The Nymphalidae are the largest family of butterflies, with more than 6,000 species distributed throughout most of the world. Belonging to the superfamily Papilionoidea, they are usually medium-sized to large butterflies. Most species have a reduced pair of forelegs and many hold their colourful wings flat when resting. They are also called brush-footed butterflies or four-footed butterflies, because they are known to stand on only four legs while the other two are curled up; in some species, these forelegs have a brush-like set of hairs, which gives this family its other common name. Many species are brightly coloured and include popular species such as the emperors, monarch butterfly, admirals, tortoiseshells, and fritillaries. However, the under wings are, in contrast, often dull and in some species look remarkably like dead leaves, or are much paler, producing a cryptic effect that helps the butterflies blend into their surroundings.
Understanding how novel complex traits originate involves investigating the time of origin of the trait, as well as the origin of its underlying gene regulatory network in a broad comparative phylogenetic framework. The eyespot of nymphalid butterflies has served as an example of a novel complex trait, as multiple genes are expressed during eyespot development. Yet the origins of eyespots remain unknown. Using a dataset of more than 400 images of butterflies with a known phylogeny and gene expression data for five eyespot-associated genes from over twenty species, we tested origin hypotheses for both eyespots and eyespot-associated genes. We show that eyespots evolved once within the family Nymphalidae, approximately 90 million years ago, concurrent with expression of at least three genes associated with early eyespot development. We also show multiple losses of expression of most genes from this early three-gene cluster, without corresponding losses of eyespots. We propose that complex traits, such as eyespots, may have originated via co-option of a large pre-existing complex gene regulatory network that was subsequently streamlined of genes not required to fulfill its novel developmental function.
Butterfly eyespots play an essential role in natural and sexual selection, yet the evolutionary origins of eyespots and of their underlying gene regulatory network remain unknown. By scoring phenotypes and wing expression of five genes in 399 and 21 nymphalid species, respectively, we tested when eyespots and expression of their associated genes evolved. We found that the origin of eyespots was concurrent with the origin of the gene expression patterns, approximately 90 million years ago. Following this event, many genes expressed in eyespot development were lost in some lineages without a corresponding loss of eyespots, indicating substantial evolution in the cluster of genes associated with eyespots. This finding suggests that complex traits such as butterfly eyespots may initially evolve by re-deploying pre-existing gene regulatory networks, which are subsequently trimmed of genes that are unnecessary in the novel context.
(A) Origin of eyespots inferred from 399 nymphalid and 29 outgroup species from phylogeny in . (B) Origin of expression in eyespot centers inferred from gene expression profiles of 23 species. Presence or absence of expression of genes in future eyespot centers indicated by black and white boxes, respectively, and grey boxes indicate species/gene combinations for which expression data are unavailable. Green bars indicate two independent origins of eyespot-associated Antp expression. In both (A) and (B), divergence times (in millions of years) are from , ; red bars on the phylogeny indicate the possible locations of the single origin of eyespots, while gold bars indicate possible locations for the single origin of gene expression for sal, Notch, Dll, and possibly en in the eyespot centers. Asterisks (*) indicate species for which expression data are from , .
