Heliconius butterflies: a window into the evolution and development of diversity

Abstract

Butterflies have become prominent models for studying the evolution and development of phenotypic variation. In Heliconius, extraordinary within species divergence and between species convergence in wing color patterns has driven decades of comparative genetic studies. However, connecting genetic patterns of diversification to the molecular mechanisms of adaptation has remained elusive. Recent studies are bridging this gap between genome and function and have driven substantial advances in deciphering the genetic architecture of diversification in Heliconius. While only a handful of large-effect genes were initially identified in the diversification of Heliconius color patterns, recent experiments have begun to unravel the underlying gene regulatory networks and how these have evolved. These results reveal an evolutionary story of many interacting loci and partly independent genetic architectures that underlie convergent evolution.

Figure 1.. Diversity and development of Heliconius butterfly wing coloration.

(A.) Example of convergence and divergence in the Heliconius genus. Left, mimicry between two clades that have diverged over 10 million years ago and converged on the same color pattern (Modified from [35]). Right, color pattern diversity within the Heliconius erato species (tree and time estimates from [27]). Dark gray shading on the map of South America indicates the distribution area of H. erato. Colored circles indicate localities in Panama where mimicking species considered in this review occur. (B.) Within the Heliconius species, four genes, including aristaless1, cortex, WntA and optix, have been functionally linked to color pattern development, with vvl being a strong candidate for an effect in the apical part of the medial forewing pattern. The genes cortex, vvl and WntA are known to be expressed in the late larval instar stage and early pupal stages [33,43,56]. The gene optix is expressed later in development and its expression pattern is dependent on the color patterns genes expressed earlier in development [31,34]. The gene aristaless1 (al1) is expressed during 5^th instar larvae and pupal development in white H. cydno [51].

Figure 2.. Introgression, divergence and selection in Heliconius species and populations.

(A.) Densitree calculated from 10 kb windows showing the reticulate (non-bifurcating) phylogenetic relations within two Heliconius clades. Blue solid lines indicate the tree supported by the majority of 10 kb windows. Red dashed lines indicate introgression events. Tree obtained from [6], supplemental Figure S21. (B.) Hybrid zone divergence (F_ST) and selective sweep signals (Sweepfinder2 Composite Likelihood Ratio (CLR) statistic) at the WntA (top), cortex (middle), and optix (bottom) gene. Butterfly cartoons highlight phenotypic differences that have been associated with these genes. Localities of hybrid zones in Central and South America are indicated as colored circles. Asterisks indicate populations for which selective sweep support is shown. Colored brackets above F_ST signals indicate proposed CRE modules. F_ST data obtained from [24], selective sweep data obtained from [4]. (C.) A hypothesis of cis-regulatory element modularity was proposed from these data as populations or species that share different combinations of pattern elements (brackets) often also share haplotypes at genomic intervals in different combinations (filled and empty rectangles indicate modular CRE variants)

Figure 3.. Functional genomics on the color pattern gene optix (A.)

ATAC-seq profile from a 3-day old pupal forewing in H. e. lativitta and CRISPR/Cas9 mutants of the obs214 CRE in H. e. lativitta and H. m. aglaope. CREs (shaded in blue) were identified based on differential accessibility between morphs, wing tissue and functional effect after excision. Wing diagram on the left shows number of identified CREs that affected red color pattern elements. Note that four out of five CREs affected both hindwing and forewing pattern. Data and images obtained from [47]. (B.) Expression of optix is dependent on patterning genes. Both divergence in these patterning genes and the cis-regulatory landscape near optix may result in morphological diversification. Diagram modified from [47]. (C.) ChIP-seq of the Optix protein identified many loci with evidence for selection (Sweepfinder2 Composite Likelihood Ratio (CLR) statistic) that are bound by Optix (orange triangles). Data obtained from [46].

Figure 4.. CRISPR/Cas9 on the color pattern gene WntA.

(A.)WntA Wildtype (WT) and CRISPR/Cas9 (KO) phenotypes on co-mimicking butterflies from two clades (left, erato/sara/sapho clade; right, melpomene/silvaniform/doris clade) that diverged over 10 million years ago. For phylogenetic relations and localities see Figure 1A. Images obtained from [35]. (B.) WT and KO phenotypes on geographic color pattern morphs of H. erato. Images obtained from [35]. (C.) Quantitative comparison of WT and KO variation between H. erato and H. Melpomene from Panama with a red forewing pattern. Heatmaps at the top demonstrate the consistency of the forewing pattern in WT and KO samples with white indicating consistent presence of forewing patterns and blue gradient indicating less consistent presence. Bottom shows inter-species differences in forewing pattern with red indicating higher presence of forewing pattern in H. erato and blue indicating higher presence in H. melpomene. The yellow and green outlines show the WntA KO area. Orange triangle indicates overlap in mismatch between WT co-mimicking populations and respective WntA KO patterning area. Images obtained from [52]. (D.) Diagram of divergence in the gene regulatory network with which WntA interacts, and subsequent selection for co-mimicking phenotypes (convergence). Tips of mountains represent peaks in the fitness landscape which are hypothesized to have been reached by independently rewiring the gene regulatory network underlying the development of color pattern traits. Diagram modified from [35].