Introduction

Animal coloration is an adaptive trait that is often diversified among closely related species. In particular, wing patterns of butterflies vary greatly and these variations have attracted public as well as researchers and have been studied since the time of Darwin. Among the various strategies used by butterflies to avoid predators, some butterflies have become unpalatable by incorporating plant alkaloids1 or producing a toxin-like cyanogenic glycoside2 as defensive chemicals. Unpalatable butterflies inform predators to their toxicity by exhibiting conspicuous wing patterns, such as the unique banding pattern of heliconid butterflies3. Some unpalatable butterflies share similar wing patterns to provide mutualistic protection called Mullerian mimicry4. In contrast, some palatable butterflies have evolved another strategy called Batesian mimicry: they mimic the conspicuous wing patterns of unpalatable species and acquire protective effects5,6,7,8. Multiple loci are involved in the expression of Mullerian mimicry phenotypes in heliconid butterflies9,10, whereas the phenotypes of Batesian mimicry species reported to date are determined by a single locus11,12,13. To elucidate the evolutionary processes of mimicry comprehensively, it is necessary to study the molecular basis of Batesian mimicry.

The swallowtail butterfly Papilio polytes is notable for exhibiting female-limited Batesian mimicry13. The females have 2 forms: the non-mimetic female (also called cyrus), almost identical to monomorphic males that have only pale yellow bands on the hindwings (Fig. 1a, b) and the mimetic female (also called polytes), which mimics the distasteful butterfly Pachliopta aristolochiae. The latter has centered pale yellow regions and periclinal red spots on the hindwings (Fig. 1c, d). This female dimorphism is controlled by a single autosomal locus H and the mimetic phenotype is dominant13. The pigmentation processes involved in Batesian mimicry of P. polytes are considered to be downstream of the H gene; however, the gene activation networks as well as the genes involved in this pigmentation are largely unknown.

Figure 1
figure 1

Adult wing color patterns of 2 species.

(a) Papilio polytes, male. (b) P. polytes, non-mimetic female. (c) P. polytes, mimetic female. (d) Model butterfly Pachliopta aristolochiae. Bars represent 1 cm. All images in this figure were captured by the author (H. Nishikawa).

Full size image

Chemical characterization of pigments is necessary for understanding the coloration processes. A few pigments in the butterfly wings have been identified to date. Pterin pigments are found in the wing scales of pierid butterflies14 and ommochrome pigments are involved in wing coloration in nymphalid and heliconid butterflies15,16. Papiliochrome is found mainly in papilionid butterflies and the NMR spectrum data indicate that the aromatic amino group of kynurenine forms a chemical bond with the β- carbon of the side chain of N-β-alanyldopamine (NBAD) in papiliochrome II17. In many papilionid butterflies, papiliochrome II forms the pale yellow pigment18, but the red pigments have not been characterized chemically. Thus, identifying these pigments both in mimetic and non-mimetic wings of P. polytes is necessary for understanding the switching between the 2 phenotypes. Moreover, it would be interesting to analyze the properties of each pigment to solve the evolutionary puzzle: the pigments in the wings of the model butterfly P. aristolochiae are used correspondingly in the mimetic females of P. polytes.

In this study, we used LC/MS to show that the pale yellow pigments of non-mimetic females consist of kynurenine and NBAD but mimetic females express other types of pigments. In addition, qRT-PCR analysis showed that kynurenine and NBAD synthesis genes are upregulated in the pale yellow regions in non-mimetic females but not in mimetic females. These data suggest that pigment synthesis can be switched on and off, resulting in either mimetic or non-mimetic females. In addition, we found that the red pigments of mimetic females are formed through the polymerization of kynurenine and NBAD compounds in a process potentially associated with novel factors such as the Toll signaling pathway.

