Conserved roles of Distal-less and spalt in regulating butterfly melanic patterns

Simple spot patterns are proposed to be ancestral to more complex eyespot patterns found on the wings of butterflies. Two genes, Distal-less (Dll) and spalt (sal), are known to be involved in two separate functions in establishing nymphalid butterfly eyespots: in the differentiation of their central signaling cells, or foci, necessary for eyespot ring formation, and in scale melanisation. However, little is known about the functions of these genes in the development of more primitive spot patterns. Here, we study the roles of Dll and sal in the development of spots and other melanic wing patterns of the Indian cabbage white, Pieris canidia, a pierid butterfly. We examined the expression pattern of both genes in developing larval and pupal wings and explored their functions using the CRISPR/Cas9 system. In P. canidia, both Dll and sal are expressed in the tips and along the margin of pupal wings, areas of future scale melanisation. Additionally, sal alone, is expressed in the future black spot patterns. CRISPR knockouts of Dll and sal showed that both genes are required for promoting melanic scales in the areas where they are expressed. Both Dll and sal also function as repressors of the pteridine pigment pathway. We conclude that both genes share a conserved role in promoting scale melanisation, across pierid and nymphalid butterflies, but are unlikely to be involved in differentiating spot centers. The genetic and developmental mechanisms that set up the location of spots and eyespots are likely distinct and independently evolved in each lineage.

Eyespots have also served as a model system in the study of the origin of novel traits. A study examining the phylogenetic distribution of eyespots across the Lepidoptera proposed that eyespots were homologous across nymphalids because they likely had a single origin close to the base of the clade (Oliver et al., 2012). A subsequent study suggested that eyespots replaced spot patterns that were already present in specific wing sectors (Oliver et al., 2014). Examining the origin of eyespots may thus benefit from investigating the development of spots in more basal lineages of butterflies, such as pierids.
Spots in pierids and eyespots in nymphalids show differences and similarities in the expression of a few candidate genes at different stages of wing development, as well as in cellular arrangements. At the late larval stage, several genes required for eyespot center differentiation in nymphalids, including the transcription factors Distalless and spalt, are absent from the presumptive spot centers of Pieris rapae butterflies (Monteiro et al., 2006;Saenko et al., 2011;Oliver et al., 2012;Connahs et al., 2019;Murugesan et al., 2021). Furthermore, these two genes are hypothesized to be part of, or downstream, of a morphogen set up by a reaction-diffusion mechanism. This mechanism is proposed to differentiate the eyespot central cells, also called the focal cells, from the surrounding cells in each wing sector bordered by veins (Connahs et al., 2019). These focal cells are more densely packed and slightly raised from the wing plane relative to other epidermal cells (Iwasaki et al., 2017). In pierids, however, no such reaction-diffusion mechanism has been proposed for spot center differentiation, and the cells at the center of these spots resemble cells elsewhere on the wing. At early pupal stages of development, however, both Dll and sal proteins are present in future black scale cells in developing wings of nymphalids. In addition to eyespot focal cells, sal protein is also associated with melanic scale patterns in both the nymphalid Bicyclus anynana and in several pierids (Monteiro et al., 2006;Saenko et al., 2011;Oliver et al., 2012;Stoehr et al., 2013). Dll proteins, however, have only been associated with melanic patterns in nymphalids, such as B. anynana (Brakefield et al.1996;Monteiro et al. 2006). CRISPR/Cas9 disruptions of both Dll and sal, and transgenic experiments in B. anynana showed that both genes are required to produce melanic scales but in slightly different ways. Both Dll and sal are required for scale melanisation (Connahs et al., 2019). However, sal has an additional role of differentiating black scales by repressing another gene, optix, that is involved in the development of orange scales of nymphalid eyespots (Banerjee et al., 2021;Murugesan et al., 2021). Ectopic laserinduced expression of Dll in the early pupa wing of B. anynana showed that Dll is sufficient to activate the melanin biosynthesis pathway in the early pupal wing .
