The authors have declared that no competing interests exist.
Conceived and designed the experiments: IVE PFV. Performed the experiments: IVE CMM. Analyzed the data: IVE CMM FMF NG. Contributed reagents/materials/analysis tools: PFV JWS. Wrote the paper: IVE PFV GD.
Historical climate changes and orogenesis are two important factors that have shaped intraspecific biodiversity patterns worldwide. Although southern South America has experienced such complex events, there is a paucity of studies examining the effects on intraspecific diversification in this part of the world.
The landscape of southern South America has been shaped by several climatic and geological processes, of which two are especially important because of the magnitude of their evolutionary consequences: orogenic changes associated to the uplift of Andes
The Pleistocene epoch was characterized by several global glacial cycles that deeply impacted polar and temperate regions. The well-studied Last Glacial Maximum (LGM), which occurred approximately 23,000–18,000 years ago (ya)
Triangles and circles show the Northern and Southern clades, respectively. Symbols sizes are proportional to sample sizes, and dashed line is the estimated limit of the ice shield during LGM according to Heusser
To date there is no strong evidence for intraglacial refugia throughout the Andes for heterothermic species. During the LGM, direct and periglacial effects may have been stronger for poikilothermic than for homeothermic species because lower temperatures would exclude the former from broad areas
Previous studies on some
The objective of this study was therefore to assess the phylogeographic structure of
We analyzed 196 specimens of
Of the 196 mitochondrial sequences analyzed, 59 were retrieved from Genbank (EU649356–EU649413, AY367791) and the other 137 were generated in this study. Total genomic DNA was extracted from muscle tissue preserved in ethanol using the Wizard SV Genomic DNA Purification kit (Promega). A fragment of the mitochondrial cytochrome b (cyt-b) gene was amplified using primers GluDGL (
Sequence alignments were performed with Clustal X
The cyt-b gene tree was constructed with Bayes Phylogenies 1.1
For the mtDNA dataset, population diversity estimates, including nucleotide diversity (
For cyt-b, the spatial structure of the genetic variation was assessed with a Bayesian approach using Geneland
Historical changes in Ne for each of the main clades were evaluated using graphic reconstructions of population sizes through time estimated with a lineage-through-time (LTT, Bayesian skyline plot analysis) plot using BEAST 1.4.7
The magnitude and spatial arrangement of gene flow was assessed with the coalescence approach implemented in Migrate 3.27
We conducted an ecological niche modelling analysis in order to identify potential refugia for
The inputs of all the analyses were deposited at Dryad (
We obtained an alignment of 644 base pairs of the cyt-b gene from 196 sequences gathered from specimens of
Main clades | N | s | h | Hd |
|
Fs | Rgg |
Northern Clade | 44 | 103/644 | 30 | 0.986 | 0.03243 | −3.442* | 0.0013 ** |
Southern Clade | 152 | 161/644 | 100 | 0.987 | 0.04042 | −3,987* | 0.0019 ** |
Total | 196 | 194/644 | 130 | 0.991 | 0.05750 | ne | ne |
N: sample size, s: number of segregating sites, h: haplotype number, ne: non estimated.
Phylogenetic reconstruction of the unique cyt-b sequences recovered two well-supported reciprocally monophyletic clades of
Branch colors as codes in the map of
The southern phylogroup (Clade S) was composed of two strongly differentiated groups (S1 and S2), the smallest of which (S1) was strongly supported statistically and included only haplotypes from a restricted area from the Chilean Coastal range located at 41°30′S. The larger clade (S2) was composed of two subclades (S2A and S2B) having a moderate probabilistic support. Subclade S2A was well supported and included haplotypes from the Western Andes (e.g. Antillanca and Puyehue) from 39°S to 41°S. These localities are within the Chilean distributional range of the subspecies
