From the formation of embryonic appendages to the color of wings: Conserved and novel roles of aristaless1 in butterfly development

Highly diverse butterfly wing patterns have emerged as a powerful system for understanding the genetic basis of phenotypic variation. While the genetic basis of this pattern variation is being clarified, the precise developmental pathways linking genotype to phenotype are not well understood. The gene aristaless, which plays a role in appendage patterning and extension, has been duplicated in Lepidoptera. One copy, aristaless1, has been shown to control a white/yellow color switch in the butterfly Heliconius cydno, suggesting a novel function associated with color patterning and pigmentation. Here we investigate the developmental basis of al1 in embryos, larvae and pupae using new antibodies, CRISPR/Cas9, RNAi, qPCR assays of downstream targets and pharmacological manipulation of an upstream activator. We find that Al1 is expressed at the distal tips of developing embryonic appendages consistent with its ancestral role. In developing wings, we observe Al1 accumulation within developing scale cells of white H. cydno during early pupation while yellow scale cells exhibit little Al1 at this timepoint. Reduced Al1 expression is also associated with yellow scale development in al1 knockouts and knockdowns. We also find that Al1 expression appears to downregulate the enzyme Cinnabar and other genes that synthesize and transport the yellow pigment, 3–Hydroxykynurenine (3-OHK). Finally, we provide evidence that Al1 activation is under the control of Wnt signaling. We propose a model in which high levels of Al1 during early pupation, which are mediated by Wnt, are important for melanic pigmentation and specifying white portions of the wing while reduced levels of Al1 during early pupation promote upregulation of proteins needed to move and synthesize 3-OHK, promoting yellow pigmentation. In addition, we discuss how the ancestral role of aristaless in appendage extension may be relevant in understanding the cellular mechanism behind color patterning in the context of the heterochrony hypothesis.


35
The diversity and complexity of butterfly color patterns is striking. What is even 36 more impressive is that this color pattern diversity within butterflies is often controlled by To study the developmental role of al1 we generated new CRISPR/Cas9 knockouts and 112 recovered both the previously described as well as novel effects. As previously 113 described, al1 knockout clones within the white band of a genetically white H. cydno 114 switched white scales to yellow (Figure 2A). However, careful observation of these 115 yellow clones in white H. cydno revealed that when these clones expanded over the 116 melanic regions of the wing, black scales became brown ( Figure 2B). Previous work 117 reported that Al1 seemed to be acting as a repressor of the yellow fate (Westerman, et 118 al., 2018). Based on this repressor activity we hypothesized that al1 knockout clones in 119 genetically yellow H. cydno would have no effect on the yellow portions of the wing. In 120 favor of this hypothesis we did not see any effects on the yellow parts of the wing, yet 121 interestingly, similar to white butterflies, clones within the melanic regions of yellow 122 butterflies also exhibited a switch from black to brown scales ( Figure 2B). 123 These results confirm the importance of al1 for the development of white wing 124 coloration. If al1 is knocked out, scales then switch to the yellow fate. However, the    134 antibody, which is known to stain homeodomain transcription factors like Al1. However, 136 this reagent is known to cross-react with similar proteins like the paralog Aristaless2 137 (Martin and Reed, 2010). In order to avoid this, we developed specific antibodies 138 against H. cydno Al1 epitopes to determine the protein subcellular localization and 139 pattern of expression in wings ( Figure S2). 140 Before looking into Al1 expression pattern in wings, we tested our antibody 141 specificity in Heliconius cydno embryos where we analyzed its relationship relative to 142 the ancestral Al function in appendages. We also aimed to provide expectations of its  (Figure 3B-D). To further elucidate our antibody specificity and determine if Al1 154 expression was causally related to appendage extension, we stained CRISPR Al1 155 knockout embryos. We observed sections of the embryos depleted for Al1, as expected 156 from a CRISPR knockout (Figure 3E-G). In addition, areas depleted of Al1 exhibited elongation defects when compared to the same appendages within the embryos that 158 had normal levels of Al1. In addition to confirming a role for Al1 in appendage extension 159 in Heliconius embryos, these data also provide evidence for the specificity of our newly 160 developed antibodies, allowing us to further probe the role of Al1 in wing color 161 patterning 162 Al1 accumulates in future white and black scale cell precursors, but not yellow 163 scale cell precursors. 164 Previous work with other nymphalid butterflies has shown that al1 expression on 165 larval wing discs resembles a modified pattern of the aristaless gene in flies (Martin & 166 Reed, 2010). Using in situ hybridization and antibody staining, we found a similar 167 pattern of expression of al1 during larval wing disc development in white and yellow H. 168 cydno ( Figure S1). This expression pattern appears to be unrelated to the white vs. 169 yellow color decision, hence we switched our attention to pupal stages.

