Frizzled2 receives WntA signaling during butterfly wing pattern formation

The Wnt-family ligand WntA instructs color pattern formation during butterfly wing development and establishes the position and limits of color fields across Nymphalidae (Martin and Reed, 2014; Mazo-Vargas et al., 2017). Several lines of evidence pinpoint WntA signaling as a driver of color pattern evolution. First, regulatory shifts of WntA spatial expression have repeatedly driven adaptive variation, as evidenced by the identification to date of more than 20 WntA alleles associated with wing mimicry in variable species of the Heliconius, Limenitis and Elymnias butterfly radiations (Gallant et al., 2014; Martin et al., 2012; Morris et al., 2020; Ruttenberg et al., 2021; Van Belleghem et al., 2017), making WntA a favored genomic locus for wing pattern variation (Martin and Courtier-Orgogozo, 2017; Van Belleghem et al., 2021). In these systems, WntA alleles explain pattern differences between populations and morphs of the same species. Additionally, mapping, expression and loss-of-function experiments show that WntA repeatedly shaped the spectacular ressemblance of Heliconius co-mimics that converged onto identical patterns, in spite of historical drift in their patterning systems (Concha et al., 2019; Van Belleghem et al., 2020). Although WntA regulation and function appear largely conserved across butterflies that follow a bauplan-like organization of wing patterns known as the Nymphalid Ground Plan (NGP), it has also evolved atypical modes of expression associated with novel pattern elements, such as in the control of white-orange intravenous markings in the monarch butterfly (Mazo-Vargas et al., 2017, 2022). A high density of open-chromatin elements at this locus, each involved in the regulatory control of pattern phenotypes, may underlie the apparent evolvability of WntA expression in nymphalids (Mazo-Vargas et al., 2022).

In addition to the insights it provides on genetic evolution, WntA wing patterning also brings the opportunity to study morphogenetic signaling in a simplified, two-dimensional developmental system, where color states provide a direct phenotypic readout of positional information during development. Although WntA evolved at the base of Metazoa (Darras et al., 2018; Kraus et al., 2016; Somorjai et al., 2018), it was lost in vertebrates as well as in Cyclorrhapha, a fly lineage that includes Drosophila (Hanly et al., 2021), making it a particularly understudied Wnt ligand. CRISPR mosaic knock-outs of Wntless (Wls) and porcupine (por) phenocopy WntA knockouts, which implies that WntA is a secreted lipid-modified ligand that is processed via the classical Wnt secretory pathway (Hanly et al., 2021). In addition to its role in Wnt maturation and secretion, the fatty acid thumb added by Por-dependent acylation is essential for steric binding of Wnt ligands into the cysteine rich domains (CRD) of receptors of the Frizzled family (Alvarez-Rodrigo et al., 2023; Hoppler and Moon, 2014; Nile and Hannoush, 2019; Povelones and Nusse, 2005). Studies of Wingless (Wg/Wnt1) in Drosophila development have shown that Frizzled1 and Frizzled2 (Fz1, Fz2) mediate short-range Wnt reception with different efficiencies, and that they are transcriptionally repressed by Wg signaling (Bhanot et al., 1999; Chen and Struhl, 1999; Moline et al., 2000). In the Drosophila wing disc, Fz1 is the sole Frizzled receptor necessary for the Wnt-independent Fz/planar cell polarity (Fz/PCP) pathway (Ewen-Campen et al., 2020); Fz2 maintains long-range activation of the Wnt pathway in cells that are beyond the reach of extracellular Wg (Chaudhary et al., 2019); Fz3 is a decoy receptor of Wg that attenuates signaling (Sato et al., 1999; Schilling et al., 2014); and Fz4 has no known function in wings (Ewen-Campen et al., 2020). In butterflies, a RNAseq study showed that the butterfly homolog of frizzled2 (fz2) is differentially expressed across color regions during early pupal wing development (Hanly et al., 2019), and in situ hybridizations of the four frizzled gene homologs showed overlap with the expression of Wnt ligand genes (Banerjee et al., 2023).

To further shed light on the mechanisms of WntA reception, we have investigated functions of the Frizzled family seven-transmembrane domain receptors during butterfly wing development. Using CRISPR-mediated mutagenesis across several butterfly species of the Nymphalidae family, we show that fz2 is the only Frizzled-family receptor required for the formation of WntA-dependent patterns. In contrast, frizzled1 mosaic knockouts indicate it mediates a conserved Fz-PCP pathway involved in scale orientation. Finally, we show that loss of function of frizzled3 (fz3) and frizzled4 (fz4) play negative roles on WntA-independent Wnt signaling during wing development.

Frizzled receptors are seven-transmembrane-domain proteins belonging to the G protein-coupled receptor family. Most non-vertebrate genomes encode four Frizzled orthology groups that originated before the Bilateria/Cnidaria split, including the four receptors annotated in Drosophila (Janssen et al., 2015; Schenkelaars et al., 2015). We recovered orthologues of these four genes in nymphalid butterfly genomes, hereafter numbered by orthology groups and abbreviated fz1, fz2, fz3 and fz4 (Fig. S1A, Table S6). Both fz1 and fz2 have a conserved KTxxxW motif (Fig. S1B), required for the recruitment of Dishevelled (Dsh), which relays two pathways in Drosophila: Fz1/2-dependent canonical Wnt signaling and Fz1/PCP signaling (Ewen-Campen et al., 2020; Gao and Chen, 2010; Mieszczanek et al., 2022; Tauriello et al., 2012). In addition, the KTxxxW is invariable between Fz2 orthologues between Drosophila and Lepidoptera, showing a conserved KTLESW peptidic chain with a predicted glutamyl-endopeptidase site, necessary for the Frizzled nuclear import (FNI) pathway in Drosophila (Mathew et al., 2005; Restrepo et al., 2022). In contrast, Fz4 lost its KTxxxW domain in both Lepidoptera and Diptera, and Fz3 shows a derived Thr>Ser substitution in position 2 of unknown functional relevance. Overall, amino acid sequences from the cytoplasmic domains suggest that nymphalid Frizzled receptors may function similarly to the dipteran model Drosophila, with Fz1/2 signaling via Dishevelled recruitment, Fz2 potentially subject to the FNI Wnt pathway and Fz3/4 playing alternate roles. In situ hybridization of the four fz orthologues of V. cardui show these four genes are expressed in wing imaginal disks from fifth instar larvae, but without an apparent association with color patterns, unlike WntA at this stage (Fig. S1C-G). These data suggest functional divergence in the use of Frizzled receptors in butterfly wing development.

