Contemporary habitat discontinuity and historic glacial ice drive genetic divergence in Chilean kelp

Abstract

Background: South America's western coastline, extending in a near-straight line across some 35 latitudinal degrees, presents an elegant setting for assessing both contemporary and historic influences on cladogenesis in the marine environment. Southern bull-kelp (Durvillaea antarctica) has a broad distribution along much of the Chilean coast. This species represents an ideal model taxon for studies of coastal marine connectivity and of palaeoclimatic effects, as it grows only on exposed rocky coasts and is absent from beaches and ice-affected shores. We expected that, along the central Chilean coast, D. antarctica would show considerable phylogeographic structure as a consequence of the isolating effects of distance and habitat discontinuities. In contrast, we hypothesised that further south--throughout the region affected by the Patagonian Ice Sheet at the Last Glacial Maximum (LGM)--D. antarctica would show relatively little genetic structure, reflecting postglacial recolonisation.

Results: Mitochondrial (COI) and chloroplast (rbcL) DNA analyses of D. antarctica from 24 Chilean localities (164 individuals) revealed two deeply divergent (4.5 - 6.1% for COI, 1.4% for rbcL) clades from the centre and south of the country, with contrasting levels and patterns of genetic structure. Among populations from central Chile (32 degrees-44 degrees S), substantial phylogeographic structure was evident across small spatial scales, and a significant isolation-by-distance effect was observed. Genetic disjunctions in this region appear to correspond to the presence of long beaches. In contrast to the genetic structure found among central Chilean populations, samples from the southern Chilean Patagonian region (49 degrees-56 degrees S) were genetically homogeneous and identical to a haplotype recently found throughout the subantarctic region.

Conclusions: Southern (Patagonian) Chile has been recolonised by D. antarctica relatively recently, probably since the LGM. The inferred trans-oceanic ancestry of these Patagonian populations supports the notion that D. antarctica is capable of long-distance dispersal via rafting. In contrast, further north in central Chile, the correspondence of genetic disjunctions in D. antarctica with long beaches indicates that habitat discontinuity drives genetic isolation among established kelp populations. We conclude that rafting facilitates colonisation of unoccupied shores, but has limited potential to enhance gene-flow among established populations. Broadly, this study demonstrates that some taxa may be considered to have either high or low dispersal potential across different temporal and geographic scales.

Figure 1

Patagonian Ice Sheet at the Last Glacial Maximum. Map of South America showing the extent of the Patagonian Ice Sheet at the LGM (after [27]). Modern oceanographic conditions around South America are also indicated. Dashed lines indicate mean annual sea surface temperatures (after [80]). Arrows show directions of major surface currents, the Antarctic Circumpolar Current (ACC), the Humboldt Current (HC), the Cape Horn Current (CHC) and the Malvinas Current (MC) (after [81]).

Figure 2

Phylogeography of Chilean bull-kelp. A) ML phylogeny of D. antarctica for COI including data from 17 new localities in Chile, as well as published sequences from 7 localities in central Chile (indicated by asterisks), and localities in New Zealand and the subantarctic [19]. Coloured triangles represent haplotypes found in Chile, with colours corresponding to those in panel B and Fig 3. Bayesian PP values are shown in black above the line, and ML bootstraps are in grey below the line. Support values < 50% are not shown, nor are values on some minor, distal branches within the major clades. Outgroups have been trimmed for clarity. B) Location of all sampled localities along the coast of Chile. Pie charts indicate distribution and proportions of haplotypes, with haplotype colours corresponding to those in panel A and Fig 3. Grey numbers to the left of pie charts show total number of samples from each locality.

Figure 3

Haplotype networks and global projection. COI haplotype network diagrams of D. antarctica for A) the 'central Chilean' clade and B) the 'Patagonian/subantarctic' clade (see also [19]), with circle size scaled approximately according to haplotype frequency, and small dots representing undetected, hypothetical haplotypes. Although the 'Patagonian/subantarctic' and 'New Zealand south' clades [19] joined parsimoniously using network analysis, the 'central Chilean' clade did not join to any other lineages at ≥90% confidence limit: a hypothetical, non-statistically supported connection between the major clades is indicated with a red dashed line. The most common (C-I/red) haplotype was found throughout Chilean Patagonia [this study] and the subantarctic [19]. C) Global projection showing locations of other (subantarctic) sites at which the Patagonian (C-I/red) haplotype has previously been detected [19]. The eastward flow of the Antarctic Circumpolar Current (ACC) is indicated.

Figure 4

Isolation-by-distance. Isolation-by-distance (Mantel) analysis, illustrating the relationship between COI genetic distance (Nei's raw average pairwise difference, D) and geographical distance for D. antarctica among central Chilean (32°-42°S) sampling locations.

Figure 5

The effect of beach length on phylogeographic structure in bull-kelp in central Chile. Maximum length of beaches between pairs of adjacent sampled localities in central Chile [data estimated from Google Earth satellite images]. Upper: locality pairs between which there is a marked mitochondrial (COI) genetic disjunction (Nei's raw average pairwise difference, D, > 1.0) are shown in black, whereas genetically-similar locality pairs (D < 1.0) are shown in grey. Lower: Nei's raw average pairwise differences (D) for COI, plotted against maximum uninterrupted beach length, between all adjacent locality pairs in central Chile.