Results

Comparison of pupal wing coloration in mimetic and non-mimetic females

To characterize the pigmentation processes, we first determined the development of pupal wing patterns. Pupal stages of P. polytes lasted for 12 days at 25°C. From immediately after pupation (P0) to 3 days after pupation (P3), epidermal cells outside the bordering lacuna (BL) were present (Fig. 2), but these cells disappeared because of cell death at stage P419. Pupal wing coloration started between P8 and P9. First, pale yellow pattern coloration was observed from P9 to P10. Notably, the pale yellow pattern of mimetic females emerged as a round shape near the wing center, whereas that of non-mimetic females appeared as an extended band. Next, the red spots appeared in the distal regions in mimetic females between P10 and P10.5. Finally, black coloration occurred and the wing pattern formation completed at stage P11. Coloration in males was almost the same as that in non-mimetic females. Based on the above observations, Fig. S1 summarizes region- and stage-specific coloration for pale yellow, red and black regions during pupal wing development in P. polytes.

Figure 2
figure 2

Pupal stages for hindwing coloration in P. polytes.

P0–P11 show each stage, in days after pupation. P0: immediately after pupation. P8: the wing is covered with scales but pigmentation is not present. P9: pale yellow coloration begins. P10: the shape of the pale yellow region appears clearly. P10.5: the pale yellow pattern appears clearer and red pigment coloration occurs in mimetic female wings. P11: black coloration occurs and all pigmentations are completed. Bars represent 5 mm. All images in this figure were captured by the author (H. Nishikawa).

Full size image

Identification of the pale yellow pigments using LC/MS analysis

Because the shape and position of the pale yellow region are different between mimetic and non-mimetic females, we first attempted to characterize the pale yellow pigment chemically. The pale yellow pigment, which was extracted from the hindwings of non-mimetic females using 70% methanol, showed 3 main peaks with retention times between 4.5 and 5.5 min on a high-performance liquid chromatography electrospray ionization accurate mass spectrometry (HPLC-ESI-MS) system (Fig. 3a). We named these peaks 1, 2 and 3 in order starting from the peak with the longest retention time. Peak 1 showed the highest intensity in total ion chromatography. The accurate mass spectrum of peak 1 was m/z 431.1929, indicating that the chemical formula of peak 1 was C21H26N4O6 (Fig. 3b). This chemical composition indicated that the peak 1 substance corresponds to a compound consisting of NBAD and kynurenine moieties: papiliochrome II17. The accurate mass spectrum of peaks 2 and 3 was m/z 653.2933 and 875.3944 respectively (Fig. 3c, d), indicating that the chemical formula was C32H40N6O9 and C43H54N8O12, respectively. MS/MS analysis of peak 2 revealed that this peak includes 2 substances: kynurenine and a molecule with m/z 445.2085 that is equivalent to an NBAD dimer, judging by the molecular weight and chemical composition (Fig. S2). In addition, the accurate molecular weight of peak 3 indicated that it was equivalent to kynurenine and an NBAD trimer (Fig. 3d). These results showed that all 3 peaks correspond to substances composed of kynurenine chemically linked to NBAD (monomer, dimer, or trimer), which mainly serve as pale yellow pigments in non-mimetic females. However, it is not clear whether all 3 substances are necessary for pale yellow coloration and if their amounts are controlled strictly.

Figure 3
figure 3

Liquid chromatography/mass spectrometry analysis of the pale yellow pigments.

(a) A chromatogram of the pale yellow pigments shows 3 main peaks 1–3. Mass spectra and chemical composition of peak 1 (b), peak 2 (c) and peak 3 (d).

Full size image

We next attempted to isolate the pale yellow pigments from mimetic females using the methods identical to those described above. The chromatogram of the extract from wings of non-mimetic females at 380 nm, the wavelength that showed maximal absorption in the spectrum of papiliochrome II, showed the same 3 peaks as discussed above and therefore, the pale yellow pigments were present (Fig. S3). However, the pale yellow pigments of mimetic females were not soluble in 70% methanol that was used for dissolving the pale yellow pigments of non-mimetic females. Therefore, these 3 pigments of non-mimetic females were not present in mimetic females (Fig. S4). This observation and further gene expression analysis (see below) indicated that the yellow pigment in mimetic females consist of compounds those were different from the papiliochrome II-related pigments of the non-mimetic females.

