optix Drives the Repeated Convergent Evolution of Butterfly Wing Pattern Mimicry

See allHide authors and affiliations

Science  26 Aug 2011:
Vol. 333, Issue 6046, pp. 1137-1141
DOI: 10.1126/science.1208227


Mimicry—whereby warning signals in different species evolve to look similar—has long served as a paradigm of convergent evolution. Little is known, however, about the genes that underlie the evolution of mimetic phenotypes or to what extent the same or different genes drive such convergence. Here, we characterize one of the major genes responsible for mimetic wing pattern evolution in Heliconius butterflies. Mapping, gene expression, and population genetic work all identify a single gene, optix, that controls extreme red wing pattern variation across multiple species of Heliconius. Our results show that the cis-regulatory evolution of a single transcription factor can repeatedly drive the convergent evolution of complex color patterns in distantly related species, thus blurring the distinction between convergence and homology.

Some of the most dramatic examples of natural selection involve Müllerian mimicry—a phenomenon in which two or more species share predator avoidance signals. This type of mimicry is one of the main forces driving the evolution of warning coloration in Heliconius (1), a neotropical butterfly genus of ~40 species famous for its extensive intraspecific wing pattern variation. The two best-studied Heliconius species are the comimics Heliconius erato and Heliconius melpomene (Fig. 1); each has more than 25 named geographic races, most of which mimic co-occurring Heliconius species, other unrelated butterflies, and day-flying moths. Mimetic wing pattern radiations like those seen in H. erato and H. melpomene have resulted in the evolution of hundreds of wing pattern races within the genus (2). Despite increasing evidence that the same genomic regions contribute to wing pattern variation in different Heliconius species, the specific genes that underlie the repeated convergent evolution of wing patterns have remained unknown (3).

Fig. 1

Mimicry rings show convergent and divergent evolution within Heliconius. (A) Comimetic Heliconius species are typically distantly related as shown in this phylogenetic tree of selected Heliconius species. Illustrated are two major mimicry groups: the blue mimics represented by the H. cydno/H. pachinus complex and the H. sapho/H. hewitsonii/H. eleuchia complex, and the red/yellow mimics represented by H. erato and H. melpomene. In addition to local convergence, wing patterns often vary markedly across the geographic range of a species to match local mimicry rings. For H. cydno, we show three of more than a dozen color pattern races. For the red/yellow mimicry ring, we show five of more than 25 different geographic races of H. erato and H. melpomene. (B) Simplified geographic distribution of the H. erato and H. melpomene color pattern forms shown in (A), as color-coded by the dots. The entire range of H. erato and H. melpomene is much larger (gray shaded region).

Three genomic regions control most wing pattern variation in Heliconius—in both comimetic species like H. erato and H. melpomene and also in highly dissimilar species such as Heliconius numata and Heliconius cydno (3). The region that controls the major red color patterns (the hindwing rays, forewing band, and basal forewing patch) maps to a 380-kb genomic interval corresponding to the D locus in H. erato and H. melpomene (4, 5) and the G locus in H. cydno (6). To pinpoint transcription units in this interval involved in wing pattern variation we screened for differential color pattern–related RNA transcription across the entire region. We examined two H. erato forewing phenotypes: one with a red midwing band, and one with a yellow midwing band and a red basal patch (Fig. 2A). For both phenotypes, we compared transcription in basal, midwing, and apical wing sections from five different developmental stages (Fig. 2A). This strategy allowed us to perform separate comparisons of each distinct red color pattern element, while also including comparisons between invariant apical forewing tips to control for population-associated expression differences unrelated to phenotype. A global analysis of all stages and tissue sections from the two phenotypes identified a single major cluster of red color pattern–associated probes at the position of optix (Fig. 2A), an intronless homeobox transcription factor gene with an 801–base pair coding region. Further analysis of the data revealed that this pattern-specific differential expression of optix begins between 12 and 60 hours after pupation and persists throughout pupal development (fig. S1). No other genes in the map interval besides optix were consistently expressed in association with color pattern variation.

