Demographic modelling with whole‐genome data reveals parallel origin of similarPundamiliacichlid species after hybridization

Abstract

Modes and mechanisms of speciation are best studied in young species pairs. In older taxa, it is increasingly difficult to distinguish what happened during speciation from what happened after speciation. Lake Victoria cichlids in the genus Pundamilia encompass a complex of young species and polymorphic populations. One Pundamilia species pair, P. pundamilia and P. nyererei, is particularly well suited to study speciation because sympatric population pairs occur with different levels of phenotypic differentiation and reproductive isolation at different rocky islands within the lake. Genetic distances between allopatric island populations of the same nominal species often exceed those between the sympatric species. It thus remained unresolved whether speciation into P. nyererei and P. pundamilia occurred once, followed by geographical range expansion and interspecific gene flow in local sympatry, or if the species pair arose repeatedly by parallel speciation. Here, we use genomic data and demographic modelling to test these alternative evolutionary scenarios. We demonstrate that gene flow plays a strong role in shaping the observed patterns of genetic similarity, including both gene flow between sympatric species and gene flow between allopatric populations, as well as recent and early gene flow. The best supported model for the origin of P. pundamilia and P. nyererei population pairs at two different islands is one where speciation happened twice, whereby the second speciation event follows shortly after introgression from an allopatric P. nyererei population that arose earlier. Our findings support the hypothesis that very similar species may arise repeatedly, potentially facilitated by introgressed genetic variation.

Divergent parasite infections in sympatric cichlid species in Lake Victoria

“…). To test whether this was repeatable and whether parasitism correlated with crevice‐dwelling in Pundamilia , but also to test for more general repeatability of parasite trends in similar cases of species divergence, we compared infections between sympatric red/blue pairs of Pundamilia at Makobe and three other islands with phenotypically similar species or morph pairs (Seehausen et al ., ; Meier et al ., ). We found that the total parasite abundance was strongly correlated between sympatric Pundamilia spp.…”

“…‘pundamilia‐like’ and P. sp. ‘nyererei‐like’, and at Luanso, it is a single genetically panmictic population with polymorphic male colouration (Meier et al ., ).…”

“…(); Meier et al . ()). Because pairwise F ST values rely on fewer assumptions than phylogenetic comparative methods, in this situation these represent a better choice for asking about how relatedness influences parasite infections among these species.…”

“…We tested these predictions by analysing the relationship between differences in parasite communities of all pairwise comparisons of the 16 sympatric cichlid species from Makobe Island. Additionally, we extended this analysis to a previously studied ‘speciation continuum’ involving replicate cases of sympatric red and blue sister species in the genus Pundamilia from three other islands to ask whether patterns of infections in replicate sister species pairs were repeatable and predictable from their degrees of genetic and habitat differentiation, which vary among islands (Seehausen et al ., ; Meier et al ., ). For testing the Janzen–Connell hypothesis of parasite‐mediated coexistence, we investigated the relationship between cichlid species abundance and parasite diversity and load.…”

Divergent mitochondrial lineages arose within a large, panmictic population of the Savannah sparrow (Passerculus sandwichensis)

“…Given the presence of linked sites in our data set (i.e., mean of 2.8 and median 2 SNPs per 89 bp locus), AIC analyses could be associated with increased error and bias towards more complex models and this analysis should be interpreted with caution (Coffman et al, ). However, other studies apply AIC methods to composite likelihoods as a first approximation in ranking models (e.g., Meier et al, ; Oswald, Overcast, Mauck, Andersen, & Smith, ). In cases where the model with the highest likelihood included more parameters than hierarchically nested models with lower likelihood values, we performed a likelihood ratio test with a GIM‐based adjustment to account for the use of composite likelihoods (Coffman et al, ).…”