Conservatism and novelty in the genetic architecture of adaptation in Heliconius butterflies

Abstract

Understanding the genetic architecture of adaptive traits has been at the centre of modern evolutionary biology since Fisher; however, evaluating how the genetic architecture of ecologically important traits influences their diversification has been hampered by the scarcity of empirical data. Now, high-throughput genomics facilitates the detailed exploration of variation in the genome-to-phenotype map among closely related taxa. Here, we investigate the evolution of wing pattern diversity in Heliconius, a clade of neotropical butterflies that have undergone an adaptive radiation for wing-pattern mimicry and are influenced by distinct selection regimes. Using crosses between natural wing-pattern variants, we used genome-wide restriction site-associated DNA (RAD) genotyping, traditional linkage mapping and multivariate image analysis to study the evolution of the architecture of adaptive variation in two closely related species: Heliconius hecale and H. ismenius. We implemented a new morphometric procedure for the analysis of whole-wing pattern variation, which allows visualising spatial heatmaps of genotype-to-phenotype association for each quantitative trait locus separately. We used the H. melpomene reference genome to fine-map variation for each major wing-patterning region uncovered, evaluated the role of candidate genes and compared genetic architectures across the genus. Our results show that, although the loci responding to mimicry selection are highly conserved between species, their effect size and phenotypic action vary throughout the clade. Multilocus architecture is ancestral and maintained across species under directional selection, whereas the single-locus (supergene) inheritance controlling polymorphism in H. numata appears to have evolved only once. Nevertheless, the conservatism in the wing-patterning toolkit found throughout the genus does not appear to constrain phenotypic evolution towards local adaptive optima.

Figure 1

Summary of crosses performed in H. hecale and H. ismenius. Geographic distribution of the subspecies used for the crosses are indicated by filling patterns, and sampling localities by circles and squares. The distribution of other H. hecale races found in Northern South America is also shown: H. h. annetta (I), H. h. rosalesi (II), H. h. anderida (III) and H. h. barcanti (IV).

Figure 2

Fine mapping of wing-patterning loci in H. hecale and H. ismenius. Grey-shaded boxes show recombinant individuals found in a total of N offspring in H. hecale melicerta × H. h. clearei (mel/cle), H. hecale melicerta × H. h. zuleika (mel/zul) and H. ismenius boulleti × H. i. telchinia (bou/tel) crosses. Annotated genes on each scaffold and candidate colour genes are represented by grey and black block arrows, respectively. Scaffolds on LG1 (top panel) are ordered according to the H. melpomene reference genome, but the order is unknown for the three scaffolds indicated on the right (HE670375, HE671246 and HE668177).

Figure 3

Phenotypic effect of Mendelian wing-patterning loci and major QTLs identified in H. hecale and H. ismenius crosses. For each type of cross (a–c), panel I (left) shows the crosses performed, the phenotypes associated with inferred genotypes at the major Mendelian loci (colour HhK; forewing melanisation HhAc/HiAc; forewing distal band layer/spot HhN/HiN; hindwing band HhBr/HiBr) and variation of the quantitative traits (dashed boxes: Hindwing spots Hspot, continuous melanisation Cm). Parental races (top left) are represented by their dorsal views, the F1 male siring the mapping families (top right) by its dorsal and ventral views and typical backcross specimens (bottom) have arrows pointing to the variable character. The name of the mapping families is written on the bottom of the panels of each cross type, with total number of offspring shown in brackets. Families labelled in bold were used to build the RAD libraries. Panel II (right) shows the genomic position and phenotypic effect of major QTLs. Coloured wing diagrams show the spatial distribution of individual QTL effects on pattern variation extracted from multivariate wing pattern analysis. Phenotypic variation is broken down into heatmaps for each of the three main colours (black, orange and yellow), representing, for every wing position, the strength of association between colour presence and allelic transition at the QTL (from blue to red). For analytical simplicity, both white and yellow elements in the H. hecale melicerta × H. h. clearei cross were considered as yellow elements. Genomic plots show genome-wide association (LOD) between wing pattern variation and markers along the 20 autosomes, with 5% (solid line) and 10% (dashed line) association thresholds. Panel aIII shows the detection of WntA transcripts by in situ hybridisation on wing imaginal discs of the last larval instar of H. h. melicerta and H. h. zuleika. WntA expression shows marked differences along the discal crossvein (arrows), in the M3-Cu2 intervein region (brackets) and in the Cu2-Cu1 intervein region (arrowheads). Colour dots indicate vein intersection landmarks. Phenotypic variation controlled by the HhAc locus is represented on the right.

Figure 4

Conservatism and novelty in the genetic architecture underlying the diversity of Heliconius wing patterns. (a) Known genetic architectures underlying pattern diversity throughout the clade mapped onto an unscaled phylogeny. Orange tree branches represent nine of the ten species in the silvaniform clade. Major colour variation loci are located on four chromosomes (top) and control variation in similar wing regions (arrows) throughout the genus. Wing phenotypes are represented based on Holzinger and Holzinger (1994). Note that the effect of the Br locus in H. cydno is shown on the ventral side. Loci with names in brackets were described based exclusively on interspecific crosses. (b) Comparative diagram of the distribution of the gene effects across the wing for toolkit loci in the silvaniform clade (excepting H. numata; left) and in the H. melpomene and H. erato clades (right), showing the general conservatism of the regions affected by homologous elements of the multilocus architecture despite some flexibility.