How butterflies match their wings

A pair of butterflies, with black wings marked by bright red stripes on the forewings and white bars on the hindwings.
Heliconius melpomene, in the “postman” pattern (Flickr: alain01789)

The following is a guest post by Ornob Alam, a graduate student in Michael Purugganan’s lab at New York University. Ornob’s PhD projects examine the demographic and evolutionary history of domesticated Asian rice in the context of past climate change and human migrations.

In 1859, the English naturalist Henry Walter Bates emerged from the rainforests of Brazil after more than a decade of recording the natural history of the region. Among his many discoveries, the adaptive strategy of mimicry in butterflies has in particular become embedded as a key area of study in evolutionary biology.

Bates noted that some butterflies that were different enough to be classified as separate species (or subspecies) closely resembled each other in wing color patterns. He concluded that certain butterflies had evolved to mimic others that were poisonous and avoided by birds. Birds learn to interpret the wing color patterns of poisonous butterflies as warning signals to avoid eating them, and any non-poisonous species that resembles a poisonous one gains some protection from predation.

Bates, however, was not able to explain why different poisonous species living in proximity also sometimes resembled each other. In 1879 Fritz Müller, a German naturalist, finally explained this as a numbers game. Teaching birds to recognize warning patterns comes at a cost, in the form of a certain number of individuals being eaten. When two poisonous species evolve to share the same wing color pattern, they share this cost and reduce it for each individual species. If a bird learns to avoid one species, the other benefits, and vice versa.

All of that helps to explain the selection pressures that favor the evolution of these two types of mimicry. But how does one species come to resemble another? Does the mimicking species independently evolve genetic variants that produce similar wing color patterns to the mimicked species, or is there occasional interspecies mating and flow of genetic variants between different species? If they do independently evolve the same patterns, do the genetic variants underlying the patterns in the different species occur in the same regions of the genome? 

There are multiple answers to these questions, depending on the colors, patterns, or species being considered. Let us attempt to answer them in the specific context of Müllerian mimicry, where different poisonous species share similar wing color patterns, in Heliconius butterflies. We will look at two (relatively) recent studies that come away with interesting but seemingly contrasting conclusions.

Same ink, different paintings

Genes involved in pigmentation, which include those encoding enzymes that produce pigments, are central to producing the actual colors of butterfly wings. Patterns formed by the colors depend on where and when these genes are expressed. Red color in Heliconius butterflies is associated with the expression of a protein called optix, a transcription factor that binds to non-coding regulatory regions to control the expression of other genes, and likely regulates the actual pigmentation genes. There is very little variation across co-mimicking species in the gene encoding optix, suggesting that the regulation of where and when optix is expressed may be key in determining where red color is produced.

Two butterflies with black wings marked by different patterns in bright red.
Figure 1. The “dennis” and “ray” patterns. (Wallbank et al. 2016)

That does turn out to be the case. Variation in regions regulating the expression of optix determines the nature of the red color patterns. A 2016 study found evidence for the existence of distinct regulatory modules that control the production of specific red color patterns. The study focused on contrasting patterns called “dennis” and “ray” (Fig. 1) that are found in subspecies of both Heliconius melpomene and Heliconius timareta. Individual subspecies can have either, both, or neither of these two patterns. By comparing the genomes of subspecies that possess either dennis or ray patterns, the authors found specific non-coding regions downstream of optix that are associated with the presence of each pattern, and interpreted the regions as enhancers – regulatory regions that enhance the expression of neighboring genes – functioning as distinct modules. They showed that the two modules had distinct evolutionary origins, and came to be shared between different species through ancient hybridization events. For instance, the ray enhancer region came into H. timareta from H. melpomene around 1.2 million years ago, even though the two species had diverged more than 2 million years ago.

Based on their findings, the authors proposed that wing color patterns can evolve by enhancer shuffling, where modular regulatory regions, once they arise, are shuffled around between populations and species to produce different combinations of elements like the ray and dennis patterns. Hybridization provides a very intuitive framework for understanding the evolution of mimicry: instead of always having to evolve the mimicked patterns from scratch, each species was sometimes able to sample existing genetic diversity from other species.

While the evidence for hybridization is robust, the inference of modular enhancers in this study is based on associations between genomic features and observed patterns, without any experimental manipulation to verify the roles of the identified regulatory regions in different subspecies. A 2019 study returned to the question of optix regulation and wing pattern evolution with a more sophisticated experimental toolset.