Expression patterns of kynurenine and NBAD pathway genes

To determine whether the pale yellow pigments are formed as a result of region-specific expression of genes involved in the synthesis of kynurenine and NBAD, we assessed the expression patterns of these genes. The kynurenine synthesis pathway is known to start with tryptophan that is converted to formylkynurenine by vermilion and further to kynurenine by kynurenine formamidase (kf)20 (Fig. S5). NBAD is known to be the pigment in the abdominal stripes of Drosophila melanogaster21 and serves as a precursor of cuticle sclerotization22. The NBAD synthesis pathway starts with tyrosine that is converted to DOPA by tyrosine hydroxylase (TH) and further to dopamine by DOPA decarboxylase (DDC). Finally, ebony converts dopamine and β-alanine to NBAD21 (Fig. S5). Therefore, we quantified the expression level of these genes in the 2 pathways by quantitative real-time PCR (qRT-PCR) using mRNAs from the pupal hindwings from P9 to P11. Based on the position of wing veins, we dissected 3 separate areas: future black region, future pale yellow region and future red spot regions in non-mimetic females (Fig. 4) and mimetic females (Fig. 5). The shapes of the pale yellow regions were different between the 2 types of females and the red region in non-mimetic females was not evident, but its corresponding region in non-mimetic females was isolated successfully.

Figure 4
figure 4

Expression levels of kynurenine and NBAD pathway genes in the late pupal wings of non-mimetic females.

Three separate regions are indicated: blue line, indicating the pale yellow region; red line, indicating the red spots-corresponding region; and black line, indicating the black region. The expression levels of all genes are shown relative to that of the rpL3 control. Error bars show standard deviation of 3 or 4 experimental replicates. (*P < 0.05, one-way ANOVA). All images in this figure were captured by the author (H. Nishikawa).

Full size image
Figure 5
figure 5

Expression levels of kynurenine and NBAD pathway genes in the late pupal wings of mimetic females.

Three separate regions are indicated: blue line, indicating the pale yellow region; red line, indicating the red spots; and black line, indicating the black region. The expression levels of all genes are shown relative to that of the rpL3 control. Error bars show standard deviation of 3 experimental replicates. (*P < 0.01, one-way ANOVA). All images in this figure were captured by the author (H. Nishikawa).

Full size image

Because the kf gene has not been identified in P. polytes, we first used degenerate PCR primers to clone a partial sequence (551 bp) of the gene (AB857712), whose deduced amino acid sequence shows 85% similarity to the gene of Heliconius melpomene. Phylogenetic analysis also indicates that this gene is the kf ortholog in P. polytes (Fig. S6). Based on this sequence and the sequences of 4 other genes that have been sequenced before, we compared the gene expression patterns in the 3 different regions between non-mimetic and mimetic females.

  1. 1

    Similar gene expression profiles in the future black regions

    In the hindwings of non-mimetic females, vermilion and kf from the kynurenine pathway were expressed at relatively low levels in the future black and red spot-corresponding regions (Fig. 4, black and red lines). Because both regions become black, we assumed that these genes are not involved in black region formation and their similar expression in the future black region in mimetic females supports this notion (Fig. 5, black line). In contrast, transcription of TH and DDC was upregulated in all future black regions at stage P10.5 and P11 (Fig. 4, black and red lines and Fig. 5, black line), suggesting that these genes when induced at these stages may participate in melanin synthesis in the black regions or in the sclerotization processes in pupal cuticle. We also observed high levels of ebony expression at stage P10 and P10.5 in the future black regions (Fig. 4, black line and Fig. 5, black line), whose functional roles are not clear.

  2. 2

    Differences in patterns of gene expression in the pale yellow regions between mimetic and non-mimetic females

    In the pale yellow region in non-mimetic females, significant changes in gene expression were observed in the kynurenine pathway: transcription of both vermilion and kf was induced at stage P10 and P10.5 (Fig. 4, blue line). In this region, all genes in the NBAD pathway were also induced at stage P10 and P10.5. These results strongly support the above conclusion that the pale yellow pigment in non-mimetic females is composed of kynurenine and NBAD. Because the pale yellow region in non-mimetic females appears clearly from P10 (Fig. 2), the increased expression level of these genes is consistent with the development timing.