The AMOVA showed that 55.5% of the total cyt-b variation was explained by differences between the two main clades (
In general, all pairwise p-distance values involving comparisons between Northern and Southern clades are higher than values from comparisons among the Southern subclades. In addition, the among-Northern subclade values are higher than among-Southern subclade values (
Mean ages and 95% highest posterior density of mtDNA phylogroups are shown in
mtDNA phylogroups | Lower HPD | Mean | Upper HPD |
All (N+S) | 0.547 | 0.751 | 0.988 |
N | 0.338 | 0.523 | 0.720 |
N1 | 0.284 | 0.440 | 0.614 |
N1A | 0.015 | 0.072 | 0.147 |
N1B | 0.181 | 0.288 | 0.411 |
N2 | 0.002 | 0.055 | 0.143 |
S | 0.424 | 0.580 | 0.753 |
S1 | 0.065 | 0.187 | 0.342 |
S2 | 0.383 | 0.524 | 0.671 |
S2A | 0.161 | 0.309 | 0.475 |
S2B | 0.352 | 0.475 | 0.608 |
S2B2 | 0.198 | 0.305 | 0.419 |
S2B3 | 0.125 | 0.235 | 0.351 |
S2B4 | 0.226 | 0.333 | 0.446 |
S2B5 | 0.152 | 0.251 | 0.363 |
The cyt-b haplotype networks are highly congruent with the Bayesian genealogy. The Northern and Southern haploclades are recovered as separate networks (
Colors are the same as in
Overall, from the 10 clades revealed in the cyt-b network, four suggested ancestral haplotypes were from the Coastal range, five from localities in the western Chilean Andes, and one from the Eastern Argentinean Andes. Among the six Andean ancestral haplotypes, two were distributed in areas outside of the LGM ice shield boundaries, whereas the other four fell within this border. For detailed distribution of ancestral haplotypes see
The nuclear gene networks (
Grey haplotypes correspond to the outgroups
Analysis of the population genetic structure using the program Geneland revealed the presence of four populations within the mtDNA Clade N (named Northern clusters A–D in the Migrate analyses) and three within Clade S (named Southern clusters A–C in the Migrate analyses). The general distributions of these clusters are detailed in the
Gene flow | (Nm) | |||||
Northern Clade clusters r | N | θ | A | B | C | D |
Coastal-Andean r (A) | 22 | 0.0235 | - | 2.103 | 1.81E-09 | 1.58E-11 |
Andean r (B) | 5 | 0.0069 | 6.46E-13 | - | 6.46E-13 | 6.46E-13 |
Coastal-Andean r (C) | 13 | 0.0124 | 1.092 | 7.51E-10 | - | 1.67E-10 |
Andean r (D) | 4 | 0.0044 | 0.119 | 1.78E-12 | 1.78E-12 | - |
|
|
|
|
|
||
Coastal mainland r (a) | 2 | 0.0026 | - | 0.4915 | ||
Andean r (b) | 20 | 0.0136 | 3.02E-03 | - | ||
|
|
|
|
|
||
Coastal mainland r (a) | 2 | 0.0030 | - | 0.0996 | ||
Andean r (b) | 10 | 0.0093 | 3.26E-07 | - | ||
|
|
|
|
|
|
|
Chiloé- Coastal mainland-Andean r (A) | 120 | 0.1374 | - | 1.403 | 1.167 | |
Coastal mainland-Andean r (B) | 22 | 0.0560 | 4.73E-09 | - | 6.313 | |
Coastal mainland r (C) | 11 | 0.0350 | 1.807 | 1.10E-07 | - | |
|
|
|
|
|
||
Coastal mainland r (a) | 23 | 0.0665 | - | 2.99E-03 | ||
Andean r (b) | 46 | 0.0215 | 16.754 | - | ||
|
|
θ |
|
|
||
Chiloé Island r (a) | 50 | 0.0408 | - | 11.636 | ||
Coastal mainland-Andean r (b) | 69 | 0.0706 | 3.45E-09 | - | ||
|
|
|
|
|
||
Chiloé Island r (a) | 50 | 0.0480 | - | 2.29E-03 | ||
Andean r (b) | 46 | 0.0223 | 15.881 | - | ||
|
|
|
|
|
|
|
West Andes r (a) | 44 | 0.0116 | - | 0.9011 | 0.9593 | |
East Andes r (b) | 10 | 0.0140 | 7.17E-13 | - | 2.72E-12 | |
Coastal mainland r (c) | 25 | 0.0260 | 1.00E-12 | 1.1936 | - |
Θ = 2Neµ. N = number of individuals in each sample. Migration rate values are in Nm. r = receiving population.
The analysis of isolation-by-distance revealed a high and highly significant positive correlation between geographic and genetic distances (
GLS was made by using a 50×50 grid and a raw genetic distance.