In search of what breaks the patterns

Four butterflies with black wings marked in red and white stripes.
Figure 2: “Radiate” versus “postman” patterns of red and white color in two Heliconius species. (Lewis et al. 2019)

The authors of the 2019 study initially focused on a species called Heliconius erato, which diverged from the clade containing H. melpomene and H. timareta around 10 million years ago, and took a different approach to identifying genomic regions that regulate optix. Active regulatory regions share several biological features. For instance, regulatory regions that are present slightly further away from, or downstream of the genes they regulate often loop over and come into contact with regions immediately upstream of those genes. The study utilized three different approaches based on detecting such features to identify several candidate regulatory regions that vary between H. erato individuals with red and white wing coloring in two patterns called “radiate” and “postman” (Fig. 2). Note that the radiate pattern, considered a single pattern here, represents the presence of both the dennis and ray patterns from the 2016 study (Fig. 1).

To understand the functional role of a particular gene or non-coding region, scientists often remove or disrupt the region, and infer its function from what they observe in its absence. The authors used CRISPR/cas9 gene-editing technology to disrupt five of the candidate regulatory regions in turn in a radiate group, and observed that loss of function of any of these elements resulted in multiple changes in wing color patterns (example in Fig. 3). The elements therefore appeared to be interdependent, as they were all necessary to form the correct pattern, and certainly not modular, as each element was not associated with the presence or absence of a specific part of the pattern. This paints a very different picture of wing pattern evolution from the modular enhancer shuffling hypothesis proposed by the 2016 study.

Comparison of butterfly wings with and without candidate gene disruption, showing that coloration is lost from some parts of the wing when the candidate has been disrupted.
Figure 3: White dotted lines trace changes resulting from disruption of a candidate regulatory region. (Lewis et al. 2019)

To address this apparent contradiction, the authors proposed an alternative model that they call enhancer shuttering. In the case of optix, this model posits that several distinct regulatory regions or enhancers work together to generate optix expression across tissues, and “shutter” proteins that bind to some regulatory regions selectively repress optix expression – and thus the appearance of red color – to generate the observed wing color pattern. Different variations of patterns could be produced through loss or gain of regulatory regions that are bound by “shutter” proteins, as well as through changes in expression of the “shutter” proteins themselves. Variation under this model is expected to be more continuous, which is in contrast to the discrete predictions of the enhancer shuffling model, where each regulatory region is responsible for a distinct, fully-formed pattern element. How is the evidence for this new model?

Most of the H. erato mutants with the disrupted regulatory regions showed varying degrees of loss of red pigmentation, suggesting that the regions are positive regulators or enhancers. However, one of the mutants showed an extension of red color, which suggests that the disrupted regulatory region in this mutant normally acts as a negative regulator that is perhaps responsive to a “shutter” factor. As another line of evidence, they showed that disruption of WntA, a gene associated with black color, resulted in the emergence of new red color patterns. WntA may thus be acting as a “shutter”, with black color representing the absence of red.

The enhancer shuttering hypothesis does not completely refute the older enhancer shuffling hypothesis. The authors offer an olive branch by suggesting that modular enhancers, like the ones reported in the 2016 study, may allow rapid switching between a few color patterns found in a particular geographical region but over evolutionary timescales, novel color patterns likely emerged through sampling a greater diversity of variants by enhancer shuttering. It would certainly be interesting to do some CRISPR-based disruptions of the dennis and ray regulatory regions proposed in the 2016 paper to see how they affect the patterns.

The authors went on to show that the radiate pattern emerged in H. erato as recently as 30,000 years ago, and spread between different H. erato subspecies through hybridization, mirroring the spread of the dennis and ray regions as reported in the 2016 study. What about the radiate pattern in H. melponene and H. timareta, that was explored as the constituent dennis and ray patterns in the 2016 study? The authors provide evidence that H. melponene and H. erato have evolved the same radiate pattern through independently undergoing changes in similar genomic regions.

Going back to our initial framing questions, it turns out that in the case of red wing patterns in Heliconius Müllerian co-mimics, mimicry can arise independently, or through hybridization, and at least sometimes, matching wing patterns evolve through parallel genetic changes.

References

Evans MA. 1965. Mimicry and the Darwinian heritage. J. Hist. Ideas. 26(2):211–220. doi: 10.2307/2708228.

Wallbank RWR et al. 2016. Evolutionary novelty in a butterfly wing pattern through enhancer shuffling,” PLOS Biol. 14(1):1002353. doi: 10.1371/journal.pbio.1002353.

Lewis JJ et al. 2019. Parallel evolution of ancient, pleiotropic enhancers underlies butterfly wing pattern mimicry. PNAS. 116(48):24174–24183. doi: 10.1073/pnas.1907068116.

About Jeremy Yoder

Jeremy B. Yoder is an Associate Professor of Biology at California State University Northridge, studying the evolution and coevolution of interacting species, especially mutualists. He is a collaborator with the Joshua Tree Genome Project and the Queer in STEM study of LGBTQ experiences in scientific careers. He has written for the website of Scientific American, the LA Review of Books, the Chronicle of Higher Education, The Awl, and Slate.
This entry was posted in adaptation, Coevolution and tagged , , . Bookmark the permalink.