    Nonetheless, in the pale yellow region of mimetic female hindwings, we did not observe the induction of the kynurenine pathway genes (Fig. 5, blue line), similar to those in the black region. Expression of NBAD pathway genes was also similar between the pale yellow and black regions in mimetic females (Fig. 5, blue and black lines). Together with the finding that the pale yellow pigments of mimetic females could not be extracted with 70% methanol (Fig. S4), these results indicate that kynurenine or its related substances are absent from in the pale yellow pigment in mimetic females. NBAD synthesis genes were slightly upregulated at P10.5, when the pale yellow pigmentation was completed. Thus, it is possible that NBAD was not involved in the synthesis of the pale yellow pigments in mimetic females. These results suggest that the yellow pigments are synthesized differently and altered between non-mimetic and mimetic females in a region-specific manner.

  3. 3

    Stage- and region-specific induction of kynurenine and NBAD pathway genes in the red spots of mimetic wings

    The red spots on the hindwings are clearly visible in mimetic females but are barely detectable in non-mimetic females (Fig. 1). Thus, elucidation of the regulation of genes in this region is crucial for understanding the molecular mechanisms underlying Batesian mimicry. In mimetic wings, the expression levels of 4 genes, vermilion, kf, DDC and ebony, but not of TH, were specifically increased in the red region at stage P10.5 (Fig. 5, red line), the red pigment synthesis stage (Fig. 2). TH was upregulated at stage P11, same as that in the pale yellow and black regions. In contrast, in non-mimetic wings, expression of these genes was not induced in the red spot-corresponding regions of non-mimetic female (Fig. 4, red line). These data suggest that both kynurenine and NBAD synthesis pathways are involved in red pigment synthesis in mimetic wings of P. polytes. To confirm this observation, we attempted to investigate the chemical composition of the red pigment; however, we were unable to dissolve the pigment in any organic solvent: methanol, acetone, chloroform and 4% hydrochloric acid in methanol, suggesting that the red pigment is highly polymerized and stable.

Melanin, known to be insoluble in any organic solvent, is a highly polymerized substance that mainly comprises an indole skeleton23 and is insoluble in hydrochloric acid. In flesh flies, melanin in the adult integument is not dissolved in hot hydrochloric acid24. Thus, we applied this hydrochloric acid treatment and found that the pale yellow regions in non-mimetic females were hydrolyzed and became transparent. In contrast, the red region in mimetic females did not change, same as the black region (melanin pigment) (Fig. S7). These results also suggest that the chemical composition of the red pigment is similar to that of melanin and may polymerize in a similar manner.

Comprehensive analysis of genes involved in red pigment synthesis in mimetic wings

To determine the substances that constitute the red pigment in mimetic wings, we performed RNA sequencing to identify genes involved in red pigment synthesis. To compare the red spots on mimetic wings and the red spot-corresponding regions in non-mimetic females at the same stage (P10.5, the red pigment synthesis stage), we prepared RNA from these regions (Fig. 4 and 5, red) of 3 pupae, respectively, for RNA sequencing. All sequence reads were assembled into 79,769 transcripts using Trinity de novo transcript assembly software package (accession number: DRZ003159). We compared the gene expression patterns of 2 regions; 373 genes were expressed more in the red spots of mimetic females (Table S2), but 169 genes were expressed more strongly in the red spot-corresponding regions in non-mimetic females (Table S3). The notable genes upregulated in mimetic wings are involved in redox reactions (6 genes), signaling pathway and transcriptional regulation (10 genes), transporter functions (26 genes) and enzymatic activity (50 genes); however, the function of these genes in the formation of the red spots is unclear.

The expression levels (reads per kilobase of exons per million mapped reads [RPKM] values) of the kynurenine pathway genes vermilion and kf in the red spots of mimetic females were higher than those in the red spot-corresponding regions in non-mimetic females (Table 1). The expression levels of NBAD pathway genes TH, DDC and ebony were upregulated in the red region in mimetic females, supporting the above results of qRT-PCR analysis (Fig. 5); however, the differences were not statistically significant (Table 2). Certain genes involved in the process of melanization, including those induced by immune responses, were upregulated in the red region in mimetic females (Table 1). Prophenoloxidase 45 (proPO 45) and yellow-d2 which are involved in the melanization process, were induced in the red region in mimetic females.

Table 1 Transcriptional changes of the selected genes in the red pigment synthesizing stage
Full size table
Table 2 Transcriptional changes of NBAD synthesis pathway genes
Full size table

We also found that genes involved in the Toll signaling pathway, such as snake, spatzle 5 and gastrulation-defective, were upregulated in the red region of mimetic females (Table 1). Toll signaling activates the melanization process during an immune response25; thus, the red pigment in mimetic females may be synthesized by a process similar to that of melanin synthesis.

