBOM et al., 1993  Elephant seal genetic variation and the use of simulation models to investigate historical population bottlenecks. J. Hered. 84:443-449[Abstract/Free Full Text].

HORAI, S., K. HAYASAKA, R. KONDO, K. TSUGANE, and N. TAKAHATA, 1995  Recent African origin of modern humans revealed by complete sequences of hominoid mitochondrial DNAs. Proc. Natl. Acad. Sci. USA 92:532-536[Abstract/Free Full Text].

HUDSON, R. R., M. KREITMAN, and M. AGUADÉ, 1987  A test of neutral molecular evolution based on nucleotide data. Genetics 116:153-159[Abstract/Free Full Text].

HUDSON, R. R., M. SLATKIN, and W. P. MADDISON, 1992  Estimation of levels of gene flow from DNA sequence data. Genetics 132:583-589[Abstract].

ISHIDA, N., T. OYUNSUREN, S. MASHIMA, H. MUKOYAMA, and N. SAITOU, 1995  Mitochondrial DNA sequences of various species of the genus Equus with special reference to the phylogenetic relationship between Przewalskii's wild horse and domestic horse. J. Mol. Evol. 41:180-188[Medline].

KOCHER, T. D., and A. C. WILSON, 1991 Sequence evolution of mitochondrial DNA in humans and chimpanzees: control region and a protein-coding region, pp. 391–413 in Evolution of Life—Fossils, Molecules and Culture, edited by S. OSAWA and T. HONJO. Springer-Verlag, Berlin.

KREITMAN, M. E. and M. AGUADÉ, 1986  Excess polymorphism at the Adh locus in Drosophila melanogaster.. Genetics 114:93-110[Abstract/Free Full Text].

KUMAR, S., K. TAMURA and M. NEI, 1993 MEGA: molecular evolutionary genetics analysis, version 1.0. The Pennsylvania State University, University Park, PA.

LAMBERT, D. M., C. D. MILLAR, K. JACK, S. ANDERSON, and J. L. CRAIG, 1994  Single- and multilocus DNA fingerprinting of communally breeding pukeko: do copulations or dominance ensure reproductive success? Proc. Natl. Acad. Sci. USA 91:9641-9645[Abstract/Free Full Text].

LESSA, E. and G. APPLEBAUM, 1993  Screening techniques for detecting allelic variation in DNA sequences. Mol. Ecol. 2:121-129.

LI, W. H. and L. A. SADLER, 1991  Low nucleotide diversity in man. Genetics 129:513-523[Abstract].

LI, W. H., M. C. TANIMURA, and P. M. SHARP, 1987  An evaluation of the molecular clock hypothesis using mammalian DNA sequences. J. Mol. Evol. 25:330-342[Medline].

LING, J. K., and M. M. BRYDEN, 1992 Mirounga leonina, No. 391, in Mammalian Species, edited by C. S. HOOD. Am. Soc. Mammalogists.

LYDEKKER, R., 1909 On the skull-characters in the southern sea-elephant. Proc. Zool. Soc. Lond. 1909: 600–606.

MCCANN, T. S., 1985 Size, status and demography of southern elephant seals (Mirounga leonina) populations, pp. 1–17 in Studies of Sea Mammals in South Latitudes, edited by J. K. LING and M. M. BRYDEN. South Australian Museum, Adelaide.

MCCANN, T. S. and P. ROTHERY, 1988  Population size and status of the southern elephant seal (Mirounga leonina) at South Georgia, 1951–1985. Polar Biol. 8:305-309.

MCCONNELL, B. J. and M. A. FEDAK, 1996  Movements of southern elephant seals. Can. J. Zool. 74:1485-1496.

MCELROY, D., P. MORAN, E. BERMINGHAM, and I. KORNFIELD, 1992  REAP: an integrated environment for the manipulation and phylogenetic analysis of restriction data. J. Hered. 83:157-158[Free Full Text].

MOORE