Community Genetics Reveal Elevated Levels of Sympatric Gene Flow among Morphologically Similar but Not among Morphologically Dissimilar Species of Lake Victoria Cichlid Fish

Abstract

We examined genetic structure among five species of Lake Victoria haplochromine cichlids in four island communities, using a full factorial sampling design that compared genetic differentiation between pairs of species and populations of varying morphological similarity and geographical proximity. We found that allopatric conspecific populations were on average significantly more strongly differentiated than sympatric heterospecific populations of morphologically similar species. Allopatric heterospecific populations of morphologically dissimilar species were most differentiated. Our work demonstrates that phenotypic divergence can be maintained and perhaps even evolve in sympatry despite considerable gene flow between species. Conversely, phenotypic resemblance among conspecific populations can be maintained despite geographical isolation. Additionally we show that anthropogenically increased hybridization does not affect all sympatric species evenly but predominantly affects morphologically similar and closely related species. This has important implications for the evolution of reproductive isolation between species These findings are also consistent with the hypothesis of speciation reversal due to weakening of divergent selection and reproductive isolation as a consequence of habitat homogenization and offers an evolutionary mechanistic explanation for the observation that species poor assemblages in turbid areas of the lake are characterized by just one or two species in each of a few morphologically distinct genera.

Figure 1

Species and sampling sites. (a) Lake Victoria, the Mwanza Gulf, and the four sampling localities, (1) Hippo Island, (2) Python Islands, (3) Luanso Island, and (4) Marumbi Island (Tanzania). (b) A neighbour joining tree for the five species investigated. Allele frequencies were estimated from AFLP data in AFLP-Surv and Reynolds genetic distance was calculated with 100 bootstrap replicates.

Figure 2

Species-abundance composition of the communities of rock-dwelling cichlids at the 4 study islands, from top to bottom: Hippo Island, Python Island, Luanso Island, and Marumbi Island. Black bars are the taxa studied in this paper. Cumulatively, they account for between 71% and 86% of the cichlids in each community.

Figure 3

Spatial structure analysis reveals a positive autocorrelation coefficient outside the 5% and 95% confidence limits (dashed lines) among individuals within about 4 km of one another, but no isolation by distance beyond this.

Figure 4

F _ST equivalent estimates for pairwise population comparisons falling into one of five different groupings: populations of the same species (always allopatric), morphologically similar taxa (M. mbipi and N. greenwoodi and P. pundamilia and P. nyererei) in sympatry and in allopatry, and nonsimilar taxa in sympatry and in allopatry. A Holm Sequential Bonferoni posteriori test was used to obtain P values. *Represents a significant difference against group 1, **represents a significant difference against group 2.

Figure 5

Individual-based and population trees. (a) An AFLP neighbor network based on Nei and Li distances. Samples are sorted and colour-coded by species. Conflicting phylogenetic signal in the center magnified bottom right. (b) A Neighbour joining tree for the 20 populations investigated. Allele frequencies were estimated from AFLP data in AFLP-Surv, and Reynolds genetic distance was calculated with 100 bootstrap replicates. The 4 different shapes indicate the 4 islands. The five different colours indicate the 5 species.