Bayesian skyline plots and mismatch analysis for the two major mtDNA clades of
The assessment of gene flow among the groups within each clade revealed a mixed pattern (
Ecological niche modeling predicts several fragmented refugia for
The color scale corresponds to the probability for suitable habitat; red and light blue indicate highest and lowest suitability values respectively. The heavy blue line delimits the maximum extension of ice shield during LGM according to Heusser
The analysis of the mitochondrial and nuclear genetic variation of
Our analyses show a complex phylogeographic history, considerable genetic structure, and signals of complex patterns of gene flow in
The deepest splits between subclades were observed in Clade N; the MRCA of the phylogroup N1A from the northernmost locality in Las Trancas was 72,000 ya (15,000–147,000 ya), but the split from its sister clade N1B was 440,000 ya (284,000–614,000 ya). Similarly, clade N2 from the coastal range in Purén diverged 55,000 ya (2,000–143,000 ya), but its split from sister clade N1 was 523,000 yr (338,000–720,000 ya). These divergence levels suggest strong fragmentation, low connectivity and reductions of Ne, especially in clades N1A and N2. Although the effects of the glacial cycles were less prominent in the North,
In addition, connectivity between Chiloé Island and the continent could have been limited because the exposed continental shelf may not have been colonized by a continuous
For samples from clade S, recent gene flow is inferred from the occurrence of shared haplotypes among several localities, all of which include Southern clade and Andean sites within the maximum limits of the LGM ice sheet. Some haplotypes are shared among nearby localities, including sites 36–40 from northwestern Chiloé Island. Antillanca (site 34) and Puyehue (site 22) share three haplotypes in the Western Andes. However, we found evidence for recent gene flow between localities more than 100 km apart, such as between Río Puelo and Bariloche (sites 27 and 35 respectively), Huerquehue and Los Llolles (11 and 17), and Chaitén and Canal Garrado (28–30). The last example suggests a more recent connection between mainland localities surrounding Chiloé Island than between this island and the continent. Prior evidence for post-LGM dispersal in the glaciated Andes has been found in other species like the freshwater crab
Despite the exhaustive sampling effort over the entire range of
Co-distribution of haplotypes from highly divergent clades suggests older migration between different areas. For example, sites 23 and 24 from the Estaquillas area (close to the mainland coast in southern Chile) included S1 haplotypes which are endemic to that area, and haplotypes from the broadly distributed clade S2B4. Similarly, sites 40 and 41 along the northern margin of Chiloé Island included haplotypes from the same clade S2B4 and from clade S2B5, which includes only haplotypes endemic to the island. If the secondary contact hypothesis is valid, and excluding the explanation of ancestral polymorphism due to incomplete lineage sorting, populations would have dispersed over long distances after an old lineage diversification. Evidence for multiclade presence was obtained previously by Vidal et al. for the localities of Las Cascadas and Los Muermos, close to Estaquillas
Several clades in the South of the mainland predate the LGM and show a restricted distribution, suggesting a “relict population” with low rates of expansion. For example, clades S1 and S2B2, which are 187,000 and 305,000 ya respectively, are restricted to a small region of the southern coast. Clade S2A, which diverged approximately 300,000 ya, has a broader distribution but is restricted to the Western Andes, suggesting that some
Although the levels of spatial structure were deeper for the mitochondrial than for both nuclear genes, the spatial trends of the genealogical relationships were similar among them and not in total conflict. The general mitochondrial pattern was strongly suggested by both nuclear genes, but the southern range showed a mixed distribution of related nuclear haplotypes and incomplete lineage sorting. Nevertheless, both nuclear genes suggest differentiation between the Northern and the Southern groups, thus corroborating the mtDNA gene tree showing the oldest split between lineages at approximately 39°S, and a more recent weaker structure among the southern populations. These kinds of differences between the patterns of nuclear and mitochondrial differentiation are expected given the larger Ne and longer sorting time to reach monophyly for the nuclear loci
Our demographic analyses of
Although research is beginning to characterize LGM refugia in southern South America
Independent evidence corroborates the occurrence of ice free areas during the LGM within the Andes, specifically the Ñuble glacial gap (36°30′S) and the Malalcahuello Valley which might have acted as refugial areas
The estimated ages for most clades suggest that their origins predate the LGM, therefore the most probable explanation for the observed distribution of the Andean clades S2A, S2B3 and S2B4 would be the long-term persistence of several refugia within the limits of the LGM, as suggested by the ENM. An alternative explanation would be that during the LGM, populations were driven into the Longitudinal Valley and later recolonized the Andean Cordillera and coastal areas. However, the latter scenario predicts less geographic structure within each clade. Some clades, such as S2B3 and S2B4, include haplotypes from both cordilleras (Coastal and Andes), but haplotypes from each cordillera tend to form reciprocally monophyletic groups within each of these clades. Additional evidence for our interpretation of persistent refugia within the LGM ice sheet comes from the geographic provenance of the inferred ancestral haplotypes of some haplogroups. Six of the 10 haplogroups in the cyt-b network were Andean in distribution and occurred in areas within the LGM ice sheet, which suggests that these phylogroups likely originated in these areas prior to the LGM. Additionally, the frequency with which Andean populations acted as sources and Coastal populations as sinks for gene flow is striking, and such patterns can only be explained by the occurrence of multiple Andean refugia during the LGM. Our estimates suggest several cases of gene flow from the Andes to non Andean areas. For clades S2B3 and S2B4, the predominant migration pattern was from the eastern (Argentinean localities in the Andes, within the ice shield boundaries) to the Chilean coastal mainland. According to these results, it is very likely that eastern Andean populations were a source of variants for Pacific coastal sink populations at similar latitudes. The recent review of Sèrsic et al.
In agreement with similar findings from other
Our phylogeographic patterns do not corroborate the current taxonomic arrangement within
(DOC)
(DOC)
(DOC)
(DOC)
(TIF)
Tania Coronado, Paulo Donoso, Andrea Iturría, Alfonso Jara, Amado Perez-Leiva, María José Sandoval, Gisela Stotz, Juan Carlos Ortiz and Marcela Vidal provided essential assistance during field work. Specimens were collected under authorizations SAG 1898 and SAG 4729 (Servicio Agrícola y Ganadero).