170
Based on our CRISPR/Cas9 results, we hypothesized that Al1 would be present 171 more widely across the wing, including the forewing band, of white H. cydno but would 172 be absent from the band in yellow H. cydno. Furthermore, quantitative real-time PCR 173 suggested that al1 is expressed at all pupal stages but generally increases over time 174 (Westerman et al., 2018). We therefore analyzed wings ranging from 2 days to 4 days 175 (before scales harden and become impermeable to antibodies, Figure 1) after pupal 176 formation (APF). We aimed our dissections to the 3 days APF mark because it allowed 177 an efficient dissection without compromising the integrity of the wing and staining before  (Figure 3). Careful observation of a side reconstruction from Z-184 stacks highlights that Al1 was concentrated within the cytoplasm of scale cells and 185 absent, at least during these time-points, within the nucleus ( Figure 4E). In contrast, Al1 186 was reduced or absent inside developing yellow scales ( Figure 4F-K). This lack or 187 lower levels of Al1 was more apparent during younger time points (day 2 to early day 3) 188 and restricted to the dorsal side of the wing ( Figure S3). Furthermore, as development 189 continued, the overall level of Al1 on the dorsal side of yellow wings faded relative to 190 that on the ventral side and this was not observed on white H. cydno wings ( Figure S3). 191 Using the vein patterns we inferred boundaries between future yellow and melanic parts  Figure S4). We found no evidence that Al1 ever localized 200 to the nucleus at 2 to 4 days APF, yet it is still possible that nuclear localization does 201 occur at a time point that we did not observe or were not able to analyze. We verified 202 antibody specificity by performing several negative controls and repeating staining in white H. cydno butterflies with antibodies against two different Al1 epitopes ( Figure   204 S5A-D). 205 These results suggest that the presence of Al1 in scale cells may be relevant for 206 scale development and pigmentation across the entire wing. Presence of Al1 in the non-207 melanic band (which has already been specified by other genes like wntA [Martin, et al 216 To test our hypothesis that reduced or absent Al1 promote the switch from white 217 to yellow, we determined Al1 levels by antibody staining in white H. cydno pupal wings 218 with al1 CRISPR/Cas9 knockouts (70% of the adult wings showed some level of mosaic 219 color switch phenotype). Pupal wings analyzed at 3 days APF exhibited a depletion of 220 Al1 in patches across the wing (Figure 6). Our observations with adult butterflies 221 suggest that these clones lacking Al1 result in the switch of white and black scales to 222 yellow and brown, respectively. We also characterized the range of CRISPR clone size 223 and shape by observing a large number of CRISPR clones across the wings of white H. 224 cydno, both in adults ( Figure S6) and by antibody staining pupal wings ( Figure S7).
As a complementary approach to test this hypothesis, we used electroporation mediated RNAi (Fujiwara and Nishikawa, 2016) to locally knockdown al1 in a specific 227 area of the wing. RNAi injections performed hours after pupation recapitulated the white 228 to yellow color switch observed on adult wings observed previously with CRISPR/Cas9 229 (Figure S8A-B). Pupal wing discs were also analyzed by immunostaining at 3 days 230 APF to determine if there was any effect in the protein localization of Al1 after RNAi 231 knockdown. As expected, we found that clones with scales lacking Al1 (Figure S8C-D) 232 were concentrated near the injection site. Water injection controls showed no effect on 233 developing scale cells from the injection or electroporation process ( Figure S5E-F). 234 Both of these results further support our hypothesis that the white scale fate is 235 associated with high levels of Al1 and by contrast lower levels or absent Al1 is 236 associated with the yellow scale fate.