We next used gene expression profiles of 21 nymphalid and two outgroup species with eyespots (Figure S2) to determine if the gene regulatory networks associated with eyespot development are homologous within nymphalids and across butterfly lineages. Networks are considered homologous in two or more taxa if all the genes and their regulatory interactions can be traced back to the same network in the most recent common ancestor . We addressed the first portion of this homology assessment by testing for gene expression in the most recent common ancestor of eyespot-bearing nymphalid species. When focal expression is reconstructed on the history of Nymphalidae, expression of sal, Notch, and Dll in future eyespot centers are all estimated to have arisen once, approximately 90 million years ago (Figure 1B). Ancestral state estimates for two genes, Antp and en, were ambiguous, with one or two origins of focal expression possible. Single-origin models had the highest likelihood for en, while Antp had a maximum likelihood estimate of two origins; likelihood ratio tests on these two genes show a better fit of the model espousing a two origins for Antp expression (green bars in Figure 1B), but cannot discern between one (identical to sal, Notch, and Dll) or two independent, more recent origins for en expression (Figure 1A, Figure S3, and Table S3). These results demonstrate that a majority of eyespot associated genes investigated here (sal, Notch, Dll and potentially en) have a single origin of expression in the eyespot centers, lending support to homology of the eyespot gene regulatory network across Nymphalidae. Eyespots in the closely related Lycaenidae (Figure S2V) and the more distantly related Papilionidae  do not express any surveyed genes at their center, suggesting an independent and developmentally distinct origin. Cursory examination of these and other butterfly lineages (Riodinidae and Pieridae) suggests that eyespots are rare in these clades and are more likely to have evolved multiple times recently, rather than once, early in the clades' evolution, as shown here for nymphalids; however, a more thorough comparative examination of eyespot evolution in these clades should be done in future.
The widespread expression of Distal-less and spalt in future eyespots of nymphalids suggests a conserved and functional role of these transcription factors in wing pattern development. Dll and sal were expressed in the future eyespot centers of all but one and two surveyed nymphalid species, respectively (Figure 1 and Figure S2). Sal expression is also associated with non-eyespot patterns in two species: in Consul fabius (Cramer), sal is expressed in larval wing discs in locations where crescent-shaped patterns develop (Figure S2F); in Siproeta stelenes (Linnaeus), wing expression of sal is associated with patches of black scales that develop on the ventral hind wings (Figure S2R).
Similarly, black patches of scales are associated with pupal stage expression of Dll and sal in B. anynana and with sal expression in the distantly related butterfly Pieris rapae (Linnaeus) (Pieridae) , . Recent transgenic experiments suggest a functional role of Dll and sal in black scale development in B. anynana during the pupal stages of development (X. Tong, in review). Functional and comparative expression data together suggest that Dll and sal may have had a prior role in wing color pattern development, before they became co-opted into the eyespot center's regulatory network. The putative previous function of Dll and sal in color patterning wings, combined with the novel genetic background provided by other co-opted genes, may have facilitated the rapid appearance of an eyespot pattern.
The correspondence of eyespot origins with an origin of expression of at least three of the five genes examined (Figure 1) is consistent with the hypothesis that eyespots originated from a gene network co-option event (Figure 3) . This hypothesis posits a complex gene regulatory network involved in differentiating some other trait in a butterfly's body became expressed, in its entirety, in the future eyespot centers, and was subsequently rewired to generate the novel eyespot patterns. Subsequent network simplification is likely to happen when genes co-opted into the novel context are not functional in producing the novel trait. Loss of gene expression in the eyespot context may happen once network genes or their cis-regulatory elements duplicate, allowing the sub-functionalization and specialization of each copy for a different function . This process of duplication and specialization provides for losses of expression in novel contexts (e.g. the eyespot), while expression is retained in the original context. Alternatively, genes from the original network may be secondarily co-opted to function in the development of the novel trait due to their fortuitous expression there. If different lineages undergo different paths of secondary co-option, this mechanism may provide an explanation for the phenomenon of developmental system drift, where networks diverge between lineages despite conservation of the final phenotype .
In summary, this study highlights the utility of the comparative approach in understanding the origins and evolution of complex traits. The differences in gene expression in eyespot centers among nymphalid species suggest considerable cryptic developmental variation in a homologous trait. This type of broad comparative survey should prove useful in identifying candidates for future functional studies within and across taxa: genes expressed in all or a majority of species likely play a necessary role in the development of a complex trait and should be the primary targets of functional experiments. Future comparative work in other systems will allow for additional tests of the co-option hypothesis, to determine how often complex traits originate via bursts of complexity in gene expression, followed by genetic streamlining of unnecessary elements. 781b155fdc