A previous RNAseq analysis in Heliconius pupal wings suggested spatial complementarity of fz2 expression levels with WntA (Hanly et al., 2019). We thus sought to test the function of fz2, and used CRISPR somatic mutagenesis to generate mosaic knockouts (mKOs) of fz2 in six species previously targeted for WntA mutagenesis (Concha et al., 2019; Mazo-Vargas et al., 2017, 2022).

In all instances, fz2 mKOs phenocopied the effects of WntA loss of function, with the removal or shifts of specific patterns on all four wing surfaces. Specifically, the phenotypes observed in V. cardui and J. coenia show (1) complete loss of the central symmetry system (CSS) – a conserved pattern element that expresses WntA (Martin and Reed, 2014; Mazo-Vargas et al., 2017); (2) shifts in peripheral patterns; and (3) specific reductions or loss of other WntA-expressing patterns, such as the V. cardui forewing eyespots and the J. coenia Basalis orange-black pattern (Fig. 1A,C). In A. incarnata, fz2 crispants reproduce all losses of silver spots known to express WntA in fifth instar wing disks, as well as an expansion of silver domains in the anterior hindwing, a phenomenon also observed in WntA mKOs (Martin and Reed, 2014; Mazo-Vargas et al., 2017). We knocked out fz2 in a pair of Heliconius butterfly co-mimics that convergently evolved a red forewing band. The fz2 mKOs recapitulated the species-specific effects of WntA removal (Concha et al., 2019; McMillan et al., 2020): a subtle proximal extension of the red band revealing a cryptic yellow dot in Heliconius melpomene rosina, versus an extensive proximal extension in Heliconius erato demophoon. Additionally, we repeated WntA mKOs in the monarch butterfly Danaus plexippus (Fig. S2), confirming that WntA evolved a role in antagonizing the formation of white patterns that outline black vein markings (Mazo-Vargas et al., 2017). Loss of function of fz2 reproduced these effects, with white patterns expanding in these fields.

As observed in WntA mKOs, fz2 mKOs generated by injection of a Cas9/sgRNA duplex in syncytial embryos resulted in healthy adult G₀ butterflies without detectable deleterious impacts on wing development (Figs S3-S6, Table S1). In particular, we did not observe the high rates of missing wings observed with por and wls mKOs (Hanly et al., 2021). This suggests that fz2 alone is dispensable for mediating Wnt functions that are essential for normal wing growth, which parallels perturbation assays in Tribolium embryos and Drosophila wings, where fz2 loss-of-function effects are phenotypically silent due to functional redundancy with fz1 (Beermann et al., 2011; Ewen-Campen et al., 2020). However, fz2 crispants showed highly efficient and penetrant wing color pattern phenotypes in all tested species (Fig. 1).

A closer look at WntA and fz2 crispants reveals interesting differences in the shape and aspect of mosaic clones (Fig. 1G-J). WntA clones always appear rounded: this is consistent with the expected non-cell-autonomous effects of a paracrine signaling ligand. In contrast, fz2 clones show elongated jagged shapes in the proximo-distal direction, consistent with expectations for cell-autonomous effects, and reminiscent of the shape of butterfly wing clones obtained with mKOs of selector transcription factors and pigment pathway genes (Livraghi et al., 2018; Tunström et al., 2023; Westerman et al., 2018; Zhang et al., 2017a,b), which we would also expect to be cell-autonomous. Taken together, these data show that Fz2 reception is necessary for mediating WntA patterning, and that WntA signals reach several cell diameters away from their source.

Together, the WntA ligand and Fz2 receptor form a paracrine signaling axis required for the induction of CSS patterns, as well as for the proper positioning of pattern elements close to the wing margin, such as the distal Parafocal elements (dPf), Marginal Band System (MBS) and the WntA-dependent forewing Border Ocelli (fBOc) in V. cardui (Fig. 2A, Figs S7 and S8A-C). In situ hybridization in a number of species had previously shown that WntA expression already prefigures pattern elements in late larval wings (Jiggins et al., 2017; Martin and Reed, 2014; Mazo-Vargas et al., 2017). However, current evidence suggests that extracellular WntA signaling mostly occurs in pupal rather than in larval wings. Indeed, injections of heparin, an extracellular Wnt interactor that results in Wnt-dependent pattern expansions (Martin and Reed, 2014; Mazo-Vargas et al., 2017; Sourakov, 2018; Sourakov and Shirai, 2020), are effective in the first 15% of pupal development – typically within 24 h after pupa formation (APF) at optimal rearing temperatures. Furthermore, scale organ precursor cell differentiation occurs around 13-15% (Dinwiddie et al., 2014), and the competency of tissue grafts to induce ectopic patterns (Nijhout, 1991) or developmental studies of eyespot formation (Brunetti et al., 2001; Monteiro, 2015), all imply the existence of paracrine signals active during early pupal development. We thus hypothesized that extracellular WntA signaling function actually occurs after the onset of metamorphosis.

