Disrupting different Distal-less exons leads to ectopic and missing eyespots accurately modeled by reaction-diffusion mechanisms

Eyespots on the wings of nymphalid butterflies represent colorful examples of the process of pattern formation, yet the developmental origins and the mechanisms behind eyespot differentiation are still not fully understood. Here we re-examine the function of Distal-less (Dll) in eyespot development, which is still unclear. We show that CRISPR-Cas9 induced exon 2 mutations in Bicyclus anynana leads to exon skipping and ectopic eyespots on the wing. Exon 3 mutations, however, lead to null/missense transcripts, missing eyespots, lighter wing coloration, loss of scales, and a variety of other phenotypes implicating Dll in the process of eyespot differentiation. Reaction-diffusion modeling enabled exploration of the function of Dll in eyespot formation, and accurately replicated a wide-range of mutant phenotypes. These results confirm that Dll is a required activator of eyespot development, scale growth and melanization and point to a new mechanism of alternative splicing to achieve Dll over-expression phenotypes.


27
The genetic and developmental origins of the bullseye color patterns on the wings of nymphalid 28 butterflies are still poorly understood. Eyespots originated once in ancestors of this butterfly lineage, 29 around 90 million years ago [1][2][3] , to most likely function as targets for deflecting predators away from 30 the butterfly's vulnerable body 1,4,5 . Eyespots may have originated via the co-option of a network of 31 pre-wired genes because several of the genes associated with eyespots gained their novel expression 32 domain concurrently with the origin of eyespots 3 . Some of these genes have since lost their expression 33 in eyespots, without affecting eyespot development, suggesting that they did not play a functional role 34 in eyespot development from the very beginning 3 . Yet, one of the genes, Distal-less (Dll), has remained 35 associated with eyespots in most nymphalid species examined so far, suggesting that it may have played 36 a functional role in eyespot origins 3,6 .

38
The function of Dll in eyespot development was initially investigated in B. anynana using transgenic   Table 1) led to a 98 variety of adult phenotypes (Fig. 2, Table 2). The most striking mutants displayed complete loss of 99 eyespots (Fig. 2a,b) followed by eyespots with significant developmental perturbations. Altered or 100 lighter scale pigmentation, associated with the eyespot mutations, appeared to correspond to the 101 extent of the mutant clones. Depending on their location, the lighter patches of wing tissue (i.e., the 102 presumptive Dll null clones) had remarkable effects on pattern formation. Eyespots vanished when 103 mutant patches covered the location of the eyespot centers (Fig. 2a,b), and mutant patches led to split 104 eyespots with mutant tissue bisecting the two eyespot centers (Fig. 2c). Some patches also had lighter 105 grey-blue scale pigmentation (Fig. 2d), lacked cover scales, or both cover and ground scales (Fig. 2e). In 106 addition to wing mutations we observed appendage defects that would be expected from a Dll 107 knockout 17,18 . A number of mutants exhibited reduced to barely noticeable stumps, legs with missing 108 tarsi ( Supplementary Fig. 2a) and deformed antennae with missing tips (Supplementary Fig. 2b).

110
Dll exon 2 mutants produced gain and loss of function phenotypes 111 Embryonic injections of guide RNAs targeting either the 5'UTR (Sg1) or the coding sequence (Sg2) of 112 exon 2 led to phenotypes similar to the ones described above (Table 2) as well as to a remarkable new 113 set of phenotypes, sometimes co-occurring on the same wing. These included ectopic eyespots along 114 the proximal-distal axis of the wing (Fig. 2f) and eyespots with a tear-drop shaped center (Fig. 2g), 115 closely resembling a spontaneous mutant variant in B. anynana known as the comet phenotype 19 (Fig.   116 2h). Ectopic eyespots were observed regardless of whether we targeted the 5'UTR or the coding 117 sequence of exon 2, as we injected each of these guide RNAs separately. Some butterflies displayed 118 both ectopic and missing eyespots on the same wing (Fig. 2i). Interestingly, ectopic eyespots were never 119 associated with changes in pigmentation in contrast to wing tissue with missing eyespots (Fig. 2i,j), 120 which always displayed the grey-blue pigmentation defects, highlighting the extent of the mutant clone 121 of cells. Similarly to exon 3 mutants, we also observed appendage mutants including truncated antenna 122 and legs or with fusion of antenna or proximal leg segments ( Supplementary Fig. 2c,d,e) .

124
Confirmation of CRISPR-Ca9 activity using next generation sequencing

125
In order to confirm that the phenotypes observed were due to genetic alterations of the targeted exons, 126 we performed next-generation amplicon sequencing of Dll to identify the entire range of mutations 127 generated from Sg1, representing exon 2 mutations, and Sg3, representing exon 3 mutations. To 128 identify mutations associated with each specific phenotype, especially in the case of exon 2 mutations 129 that produced both ectopic as well as missing eyespots, we isolated DNA from the adult wing tissue by 130 carefully dissecting around regions corresponding to missing, ectopic, or comet eyespots (see Fig 2f,g,i).

