Single locus affects embryonic segment polarity and multiple aspects of an adult evolutionary novelty

Eyespot mutations are embryonic recessive lethal alleles at a single locus

Three spontaneous mutants with disturbed eyespot size and/or colour composition have been isolated from laboratory populations of B. anynana. Relative to 'wild-type' (WT) butterflies, Bigeye (BE) mutants have dramatically enlarged eyespots [21, 41], Frodo (Fr) have eyespots with a broader outer golden ring, and Spread (Spr) have very large eyespots with gold scales almost completely replacing the black scales (Figure 1a).

The aberrant embryos in the BE × BE, Fr × Fr, or Spr × Spr crosses showed severe and very similar morphological defects, suggesting that all three mutations are alleles of the same locus. This was confirmed by complementation tests: crosses between BE and either Fr or Spr mutants yielded embryonic lethality in approximately one fourth of the offspring (23.9% of 4645 progeny in 33 families), with identical morphological aberrations as those found in the offspring of crosses between mutants of the same phenotype. In contrast, no embryonic lethality or aberrant morphology was observed in 1756 progeny of 13 crosses between BE and Goldeneye individuals. This shows that these two mutations occurred in different genes, which is also consistent with the fact that the embryonic phenotype caused by the Goldeneye allele (i.e., disturbed blastokinesis [18]) is different from the embryonic effects produced by the alleles underlying the BE, Fr or Spr phenotypes (see "Analysis of aberrant embryos reveals defects in segment polarity" section).

Different BE phenotypes and variation in lethality of underlying alleles

Dramatic effects on eyespot morphology in BE, Fr and Spr mutants seem to be caused by different dominant alleles at the same locus, each disturbing embryonic development in homozygotes. However, the inheritance mode of the Spr phenotype appears to be more complex. The effect of the underlying allele on eyespot colour composition but not size is recessive, since offspring from Spr × WT crosses either have 'wild-type' eyespots or large eyespots with 'normal' colour scheme (46.7 and 53.3%, respectively, in 1272 adult offspring from nine families; Table 1). The latter phenotype is indistinguishable from that of BE individuals and is hereafter referred to as BE². A similar phenotype, hereafter called BE³, is found in the 'non-Spr' progeny from Spr × Spr crosses (34.8% of 440 adults in 13 families). Single-pair crosses were set up to determine whether enlarged eyespots in BE individuals and in the offspring of Spr × WT (BE² individuals) and Spr × Spr crosses (BE³ individuals) were caused by the same or by different alleles (Table 1).

Similarly to BE × BE, crosses between two BE² individuals yielded embryonic lethality in approximately one fourth of the progeny (22.9% of 1488 progeny in eight families) and enlarged eyespots in two thirds of the adults (64.8% of 429 adult offspring). These results are consistent with BE and BE² phenotypes being due to the same embryonic lethal allele with a dominant effect on eyespot size. In contrast, only a small fraction of embryos from crosses between two BE³ individuals died before hatching and showed typical morphological defects (Table 1). This ratio was significantly lower than the expected 25% (χ ² test of homogeneity among four families = 10.78, P < 0.05; χ ² goodness-of-fit for each family = 53.05, 22.00, 29.37, 42.29, all with P < 0.05) and varied between 0 and 5.3%, perhaps due to modifier loci or incomplete penetrance. No embryonic lethality was observed in five of the seven crosses between BE³ and BE individuals. The fraction of aberrant embryos in the other two families (10.8 and 1.1%) was significantly lower than the 25% expected if the BE³ and BE phenotypes were produced by the same lethal allele (χ ² test for homogeneity among seven families = 90.98; χ ² goodness-of-fit in two families = 20.79 and 7.55; P < 0.01). This suggests that the mutation underlying the BE³ phenotype is a different allele. It has a mildly deleterious effect on embryogenesis (i.e., low incidence of embryonic mortality) and a recessive effect on eyespot size, since all offspring of the BE³ × WT crosses (i.e., BE³ heterozygotes) have 'wild-type' eyespots (Table 1).