We examined WntA and fz2 mRNA expression in early pupal wings in order to gain further insights into WntA/Fz2 patterning (Fig. 2, Fig. S8). Consistent with an instructive role for pattern induction in pupal wings, WntA expression precisely prefigured the position and shape of WntA-dependent pattern elements, including delineations of the adult patterns and graded expression levels that prefigure their textural details in a more highly correlated way than in larvae, in all species examined. (e.g. Fig. 2D, Fig. S8F). On the other hand, fz2 is generally expressed in an anti-correlated pattern with WntA. In A. incarnata, as in V. cardui, we found complementary expression of fz2 and WntA in pupal stages in wild-type wings (Fig. 2F-K). In summary, fz2 showed complementary expression to WntA, with low staining in WntA-positive regions across all the wild-type tissues we assayed. This is indicative of WntA-dependent repression of fz2 expression in the nymphalids, a phenomenon observed with Wg and fz2 in larval D. melanogaster wing discs (Cadigan et al., 1998; Zhang and Carthew, 1998).

In order to test whether WntA represses expression of fz2, we performed fz2 mRNA in situ hybridization in WntA G₀ mosaic knockouts of V. cardui and A. incarnata. In both species, patterns of fz2 expression were markedly affected by WntA KO, and all WntA-expressing patterns lost the local repression of fz2 (Fig. 2E-H, Fig. S8H,I,L,M). The local repression of fz2 requires a functional WntA. Thus, although Fz2 is a positive regulator of WntA signaling, fz2 expression is also repressed by WntA, and this mechanism is likely to enforce a transcriptional negative-feedback loop on this pathway.

The mRNA signal for fz2 remained low in D₁ and D₂ Discalis patterns in the WntA-deficient wings of the two species we assayed (Fig. 2H,P), implying that a WntA-independent signal repressed fz2 in these patterns. In parallel, a closer examination of these phenotypes reveal that fz2 KOs consistently induced black pattern expansions in V. cardui forewing D₁ and D₂ (Fig. S9A). This effect is not detected in WntA KOs, implying that Fz2 receives additional patterning signals from other ligands in this part of the wing. In the V. cardui forewing, this may include Wg, as D₁ and D₂ express both wg and WntA (Fig. S9A-G). We infer that Fz2 mediates the patterning functions of Wg or another Wnt, and may be under additional negative transcriptional feedback from this input, similar to WntA in other color patterns, or to the Wg/Fz2 pair in Drosophila (Cadigan et al., 1998; Chaudhary et al., 2019; Schilling et al., 2014). On the other hand, in the two nymphalids A. incarnata and J. coenia (Figs S9H,I and S10), wg-positive patterns (Martin and Reed, 2010, 2014), such as the D₂ element, are unaffected by the fz2 KO, suggesting that if Fz2 is necessary for Wg color patterning, it must be redundant with another receptor in some contexts or species.

In addition to the transcriptional negative feedback on fz2, our data also suggest that WntA expression is under the control of positive feedback. This effect is most visible in the CSS patterns of V. cardui, where the spatial complexity of WntA pupal expression disappears when WntA signaling is removed. For example, WntA expression in wild-type hindwings prefigures textural details of the CSS, with sharp outer boundaries, and a complex composition of undulating patterns in between (Fig. 2D, bottom row). These features of WntA expression are lost upon WntA KO (Fig. 2H, bottom row), meaning that a positive feedback of WntA signaling likely refines the sharply delineated and undulatory aspect of the ventral hindwing CSS in V. cardui. Similarly, in the wild-type posterior forewing, WntA is expressed in wide blocks that mark the presumptive orange patterns (Fig. S8G). In WntA-deficient pupal forewings, i.e. in the absence of a WntA positive-feedback loop, WntA expression is depleted from the inside of these patterns but remains high in two continuous lateral lines (arrowheads in Fig. S8G).

Elsewhere, WntA in situ hybridization shows reduced staining following WntA mKO in V. cardui and A. incarnata marginal patterns, as well as in the A. incarnata CSS (Fig. 2, Fig. S8), indicating the existence of feedback mechanisms. As an exception, WntA expression in the V. cardui forewing eyespots appears unchanged upon mKO and may be largely WntA independent (Fig. 2D,H, upper rows). In summary, the shifts in WntA expression observed in loss-of-function crispants indicate that positive feedback is necessary for the elaboration of CSS features in V. cardui wings. The importance of this feedback may vary across regions and species.

The patterns closest to the wing margin consist of stripe patterns that form the MBS as well as the dPf, a pattern often taking the shape of individuated arcs or chevrons between each wing vein compartment (Nijhout, 2017; Otaki, 2012; Taira et al., 2015). WntA is strongly expressed at the vein tips of V. cardui pupal hindwings (Fig. 3A). Transcripts of fz2 are repressed across a larger domain across the periphery, suggesting local repression by negative feedback from marginal signals (Fig. 3B). WntA mKOs leave border ocelli (eyespots) and MBS patterns unaffected in this system (Hanly et al., 2021; Mazo-Vargas et al., 2017), but result in a visible shift and reversal of the dPf blue chevron shapes (Fig. 3C,D). These phenotypes are reproduced in fz2 crispant butterflies, with dPf elements showing distalization and thickening (Fig. 3E). Unlike WntA-dependent continuous clonal effects of dPf elements, clonal mosaics of fz2 KO result in apparent cell-autonomous gains of blue scales (Fig. 3D′,E′). Similar effects are visible in monarchs (Fig. 1H), where WntA and fz2 are required for establishing the black contours of marginal white spots. Together, these expression and loss-of-function data suggest that WntA diffuses away from the vein tips and requires fz2 for proper positioning and shaping of the dPf elements.