To date, no studies have managed to functionally identify the up-stream signals that activate Dll and sal in the black disc of an eyespot. These signals are likely candidate long-range morphogens produced by the central signalling cells, the foci. Several promising candidates have been studied over the years: i) members belonging to the Wnt signalling gene family like wingless (Monteiro et al., 2006;Özsu et al., 2017), and ii) a signal transducer of the BMP pathway, mothers against decapentaplegic (smad) were previously shown to be expressed in eyespot foci (Monteiro et al., 2006), and RNAi against Wg led to smaller eyespots (Özsu et al., 2017). This is particularly exciting as in Drosophila, Dll is a target gene of both dpp and wg signalling in the establishment of the proximo-distal (PD) axis during limb and wing development (Cohen et al., 1993;Estella et al., 2008). Likewise, the expression domains of spalt are also strictly defined by dpp signalling during the development of the Drosophila wing blade (de Celis et al., 1996;Barrio & de Celis, 2004).
Given that both Dll and sal proteins are associated with nymphalid eyespot foci, as well as other elements such a black scales and parafocal patterns along the wing margin of numerous nymphalid species (Monteiro et al., 2006;Zhang & Reed, 2016;Connahs et al., 2019;Reed et al., 2020), here we ask whether these transcription factors have a similar role in establishing melanic wing patterns in a species belonging to a more basal butterfly lineage, the Pieridae. We also ask whether the same candidate morphogens suspected of setting up eyespot centers in nymphalid butterfly larval wings might be setting up spot centers in pierid larval wings or defining the expression of sal or Dll in black scales, in pupal wings. To answer these questions, we used immunofluorescence and in-situ hybridisation to study the protein domains of Dll, sal, and armadillo (arm), a signal transducer of the Wnt signalling pathway, and the mRNA expression of dpp, a BMP ligand in Pieris canidia, the Indian cabbage white, B. anynana, and in a few other nymphalid species. We then studied the function of Dll and sal in P. canidia using CRISPR/Cas9 genome editing techniques. The goal of our study is to shed light on possible conserved or divergent mechanisms that differentiate spots and eyespots in two main lineages of butterflies.

Figure 1. Immunostainings of Distal-less (green) and Spalt (red) proteins in
Bicyclus anynana and Pieris canidia larval and pupal wings. A, A', B, B', C, C', D and D') Distal-less protein is present in late fifth instar larval and 24-26h pupal wing discs. A, A', C & C') In B. anynana larval and pupal wings, Dll is observed between veins as finger-like projections from the wing margin, ending with a discrete focus at the proximal tip of the fingers, that corresponds to the eyespot centres. In pupal stages of development, Dll becomes additionally observed in cells that correspond to the black scales of the eyespot pattern. B, B', D & D') In P. canidia, intervein finger-like projections of Dll protein are observed but with no discrete foci at the tips of the fingers. E, E', F, F', G, G', H & H'). Spalt protein is present in late fifth instar larval and 24-26h pupal wings discs. E, E', G, & G') In B. anynana, spalt protein is observed in eyespot foci during the larval stage. Like Dll, sal becomes additionally observed in the cells that map to the black scales in the eyespots during pupal wing development. F, F', H, & H') In P. canidia, there is no cluster of cells in the middle of the spot pattern that is expressing higher levels of sal proteins in larval wings, and sal is present in the cells that map to the black scales in spots in 24h pupal wings. I, I', J, J') Distal-less and sal proteins are also observed in cells that will become black scales located along the wing margin at both the wing tips and in the chevron patterns along the wing margin in P. canidia. Note the strong punctate nuclear staining of scale-building cells taken at 20x magnification. Scale bars for (C, D, G, and H -500µm)

Presence of Distal-less and Spalt proteins in B. anynana and P. canidia
We examined the distribution patterns of Dll proteins for both larval and 24h pupal wings of B. anynana and P. canidia (Fig 1). Larval wing discs of both species showed strong levels of Dll along the wing margin, and in midline finger-like projections from the margin, between developing veins (Fig 1A & 1A'). Levels of Dll protein were higher in a cluster of cells at the end of these fingers in B. anynana larval and pupal wings but not in P. canidia (Fig 1A & Fig 1C). In P. canidia larval and pupal wings, Dll levels continue to be high in mid-line projections in individual wing sectors (Fig 1B', D'). These findings are consistent with previous studies done in a closely related species, Pieris rapae (Reed & Serfas, 2004;Monteiro et al., 2006). A novel observation, however, is that Dll is also present in areas along the wing margin containing the black chevrons, and in the wing apex, mapping to the areas of melanised scales at these two locations (Fig 1I & 1I').