Fig. 2

optix expression prefigures red wing patterns in Heliconius. (A) H. erato genomic expression tiling microarrays spanning the entire red color pattern genomic map interval pinpoint optix as the only gene strongly differentially expressed in association with red forewing pattern elements. Details are provided in the SOM and fig. S1. (B) optix mRNA expression in 72-hour pupal wings of different species and races of Heliconius shows spatial association with all red wing patterns and (C) discrete patches of acute scales. (D) In V. cardui, optix expression is associated with acute scale patches but not color patterns.

The finding that optix is differentially expressed in association with red color patterns in two H. erato phenotypes raised the question of whether optix expression is also associated with red patterns in other H. erato races and other Heliconius species. We thus examined in situ spatial expression patterns of optix mRNA in 72-hour pupal forewings and hindwings across a diversity of phenotypes and species. Across three geographic races of H. erato, three races of H. melpomene, and one race of H. cydno, we found that all red wing patterns, including red hindwing rays in both H. erato and H. melpomene, were perfectly prefigured by optix expression (Fig. 2, B and C). optix also prefigured forewing bands in H. melpomene plesseni that consist of both red and white scales (Fig. 2B), consistent with the activity of an unlinked modifier locus that modulates the amount of red in forewing bands (7). Overall, optix expression predicted color pattern shapes with such precision that it was possible to identify samples to species and race by optix expression alone.

The concurrence of mapping and expression data led us to hypothesize that optix is the major gene that controls red color pattern variation in H. erato, H. melpomene, and H. cydno. To test this, we examined optix genotype-by-phenotype associations in hybrid zones where there is natural variation for optix-associated red color patterns (Fig. 3). These hybrid zones consist of populations that represent thousands of generations of recombination between wing pattern types; therefore, DNA sequence associations are due to linkage with functional variation and not other phenomena like population structure (8, 9). In a Peruvian H. erato hybrid zone where linkage disequilibrium is reported to decay to background levels within 1 kb (8), we examined DNA sequence variation in a series of red color pattern interval genes from 73 individuals (Fig. 3B). optix showed by far the strongest association, with 13 of 30 variable nucleotides having significant associations of P < 1 × 10−15, with the strongest being P < 1 × 10−32 (Fig. 3B and table S2). Similarly, we sampled 62 individuals from the parallel Peruvian H. melpomene hybrid zone and found that optix showed the most significant associations of any sampled marker, with 5 of 29 variable nucleotides having significant associations of P < 1 × 10−10, with the strongest being P < 1 × 10−32 (Fig. 3C and table S3). In H. melpomene, a cluster of genes (slu7, kinesin, GPCR) ~150 kb away from optix showed a total of six nucleotide associations with P < 1 × 10−10 (Fig. 3C), though none of these genes displayed color pattern–related spatial expression and they showed only marginal associations in H. erato. This broad spread of associated nucleotides in H. melpomene is consistent with the observation that linkage disequilibrium in this hybrid zone can extend 100 kb or more at wing pattern–associated sites (9). Lastly, we assessed optix genotype-by-phenotype associations in 32 individuals collected from the Costa Rican H. cydno/H. pachinus hybrid zone wherein, as with H. erato, highly significant associations are only expected within 1 kb of functional variation (6). Of all the sampled genes, optix had the most significant association with a nucleotide that showed P < 1 × 10−12 (Fig. 3E and table S4). Together, these three independent population genetic analyses all pinpoint optix as the major gene that controls red color pattern variation.

Fig. 3

optix shows highly significant genotype-by-phenotype associations in three independent Heliconius hybrid zones. (A) Overlapping H. erato and H. melpomene hybrid zones in Peru, where optix shows the strongest genotype-by-phenotype association across all the genes surveyed in (B) H. erato and (C) H. melpomene. (D) A Costa Rican H. cydno/H. pachinus hybrid zone where (E) optix shows the most significant genotype-by-phenotype association out of all the surveyed genes. The horizontal lines mark a false discovery rate of 0.001.

Because optix alleles from different geographic races of Heliconius show clear differences in spatial expression patterns (Fig. 2B), it follows that there are cis-regulatory differences between alleles. To rule out allelic variation in optix protein amino acid sequences, we sequenced the entire coding region of optix from H. erato emma, H. erato favorinus, H. erato petiverana, H. melpomene rosina, H. melpomene melpomene, H. pachinus, and H. cydno galanthus. All variation that we observed was synonymous, meaning that despite 12 to 25 million years of evolution (10) the optix amino acid sequence has remained invariable. This was true between highly divergent races of the same species (within H. erato), between divergent phenotypes of closely related species (H. cydno versus H. melpomene), and between convergent phenotypes of distantly related species (H. erato versus H. melpomene). Thus, when considered in the context of the mapping work, we deduce that the functional variation between optix alleles is cis-regulatory.

