Convergence in sympatric swallowtail butterflies reveals ecological interactions as a key driver of worldwide trait diversification

Abstract

Ecological interactions can promote phenotypic diversification in sympatric species. While competition can enhance trait divergence, other ecological interactions may promote convergence in sympatric species. Within butterflies, evolutionary convergences in wing color patterns have been reported between distantly related species, especially in females of palatable species, where mimetic color patterns are promoted by predator communities shared with defended species living in sympatry. Wing color patterns are also often involved in species recognition in butterflies, and divergence in this trait has been reported in closely related species living in sympatry as a result of reproductive character displacement. Here, we investigate the effect of sympatry between species on the convergence vs. divergence of their wing color patterns in relation to phylogenetic distance, focusing on the iconic swallowtail butterflies (family Papilionidae). We developed an unsupervised machine learning-based method to estimate phenotypic distances among wing color patterns of 337 species, enabling us to finely quantify morphological diversity at the global scale among species and allowing us to compute pairwise phenotypic distances between sympatric and allopatric species pairs. We found phenotypic convergence in sympatry, stronger among distantly related species, while divergence was weaker and restricted to closely related males. The convergence was stronger among females than males, suggesting that differential selective pressures acting on the two sexes drove sexual dimorphism. Our results highlight the significant effect of ecological interactions driven by predation pressures on trait diversification in Papilionidae and provide evidence for the interaction between phylogenetic proximity and ecological interactions in sympatry, acting on macroevolutionary patterns of phenotypic diversification.

Keywords: Papilionidae; machine learning; macroevolution; mimicry; wing color pattern.

Fig. 1.

Color pattern variations and phylogenetic relationships across Papilionidae butterflies. First row: Phenotypic variations captured by the unsupervised machine learning–based methods applied to our 2,716 pictures of Papilionidae. Independent PCA was carried out on (A) males and (B) females. Note that we used the mean phenotype by sex and by species and represented only the first two axes, explaining only 19.9% of the phenotypic variance. We display the actual picture of butterflies for randomly sampled species (with at least one species per genera) on the morpho-space to observe how actual color pattern variation was separated by our method of phenotypic discrimination. Second row: Phylo-morpho space computed on the mean phenotype for each species in (C) males and (D) females. The phylo-morpho-space is the projection of the morpho-space coordinates on the first two axes of the morpho-space principal components. Each colored dot represents the phenotype of a given species, and the black lines show the projection of the phylogenetic relationships among the species. The color code corresponds to the different genera, the color gradient corresponding to the location of the genus on the phylogeny.

Fig. 2.

Discriminant features used to build the morpho-space obtained using Grad-CAM mapping. We selected one or two species for several genera to give examples that cover different parts of the phenotypic space. Redder color indicates pixels that weigh the most in the activation of the neural network. From left to right, up to down: (A) Allancastria cerisyi, (B) Atrophaneura priapus, (C) Battus philenor, (D) Graphium weiskei, (E) Mimoides euryleon, (F) Ornithoptera priamus, (G) Pachliopta aristolochiae, (H) Papilio caiguanabus, (I) Parides photinus, (J) Parnassius nomion, (K) Trogonoptera brookiana, (L) Troides aeacus, (M) Lamproptera curius, and (N) Troides andromache.

Fig. 3.

Complex interactions between the effect of phylogenetic distance and overlap on phenotypic distance. Heatmaps show predicted values of phenotypic distance in function of phylogenetic distance and percentage of overlap are shown for (A) males and (B) females. Example curves of predicted phenotypic distance in function of percentage of overlap for three arbitrarily chosen phylogenetic distances are shown for (C) males and (D) females. The three gray dotted lines for each heatmap show the phylogenetic distance chosen for the examples of curves of predicted phenotypic distance in function of percentage of overlap—1 and 4 are example curves for a phylogenetic distance of 12.1 Myr, 2 and 5 for 51.87 Myr, and 3 and 6 for 91.63 Myr (sum of branch lengths).

Fig. 4.

Detection of convergence and divergence events between sympatric and allopatric species pairs. Distribution of residual medians for (A) males and (B) females after 100,000 permutations for sympatric pairs in blue and allopatric pairs in pink. The dashed blue vertical line corresponds to the observed median for the residuals of pairs of sympatric species, and the dashed pink vertical line corresponds to the observed median for the pairs of allopatric species. (C–F) Phenotypic convergence and divergence associations between pairs represented on the swallowtail phylogenetic tree with male convergence (C) and divergence (D) between pairs, and female convergence (E) and divergence (F) between pairs. Examples of pairs of convergent and divergent species are plotted along the phylogeny: C-1: Mimoides lysithous, C-2: Papilio erostratus, C-3: Papilio hectorides, C-4: P. photinus, C-5: Parides bunichus. D-1: Graphium antiphates, D-2: Graphium xenocles, D-3: P. nomion, D-4: Parnassius stubbendorfii, D-5: Papilio pelaus, D-6: Papilio aristodemus. E-1: Graphium xenocles, E-2: Papilio protenor, E-3: P. erostratus, E-4: Papilio clytia, E-5: P. photinus. F-1: Graphium antiphates, F-2: Graphium xenocles, F-3: P. nomion, F-4: Parnassius stubbendorfii, F-5: Papilio protenor, F-6: Papilio hipponous, F-7: Parides bunichus, F-8: Euryades corethrus.

Fig. 5.

Evolution of sexual dimorphism in color pattern throughout the family Papilionidae: contribution of divergence of male phenotype vs. female phenotype in the evolution of sexual dimorphism was estimated by comparing sister species. (A) Swallowtail phylogeny showing sister species with the level of dimorphism (grayscale) and ratio of raw contrasts: (male/male distance over female/female distance) - 1 for positive values and – (female/female distance over male/male distance) - 1 for negative value for sister species so that relative divergence becomes more pronounced as the value diverges from 0. (B) Distribution of the ratio of raw contrasts for sister species with the median indicated. (C) Male and female specimens are shown for the two-sister species presenting the highest ratio: 1) Euryades corethrus above and Euryades duponchelii below, and for the two sister species presenting the lowest ratio: 2) Papilio meriones above and Papilio dardanus below.