The wing patterns of A. incarnata present an interesting conundrum on the ventral hindwing. As described above, WntA and Fz2 are both necessary for the induction of most CSS and peripheral silver patterns (Fig. 3F-J′). However, knockouts of both genes result in an expansion and gain of silver spots (Fig. 1E) – for example, the large silver element situated in the anterior compartment (anterior to the M₂ vein) is normally split into two disjunct silver spots (dotted lines in Fig. 3M), whereas these spots fuse and extend basally in WntA and fz2 KOs. Similarly, the M₁-M₂ region (above the green line in Fig. 3M) normally lacks silver elements, but shows an ectopic silver pattern in these crispants. Why are there two opposed effects of WntA on silver spots? Surprisingly, we found that WntA expression is absent from the anterior silver spots, and is instead expressed in their brown contouring regions in pupal wings (Fig. 3K-N). WntA signaling is notably active in the domain that splits the M₁-M₂ region into two spots, as observed by strong WntA and low fz2 mRNA patterns, and decreased signaling results in silver spot extensions. In other words, although WntA/Fz2 generally activates silver and black outlines in A. incarnata ventral surfaces, it can also locally shift to a complementary expression mode that induces the brown contours of similar patterns instead. These data establish WntA/Fz2 signaling as a versatile developmental tool for setting boundaries in a context-dependent fashion, with the capacity to induce elements that are variably perceived as ‘pattern’ or ‘background’ (Nijhout and Wray, 1988).

Next, we assayed the other three Frizzled receptors for effects on development and wing patterning. First we generated fz1 mKO across a large sample of J. coenia embryos (n=5273), varying injection time and CRISPR duplex concentration, and testing two different sgRNA targets. We did not observe pattern effects in these butterflies, but instead, crispants showed disorganized arrays of scales that were most conspicuous on the wings and antennae (Fig. 4A-C′, Fig. S11). These effects are clonal (i.e. with visible boundaries between mutant and wild-type scale fields), and reminiscent of Drosophila PCP phenotypes (Adler, 2012; Lawrence and Casal, 2018): similarly to wing hairs in PCP mutant flies, fz1-deficient butterfly scales are globally misoriented, forming streaks of scales that deviate from the normal proximal-to-distal orientation. Mutant scales also showed upright positions relative to the wing plane, indicative of a loss of the asymmetry normally observed in the scale protrusion axis during early pupal development (Dinwiddie et al., 2014). Mosaic knockouts of fz1 in V. cardui did not generate color pattern or scale defects among emerged adult butterflies (Table S1). However, in contrast to J. coenia, some pupae from this experiment showed wing growth defects and failed to form a cuticular seal with the abdomen, resulting in pupal death or failure of imago emergence due to desiccation (Fig. S12). Additionally, in the J. coenia crispants that showed PCP phenotypes on one side only (Fig. S13), mutant forewings were shorter and rounder than their wild-type counterparts (Wilcoxon signed rank test, P<0.05). This suggests that although fz1 deficiencies can also affect wing growth in J. coenia, pupal viability is less sensitive to perturbation in this species than in V. cardui.

To further verify its link to scale polarity, we used siRNA electroporation to generate knockdowns of fz1 at the pupal stage (Fujiwara and Nishikawa, 2016; VanKuren et al., 2023), thus bypassing the wing growth phase from larval stages. Electroporated siRNA knockdowns consistently recapitulated PCP phenotypes in J. coenia and A. incarnata (Fig. 4C,G-J). In each butterfly, these effects were specific to the electroporated surface, and were not observed in sham-treated controls targeting a GFP sequence (Fig. S14). In J. coenia, we also reproduced hindwing-to-forewing pattern homeoses by knocking down the Ultrabithorax Hox gene (Tendolkar et al., 2021), as well as by local removal of the CSS using fz2 knockdowns (Fig. S15). These two positive controls both reproduced expected effects and show that the PCP-like effect is not induced as an artifact of electroporation. Thus, the combination of CRISPR KOs and siRNA knockdowns indicate the conservation of a Fz1-PCP pathway required for the proper orientation of epithelial features, similar to Drosophila (Ewen-Campen et al., 2020; Joyce et al., 2020).

Next, we found that mosaic knockouts of both fz3 and fz4 in V. cardui, J. coenia and A. incarnata generated similar effects on vein and color patterning (Fig. 5A-C, Figs S16-S20). Most notably, disruption of fz3 and fz4 results in ectopic veins across all wing surfaces, sometimes forming complete veins and sometimes forming short spurs or loops (Fig. 5D), showing that these two genes have non-redundant roles in the repression of vein formation. As the lacunae, which prefigure the veins, form during larval wing development, this vein inhibition function is likely to occur before metamorphosis (Banerjee and Monteiro, 2020; Nijhout, 1991; Reed and Gilbert, 2004; Reed et al., 2007). Consistent with this idea, fz3 and fz4 are expressed in the intervein regions of V. cardui larval wing discs (Fig. S1). In places where a supernumerary vein interacted with the eyespot field in V. cardui and J. coenia, we observed alterations to the shape and size of eyespots, including the formation of additional eyespot foci (Fig. 5E, Figs S16-S19). Veins act as important signaling centers and domain boundaries (Connahs et al., 2019; Nijhout, 2017), and ectopic foci may thus be a secondary consequence of ectopic veins, rather than a direct interaction between Fz3 and/or Fz4 and eyespot-inducing morphogens. Likewise, disruptions of CSS elements were accompanied by ectopic venation (i.e. Fig. 5A″), suggesting a collateral effect of venation on field boundary conditions rather than on the WntA morphogen itself.