To make an initial assessment of whether optix is associated with color pattern determination in non-Heliconius butterflies and moths, we examined in situ optix mRNA expression in 72-hour pupal wings of the nymphalid butterflies Vanessa cardui (Fig. 2D) and Agraulis vanillae (fig. S2), and the pyralid moth Ephestia kuehniella (fig. S2). None of these species showed color pattern–related expression of optix, even though both butterflies have red color patterns and A. vanillae is from a heliconiine genus closely related to Heliconius. In all lepidopteran species sampled, however, optix precisely prefigured small, discrete patches of specialized acute (pointed) scales (Fig. 2, C and D). We have identified several H. melpomene mutants that show homeotic transformations of normal red scales into acute scales (fig. S2), suggesting a simple regulatory switch between “red” and “acute” identities and demonstrating how optix might serve a dual function in determining these two scale types. In sum, our observations suggest that optix is ancestrally associated with acute scale morphology in lepidopteran pupal wings and that its color patterning function in Heliconius may be derived.

optix provides a compelling example of a gene that drives adaptation because its various alleles are regulatory variants that have pronounced effects on complex large-scale patterns. Because red color pattern variation is primarily achieved through optix cis-regulatory variation, we hypothesize that the different races within a given Heliconius species share a common wing prepattern that is interpreted by different optix alleles in different ways. Although the nature of this prepattern remains a mystery, the ability of optix to produce a complex variety of readouts from a common regulatory background suggests that it acts as a signal-integrating “input-output” regulatory gene (11). On the basis of this model, we speculate that Heliconius color pattern variation evolves primarily through the gain and loss of pattern-specific cis-regulatory elements, as do Drosophila wing melanin patterns (12, 13), although we have yet to identify any specific optix cis-regulatory elements. That a single gene drives the evolution of both convergent and divergent color patterns across a range of species also presents a conceptual conundrum: Convergence and homology are often presented as contrasting explanations of similarity, yet the color patterns of comimetic Heliconius species could be considered both convergent and homologous under many modern definitions (14, 15). The challenge now is to elucidate the functional role of optix in color pattern formation in order to understand how it was co-opted and why it served as such an efficient catalyst for the evolutionary radiation of Heliconius wing patterns.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Tables S1 to S5


References and Notes

  1. Acknowledgments: We thank M. A. Flores and A. Portugal for help rearing butterflies, M.-J. Sung for help with electron microscopy, J. Miller for help with microarray analysis, and J. Mallet for sharing samples. Supported by funding from NSF (R.D.R., W.O.M., M.R.K.), NIH (H.M.H., M.R.K.), and Centre National de la Recherche Scientifique–Guyane (B.A.C.) Sequences are available on GenBank (JN102349–JN102354). Microarray data are available on Gene Expression Omnibus (GSE30221). Permission to collect and export butterfly tissue was provided by the Peruvian Ministerio de Agricultura and Instituto Nacional De Recursos Naturales (004-2008-INRENA-IFFS-DCB and 011756-AG-INRENA); the French Guiana Ministere de L’Ecologie, de L’Energie, du Developpemet Durable et de la Mar (BIODAD-2010-0433); the Ecuadorian Ministerio del Ambiente Ecuadorian (013-09 IC-FAU-DNB/MA); and the Panamanian Autoridad Nacional del Ambiente (SC/A-7-11). W.O.M., F.N, and R.D.R. designed experiments and conducted initial work. R.P. produced butterfly crosses and generated and analyzed the microarray data with H.M.H. A.M. generated in situ hybridization data. B.A.C., C.P.-D., C.D.J., N.L.C, and M.R.K. contributed nucleotide association data. R.C. and G.H. assisted with positional cloning. R.P., A.M., B.A.C., and R.D.R. prepared figures and R.D.R. wrote the manuscript with support from R.P., A.M., and W.O.M. The authors declare that they have no competing financial interests.

Stay Connected to Science

Navigate This Article