Discussion

In this study, we found that the pale yellow pigments in the wings of non-mimetic females of P. polytes are composed of kynurenine and NBAD (Fig. 3), which appear to form papiliochrome II (peak 1) and its related pigments (peaks 2 and 3). In papiliochrome II, the beta carbon of the side chain of NBAD is chemically linked to the amino group of kynurenine17. On the basis of their molecular weights, peak 2 (m/z 653.2933, Fig. 3c) was assumed to be a compound comprising kynurenine and an NBAD dimer and peak 3 (m/z 875.3944, Fig. 3d) was assumed to be a compound comprising kynurenine and an NBAD trimer. The pale yellow regions were also observed in the wings of mimetic female, but papiliochrome II and its related pigments were not observed (Fig. S4). Additionally, the expressions of genes involved in kynurenine and NBAD synthesis was upregulated in the pale yellow regions in non-mimetic females, but not in the pale yellow regions of mimetic females. These results suggest that papiliochrome II and its related pigments and the relevant biosynthetic genes are activated only in the pale yellow regions in non-mimetic females.

In P. polytes, male butterflies have been suggested to prefer non-mimetic females over mimetic females26. In another butterfly, Pieris rapae, Obara and Hidaka (1968)27 reported that the difference in the reflection of ultraviolet light affects the mate choice. Papiliochrome II is known to reflect ultraviolet light28 and thus, we can speculate that papiliochrome II and its related pigments in non-mimetic female wings contribute to mate choice in P. polytes. In mimetic females, papiliochrome II synthesis appeared to be reduced, which may result in reducing the male preference. Despite reduced male preference, mimetic females that switch the region-specific expression of the pale yellow pigments would succeed in enhancing the mimicry effect and protecting themselves from predators. It is intriguing that this type of trade-off can be explained by a switch of pigment synthesis.

In P. polytes, the H gene is known to be responsible for the development of the mimetic form13 and controls not only the switch in the wing pattern but also the flight behavior of mimetic females29. Although the H gene has not been identified yet, this study demonstrated that expression of kynurenine and NBAD pathway genes (Fig. 4 and 5) and several other genes (Table 1, 2, S2 and S3) is strongly altered. In heliconid butterflies, the black region at the center of the forewing is determined by WntA30,31 and the forewing band pattern is determined by optix32. In P. polytes, the H gene itself or transcription factors downstream of the H gene may control the pigment synthesis pathway and its region-specific expression.

The qRT-PCR and RNA sequencing analyses demonstrated that NBAD and kynurenine pathways are involved in red pigment synthesis in mimetic wings. We found that the chemical property of the red pigment was similar to that of the black pigment melanin, which is insoluble in hydrochloric acid (Fig. S7) and many organic solvents (data not shown). Melanin is considered to be a mixed polymer comprising indoles such as 5,6-dihydroxyindole-2-carboxylic acid and 5,6-dihydroxyindole in addition to variable amounts of other precursors in the synthesis pathway23. Unlike melanin synthesis, red pigment synthesis may use NBAD and kynurenine as the precursor substrates for polymerization.

Furthermore, multiple genes involved in the Toll signaling pathway, such as snake, spatzle 5 and gastrulation-defective, were upregulated in the red spots in mimetic females. Toll signaling pathway is involved in the immune response to bacterial or fungal infection33,34 and in the formation of the dorsoventral axis35. In D. melanogaster, 6 types of spatzle genes have been identified; however, the functionis known only for spatzle 1. Spatzle 1 function is known as a ligand of the Toll receptor. It is processed into its active form by a serine protease cascade involved in gastrulation-defective and snake. After dimerization, the homodimer of spatzle 1 binds to the Toll receptor and activates the Toll signaling pathway. In addition to Toll signaling, activation of phenoloxidase and yellow genes is required for the melanization reaction during the immune response25. The upregulation of Toll signaling and melanization genes in the red regions in mimetic females implies that the formation of the red pigments in mimetic females mediated by a biosynthetic pathway analogous to melanization. First, a Toll signaling-like pathway is activated by the ligand molecule Spatzle 5 and its partner, an unknown Toll receptor from the dozen or so members of the Toll family in a region-specific manner. This process induces the expression of proPO 45 and yellow-d2, which promotes polymerization of kynurenine and NBAD substrates that are activated simultaneously. In heliconid butterflies, an ocular pigment biosynthetic pathway (e.g., the one involving ommochrome synthesis genes such as vermilion and cinnabar20) produces the red or yellow coloration of the wings32. In contrast, in the red pigment synthesis in P. polytes wings, a different gene network, analogous to the immune response pathway, may have been recruited for red pigmentation during evolution.

