Potential Contribution of Ancient Introgression to the Evolution of a Derived Reproductive Strategy in Ricefishes

Abstract

Transitions from no parental care to extensive care are costly and involve major changes in life history, behavior, and morphology. Nevertheless, in Sulawesi ricefishes, pelvic brooding evolved from transfer brooding in two distantly related lineages within the genera Adrianichthys and Oryzias, respectively. Females of pelvic brooding species carry their eggs attached to their belly until the fry hatches. Despite their phylogenetic distance, both pelvic brooding lineages share a set of external morphological traits. A recent study found no direct gene flow between pelvic brooding lineages, suggesting independent evolution of the derived reproductive strategy. Convergent evolution can, however, also rely on repeated sorting of preexisting variation of an admixed ancestral population, especially when subjected to similar external selection pressures. We thus used a multispecies coalescent model and D-statistics to identify gene-tree-species-tree incongruencies, to evaluate the evolution of pelvic brooding with respect to interspecific gene flow not only between pelvic brooding lineages but also between pelvic brooding lineages and other Sulawesi ricefish lineages. We found a general network-like evolution in Sulawesi ricefishes, and as previously reported, we detected no gene flow between the pelvic brooding lineages. Instead, we found hybridization between the ancestor of pelvic brooding Oryzias and the common ancestor of the Oryzias species from the Lake Poso area. We further detected signs of introgression within the confidence interval of a quantitative trait locus associated with pelvic brooding in O. eversi. Our results hint toward a contribution of ancient standing genetic variation to the evolution of pelvic brooding in Oryzias.

Keywords: Sulawesi ricefishes; ancient gene flow; convergent evolution; pelvic brooding; reproductive strategy.

Fig. 1

Ancestral state reconstruction supporting convergent evolution of pelvic brooding in Sulawesi Oryzias (PBO) and Adrianichthys (PBA). Shown is the maximum likelihood tree generated in this study based on 1,907 orthologous single-copy genes. Suggested node ages for the pelvic brooding lineages are taken from Mokodongan and Yamahira (2015). Black dots below nodes indicate maximal bootstrap support (BS = 100). One node (gray dot), supporting the sister group relationship of the PBO and O. dopingdopingensis lineage with the Malili Lakes species, has a bootstrap support of BS = 78. The distribution centers of the Central Sulawesi ricefishes included in the tree are indicated on the map in the upper left corner; symbols on the map correspond to those indicated in the phylogeny. In the lower right corner, brooding females of a representative of both pelvic brooding ricefish lineages PBO (O. eversi) and PBA (A. oophorus) are shown. Map of Sulawesi including water shed information was obtained from open-source.

Fig. 2

Results of multispecies-coalescence approach to investigate gene-tree–species-tree incongruencies. (A) Astral species tree with visualizations of gene tree discordances based on Discovista. Branch quartet frequencies are coded at each node with different size and color of the respective node number: numbers in black depict quartets in which >80% of all gene trees follow the shown topology. Numbers in turquoise (branch 2,5,6 and 13) indicate that quartets following the shown topology are present in 50–79% of all gene trees, and red numbers (branch 3, 4 and 7) indicate splits where the shown topology of the quartet was found in <50% of all gene trees. The pie chart in the upper left corner shows the frequency of each well (black, >80%), moderately (turquoise, 50–79%), and poorly (red, <50%) supported quartet in the Astral species tree. (B) Densitree based on 1,907 single gene trees generated by IQtree. The thin blue line shows the consensus phylogeny (species tree). Blurry areas indicate gene-tree–species-tree incongruencies. (C) Quartets depicted in the Astral species tree (t1) and alternative quartets (t2 and t3) of splits with poorly supported quartet scores (<50%). All quartet scores and alternative quartets for other splits are shown in detail in supplementary figure S2, Supplementary Material online.

Fig. 3

Signatures of introgression between Central Sulawesi ricefishes. (A) F-branch [fb(C)] statistics across the Central Sulawesi ricefishes. The excess of allele-sharing is shown between tips of the tree (horizontally arranged along the x axis) and each other tips (solid lines) and nodes (dashed lines) in the phylogenetic tree (vertically arranged on the y axis). The tree that was used as a basis for the branch statistic was a Neighbor-Joining tree generated based on the SNP dataset (DS2). The redness of each cell in the matrix indicates the degree of excess allele sharing between each tree tip (C) and each branch (b). Excess of allele-sharing was significant when the z-score was >4 (equivalent to Bonferroni multiple-testing corrected P-value of 0.001), which is indicated by small black squares. Gray cells in the matrix correspond to tests that are not consistent with the phylogeny. Colors in the tree correspond to phylogenetic lineages as in figure 1. (B) Phylogenetic network revealed by SNaQ including representative of all ricefish lineages, indicates gene flow between ancestor of Lake Poso Oryzias and ancestor of PBO. Another hybrid edge is indicated between O. dancena and O. celebensis and Sulawesi lineages, though the proportion consisting of introgressed O. dancena genome was low, with 0.5%. The network with two hybrid edges was chosen due to its low network score, which did not improve with more hybrid edges. Networks with three and four hybrid edges are shown in supplementary figure S4, Supplementary Material online. (C) Region on the O. celebensis reference genome OCchr1_17_19 showing high introgression signal (top 1% of f_d in color) and overlapping with one (QCon1) of the three QTL windows correlating with the extent of the ventral concavity in O. eversi (Montenegro et al. 2022). The asterisk marks the location of the QTL within the QTL confidence interval (Montenegro et al. 2022).