We further examined possible color patterning functions of fz3 and fz4, with their confounding effects on venation in mind. In J. coenia and A. incarnata, Discalis elements are bicolor symmetric systems and express Wnt ligand genes such as wg (Figs S4I and S5B), Wnt6 and Wnt10 (Martin and Reed, 2014), and WntA, although the last only contributes to the patterning of A. incarnata D₁ (Mazo-Vargas et al., 2017). Both fz3 and fz4 showed expansions of Discalis elements (Fig. 5F-G). These effects of fz3 perturbation are again likely to derive from ectopic venation, as they are accompanied by visible epithelial thickening in these regions or as they formed close to other veins that may have formed spurs. Interestingly, fz4 mKOs had distinct effects that suggest a direct modulation of pattern-inducing morphogenetic signals, rather than a collateral effect of ectopic venation. In fz4 crispants, D₁ and D₂ had expanded inner and outer elements on the ventral side, whereas on the dorsal side, only the outer black element was expanded (Fig. 5F,G). Unlike fz3 phenotypes, these expansions appear broader and always include a graded, fuzzy effect at their borders, suggesting a possible non-cell-autonomous effect on extracellular signaling.

Observation of the wing peripheral regions of J. coenia and V. cardui provide further insights (Fig. 6), as they juxtapose a sequence of multicolor patterns (dPf and MBS) with possible contributions of WntA and Wg, Wnt6 or Wnt10 (Martin and Reed, 2014; Mazo-Vargas et al., 2017). Perturbation of fz4 yields specific effects on MBS patterning, seen as a loss of the sub-marginal band in J. coenia and an MBS distal shift or reduction in V. cardui, as well as a distalization of eyespots observed in V. cardui only (Fig. 6A,B,D,E,G,H). In contrast, the color pattern arrangements of fz3 crispants are similar to wild type in complexity (i.e. similar sequences and sizes of color layers), but instead show distortions that indicate two kinds of tissue distortions: ectopic venation and ectopic wing margins. These secondary wing margins are visible as distal outgrowths that feature the sequence of scale types normally found at the primary margin, including linear streaks of elongated margin scales (Fig. 6C,F,I). Together with the strong expression of fz3 in the peripheral tissue of larval wing disks (Fig. S1F), these results suggest that fz3 is necessary for proper margin specification.

Here, we showed that Fz2 is the sole Frizzled-family member necessary for WntA reception in butterflies. Importantly, both WntA and fz2 knockouts result in viable embryos, larvae and adults, without deleterious effects on survival or morphology, in spite of complex expression patterns (Hanly et al., 2021; Janssen et al., 2015). These results are in agreement with loss-of-function experiments in Tribolium, where knockdowns of WntA and fz2 alone have no effect on embryogenesis (Beermann et al., 2011; Bolognesi et al., 2008), or in Drosophila, where null-homozygous fz2 mutant adults are sterile but viable (Chen and Struhl, 1999; Cohen et al., 2002).

In addition, we gained insight on how WntA-dependent elements establish symmetric and complex arrays of color scales, such as the undulating brown patterns of the V. cardui CSS or the A. incarnata silver spots and their black contours (Figs 2,3). We propose that WntA acts as a morphogen, inducing scale cell differentiation in a graded and concentration-dependent manner. First, WntA acts as a paracrine factor, as indicated by the non-cell-autonomous nature of WntA crispant clones, with rounded boundaries that contrast with the expected cell-autonomous feature of receptor-deficient clones in all species (Figs 1G-J and 3). Second, mRNA staining intensities prefigure the differentiation of scales into multiple different types. For example, we see spatially intricate expression signals of WntA in the V. cardui hindwing ventral CSS, where wild-type patterns display symmetric arrangements of beige, black, white and brown scales that disappear in WntA KOs (Fig. 3C-E). In A. incarnata WntA KOs, the non-cell-autonomous clonal effects we observed, such as the contraction of black-contoured silver spots into black spots, indicate a lower signaling level (Fig. 1G,I). This contraction is not observed in the corresponding patterns carrying cell-autonomous fz2 mKO clones, where dislocated patterns display both silver and black scales. These distinct phenotypes are consistent with receiving cells responding to specific thresholds in extracellular levels of WntA, or to duration of exposure, and rule out sequential induction models involving intermediate signals (Iwata and Otaki, 2019). Together, the ability of WntA to act at a distance and to cause differentiation into multiple concentration-dependent cell types indicate that it is a bona fide morphogen.

We also observed localized depletion of fz2 expression in WntA-dependent patterns, and a reduction of this effect upon WntA KO, showing that WntA/Fz2 high-level signaling triggers a negative feedback on fz2 mRNA expression. We found that WntA expression is under positive feedback, a phenomenon that may locally amplify signal and notably refine expression levels that prefigure color outputs in the V. cardui CSS (Fig. 2A-H). This pathway-intrinsic dual feedback mirrors aspects of Wg/Fz2 regulation in the Drosophila wing disk, and is generally thought to mediate precise and robust developmental signaling (Lander, 2007; Mehta et al., 2021; Sagner and Briscoe, 2017). These results lay out an exciting avenue of research for studying the formation of discrete responses for graded input in the butterfly wing, an epithelial tissue that effectively ‘displays’ differentiated cell states as scale colors.