The presence of sal proteins was also examined for both species at the same time points in larval and pupal wings. In a similar manner to Dll, sal proteins were present in the eyespot foci in late larval wings of B. anynana (Fig 1E, E') but absent from spot centers in P. canidia (Fig, 1F, F'). In 24h pupal wings, sal was additionally observed in the scale-building cells that map to the black scales of an eyespot ( Fig  1G'). In P. canidia, sal was observed in the scale-building cells that map to all the densely melanised areas on the wing, including the black spots, the chevrons at the margin of the wing, and the apex of the wing (Fig 1H', 1J & 1J'). However, spot centers did not have elevated levels of sal, nor did these central cells appear distinct from surrounding spot cells, as they do in eyespots. These results are similar to those previously described for other pierids (Monteiro et al. 2006;Stoehr et al., 2013).
The protein localizations of Dll and sal in three other nymphalid species was like those observed in B. anynana. Dll and sal were present in the focal cells of future eyespots and spots (of Hypolimnas bolina jacintha) during the larval stage (Fig 2). This pattern persisted in 24h pupal wings but the two proteins were additionally present in a few surrounding scale-building cells that map to black pattern elements in an eyespot or spot. The simple white spots of Hypolimnas bolina are likely equivalent to central cells of an eyespot that have become reduced to a single ring/spot of color with just a few black cells around them.

Presence of armadillo (arm) and expression of decapentaplegic (dpp) in B. anynana and P. canidia
In the Drosophila wing margin, Dll is a downstream target of Wnt signalling, whereas in the center of the wing, sal is a target of Dpp signalling. To investigate whether Wnt and Dpp signalling could be upstream of the melanic patterns in P. canidia, we performed immunostainings targeting the protein armadillo (arm), a signal transducer of canonical Wnt signalling, and performed in situ hybridizations with a probe against dpp. We found arm present in the wing margin and in finger-like patterns from the wing margin in both B. anynana (as previously described in Connahs et al. 2019) and P. canidia (Fig 3A & 3B). However, arm is present in the eyespot centers in B. anynana but not in spot-like patterns in P. canidia during both larval and pupal stages (Fig 3A', 3B', 3C' & 3D'). This suggests that Wnt signaling is stable and active in B. anynana eyespot centers but not in P. canidia spot centers. In B. anynana, dpp is present in cells flanking the veins and also along the anterior-posterior (AP) boundary (as previously described in (Connahs et al., 2019;Banerjee & Monteiro, 2020b)), and later in eyespot centers in 16h pupal wings (Fig 3E & 3G). In P. canidia larval wings, dpp is expressed strongly along the veins and the border lacuna, parallel to the wing margin. No dpp was detected in areas mapping to the spot pattern in 17h pupal wings (  To test the function of Dll in spot development and melanisation, we targeted both exons 2 and 3 using the CRISPR/Cas9 system ( Fig 4A). Consistent with the immunostaining results for Dll, melanic wing patterns located along the wing tip and in chevrons along the wing margin were disrupted (Fig 4C). The black spots were not affected. In the affected areas, black scales were transformed into white scales. In two of the crispants, both ground and cover scales were missing from the affected regions (Fig 4D & 4F). To test the function of sal in spot development and in scale melanisation, we targeted exon 2 with the CRISPR/Cas9 system. The resulting mosaic phenotypes support a role for sal in scale melanisation in the spots, wing tips, and chevrons along the wing margin. We observed missing spots on both dorsal and ventral surfaces of forewings, fragmented spots, and a missing black wing marginal chevron in a single individual (Fig 5C, M8). Black scales in these areas were transformed into white scales. In addition, we saw one individual with less melanised scales (Fig 5C,  M9).