On the basis of the segregation of eyespot and embryonic mutant phenotypes in the experimental crosses, we propose that the three mutations occurred in the BE/Fr/Spr (BFS) locus, and that two of them are embryonic recessive lethal. We suggest that different combinations of the wild-type (BFS ⁺ ) and the three mutant alleles underlie the observed eyespot and embryonic phenotypes (Figure 2). The BE/BE² phenotype (each obtained in a different cross; see Table 1) is due to a single copy of the recessive lethal BFS ^B allele, the identical BE³ phenotype is caused by the recessive nonlethal BFS ^a allele, and the Spr phenotype, by a combination of both. The Fr phenotype is produced by a single copy of another recessive lethal allele, BFS ^C . In this model, alleles BFS ^a and BFS ^C each carry a mutation at a single site, while the BFS ^B allele carries two, one shared with BFS ^a . This is consistent with the fact that the Spr mutant (genotype BFS ^a /BFS ^B ) was originally isolated from the BE stock (see Materials and Methods), and it explains the data from all our crosses, including the more complex inheritance we described for Spr, and the different "enlarged eyespots" phenotypes we characterized (see previous section). Specifically (see Table 1), (1) crosses between two Spr individuals segregate for Spr and BE³ phenotypes (and not for Spr and WT, as would be the case if the Spr phenotype was due to a single, independent mutation at the BFS locus), and (2) the Spr phenotype is lost in progeny from Spr × WT crosses (showing that it is recessive), but recovered in progeny from BE³ × BE crosses (showing that BE individuals carry the relevant mutation) (see Figure 2).

Analysis of aberrant embryos reveals defects in segment polarity

Compared to WT embryos of the same age, homozygotes for each of the lethal alleles displayed severe and similar abnormalities. They were much shorter and thicker than WT embryos (Figure 1b). The typical 3 thoracic and 10 abdominal segments were all present, as was clear from the number of thoracic and abdominal appendages, but the segments were compressed and their borders poorly defined. Dorsal closure was not completed in approximately 30% of the embryos. The thoracic legs and the mouthparts (arrows and arrowheads in Figure 1b) were short; some embryos were missing one leg while one of the remaining legs was branched. We observed variation in timing of death among mutants. Typically, embryos from Spr × Spr crosses died before the stage when bristles appear (~70% of developmental time, DT [42]), and looked more compacted, while embryos from crosses between BE or Fr butterflies were fully sclerotized and had melanized head capsules, and died at approximately 90% DT. The severe shortening of segments and poorly defined segment boundaries observed in all mutant embryos suggest that the mutations affect the structure of each segment, rather than their establishment, possibly by interfering with the normal function of segment polarity genes.

Analysis of expression patterns of key segment polarity genes engrailed (en) and wingless (wg) revealed substantial differences between WT and mutant embryos. In early embryos (Figure 3a), en was expressed in the posterior compartment of each segment, as is typical for insects [43]. No aberrations in morphology or en expression were detected in 85 embryos dissected at 12-13% DT, revealing no differences among mutant and WT embryos. At 15% DT, the differences became obvious with the En protein detected not only in the posterior compartment, as in WT, but also in the anterior cells of each segment in mutant embryos (Figure 3b). Defects in segmentation became even more apparent at 25% DT, when mutant embryos appeared shorter due to compacted segments, and the En stripes were almost twice as broad as those of WT (Figure 3c). At 40% DT, mutant embryos looked almost spherical, with En clearly visible in a posterior and an anterior stripe within each segment, thoracic appendages, and abdominal prolegs (Figure 3d). Differences in wg expression were also very clear at 25-40% DT. In WT, wg mRNA was detected in each segment in stripes positioned just anterior to the En domain, while in mutant embryos two stripes of wg were observed in each segment, with the additional stripe positioned posterior to the expanded En domain (Figure 3e).

These results show that the lethal alleles at the BFS locus do not affect the specification of segment number or the establishment of en and wg expression domains in embryonic segments, but rather disturb their correct maintenance and generate a typical segment polarity phenotype, i.e., the 'replacement' of the anterior part of each segment by a mirror image duplication of the posterior part [44]. In Drosophila melanogaster, identical embryonic defects are caused by loss-of-function mutations, or experimental knockdown, of one of the negative regulators of Wg signalling; zeste-white 3 [45], naked cuticle [46], axin [47], APC2 [48], or Bili [49].

Epidermal response properties are affected in the Spr mutant

The formation of eyespots in developing pupal wings involves the production and diffusion of a morphogen from the cells at the center (stage called 'focal signalling') and the response of the surrounding epidermal cells [31, 32]. Different levels of the signal induce expression of different regulatory genes, e.g., Distall-less (Dll) and spalt (sal) in the inner disc, and en in the outer ring [17], which then directly or indirectly control biosynthesis of different pigments. Changes in the focal signal (e.g., morphogen concentration, stability or diffusion coefficient) have been shown to explain most of the variation in eyespot size between artificial selection lines [50], while the sensitivity levels of the epidermal cells to that signal determine eyespot colour composition [20, 51] and, to a lesser extent, eyespot size [21, 50]. Genetic correlations between eyespot size and colour composition are typically low [50–52], suggesting that different sets of genes determine these two aspects of eyespot morphology, and their underlying components of pattern induction.

