Butterfly eyespots evolved via cooption of an ancestral gene-regulatory network that also patterns antennae, legs, and wings

Abstract

Butterfly eyespots are beautiful novel traits with an unknown developmental origin. Here we show that eyespots likely originated via cooption of parts of an ancestral appendage gene-regulatory network (GRN) to novel locations on the wing. Using comparative transcriptome analysis, we show that eyespots cluster most closely with antennae, relative to multiple other tissues. Furthermore, three genes essential for eyespot development, Distal-less (Dll), spalt (sal), and Antennapedia (Antp), share similar regulatory connections as those observed in the antennal GRN. CRISPR knockout of cis-regulatory elements (CREs) for Dll and sal led to the loss of eyespots, antennae, legs, and also wings, demonstrating that these CREs are highly pleiotropic. We conclude that eyespots likely reused an ancient GRN for their development, a network also previously implicated in the development of antennae, legs, and wings.

Keywords: evolutionary biology; evolutionary developmental biology; genetics.

Fig. 1.

Tissues used for RNA-seq analysis and character tree constructed using DE genes. (A) We used 16 tissue groups from three separate developmental stages of B. anynana for RNA extractions. Embryos at 3 h, 12 h, and 24 h after egg laying. Larval forewings, T1 legs, horns, and prolegs. Pupal antennae, T1 legs, forewings, eyes, wing margins, eyespots, and two eyespot control tissues, all dissected at 3 h after pupation, and a wounded wing dissected at 24 h after pupation. (B) PCA using 10,281 DE genes obtained from pairwise comparisons between different tissues. Tissues are clustered according to their developmental stages. (C) Character tree constructed using 10,281 DE genes showed eyespot tissue clustered with antenna tissue first, and next with tissues from the same developmental stage, except for a 24-h pupal wounded wing (♦), which clustered with larval wing tissue. (D) Character tree constructed using 7,133 DE genes from 3-h pupal stage with 3-h embryos as outgroup showed eyespot tissue clustered with antenna tissue. **100 unbiased (AU) P value; *90 to 99 unbiased (AU) P value.

Fig. 2.

Function of sal and regulatory interactions between Dll, sal, and Antp inferred with CRISPR and immunohistochemistry (A). WT female forewing. (B) sal crispant female forewing. (C) WT female hindwing. (D) sal crispant female hindwing. (E) Levels of Dll and Antp proteins in WT forewing. (F) Levels of Dll and Antp proteins in Dll crispant forewing. (G) Levels of Dll and Antp proteins in an Antp crispant forewing. (H) Levels of Dll and Sal proteins in WT forewing. (I) Levels of Dll and Sal proteins in Dll crispant forewing. (J) Levels of Dll and Sal proteins in sal crispant forewing. (K) Levels of Sal and Antp proteins in WT forewing. (L) Levels of Sal and Antp proteins in Antp crispant forewing. (M) Levels of Sal and Antp proteins in sal crispant forewing. White square regions were highly magnified. (N) Schematic diagram of genetic interaction among Dll, sal, and Antp in the eyespot region of a developing forewing. (Scale bars in A–D: 5 mm for whole wings and wing details.) (Scale bars in E–M: 100 μm in low and 50 μm in high magnification.)

Fig. 3.

Multiple traits are affected by disruptions of a single Distal-less CRE. (A) B. anynana Dll locus (previously cloned into a BAC) visualized using IGV showing all 18 FAIRE-seq open-chromatin regions at 24 h postpupation (short pink lines). First exon (untranslated region [UTR] in blue) shows the open-chromatin region (highlighted by a short pink line) at position 54 kb at the transcriptional start site of Dll. The FAIRE peak at position 150 kb (Dll319; highlighted with a purple bar) is open in the B. anynana forewing and was targeted with CRISPR. Four RNA guides were used simultaneously to target this region. (B) FAIRE-seq results showing an open region of chromatin in the distal forewing (FWD) at position 150 kb on the Dll BAC (blue peak). (C–E) Crispant phenotypes from the same individual: With a missing thoracic leg as a caterpillar, and the same missing thoracic leg, and also missing hindwing, as an adult. (F and G) Crispant wing phenotypes showing loss of eyespots and pigmentation defects. (H) Crispants showing antennal defects. (I and J) Dll319 CRE driving EGFP in eyespot centers in a 24-h pupal wing and a fifth instar larval wing, respectively, in transgenic animals. (K) Transgenic embryo showing EGFP expression driven by the Dll319 CRE in mouthparts, antennae, legs, and pleuropodia (white arrows from Left to Right). HWD, distal hindwing; HWP, proximal hindwing.

Fig. 4.

Visualization of open chromatin around Dll and sal genomic regions for different tissues, and identification of a sal pleiotropic CRE. (A) ATAC-seq reads around the Dll genomic region with highlights in the open regions shared across different tissues (orange) and the targeted Dll319 (blue). (B) ATAC peak regions around the sal genomic region with the sal740-targeted region highlighted in blue. (C–F) sal740 crispant phenotypes: Missing and reduced eyespots (C), split horn (D), thinner and discolored antenna compared to wild type (E), lost chevrons in the wing margin and ectopic vein in the Cu2 sector (F). (G) Table with the total number of open peaks associated with eyespot DE genes and number of peaks shared between eyespots and different tissues. (H) Venn diagram showing the number of open-chromatin regions shared between different tissue groups. (I) Schematic illustrating the hypothesis that eyespots evolved via cooption of an ancestral appendage GRN with genes (Dll and sal) in the GRN reusing the same CREs in both appendages and eyespot development.