Geographic contrasts between pre- and postzygotic barriers are consistent with reinforcement in Heliconius butterflies

Abstract

Identifying the traits causing reproductive isolation and the order in which they evolve is fundamental to understanding speciation. Here, we quantify prezygotic and intrinsic postzygotic isolation among allopatric, parapatric, and sympatric populations of the butterflies Heliconius elevatus and Heliconius pardalinus. Sympatric populations from the Amazon (H. elevatus and H. p. butleri) exhibit strong prezygotic isolation and rarely mate in captivity; however, hybrids are fertile. Allopatric populations from the Amazon (H. p. butleri) and Andes (H. p. sergestus) mate freely when brought together in captivity, but the female F1 hybrids are sterile. Parapatric populations (H. elevatus and H. p. sergestus) exhibit both assortative mating and sterility of female F1s. Assortative mating in sympatric populations is consistent with reinforcement in the face of gene flow, where the driving force, selection against hybrids, is due to disruption of mimicry and other ecological traits rather than hybrid sterility. In contrast, the lack of assortative mating and hybrid sterility observed in allopatric populations suggests that geographic isolation enables the evolution of intrinsic postzygotic reproductive isolation. Our results show how the types of reproductive barriers that evolve between species may depend on geography.

Keywords: Butterflies; gene flow; hybrid sterility; prezygotic isolation; speciation.

Figure 1

Left panel: The geographic distributions of H. elevatus and H. pardalinus (all Amazonian subspecies) at a continental scale, with the range of subspecies H. p. sergestus shown in yellow. Right panel: Local map showing the fine scale distributions in northern Peru, centered on the range of H. p. sergestus. In this map, the red triangles correspond to the subspecies H. p. butleri, which intergrades into other, similarly patterned subspecies in lowland Amazonia. Data are taken from Rosser et al. (2012) and supplemented with newer field collections made by the authors (see Methods and Results sections).

Figure 2

(A) Maximum‐likelihood phylogeny for the Peruvian silvaniform taxa, with H. melpomene aglaope as the outgroup, based on restriction site associated DNA (RAD) sequences (Supplementary Information S1). The scale bar refers to the number of substitutions per site, and node values are bootstrap support. Figures in brackets indicate the number of samples. Heliconius p. butleri clustered with the subspecies H. pardalinus dilatus from central Peru; the two are very similarly patterned and gradually intergrade. (B) Color patterns of the three parental taxa and their F1 hybrids, together with a summary of their relative geographic distributions and reproductive compatibility.

Figure 3

Observed climatic niches of H. elevatus (blue squares), H. p. butleri (red triangles), and H. p. sergestus (yellow circles) along rainfall (mm) and temperature (°C) gradients.

Figure 4

Host plant preference of the three taxa. (A) Preference measured as the proportion of eggs laid by multiple females on 21 species of Passiflora (Table S1) commonly occurring near Tarapoto and representing potential host plants. Seven plant species were not oviposited on and are not shown. (B) Preference measured as the proportion of eggs laid on size/quality matched shoots of four Passiflora species. In brackets is the total number of eggs laid by each taxon. 12 H. elevatus, 12 H. p. butleri, and 10 H. p. sergestus females were tested. Numbers within each column show the number of eggs laid on each host plant.

Figure 5

Taxa (represented by symbols) and compounds (represented by letters) along the first two dimensions of the NMDS ordination of 25 putative sex pheromone compounds found in hind‐wing androconia of males. Axes represent gradients of similarity among samples (similarity in compound composition) and among compounds (similarity in relative abundance across samples). A, homovanillyl alcohol; B, hexahydrofarnesylacetone; C, ?‐eicosene; D, ??‐heneicosadiene; E, (Z)‐9‐heneicosen; F, heneicosane; G, ?‐docosene; H, oleyl acetate; I, octadecyl acetate; J, phytol; K, (Z)‐9‐tricosene; L, tricosane; M, (Z)‐11‐eicosenyl acetate; N, tetracosane; O, (Z)‐11‐eicosenyl propionate; P, pentacosane; Q, 11‐methylpentacosane; R, (Z)‐13‐docosenyl acetate; S, hexacosane; T, 11‐methylhexacosane; U, heptacosane; V, 11‐methylheptacosane; W, octacosane; X, nonacosane; Y, octacosanal (? indicates unknown position of double bond).

Figure 6

Assay of hovering courtship behavior within and between taxa. Single female virgins were presented to groups of 15 males (five of each taxon) and hover courtship toward the female was recorded. The expected number of hover courtship behaviors per trial by males toward the female taxa and the statistical significance of any differences were obtained from GLMM model outputs. Error bars are 95% Wald confidence intervals; n is the number of virgin females tested of each taxon; E, H. elevatus; Pb, H. p. butleri; Ps, H. p. sergestus. Results of the other courtship behaviors measured (approaches and alightings) are shown in Figure S4, and details of the significance values following Bonferroni correction are provided in Table S3.