Postglacial recolonization of the Southern Ocean by elephant seals occurred from multiple glacial refugia

The Southern Ocean is warming more rapidly than other parts of our planet. How this region’s endemic biodiversity will respond to such changes can be illuminated by studying past events, through genetic analyses of time-series data sets including historic and fossil remains. Archaeological and subfossil remains show that the southern elephant seal (Mirounga leonina) was common along the coasts of Australia and New Zealand in the recent past. This species is now mostly confined to sub-Antarctic islands and the southern tip of South America. We analysed ancient seal samples from Australia (Tasmania), New Zealand, and the Antarctic mainland to examine how southern elephant seals have responded to a changing climate and anthropogenic pressures during the Holocene. Our analyses show that these seals formed part of a broader Australasian lineage, comprising seals from all sampled locations from the south Pacific sector of the Southern Ocean. Our study demonstrates that southern elephant seal populations have dynamically altered both range and population sizes under climatic and human pressures, over surprisingly short evolutionary timeframes for such a large, long-lived mammal.

Methods
Control Region Sequence Analysis: Partial control region (HVR1) sequences were successfully amplified from modern and ancient samples. A data set with representative sequences from various populations was constructed. Genetic diversity metrics were calculated, and FST values were used to measure genetic differentiation. AMOVAs were performed to infer geographic structure. A time-sensitive haplotype network was constructed.

Mitochondrial Genome Assembly: DNA extraction was performed on selected ancient samples for mitochondrial genome assembly. Libraries were prepared, quantified, and enriched using hybridization capture with bait from a closely related species. PCR amplification and pooling of samples were performed and sequenced using the Illumina MiSeq platform.

Data Processing: Sequence data were processed using specific scripts to trim adapters, merge overlapping read pairs, and filter low-quality reads. The processed reads were mapped to a reference mitochondrial genome using BWA and filtered for quality and duplicates. Consensus sequences were computed, and patterns of post-mortem DNA damage were assessed using MapDamage (SI Appendix).

Phylogenetic Analysis of Mitogenome Sequences: Newly generated mitogenome sequences were aligned with global southern elephant seal sequences.

 We also used Bayesian analysis in BEAST v1.8.2 (Suchard et al. 2018) to compare three models of post-mortem sequence damage: no damage, age-dependent damage, and age-independent damage (Ho et al., 2007; Rambaut et al., 2009; Ho, 2012). Models were compared using marginal likelihoods estimated by stepping-stone sampling (Xie et al., 2010). Posterior distributions of parameters were estimated using Markov Chain Monte Carlo sampling over 75 million steps, with samples drawn every 104 steps.

 To infer a dated phylogenetic tree, we performed a Bayesian molecular dating analysis in BEAST. We used a Skyride coalescent tree prior (for species with a dynamic population history) and a strict molecular clock, selected by comparison of marginal likelihoods calculated by stepping-stone sampling. Estimates of node times were calibrated using the sample dates for the ancient Tasmanian, New Zealand, and Victoria Land Coast sequences. We also fixed the mutation rate of HVR1 based on an estimate from a previous study of the southern elephant seal (de Bruyn et al., 2009). Posterior distributions of parameters were estimated using Markov Chain Monte Carlo sampling over 75 million steps, with samples drawn every 104 steps. We ran the analysis in duplicate to check for convergence and combined the samples after discarding the first 10% as burn-in. Sufficient sampling was confirmed by inspecting the samples in Tracer v1.7.1 (Rambaut et al., 2018).

See uploaded supplementary information to Zenodo entry for full methodologies.