Here we showed that the gene networks involved in color pattern formation are altered even between the similar pale yellow regions in mimetic and non-mimetic females. In addition, we demonstrated that both kynurenine and NBAD pathway networks are used in the pale yellow region of non-mimetic wings and the red region of mimetic wings, although their temporal and spatial expression profiles are controlled differently. These drastic changes are possibly regulated directly or indirectly by the H gene, the most upstream gene. Because H/h males do not show mimetic phenotypes, the H gene appears to be functional in producing mimetic wing coloration only in females. This observation is suggestive of the involvement of the H gene in sex differentiation pathways in later developmental stages. Identifying the H gene itself should help address the crosstalk between the sex differentiation pathway and the gene networks involved in changes of wing coloration in mimetic females.

Methods

Experimental animals

P. polytes was purchased from Eiko-Kagaku (Osaka, Japan) or Chokan-kabira (Okinawa, Japan). Larvae were reared on the leaves of Citrus unsyiu (Rutaceae) or on an artificial diet (Insecta LFM, NOSAN) including Citrus leaves powder under long day conditions (L:D = 16 h:8 h) at 25°C. Pupal samples were staged by the length of time after pupal ecdysis.

Extraction of the pale yellow pigments from female hindwings

The pale yellow regions were dissected from the hindwings of 10 adults and added to 500 μL of 70% methanol. The supernatant containing the pale yellow pigment was completely evaporated and re-dissolved in water. This sample was centrifuged at 15000 rpm for 10 min to remove insoluble components and the supernatant was used in the experiments described below.

HPLC-ESI-MS

An HPLC system (Ultimate 3000, Dionex) was used for analysis. The autosampler tray and the column oven were maintained at 4°C and 45°C respectively. Chromatographic separation was performed on a 2.0 mm i.d. × 100 mm column (CD-C18, Cadenza). The elution gradient was based on a binary solvent system [time in min (% eluent A/% eluent B): 0 min (95/5); 1 min (95/5); 5 min (5/95); 10 min (5/95); 10.1 min (95/5); 20 min (95/5)]. Solvent A consisted of 0.1% formic acid in water and solvent B consisted of 0.1% formic acid in methanol. The flow rate was constant at 0.25 mL/min. The mobile phase was directly delivered to the ESI ion source. The compounds were analyzed on an ESI-Qq-TOF mass spectrometer (micrOTOFQ II, Bruker Daltonics). The source was operated in the positive ion mode.

Cloning of genes involved in kynurenine and NBAD synthesis

The sequences of vermilion, TH, DDC and ebony were obtained from GenBank. Their accession numbers are AK405309, AB525743, AB525744 and AB525746, respectively. Degenerate oligonucleotide primers were designed on the basis of sequences of other insects such as D. melanogaster, Tribolium castaneum and H. melpomene. The primers were 5′-CTNGAGAGGGARYACTCTCCAAGCATGTGGTC-3′ and 5′-AATTCSCGCAAATCASANACGCCNGATAT-3′ for kf. The cycling conditions for PCR were 35 cycles of 95°C for 2 min, 49°C for 0.5 min and 72°C for 1 min. PCR products were isolated, subcloned into a TA cloning vector (pGEM-T Easy vector, Promega) and sequenced on an ABI PRISM® 3100 Genetic Analyzer (ABI, USA).

