Adaptive divergence in shell morphology in an ongoing gastropod radiation from Lake Malawi

Abstract

Background: Ecological speciation is a prominent mechanism of diversification but in many evolutionary radiations, particularly in invertebrates, it remains unclear whether supposedly critical ecological traits drove or facilitated diversification. As a result, we lack accurate knowledge on the drivers of diversification for most evolutionary radiations along the tree of life. Freshwater mollusks present an enigmatic example: Putatively adaptive radiations are being described in various families, typically from long-lived lakes, whereas other taxa represent celebrated model systems in the study of ecophenotypic plasticity. Here we examine determinants of shell-shape variation in three nominal species of an ongoing ampullariid radiation in the Malawi Basin (Lanistes nyassanus, L. solidus and Lanistes sp. (ovum-like)) with a common garden experiment and semi-landmark morphometrics.

Results: We found significant differences in survival and fecundity among these species in contrasting habitats. Morphological differences observed in the wild persisted in our experiments for L. nyassanus versus L. solidus and L. sp. (ovum-like), but differences between L. solidus and L. sp. (ovum-like) disappeared and re-emerged in the F₁ and F₂ generations, respectively. These results indicate that plasticity occurred, but that it is not solely responsible for the observed differences. Our experiments provide the first unambiguous evidence for genetic divergence in shell morphology in an ongoing freshwater gastropod radiation in association with marked fitness differences among species under controlled habitat conditions.

Conclusions: Our results indicate that differences in shell morphology among Lanistes species occupying different habitats have an adaptive value. These results also facilitate an accurate reinterpretation of morphological variation in fossil Lanistes radiations, and thus macroevolutionary dynamics. Finally, our work testifies that the shells of freshwater gastropods may retain signatures of adaptation at low taxonomic levels, beyond representing an evolutionary novelty responsible for much of the diversity and disparity in mollusks altogether.

Keywords: Adaptive radiation; Ampullariidae; Common garden experiment; Differential fitness; Geometric morphometrics; Local adaptation; Phenotypic plasticity.

Fig. 1

Comparison of nominal species and molecular groups in Lanistes from the southern Malawi Basin. A) Lanistes sp. (ovum-like); B) L. solidus; C) L. nyassanus. One molecular group (group A) contains exclusively specimens of L. solidus and L. nyassanus, whereas the other (group B) consists predominantly of L. sp. (ovum-like) and L. solidus. Blue spheres indicate the number of specimens belonging to a certain nominal species and molecular group, whereas colored connections link specimens that come from the same locality but occupy different spheres, i.e. green for group A, orange for group B. Modified from [31]

Fig. 2

Fecundity of morphospecies during the common garden experiment. Fecundity represents the number of offspring (mean ± standard error) generated by morphospecies averaged over the F₁ and F₂ generations. Statistically significant differences are indicated with asterisks (0.05 > * > 0.01 > ** > 0.001)

Fig. 3

Pairwise comparisons of centroid size by morphospecies for parents (wild-caught and F₁) and for offspring (F₁ and F₂) in our common garden experiment. For parents the centroid size reflects differences in adult body size among morphospecies whereas for offspring it reflects growth rate (all offspring individuals were photographed at an age of 6 months; see material and methods). Black error bars indicate the mean ± standard error, whereas the spread of the data is indicated by the grey bars (mean ± standard deviation). Statistics result from Wilcoxon rank-sum tests with Bonferroni correction. Statistically significant differences are indicated with asterisks (0.05 > * > 0.01 > ** > 0.001 > ***)

Fig. 4

Morphospace occupation for all specimens in our common garden experiment. a wild-caught parents; b F₁ offspring; c F₁ parents; d F₂ offspring. Colors indicate morphospecies (blue = L. nyassanus, red = L. solidus, black = L. sp. (ovum-like)), whereas symbols (circles, triangles, crosses) indicate replicates. The biplot is represented at 75% of its actual size and indicates the contribution of morphometric components to the morphospace. Grey spheres and solid separation lines indicate the best-supported solutions of model-based clustering with Gaussian mixture models (see Fig. 6); for F₂ offspring (d) the three-group model is added with dashed lines

Fig. 5

Morphospace changes in replicates for the F₁ and F₂ generations. Morphospace changes (indicated with arrows) are reconstructed from the morphospace occupation of populations (mean ± one standard deviation). All arrows for replicates point in a similar direction and are quasi-parallel, except for the slightly different trajectory of L. solidus rep. 1 in comparison to the other replicates in the F₂ generation. Replicates thus show overall very similar changes, indicating that the design of our experiment was robust

Fig. 6

Model support for the various normal mixture models in function of the number of clusters (1–9) considered. A Bayesian information criterion (BIC) was used to examine the fit of clustering solutions proposed by 6 spherical and diagonal normal mixture models to the morphospace occupation for all groups in the experiment: a wild-caught parents; b F₁ offspring; c F₁ parents; d F₂ offspring. Scenarios with 1 to 9 clusters were considered; models are explained in Additional file 1: Supplementary text

Fig. 7

Heritability of shell morphology as inferred from regressions of mid-parents versus offspring. The axes of nmMDS were used as proxy for shell morphology, each of which represents a module of which shape variation is illustrated in Additional file 1: Figure S2. Individual points represent randomly constructed associations of mid-parents and offspring for a single morphospecies in a single bootstrap replicate. Each point represents thus either an association of wild-caught parents and F₁ offspring or F₁ parents and F₂ offspring. The regression represents the summary statistics from 10,000 bootstraps on parent-offspring associations, with the mean in black and the 95% confidence interval in grey