237
Ommochrome pathway genes are differentially expressed between white and 238 yellow wings. 239 To infer the potential downstream consequences of differential al1 expression,

285
Our data showed that exposing the pupal wing to the Wnt signaling inhibitor 286 iCRT3 did produce a white to yellow switch as predicted ( Figure 8B-C). In parallel, 287 when the Wnt inhibitor was used on melanic parts we observed the change from black 288 to a paler color as expected from a WntA knockdown (Figure 8D-E). Furthermore, 289 wings exposed to the inhibitor also showed depleted levels of Al1 when comparing the 290 dorsal (in closer contact to iCRT3) and ventral sections on the wing (Figure 8E-G). 291 DMSO/PBS controls showed normal Al1 levels, highlighting that the procedure itself did 292 not cause the observed effect ( Figure 8F). Furthermore, the untreated wing of the same 293 butterfly showed normal levels of Al1 as well. Yellow wings that were treated with the GSK3 inhibitor CHIR99021, which promotes Wnt signaling, developed white scales as 295 hypothesized ( Figure 8J-K). Finally, we also observe several melanic scales within 296 yellow band region as expected by a Wnt gain of function ( Figure 8L-M). 297 Following exposure to iCRT3, some wings exhibited zones with peculiar scale 298 phenotypes ( Figure 8H). Examination of these zones showed that some of the scales 299 were normal size and had normal Al1 levels but others were smaller and exhibited lower   we observe that Al1 is present in the entire wing and represses the yellow scale fate. It 338 is the absence of that repression which ultimately results in the color switch and pattern establishment we observe in the adult. While repression is a well-described 340 developmental phenomenon, the color pattern variation achieved via repression of al1 341 makes this a unique mechanism relative to other Heliconius color patterning genes.

342
Considering al1 along with wntA, optix, and cortex it becomes clear that even 343 though all of these genes control wing color patterning, they do so by very different 344 mechanisms. For example, WntA is a signaling ligand that has its effect early within the

387
This is an interesting case considering that both Exd and Al1 are homeodomain proteins 388 and similar accumulation is visible in our data. Therefore, it is possible that Al1 could act 389 as a direct regulator (by an un-observed nuclear translocation or a cleavage event) of 390 the differentially expressed genes needed for yellow pigmentation.

391
An alternative possibility is that Al1 regulates wing pigmentation indirectly via a 392 process known as the Heterochrony hypothesis (Koch, et al. 2000). This is an   Furthermore, outside of insects the Aristaless-like Homeobox (ALX) protein is a key regulator of rodent pigmentation (Mallarino, et al., 2016). Such regulation in principle is controlled by adjusting the rate of maturation of melanocytes, which are the pigmented 410 cells that ultimately carry out the pigment synthesis of the hairs on the rodent body 411 (Mallarino, et al., 2016). These observations support the idea that Al1 could be

Imaging of wild type and CRISPR adult wings
Butterflies were pinned to flatten the wings and dry the tissue allowing for better  Al1 antibody staining of embryos, larval, and pupal wings. 528 We raised polyclonal antibodies against two Al1 peptides using GenScript (New and 30% identity with Al2. Polyclonal antibodies were affinity purified after harvesting 532 and tested for specificity by performing Dot blot tests as described in Figure S2. Raw 533 data is available in Source Data 1-2. 534 We performed antibody staining in larval and pupal wings following Martin et al.    the pupa was left resting without hanging for 24 hours to allow for healing and recovery.

605
If the wing was going to be imaged the dissection and staining was carried out as 606 described above. In the case where the butterfly was going to be allowed to develop to 607 adulthood it was hung again between 24 to 48 hours after exposure and taken back to 608 the greenhouse. Approximately 60 pupae of white H. cydno were treated with ICRT3. 609 We used wings between 2 to 4 days APF from 10 individuals for Al1 antibody stainings   Table: 933