Phylogenetic analysis

Sequences were aligned using ClustalX36. Phylogenetic trees were constructed by the neighbor-joining method in the MEGA5 software37. The confidence level of various phylogenetic lineages was assessed by bootstrap analysis. We also compared several related genes using phylogenetic analysis. The following sequences were used to create the phylogenetic tree: Hm, H. melpomene (GQ183897); Tc, T. castaneum (XM_962551); Dm, D. melanogaster (FBgn0031821); Aa, Aedes aegypti (DQ440173); and Cq, Culex quinquefasciatus (XM_001842738). The sequences of kf of Hm, Tc, Aa and Cq were obtained from GenBank and that of kf and Dm was obtained from Flybase. Accession number of the kf gene in Papilio polytes was registered as AB857712.

qRT-PCR

The pupal wings were dissected (in cold phosphate-buffered saline, pH = 6.8) to isolate distinct parts: future black regions, future red regions and future pale yellow regions in non-mimetic (Fig. 4) and mimetic (Fig. 5) females at the stages that correspond to 9–11 days after pupation (P9, P10, P10.5 and P11). We used 3–4 pupae for qRT-PCR assays at each stage: 3 pupae from each stage except for stage P10 of non-mimetic females (4 pupae). Wing tissue was either immediately transferred to cold TRI reagent (Sigma) and homogenized for RNA extraction or stored at −20°C before RNA extraction. The samples were treated with DNase I (TaKaRa) to remove the genomic DNA, followed by phenol-chloroform extraction and ethanol precipitation. Thereafter, RNA was dissolved in water and reverse-transcribed with random primers (N6) using a first-strand cDNA synthesis kit (GE Healthcare). Real-time PCR was performed using Power SYBR Green PCR Master Mix on a StepOne system (Applied Biosystems Inc.). PCR was performed using these cycling condtions: at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 60 s. Ribosomal protein L3 (rpL3) was used as an internal control. Expression levels of these genes were normalized to that of rpL3. Statistical analysis was performed using one-way ANOVA at each stage. To evaluate significant differences, the region in which the gene was differentially expressed was tested using Bonferroni correction. The sequences of the primer sets are listed in Table S1.

RNA sequencing

We used 3 pupae of each type (mimetic and non-mimetic females). The pupal hindwings were dissected to prepare the red region in mimetic females (Fig. 5, red region) and the red spot-corresponding region in non-mimetic females (Fig. 4, red spot-corresponding region) at P10.5. Total RNA was isolated as described above and purified using an RNeasy column (Qiagen). We prepared cDNA libraries ligated using different adaptor indices (total 6 libraries were created) according to the vendor's instructions (TruSeq RNA sample preparation, the low throughput protocol, Illumina). In brief, magnetic beads containing poly-T molecules were used to purify mRNA from 1 μg of total RNA. Purified mRNA was then fragmented chemically and reverse-transcribed into cDNA. Finally, end repair and A-tailing were performed before Illumina adapters were ligated to the cDNA fragments. These samples were used to amplify cDNA in 10 PCR cycles. The amplicons were validated using an Agilent 2100 bioanalyzer and Agilent High Sensitivity DNA kit (Agilent Technologies) and quantified using KAPA quantification. Each cDNA library was diluted to 2 nM. cDNA libraries were sequenced on the Illumina HiSeq 2000 using HiSeq v1 flow cells and sequencing chemistry. The accession numbers of all raw reads in RNA sequencing were as follows: the red regions in mimetic females (DRR014115, DRR014116 and DRR014117), the red spot-corresponding regions in non-mimetic female (DRR014118, DRR014119 and DRR014120).

To construct the assembled transcriptome, all raw reads were assembled using the Trinity (http://trinityrnaseq.sourceforge.net/)38 de novo transcriptome assembly software packages. All reads of each library were mapped separately against the transcriptome sequences using the Bowtie 2, version2.1.0, (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml) short-read aligner39. Transcript expression levels were estimated to calculate RPKM values. To identify genes differentially expressed in the 2 regions, we performed edgeR with iDEGES/edgeR using TCC version 1.0.0 of the R package. Assembled transcripts that were upregulated on the basis of edgeR with iDEGES/edgeR in both regions were annotated in the BLASTX software against D. melanogaster transcriptome sequences in Flybase (http://flybase.org/blast/) with an E value cut off of 0.001. Genes with E values of > 0.001 were annotated again in BLASTX against the NCBI nonredundant protein sequence database, with an E value cutoff of 0.001.