131
To characterize mutations we used CRISPResso, a software pipeline for analyzing next generation  (Table 2). For Sg3, we sequenced two individuals (Fig. 2c,d) and identified a range of 137 mutations with the most frequent representing a 42 bp and 4 bp deletion, respectively ( Fig. 2k and 138 Supplementary Fig. 3a). For Sg1 we sequenced 3 individuals (Fig. 2f,g,i). A large 72bp deletion was 139 observed in a mutant displaying ectopic eyespots (Fig. 2f, Supplementary Fig. 3b). In contrast, relatively 140 small indels were observed for another ectopic eyespot mutant (Fig. 2i,k, Supplementary Fig. 3c), and 141 surprisingly, the same 7 bp insertion emerged as the most dominant mutation from wing tissue either 142 with ectopic or missing eyespots (Fig. 2i,k). The most dominant mutation observed for the comet 143 eyespot phenotype represented a single base pair deletion (Fig. 2g,k, Supplementary Fig. 3b). Overall,

144
CRISPResso identified only a very small proportion of mutations as disruptions to potential splice sites 145 we decided to explore whether perhaps mutations that targeted each of the exons led to modifications 147 in the way that Dll was transcribed.   Cas9) as well as from wild-type non-injected embryos. PCR amplification from cDNA using primers 171 spanning exon 1 to exon 6 revealed that embryos injected with either Sg1 or Sg2, targeting exon 2, 172 produced a novel product approximately 500 bp shorter than the wild-type product. Sequencing this 173 short product revealed a deletion of 492 bp representing exon 2, suggesting that this exon had been 174 completely spliced out. In contrast, we did not observe any alternative splicing for cDNA obtained from 175 wild-type embryos or embryos injected with Sg3 ( Supplementary Fig. 5).

176
To examine whether targeting exon 2 resulted in ectopic eyespots due to Dll overexpression we 177 performed qPCR on cDNA from embryos injected either with Sg1 or Sg3, using primers designed to

250
The system we modeled is described by the following reaction-diffusion equations for the

302
until at some threshold value the spot splits vertically into two smaller spots. This phenotype was very 303 similar to the phenotypes observed in Fig. 2f,i, Fig.5b,c. Further increasing K resulted in the double spot 304 phenotype turning into an extended finger pattern, close to the observed comet phenotype (Fig. 2g,

307
which corresponds to reduced Dll production (Fig. 5a). The simulations support these experimental finding by showing that reducing K also results in smaller eyespots (Fig. 5d). where eyespots will develop, and is absent from the wing sectors where eyespots will not develop 7,8 .

337
In a recent study, Zhang and Reed (2016) 8

401
In addition to eyespot center differentiation, we confirmed that Dll has an additional role in wing

445
Animal husbandry 446 B. anynana were reared at 27 o C and 60% humidity inside a climate room with 12:12hrs light : dark cycle.

447
All larvae were fed young corn leaves until pupation. Emerged butterflies were frozen and then the 448 wings were cut from the body for imaging using a Leica DMS1000 digital microscope.

450
Guide RNA design

451
Guide RNAs corresponding to GGN20NGG (Dll) were designed using CRISPR Direct 47 . We separately 452 targeted three sites in Dll with two guides targeting exon 2, (in the 5'UTR and coding sequence) and a 453 third guide targeting the homeobox of exon 3 (Fig. 1a)

462
In vitro cleavage assay

463
The guide RNAs were tested using an in vitro cleavage assay. Wildtype genomic DNA was amplified

517
The reaction was set up following the manufacturer's instructions and run on a BIORAD thermocycler.

518
Relative expression software tool (REST) was used to analyze the expression data 51 .

521
In-situ hybridization was performed on 5 th instar larval wing discs.

540
Modeling details

541
Parameter estimation: We modeled a wing sector bordered by veins and containing a single eyespot as 542 a rectangle with typical width = 150 and length = 262 (Fig. 3d), 16 . We used degradation 543 and diffusion rates for both A1 and A2 close in magnitude to those measured for Wg and Dpp 544 respectively in the Drosophila wing disc 30 . Due to the longer time scales involved in eyespot patterning,

545
both degradation and diffusion rates were assumed to be smaller than in Drosophila (therefore, we observed a decrease in dpp ( Supplementary Fig. 6) at late larval stage, we decreased α by 25% at time 548 t = 60h in the simulation.

549
We present in Fig. 4

556
Boundary conditions: Boundary conditions were implemented based on the in situs and 557 immunostainings for dpp and Arm (Fig. 3a,b). The wing margin was modeled as a source term of Wg as 558 Arm is present along the wing margin of B. anynana and wg is also present along the wing margin of 559 other butterflies 53 . As dpp is absent along the wing veins (Fig. 3a), we modeled the veins as sinks for 560 both Wg and Dpp, which helped to confine the activator and substrate to the central part of the wing 561 sector in a finger-like pattern (Fig.3d,f). These conditions differ from those used in 15,16 where the     *These results are based on easily visible phenotypes and are likely an underestimation particularly from individuals that partially or fully eclosed but with highly crumpled and folded wings making it difficult to evaluate the extent of the mutations.