In order to achieve the modular organization of pattern, butterfly wings are thought to deploy distinct cell signaling pathways and selector genes that control discrete sets of patterns in a region-specific manner (Martin and Reed, 2014; McKenna et al., 2020). As multiple Wnt ligand genes are expressed in butterfly wing transcriptomes (Fenner et al., 2020; Hanly et al., 2019), it is important to elucidate how many Wnt pathways are active in color patterning or, in other words, whether different branches of the Wnt transduction mechanisms can mediate separate cellular responses. Strikingly, WntA and fz2 KOs remove the brown CSS in J. coenia but leave intact the overlapping, distinctively colored and wg positive D₁ and D₂ (Fig. S5) that are thought to be induced by Wg, Wnt6 and Wnt10 (Martin and Reed, 2014; Mazo-Vargas et al., 2017). This suggests that different Wnt signals diverge in their effects across the wings and induce distinct responses, with WntA most noticeably inducing the CSS and dPf, and Wg, Wnt6 and Wnt10 patterning D₁-D₂ and MBS elements. However, the distinctiveness of these elements would be difficult to explain if color-patterning Wnt pathways converge on the same transduction mechanism. This leads us to the working hypothesis that the WntA/Fz2 pathway is non-canonical, potentially separating its signal from canonical Wg activity in different sets of patterns. Non-canonical Fz2 pathway branches have been identified in Drosophila and will deserve further examination in butterflies to test this possibility, with candidate mechanisms including PTK7 family co-reception (Linnemannstöns et al., 2014; Peradziryi et al., 2011), ROR-family co-reception (Ripp et al., 2018; Shi et al., 2021), Fz2/Ca²⁺ signaling (Agrawal and Hasan, 2015) and the FNI pathway (Mathew et al., 2005; Restrepo et al., 2022).

Individual knockouts of Fz1 and Fz2 failed to disrupt Wg-positive patterns, suggesting that Fz1 and Fz2 act redundantly to mediate canonical Wg functions in butterflies, similarly to Drosophila and Tribolium (Beermann et al., 2011; Cadigan et al., 1998; Ewen-Campen et al., 2020). Nonetheless, fz1 knockouts and knockdowns in several species showed that it is necessary for epithelial polarity, with phenotypes displaying disorganized scale arrays, akin to the wing trichome phenotypes characteristic of PCP in fruit flies (Adler, 2012). As in Drosophila (Ewen-Campen et al., 2020; Joyce et al., 2020), Fz1/PCP is independent of Wnt activity in butterflies, because knockouts of the Wnt secretion factors Wls and Por do not yield scale polarity defects (Hanly et al., 2021). These data validate previous inferences that this pathway is functionally conserved across Bilateria and Cnidaria (Momose et al., 2012; Yang and Mlodzik, 2015), and prompt further studies of Wnt-independent Fz1/PCP in a wider variety of arthropods and tissue types. Future investigations of butterfly wing PCP in particular could look at the effect of fz1 mutation on scale and socket precursor orientation, cytoskeleton distribution, and asymmetry of the PCP marker Prickled during the first phases of pupal wing development (Adler, 2012; Day et al., 2019; Dinwiddie et al., 2014).

The expression patterns of fz3 and fz4 in arthropod embryos are diverse, but imply ancestral roles in nervous system and appendage patterning (Janssen et al., 2015). In Drosophila, Fz3 acts as a decoy receptor, i.e. as a competitive-antagonist that attenuates Wnt signaling and contributes to establishing the Wg gradient in wing discs (Sato et al., 1999; Schilling et al., 2014). Of note, fz3 is lost from the genomes of Tribolium beetles and Parasteatoda spiders (Janssen et al., 2015), and the developmental functions of Fz4 remain elusive in Drosophila and Tribolium (Beermann et al., 2011; Ewen-Campen et al., 2020).

Here, we show that both fz3 and fz4 play key roles in the formation of normal venation in the developing butterfly wings. Clonal knockouts had sporadic ectopic vein branching, implying that these genes are acting to repress vein development. No such role has been described in the literature for Fz3 or Fz4 in D. melanogaster wings (Ewen-Campen et al., 2020; López-Varea et al., 2021a). Despite this, activation of canonical Wnt signaling via misexpression of wg or of a constitutively active Arm induces ectopic veins in flies (Brennan et al., 1999; Lunde et al., 2003; Zeng et al., 2008), and knockdowns of the Wnt pathway genes pegasus, Earthbound, pygopus and nemo result in wing venation defects (López-Varea et al., 2021a,b), overall suggesting that the Wnt pathway is indeed a modulator of vein formation in flies.

Similarly, we found that fz3 is strongly expressed in the wing periphery and that its loss-of-function results in ectopic wing margin duplications (Fig. 6). Lepidopteran wing disks develop with their dorsal and ventral epithelia in apposition, and form a wing margin that separates pre-apoptotic peripheral tissue from the wing epithelium (Macdonald et al., 2010; Nijhout, 1991). The Wnt ligand genes wg, Wnt6 and Wnt10, as well as the transcription factor Cut, mark the peripheral tissue, analogous to the dorso-ventral (D-V) wing boundary characterized in flies (Gieseler et al., 2001; Macdonald et al., 2010; Swarup and Verheyen, 2012). Analogous to the role of canonical Wnt signaling in specifying the D-V boundary in flies (Buceta et al., 2007; Swarup and Verheyen, 2012), it is plausible that the fz3-deficient secondary margins we observed in butterflies result from a Wnt signaling gain of function.

