Shifting barriers and phenotypic diversification by hybridisation

Abstract

The establishment of hybrid taxa relies on reproductive isolation from the parental forms, typically achieved by ecological differentiation. Here, we present an alternative mechanism, in which shifts in the strength and location of dispersal barriers facilitate diversification by hybridisation. Our case study concerns the highly diverse, stenotopic rock-dwelling cichlids of the African Great Lakes, many of which display geographic colour pattern variation. The littoral habitat of these fish has repeatedly been restructured in the course of ancient lake level fluctuations. Genetic data and an experimental cross support the hybrid origin of a distinct yellow-coloured variant of Tropheus moorii from ancient admixture between two allopatric, red and bluish variants. Deficient assortative mating preferences imply that reproductive isolation continues to be contingent on geographic separation. Linking paleolimnological data with the establishment of the hybrid variant, we sketch a selectively neutral diversification process governed solely by rearrangements of dispersal barriers.

Keywords: Carotenoid coloration; Cichlidae; Lake Tanganyika; Tropheus; climate change; colour pattern; environmental fluctuations; genetic admixture; hybrid speciation; mate choice.

Figure 1

Phenotypic diversification by hybridization in the shifting barriers model. Initially, two phenotypically distinct, ecologically equivalent taxa are separated by a dispersal barrier (1). In the following, rearrangements of dispersal barriers may be associated with changes in the spatial distribution of taxa, such as shifts along mountain or lake slopes, driven by environmental fluctuations. In stage (2), a rearrangement of dispersal barriers creates a spatially confined zone of secondary contact, in which hybridization gives rise to a novel phenotype (3). The new dispersal barriers provide for isolation between hybrid and parent taxa, allowing each of them to retain their original ecological niche within their respective distribution ranges. Following stage 3, further developments depend on the occurrence, timing and extent of further habitat rearrangements and on the evolution of assortative mating. Possible outcomes of renewed secondary contact between hybrid and parent taxa (4) include sympatric coexistence after completed speciation (5a), further phenotypic diversification by hybridization (5b) or the complete assimilation of one taxon if the admixture zone extends across this taxon’s full range (not depicted). For simplicity, the graph does not include minor barriers within each taxon’s range, which are expected to create additional structure. The linear distribution of taxa, as drawn in this example, is not a condition for the application of the model.

Figure 2

Tropheus moorii populations sampled from southern Lake Tanganyika. Along the studied shoreline, rocky sections alternate with sandy beaches. A particularly strong habitat barrier is imposed by the wide sandy estuary of the Lufubu River, which separates the distribution ranges of the bluish and the yellow-morph Tropheus. Note that T. moorii are sexually monomorphic. T. m. ‘Chipimbi’ were included only in carotenoid analysis. T. m. ‘Funda‘ (no photo) are indistinguishable from T. m. ‘Nakaku‘; T. m. ‘Katete’ and T. m. ‘Ndole’ (no photo) resemble T. m. ‘Moliro’ and T. m. ‘Chimba’, respectively. Photographs: Ad Konings, Peter Berger, Wolfgang Gessl.

Figure 3

Genetic admixture in the yellow morph Tropheus and experimental recreation of the hybrid color pattern. (a) AFLP-based assignment probabilities for K= 2 genetic clusters, determined in STRUCTURE. (b) Frequency distribution of the G/T polymorphism (red/blue) at a SNP locus in TMO-4C4. (c) Multidimensional scaling plot of AFLP variation. (d) Resemblance between the genetically admixed Tropheus moorii ‘Ilangi’ and experimental F1 hybrids between bluish- and red-morph T. moorii.

Figure 4

Carotenoid chromatograms. This schematic representation depicts the relative signal intensity per HPLC peak, averaged across samples, against the average retention time of the peak. Standard deviations are indicated by the thin vertical lines extending from each bar. The arrow marks a signal restricted to the ‘Ilangi’ population.

Figure 5

Bayesian inference tree based on mtDNA sequence data.

Figure 6

Female courtship preferences in two-way choice experiments. Relative female courtship (RFC) values indicate the rate of female courtship directed towards own-population males in relation to courtship directed towards the alternative male. Each bar represents one mate choice trial. The results of the intercept-only models show that across trials within an experiment, RFC did not differ significantly from zero.