Individual scales of Dll and sal mutants and wild-type butterflies were then closely examined using Scanning Electron Microscopy (SEM) to look for any changes in scale structure that might be under the regulation of either gene. Wild-type black scales had little to no pigment granules present, in contrast to white scales ( Fig 6A). In both Dll and sal mutants, black scales that converted into white contained dense rows of ovoid-like pigment granules deposited along the cross-ribs (6B, C), resembling Wt white scales. The scales of the spalt crispant that displayed less melanised scales in the black spot region ( Fig 6D) were intermediate in color and in morphology -the windows were not completely open and remnants of upper lamina were observed along the crossribs as compared to Wt black scales (Fig. 6D). Pigment granules are also scattered within the scale lumen. Pigment granules are present in great numbers in white scales. B) SEM images of black scales and white scales removed from a Dll crispant. This crispant had greatly reduced spots on its hindwing. Scales that lost melanin pigments showed a morphology resembling that of wildtype white scales. C) SEM images of black and white scales removed from a spalt crispant. The SEM image labelled as spalt mKO showed a close-up view of a scale (originally black) removed from the CRISPR/Cas9 mosaic knockout area. Black scales converted into white scales with pigment granules, resembling those of wild-type white scales. D) SEM images of a spalt mutant that displayed an intermediate scale phenotype with less melanised scales in the black spot region. The morphology of these grey scales resembles that of WT black scales but windows of these scales were not fully opened and there remains residues of the upper lamina. Scale bars: 2μm

Discussion
Most of the work done on unravelling the genetics of butterfly wing patterns has remained overwhelmingly focused on nymphalid butterflies. However, we do not know if genes involved in the development of nymphalid-specific wing patterns, such as eyespots, also play a role in the development of other more primitive patterns in other butterfly lineages. Here, we present three main findings. First, we provide functional evidence for a deeply conserved role of two transcription factors, Distalless and spalt, as being required for melanin pigment production in both pierids and nymphalids. Second, we illustrate the possible dual role that both Dll and sal play in repressing the development of pterin pigment granules, located within individual scales of pierids butterflies, and in the activation of the melanin biosynthesis pathway. Lastly, we propose that unlike eyespot center differentiation, spot differentiation does not depend on the expression of either Dll or sal at the center of the pattern during the larval stages of development.
Previous research suggested that eyespots may have derived from simpler spot patterns (Oliver et al., 2014), but genes previously associated with eyespot patterns were not found in spots, with the exception of sal (Monteiro et al., 2006). Here we show that both Dll and sal have deeply conserved roles in regulating melanic patterns in lepidopteran wings, that predates the divergence of nymphalid and pierid butterflies (Fig 7). sal knockouts showed disrupted black spots and marginal markings, whereas Dll knockouts showed reductions of melanic patterns located along the wing tip and wing margins of both forewings and hindwings. We propose that the developmental mechanism underlying the evolution of melanic spots and eyespots is homologous in this context, with the same genes performing a similar function i.e., differentiating black scales, in both traits. We still do not know how melanin pathway genes are being regulated by either Dll or sal nor do we know the upstream signal(s) that both genes are responding to in lepidopterans. Future studies could try to unravel the regulatory connections between Dll and sal and downstream melanin biosynthesis genes.

Figure 7. The evolution of novel expression domains and functions of Distalless and spalt mapped onto a simplified phylogeny of butterflies. Expression of
Dll and sal associated with black scales in the pupal stage is mapped to a branch ancestral to the divergence of pierid and nymphalid butterflies. Because sal is expressed in late larval wing discs in areas corresponding to black wing spot elements in Idea leuconoe (Danainae; (Bhardwaj et al., 2020)), the origin of expression of that gene in the larval stage was mapped independently in the lineage leading to Danainae. In nymphalids, both Dll and sal likely gained derived roles in the differentiation of eyespot centers.
Interestingly, while both genes are required for the formation of black marginal chevrons and wing tips, sal alone is required for the development of wing spots in P. canidia. We postulate that Dll is likely working upstream of sal in areas where the two genes are co-expressed, but not in the black spot area of P. canidia. The regulatory interaction between sal and Dll in setting up melanic wing patterns has been inferred from mutants and from functional work in B. anynana. In wildtype B. anynana, both spalt and Dll are co-expressed in the black scales of an eyespot. A recent study showed that sal is a repressor of optix, a gene that is expressed in the orange scales of eyespots (Banerjee et al., 2021). Disruption of sal led to loss of the black scales and to their replacement by optix-expressing orange scales (Banerjee et al., 2021;Murugesan et al., 2021). Dll, on the other hand, is required for melanin pigment production in the black scales and in background brown wing scales (Connahs et al., 2019). Further, in Goldeneye B. anynana mutants, which had its black scales replaced by orange scales within the eyespot pattern, Dll proteins persisted while sal proteins were absent (Brunetti et al., 2001). This suggests that Dll is upstream of sal and that, at least for B. anynana, both Dll and sal are required for the development of black scales in eyespots. Recent work showed that Dll upregulates sal in the eyespot centers of B. anynana during the larval stage, whereas sal does not up-regulate Dll (Murugesan et al., 2021). This circuit might also be deployed during the pupal stage in the black disc region of nymphalid eyespots and in the tips and black chevrons of P. canidia wings, but additional work will be necessary to confirm this.