A recent study suggested that butterfly Fz4 represses canonical Wnt signaling intracellularly in larval wings, and that it activates canonical Wnt signaling as well as Fz/PCP at pupal stages (Banerjee et al., 2023). Our data support the proposed Wnt antagonist role. However, we also found that Fz4 lost its KTxxxW domain in Diptera and Lepidoptera (Fig. S1), which makes it unlikely to recruit Dsh and thus to relay canonical Wnt and PCP signaling at later stages (Butler and Wallingford, 2017; Gao and Chen, 2010; Strutt et al., 2012). Instead, we propose that Fz4 acts as a decoy receptor that attenuates the amount of free Wnt ligands, by binding them without being structurally capable of transmitting a signal. In this model, Fz4 mKO leads to higher Wnt pathway activity and thus the induction of ectopic veins, and, in Bicyclus, of ectopic eyespots (Banerjee et al., 2023). A similar mechanism may explain the effects of fz4 knockout on Discalis and MBS pattern elements. Specifically, the inner Discalis patterns (orange in J. coenia, silver in A. incarnata) expanded on ventral surfaces, while the outer black contours expanded in these same patterns on dorsal sides (Fig. 5F,G). In the peripheral patterns of both V. cardui and J. coenia, we observed a distal shift of eyespots accompanied by a simplification of the MBS patterns. Both Discalis and peripheral pattern aberrations from fz4 crispants in three species showed rounded rather than jagged, irregular clones with sharp color breaks at boundaries (Figs 5F,G and 6). We interpret these effects as non-cell-autonomous, contrary to what would be expected from a classic receptor, but consistent with a Wnt decoy function occurring upstream of signal transduction, in the extracellular space. It is worth noting that fz4 mKOs resemble the effect of extracellular heparin injections in early pupae, a treatment that emulates Wnt gain of function, including Discalis element expansions, as well as MBS pattern fusion and expansions in J. coenia, V. cardui and A. incarnata (Martin and Reed, 2014; Serfas and Carroll, 2005; Sourakov, 2018; Sourakov and Shirai, 2020). Importantly, heparin likely enhances the effects of WntA in the CSS, but fz4 mKOs only affect other heparin-sensitive patterns, implying that Fz4 is in fact unable to bind WntA. Instead, Fz4 is likely counteracting one or a combination of Wg, Wnt6 and Wnt10, which are all co-expressed in the Discalis elements and wing-peripheral region margins in larval wings (Banerjee et al., 2023; Martin and Reed, 2014). Although expression and functional assays of Wnt family members at the pupal stages will be needed to test this model, the current data suggest that sub-functionalized Fz3 and Fz4 antagonize distinct aspects of Wnt signaling in a context-dependent manner.

The WntA/Fz2 signaling axis has heretofore remained unstudied due to the loss of WntA in classical model systems. However, it could well prove to be evolutionarily conserved, and it will be interesting to test whether its role as a morphogen is involved in epithelial patterning elsewhere, or to identify other functions across Bilateria and Cnidaria (Hanly et al., 2021; Li et al., 2017; Schenkelaars et al., 2015). The Fz1/PCP pathway and the Wnt-antagonist role of Fz3 may both date back to pre-bilaterian ancestors (Momose and Houliston, 2007; Momose et al., 2012; Schenkelaars et al., 2015), whereas our sequence analysis suggests Fz4 only recently lost its Dsh-interacting domain during the course of insect evolution (Fig. S1B, cyan box). Overall, functional analyses of Frizzled receptors, seen under the colorful features of butterfly wings, enrich our understanding of their diverse functions in mediating cell signaling and tissue patterning.

Vanessa cardui (Linnaeus 1758) were purchased from Carolina Biological Supplies, and Junonia coenia (Hübner 1822), originating from the laboratory of Fred Nijhout (Duke University, NC, USA) were reared on an artificial diet, in a growth chamber at 25°C, 40-60% relative humidity and under a 14:10 h light:dark cycle, following previously described procedures (Martin et al., 2020; Tendolkar et al., 2021) with the following modification: to avoid the spread of viral disease, eggs are surface decontaminated for 2 min (V. cardui) or 4 min (J. coenia) with a benzalkonium chloride 5% solution, before being dried and left to hatch on double-tape placed under the cup lids, with no more than 40 eggs per container. In these conditions, the mean pupal developmental time is 160 h for V. cardui and 190 h for J. coenia.

Agraulis incarnata (Riley 1926) is the revised species name for subspecies of (formerly named) Agraulis vanillae found in the USA (Núñez et al., 2022). A. incarnata adult butterflies were maintained in a closed enclosure situated in a rooftop greenhouse in Washington DC, with fluctuating temperatures 23-28°C and a periodic misting system for humidity. Gatorade 50% cups, Lantana spp. and Buddleia spp. provided nectaring sources. Normal feeding and sexual behavior during the day required light supplementation with six Repti-Glo 10.0 Compact Fluorescent Desert Terrarium Lamp bulbs (Exo Terra). Eggs were collected on Passiflora biflora or Passiflora incarnata and were washed for 1 min with benzalkonium chloride 5%. Larvae were reared in an incubator at 28°C, 40-60% relative humidity, and under a 14:10 h light:dark cycle. This species is prone to viral disease when crowding occurs. To remediate this, larvae were reared on artificial diet, following the V. cardui rearing procedure with the following modifications. Smaller batches of embryos were hatched and reared during the initial stages (n<8 individuals per cup), and larvae were moved to individual cup containers at the third larval instar. The Passionvine Butterfly artificial diet (Monarch Watch Shop), supplemented with dried P. biflora or Passiflora incarnata was used for larval feeding. Larvae required frequent movement to fresh diet cups, removal of excess silk in the fifth instar larvae and addition of a horizontal toothpick section under the cup lid to induce pre-pupal silk pad spinning.

Heliconius erato demophoon butterflies (Ménétriés 1855) and Heliconius melpomene rosina (Boisduval 1870) were reared at the Heliconius insectaries in Gamboa (Panama) and at the LMU Munich (Germany) using previously described methods (Concha et al., 2019; Rossi et al., 2020). Danaus plexippus (Linnaeus 1758) butterflies were reared on milkweed in greenhouse cages under standard conditions at UC San Diego (USA).

