Phylogenomics uncovers early hybridization and adaptive loci shaping the radiation of Lake Tanganyika cichlid fishes

Abstract

Lake Tanganyika is the oldest and phenotypically most diverse of the three East African cichlid fish adaptive radiations. It is also the cradle for the younger parallel haplochromine cichlid radiations in Lakes Malawi and Victoria. Despite its evolutionary significance, the relationships among the main Lake Tanganyika lineages remained unresolved, as did the general timescale of cichlid evolution. Here, we disentangle the deep phylogenetic structure of the Lake Tanganyika radiation using anchored phylogenomics and uncover hybridization at its base, as well as early in the haplochromine radiation. This suggests that hybridization might have facilitated these speciation bursts. Time-calibrated trees support that the radiation of Tanganyika cichlids coincided with lake formation and that Gondwanan vicariance concurred with the earliest splits in the cichlid family tree. Genes linked to key innovations show signals of introgression or positive selection following colonization of lake habitats and species’ dietary adaptations are revealed as major drivers of colour vision evolution. These findings shed light onto the processes shaping the evolution of adaptive radiations.

Adaptive radiation of barbs of the genus Labeobarbus (Cyprinidae) in an East African river

“…125 species sensu Vreven et al., ). Complex genomes that have experienced past hybridisation may contribute to rapid adaptive radiation, as established in cichlids from African Great lakes (Irisarri et al., ; Meier et al., ). Ancient introgressions have also been detected in Arctic charr ( Salvelinus ; Lecaudey et al., ), which are classic examples of sympatric divergence in high latitudes (Alekseyev et al., ; Hindar & Jonsson, ; Knudsen, Klemetsen, Amundsen, & Hermansen, ; Taylor, ).…”

Evolutionary transitions to cooperative societies in fishes revisited

“…Unfortunately, these five nuclear genes are available for only a small number of species, and, in contrast to the mitochondrial data, contain only little phylogenetic information (e.g., Koblmüller et al, 2017). Therefore, Dey et al's (2017) tree topology mainly reflects mitochondrial relationships, contradicting recent phylogenomic evidence (McGee et al, 2016;Takahashi & Sota, 2016;Irisarri et al, 2018) and concordant morphological classifications. This causes an erroneous increase in the number of transitions to cooperative breeding by up to 50%.…”

“…In the tree based on these nuclear markers, some of the phylogenetic relationships among non-ossified lamprologines are not recovered according to current knowledge. For instance, Neolamprologus brichardi, N. pulcher, and N. savoryi, which appear closely related based on nuclear multilocus and genomic data (e.g., Sturmbauer et al, 2010;McGee et al, 2016;Irisarri et al, 2018), as well as by morphology and behavior, do not cluster together in Dey et al (2017). This discrepancy is most probably an artifact resulting from sampling different loci in the different non-ossified group lamprologines (see Suppl.…”

“…The phylogenetic reconstruction of Dey et al (2017) Neolamprologus christyi, N. modestus, N. mondabu, N. petricola, and N. tetracanthus; the genus Telmatochromis; the Neolamprologus pulcher complex; Koblmüller et al, 2007;Sturmbauer et al, 2010;McGee et al, 2016;Irisarri et al, 2018) were not recovered. Dey et al applied topological constraints to amend some of the phylogenetic peculiarities shown by their tree, which we find problematic for reasons explained further below.…”

“…Phylogenetic tree inference typically aims at an objective reconstruction of phylogenetic relationships based on available data, and topological constraints should be employed only for the testing of alternative tree topologies. In Dey et al, however, topological constraints were applied to two clades in order to enforce consistency with published phylogenies (Irisarri et al, 2018;McGee et al, 2016;Sturmbauer et al, 2010). This not only resulted in a tree that is not backed up by the input data, but also affected branching order in other parts of the tree (e.g., compare the constrained tree of Dey et al (2017) to our unconstrained tree in Figure A3).…”