Similarly to a previously reported gene, optix (Zhang et al., 2017), spalt may be functioning as a 'switch' gene that represses the pterin biosynthesis pathway (white) while activating the melanin biosynthesis pathway (black). If spalt was purely an upstream activator of genes involved in melanin synthesis, we would expect to see scale morphology of mutant scales resembling those of the flanking black scales that were unaffected by the CRISPR/Cas9 knockout. However, when spalt mutant scales were examined using SEM, we observed numerous pigment granules densely arranged along the cross-ribs, closely resembling the structures found in wildtype white scales. White scales of pierid butterflies differ from those of other butterfly species in that the cross-ribs of each scale are attached to many spheroidal beads (Ghiradella et al., 1972;Stavenga et al., 2004;Wilts et al., 2017). These beads contain leucopterin, a class of heterocyclic pigment that absorbs exclusively in the ultraviolet range. When coupled with the strong light-scattering properties of these beads, leucopterin filled granules cause scales to appear white (Wilts et al., 2011). Our examination of the poorly melanised spot that was likely derived from a hypomorphic allele of sal, or perhaps a heterozygote crispant clone, suggests that intermediate scale colors (grey) and morphologies are possible (Fig 7D). This mutant suggests that intermediate levels of sal protein might be insufficient for complete downregulation of the pteridine pathway and for complete up-regulation of the melanin pathway.
Dll mutant clones displayed two phenotypes, loss of all scales and a change in scale color from black to white. The loss of both cover and ground scales, lends further support to butterfly scales being a derived form of a sensory bristle (Galant et al., 1998) that requires Dll for its development (Panganiban, 2000). This corroborates a previous finding by Connahs et al. (2019) whereby loss of scales was also observed in Dll crispants in B. anynana. The transformation of black to white scales may be connected to hypomorphic alleles of Dll, or perhaps to heterozygote crispant clones. It is tempting to speculate that like sal, Dll might also regulate two different pigment pathways simultaneously. However, it is more likely Dll was working upstream of sal in the wing marginal patterns and that the knocking out Dll resulted in the downregulation of sal, leading to the formation of ectopic pigment granules. This is also supported by the observation that knockouts of sal alone, in spots, produces the scale color switch phenotype.
Thus, we propose that prior to the split of nymphalid and pierids butterflies, both Dll and sal had roles in activating downstream melanin pigment enzymes to produce melanic scales. These transcription factors were regulating the development of melanic wing patterns on different areas of the wings, with sal required for black scale differentiation in some domains, and Dll required for the same process in other domains, perhaps performing this function via the up-regulation of sal.
However, after the divergence of nymphalids from pierids, nymphalid eyespot evolution relied on the novel larval expression of Dll and sal in the foci at the tips of intervein midlines. This novel expression may have taken place through a gradual increase of Dll expression that promotes a stable expression of Dll at the foci via a reaction-diffusion mechanism (Connahs et al. 2019 ; Fig 8). Higher Dll levels, in turn, may be dependent on Wnt and dpp signals which become anti-colocalized at late stages of eyespot focus differentiation, again via the same reaction-diffusion process (Connahs et al., 2019) (Fig 8). In P. canidia, armadillo protein patterns were quite similar to those observed in B. anynana but again, no arm foci were detected at the end of the intervein fingers (Fig 3B'). The dpp pattern was also different in P. canidia and was not anti-colocalized with the arm pattern (Fig 3F'). This suggests that a reaction-diffusion mechanism like that proposed for B. anynana is not taking place in P. canidia during mid larval development.