Egg microinjections were performed as previously described to deliver equimolar mixes of Cas9/sgRNA into butterfly syncytial embryos (Martin et al., 2020; Thulluru et al., 2022). In brief, sgRNAs were designed to target the transmembrane domain of the Frizzled proteins (sequences in Table S2), ordered as synthetic sgRNAs (Synthego), mixed in 1:2 mass ratio with recombinant Cas9-2xNLS (PNABio, or UC Berkeley Macrolabs-QB3) and microinjected in embryos at 1-6 h AEL. Injection conditions varied along two main parameters: time of injection after egg laying and Cas9:sgRNA duplex concentration (between 250:125 and 500:250 ng/μl). More-concentrated early injections result in large mutant clones, but can be deleterious and result in low survival. The level of mosaicism can be increased using a more-dilute mix injected at least 4 h AEL, overall resulting in higher survival, but smaller wing clones.

Detection of mRNA expression was performed following a standard method (Martin and Reed, 2014), with minor modifications for early pupal wings. Pupal wings at 13-17% development tissues were dissected in ice-cold 1× phosphate-buffered saline (PBS). A scalpel or straight Vannas Spring Scissors (8 cm/2.5 mm tip) were used to cut the contour of forewing, which was then detached from the pupal case by first grabbing the base then pulling the wing carefully towards the distal edge; forceps were used to remove part of the peripodial membranes while keeping for forewings and hindwings embedded at their edges. Forewings were placed back to the cuticle and transferred to fixative (1×PBS, 10 mM EGTA and 9.25% formaldehyde) for 30-40 min at room temperature, together with the entire pupal case with the hindwings attached. After fixing, wings were washed with ice-cold PBT (1×PBS and 0.1% Tween20) twice, while still embedded in their cuticle or pupal case, which was then removed with fine forceps. Subsequent PBT washes and the rest of the published protocol were carried out with two pupal wings per tube. For the final development of stains, wings were incubated in BM Purple (Roche Applied Science) for 10-15 h at room temperature. Riboprobes for WntA and wg have been previously described (Martin and Reed, 2014). Riboprobes for the four frizzled genes were PCR amplified from V. cardui cDNA (Table S4), transcribed with a Roche T7 DIG RNA labeling kit, purified with Ambion MEGAClear columns and stored at −80°C. Images were taken using a Nikon D5300 camera mounted to a Nikon SMZ800N trinocular dissecting microscope, equipped with a P-Plan Apo 1×/WF 0.105 NA 70 mm objective and a Nikon C-DS stand for diascopic mirror illumination.

Dicer-substrate siRNAs (DsiRNAs) were designed against nymphalid fz1- and fz2-coding sequences (Table S5) using both the online IDT siRNA designer tool and the siRNA iScore Designer web services (Ichihara et al., 2007), ordered at the 2 nmol or 10 nmol scales as Custom DsiRNAs with standard purification (IDT DNA Technologies), resuspended at 70-100 μM in 1× Bombyx injection buffer (pH 7.2; 0.5 mM NaH₂PO₄, 0.5 mM Na₂HPO₄ and 5 mM KCl), and stored as frozen aliquots at −80°C until use. Electroporations were conducted within 30 min of the onset of pupation using a BTX ECM 830 electroporator with gold electrodes (Harvard Apparatus, Genetrodes bent gold tip, 5 mm). Electroporation procedures followed a previously described procedure (Fujiwara and Nishikawa, 2016). In brief, pupal forewings were lifted over a thin pad of 1% agarose prepared in 10×PBS, injected in their ventral surface with 1-4 µl of annealed siRNA mix (distal to the discal vein) and covered with a 20-30 µl droplet of 1×PBS. The positive electrode was placed in contact with the droplet, taking great care to not touch the wing epithelium (see below), while the negative electrode was in contact with the agarose pad on the dorsal side of the wing, before electroporation with five square pulses of 280 ms at 9-15 V, separated by 100 ms intervals. The PBS droplet was removed before reinsertion of the forewing into its pupal case. Pupae were sprayed with water every 2 days before emergence to avoid desiccation. In order to control for potential artifacts of the procedure, sham controls using GFP siRNA injections and injection dye were run. GFP siRNAs were injected into forewings or hindwings and electroporated with five square pulses of 280 ms at 8-15 V, separated by 100 ms intervals. Similarly, Phenol Red (2-4 μl) was injected at a concentration of 0.05% to monitor injection spread across the forewing. In the red dye injection experiments, to test for potential effects of excessive current, the positive electrode was directly placed in contact with the wing tissue within a PBS droplet, before electroporation using 6 V to 9 V at 280 ms, 5 pulses with 100 ms interval each.

Wings were de-scaled by washing in 90% ethanol for ∼3 min, then switching to 10% Clorox bleach for 3 min, and repeating until the scales and pigments were removed from the wings. Washed wings were then imaged on a Keyence VHX-5000 microscope mounted with a VH-Z00 T lens.

Sanger sequencing of G₀ crispants was performed to verify the induction of indels at the predicted cut sites for several sgRNAs (Fig. S21). Thoracic muscle from adult fz2 crispants was amplified by PCR, followed by clean-up and sequencing (V. cardui), or isolated in single bacterial colonies by TA cloning before sequencing (D. plexippus). For fz1, fz2 and fz4 (V. cardui), CRISPR injections were performed 30-45 min AEL to limit mosaicism, targets from single crispant embryos (24 h or 48 h AEL) were amplified using Phire Tissue Direct PCR Master Mix (Invitrogen, diluted protocol), and 45 μL reactions were purified using the PureLink PCR purification kit (Invitrogen) and sequenced. V. cardui fz1 and fz3 crispant chromatograms contained multiple indel alleles and were analyzed using TIDE analysis (Brinkman et al., 2014). Primer sequences are provided in Table S3.

We thank Rachel Canalichio and the staff of the Wilbur V. Harlan Greenhouse at the GWU for butterfly rearing and host plant resources; Bob Reed, Brian Counterman and Donya Shodja for stimulating discussions and comments on the manuscript; Chip Taylor and Ann Ryan for providing Passionvine Butterfly artificial diet; and Richard Merrill for facilitating H. melpomene experiments.