Damage applied to the developing wing in early pupa typically induces the formation of ectopic patterns [53], and can be used to probe variation in epidermal response sensitivities (see [20]). Damage-induced ectopic eyespots were produced in more than 60% of cauterized WT and Spr pupae (32 of 50 for WT, and 56 of 70 for Spr), and typically resembled the native eyespots on the same wing surface (Figure 4a). Damage-induced eyespots consisted of a black inner disc and an outer golden ring in WT, but were almost entirely made up of gold scales in Spr. These results show that the ability of the entire wing epidermis to respond to eyespot-inducing signals, such as damage, is altered in Spr mutants. We also show that the expression patterns of 'response' eyespot ring genes were altered in Spr wings: en was detected in almost all cells of the eyespot field (Figure 4b), while the numbers of scale-building cells expressing Dll and sal were strongly reduced (Figures 4c and 4d). Taken together, our results are consistent with an effect of the target gene downstream of the eyespot-inducing focal signalling, but upstream of the response patterning genes.

The Wnt/Wg signalling pathway and butterfly eyespot formation

Our study of embryonic defects and of en/wg expression in the target B. anynana mutants suggests that the BFS locus is a negative regulator of the Wnt/Wg signalling pathway. Analysis of eyespot formation in Spr wings further suggests that the product of this gene acts upstream of eyespot ring patterning transcription factors during the 'response to focal signal' stage. The Wg morphogen, implicated in the evolution of wing pigmentation spots in Drosophila [54], has been proposed as a candidate focal signal in butterflies [35]. It is produced by eyespot centres in early pupal wings of B. anynana [35], and is known to upregulate en and Dll in insect embryos and imaginal wing discs, respectively [45, 55]. In butterfly wings, Wg could act as the inducer of the circular Dll and en domains corresponding to eyespot rings in adults [17]. Mutations in a negative regulator of the Wnt/Wg pathway, as could be the BFS gene, would affect the response of epidermal cells to this signal, altering en and Dll expression and therefore the distribution of black and gold scales. Molecular identification and functional characterization of the BFS locus, including the analysis of expression of key components of Wg signalling in both normal and altered eyespot development, will be necessary to assess its role in the regulation of this signalling pathway.

In a previous study, we mapped the BFS locus to approximately a 1.3 Mb interval on B. anynana chromosome 17, which has a high level of synteny with its ortholog in Bombyx mori [37], the model lepidopteran with the fully sequenced and annotated genome. Curiously, the orthologous chromosome in the butterfly Heliconius melpomene carries colour pattern loci implicated in intra- and interspecific divergence [56]. Inspection of gene content in the genomic regions of B. mori (nscaf2829 [57]) and H. melpomene [56], orthologous to the BFS interval, revealed that they do not contain any of the candidate genes generated via the comparison of embryonic phenotype to Drosophila mutants (namely, Axin, zeste-white 3, APC2, naked cuticle, and Bili), or any other gene with a known role in the regulation of Wg signalling. The high levels of synteny between these lepidopterans [37] suggest that the implicated interval in B. anynana, even though its exact genetic content is still unknown, probably also does not contain any of the described negative regulators of the Wnt/Wg pathway. This pathway is crucial for embryonic development and numerous other processes, and is remarkably conserved and extensively studied (reviewed in [58–60]). Nevertheless, sophisticated new genetic and genomic approaches continue to identify new Wnt/Wg regulators [61, 62], some of which are taxon-restricted (e.g., WTX [63]). The BFS locus could be another, yet unknown, negative regulator of this pathway, potentially specific to this lepidopteran lineage.

Butterfly stocks and crosses

All butterfly stocks were at reared 27°C as in [36]. The BE and Fr stocks were each set up from a single individual isolated from different laboratory populations in 1994 and 2007, respectively. The Spr stock was founded from a single individual isolated from the BE stock in 2003. All mutant stocks have been maintained with selection in favour of the mutant phenotype and, when necessary to avoid inbreeding depression, were outcrossed to the laboratory outbred WT stock.