The mechanism that sets up spots and black discs of color around eyespots, during the pupal stage, may also be distinct. During early pupal stages, no discernible arm nor dpp signals were observed in spot centers (Fig 3D' & 3H') as they were in eyespot centers (Fig. 3C', 3G'). It is possible that sal in P. canidia may be responding to a gradient of BMP ligands such as dpp that is emanating from the wing margin. High levels of dpp expression were present along the wing margin of P. canidia larval wings (Fig 3F'), but not in B. anynana (Fig 3E'). Thus, we speculate that the role of Dll and sal in establishing nymphalid eyespot foci is novel and derived as compared to pierid spot development.
This derived role of Dll and sal as eyespot center organisers is supported by the fact that in late larval wings, the expression of both Dll and sal in the presumptive eyespot centers in nymphalid species is essential for eyespot development (Zhang & Reed, 2016;Connahs et al., 2019;Murugesan et al., 2021). Knockouts of Dll and sal in B. anynana that affected cells located in the eyespot center always led to the complete disappearance of an eyespot (Connahs et al., 2019;Murugesan et al., 2021). The expression of both genes, however, is absent from spot centers in pierid species during the larval stage (Reed & Serfas, 2004;Stoehr et al., 2013). Correspondingly, when scale cells located in the spot center were affected in P. canidia spalt knockout mutants, we did not observe entire spots disappearing. Instead, scattered areas of the spot retained melanised scales (Fig 5C).
The developmental mechanism of pierid spot differentiation is not yet fully understood. Pierid spots may rely on differentiated cells at their center to signal to surrounding cells to differentiate the complete spot pattern, as previously proposed (Stoehr et al. 2013). Alternatively, spots may be fragments of an anterior-posterior banding system that relies instead on both activator or inhibitory signals spreading from the wing margin and wing veins (Monteiro et al. 2006). Regardless of the exact mechanism of spot development, our current experiments show that spots do not rely on Dll and sal being expressed at their center during the larval stages to differentiate. In late larval wing discs of B. anynana, both Dll (green) and sal (red) are co-expressed at high levels in the center of eyespots. However, in late larval wing discs of P. canidia, Dll and sal are not expressed in spot centers. Only Dll is expressed in mid-line fingers encroaching inwards from the wing margin. Eyespot centers in B. anynana are likely established through a reaction diffusion mechanism involving Wnt and BMP signaling (Connahs et al., 2019). The absence of arm proteins and dpp expression in P. canidia spot centers suggests that spots may not develop through the same mechanism. In nymphalid eyespots, Dll and sal respond to signals emanating from the foci. However, in early pupal stages, both arm and dpp are absent in spot centers. There may be no central signaling cells that are present in spot patterns that are activating downstream genes (i.e sal). An alternative explanation would be that sal may be responding to a gradient of BMP ligands and a band of an unidentified inhibitory molecule (orange) secreted from the wing margin to give rise to the black markings of P. canidia during late pupal development.

Conclusion
Simpler monochromatic spot-like patterns have been proposed to be a primitive form of more complex eyespot patterns. In this study, we tested the function of two transcription factors expressed in nymphalid eyespots, Dll and sal, in a basal butterfly lineage with primitive spots and other melanic patterns on its wings, P. canidia. Our work suggests that each transcription factor is required for activating distinct melanic elements in this species, including the spots, but these genes have likely no role in positioning spots on the wing. This suggests that the mechanism of melanin pathway regulation is shared between both spots and eyespots but the mechanism of setting up the position of these two pattern elements might have independently evolved in each lineage. Future work involving functional knockouts of other candidate genes will be able to shed more light on spot development and its connections with eyespot development.

Animals
Pieris canidia used in this study were the descendants of wild-caught individuals from Singapore. Larvae were fed on potted Brassica chinensis var. parachinensis plants and adults on 10% sucrose solution. Bicyclus anynana larvae were fed on potted corn and adults on mashed banana. Both species were reared at 27°C and at 60% humidity under a 12:12h light/dark photoperiod. All other species of butterflies used for comparative immunostainings work were reared at Entopia, a butterfly farm (Penang, Malaysia) under outdoor conditions.

Immunostainings
Immunostainings were performed on 5 th instar larval wings and 16-30h pupal wings dissected based on a protocol previously described by (Banerjee & Monteiro, 2020a) in 1xPBS at room temperature. Wings were fixed with 4% formaldehyde for 30 mins, washed with 1xPBS for four times at 10 mins, and transferred to 2ml tubes filled with block buffer for blocking at 4°C for up to several months to reduce non-specific binding of the antibodies. Wing discs were then incubated in primary antibodies against Distal-less (1:200, mouse, a gift from Grace Boekhoff-Falk), and spalt (1:10000, guinea-pig Sal GP66.1) overnight at 4°C, washed with multiple rounds of wash buffer, and stained in secondary antibodies anti-mouse AF488 (Invitrogen, #A28175) and anti-guinea pig AF555 (Invitrogen, #A-21435) at a concentration of 1:500. Stained wings were then washed with multiple rounds of wash buffer, away from light, and mounted on glass slides with an in-house mounting media. Images of the wings were taken with an Olympus FV3000 Confocal Laser Scanning Microscope. All buffer compositions are summarised in Table S2.

Whole-mount in-situ hybridisation
In-situ hybridisations were performed on early to late 5 th instar larval wings and 16-18h pupal wings dissected in 1xPBS at room temperature to prevent the crumpling of wings. The wings were fixed with 4% formaldehyde in PBST for 30 mins, digested with 1.25μl of Proteinase-K in 1ml of 1x PBST for 5 mins on ice. The digestion reaction was stopped with a 2 mg/ml glycine solution in 1X PBST and followed with 3 washes of 1X PBST. Larval wings were removed from ice briefly for 5 mins and placed right back on ice to induce 'puffing' of the peripodial membrane for easier removal of the membrane using fine tip forceps. After removing the peripodial membrane, the wings were transferred to increasing concentrations of prehybridisation buffer in 1X PBST and incubated at 60°C for at least 1h in prehybridisation buffer. Incubated wings were hybridised at 60°C with the probe (100ng/ml) in a hybridisation buffer for 16-24h. The next day, after incubation with the riboprobe, wings were washed with pre-hybridisation buffer for 5 x 10mins at 60°C. The wings were then brought back to room temperature and transferred to 1x PBST gradually. 1x PBST was used to wash the wings for 2 x 5mins, and wings were subsequently transferred for blocking for 1h. Anti-digoxygenin was diluted in block buffer at a ratio of 1:3000 for incubation with the wings for 1h. Once completed, the wings were washed with block buffer for 5 x 5mins on a rotary shaker and transferred to an alkaline phosphatase buffer containing NBT-BCIP. Wings were left to incubate in the dark to develop color signal to the required intensity. A Leica DMS1000 microscope was used to image the stained wings. All buffer compositions are summarised in Table S3.

CRISPR-Cas9
Knock-outs of the genes Dll and sal in P. canidia, were generated using the methods outlined in a previously published protocol (Banerjee & Monteiro, 2018). Single guide RNAs (sgRNAs) targeting the genomic regions of exons 2 and 3 of Dll and exon 2 of sal were designed using the webtool CHOPCHOP (Labun et al., 2019). For the gene sal, a total of 575 embryos were injected with a mixture containing 300ng/µl of sgRNA (one guide) and 600ng/µl of Cas9 protein (NEB, M0641) while for Dll, 357 embryos were injected with a mixture containing 100ng/µl of sgRNAs (2 guides) and 300ng/µl of Cas9 protein (Table S4).
Wild-type P. canidia laid eggs on a piece of parafilm that was wrapped around a small container that had its top covered with a piece of fresh cabbage leaf. The container was placed within the butterfly cage for up to 6 hours at a time to maximise the number of eggs collected. The parafilm and leaf were then removed from the container and transferred to a petri-dish for injection with the Cas9 injection mixture. Pieces of moist cotton wool were placed in each petri-dish post injection to avoid desiccation of injected eggs. Hatchlings were then directly transferred to Brassica sp plants and reared to adult eclosion. Upon emergence, the butterflies were frozen immediately in separate glassine envelopes and examined under the microscope for asymmetrical (left-right wing) phenotypic defects. Genomic DNA was isolated from the affected mosaic areas from CRISPR mutants, and indels were identified through Sanger and NGS sequencing.