Different crosses were set up to determine the mode of inheritance of the mutant alleles (Table 1). To test whether the mutations behave as embryonic recessive lethal alleles, butterflies from each mutant stock were crossed to individuals of the same phenotype and to WT stock butterflies. Crosses between butterflies of different mutant phenotypes were performed to determine whether the mutant alleles occur in the same locus (complementation tests). Eggs were collected from individual mating pairs, hatched larvae were reared through to adulthood, and the eclosed butterflies were frozen and scored for eyespot phenotype. Unhatched eggs were counted, dechorionated in 50% bleach solution for 1 min, rinsed with water, and fixed in 4% formaldehyde solution in 1 × phosphate-buffered saline (PBS). Embryos were dissected under a light microscope. Unfertilized eggs, identified as those lacking large blastoderm cells, were excluded from the analysis.

Statistical analysis

χ² tests were used to compare observed proportions of mutant embryos and adults to those expected if the alleles underlying the BE, Fr and Spr phenotypes were embryonic recessive lethal with dominant effect on eyespot pattern. First, we tested whether the ratios varied among families using the χ² test of homogeneity. If ratios were not significantly different, the numbers of progeny were pooled over all families belonging to the same cross type and mutant phenotype; these overall frequencies were then tested against the expected ratios using the χ² goodness-of-fit test. If the χ² homogeneity test revealed significant variation among families, the frequencies of aberrant embryos and mutant adults were tested against the expected values separately in each of the families (the case for only BE³).

Embryo morphology and gene expression analysis

Eggs were collected at different times after laying, dechorionated in 50% bleach solution for 1 min, rinsed with water, fixed in 10% formaldehyde in PBS/50 mM EGTA for 30 min, dehydrated in methanol in four steps and stored at -20°C until use. For scanning electron microscopy studies, 10-20 fixed embryos of each mutant phenotype were rehydrated in PBS in four steps and mounted on specimen holders, air-dried, coated with gold and viewed in a JEOL JSM 6400 scanning electron microscope.

For the analysis of gene expression patterns, 50-100 embryos were examined for each mutant line per developmental stage, defined according to the system developed for M. sexta [42]. Fixed embryos were rehydrated in PBS, blocked in PBT (5 mg/ml bovine serum albumin in PBS) for 2 hr, incubated with primary antibody overnight at 4°C (anti-Engrailed 4F11 [43]), washed 10 times in PBS with 0.1% Triton X-100 and incubated for 2 hr at 4°C in a 1:200 dilution of the secondary antibody (Alexa Fluor 488; Molecular Probes). After 10 washes in PBS, embryos were incubated in 100% glycerol for 1 hr and mounted on glass slides. Images were collected on a Bio-Rad MRC 1024 ES laser scanning confocal microscope.

A 315-bp fragment of B. anynana wingless gene (AY218276) was amplified from embryonic cDNA with primers 5'-GTCATGATGCCCAATACCG and 5'-GCAGTTGCATCGTTCCACTA and cloned into pCRII-TOPO dual-promoter vector using the TOPO TA cloning kit (Invitrogen). Plasmids were isolated with QIAprep Spin Miniprep Kit (Qiagen) and used as template for PCR reactions with vector primers M13F and M13R. The amplified products were cleaned with Wizard SV Gel and PCR Clean-Up System (Promega) and used for SP6 or T7 transcription. Sense and antisense digoxygenin-labeled riboprobes were synthesized using SP6 and T7 RNA polymerases and DIG RNA labeling mix (Roche Applied Science). The probes were run on an agarose gel and measured with NanoDrop spectrophotometer (Thermo Scientific) to verify their quality and concentration. In situ hybridization with sense and antisense probes was performed at 55°C for 48 hr following the protocol described previously [36]. The probes were detected with NBT/BCIP (Roche Applied Science). Stained embryos were mounted on slides and photographed with a Leica DC 200 digital camera attached to a Leica MZ 125 microscope.

Induction of ectopic eyespots

Dorsal surface of left pupal forewings was damaged at 12-18 h (± 30 min) after pupation, when ectopic eyespots are produced most often [53]. Wings were pierced with a finely sharpened tungsten needle (cat. no. 501317; World Precision Instruments) at a site next to the small anterior eyespot, approximately halfway between the wing margin and the normal location of the eyespots. Operated pupae were returned to 27°C; adults were frozen soon after emergence, and their wings were photographed with a Leica DC 200 digital camera attached to a Leica MZ 125 microscope.

Immunohistochemistry of pupal wings

Pupal wings of WT and Spr individuals were stained at 16-18 h after pupation according to the protocol described previously [17] with antibodies against Engrailed (4F11 [43], 1:50 dilution), Distal-less ([71], 1:200 dilution) and Spalt ([72], 1:500 dilution). Alexa Fluor 488 and Texas Red (Molecular Probes) were used as secondary antibodies in 1:200 dilutions.