Lepidopteran wing scales contain abundant cross-linked film-forming histidine-rich cuticular proteins

Scales are symbolic characteristic of Lepidoptera; however, nothing is known about the contribution of cuticular proteins (CPs) to the complex patterning of lepidopteran scales. This is because scales are resistant to solubilization, thus hindering molecular studies. Here we succeeded in dissolving developing wing scales from Bombyx mori, allowing analysis of their protein composition. We identified a distinctive class of histidine rich (His-rich) CPs (6%–45%) from developing lepidopteran scales by LC-MS/MS. Functional studies using RNAi revealed CPs with different histidine content play distinct and critical roles in constructing the microstructure of the scale surface. Moreover, we successfully synthesized films in vitro by crosslinking a 45% His-rich CP (BmorCPR152) with laccase2 using N-acetyl- dopamine or N-β-alanyl-dopamine as the substrate. This molecular study of scales provides fundamental information about how such a fine microstructure is constructed and insights into the potential application of CPs as new biomaterials.

S cales are the most remarkable and symbolic characteristics of Lepidoptera. The brilliant and patterned colors underwritten by scales play important roles in mating behavior 1 , camouflage and warning coloration [2][3][4] , temperature regulation [5][6][7] , and protection against predators 8,9 . Butterfly wing scale geometry and surface patterning are also reported to improve flight efficiency 10 . Having scales that cover the wings and body of butterflies and moths increases their chances of survival and reproduction in diverse environments.
A single layer of epidermal cells (resting, as always, on a sheet of basement membrane) is responsible for producing and shaping the cuticle. Epidermal cells can undergo a series of divisions culminating in a final differential division that yields a scale and a socket cell [11][12][13] . Cuticular proteins (CPs) and chitin synthesized de novo by scale cells accumulate around them and gradually form very hard scales 13,14 . Varied types of scales are precisely arranged in different regions of wings and body to produce unique color patterns 15 .
Having various shapes from cylindrical to flat, the scale is basically an envelope with "upper" and "lower" surfaces. The standard (unspecialized) scale is typically characterized by a series of ridges, crossribs, lamellae, microribs, and pigment 14,16 . Differently colored lepidopteran wing scales have different cuticular micro structures [17][18][19] , and pigmentation of scales 20,21 coincides with the development of their fine morphological features 16 . A study in the Tiger Swallowtail Papilio glaucus first suggested that sclerotization is linked to final scale color 20,22,23 . Matsuoka and Monteiro reported that melanin pathway genes play a dual role in scale construction and pigmentation by knockout of the pigment enzymes in the butterfly Bicyclus anynana 24 . Histochemical studies of butterfly wing scales show that F-actin plays an essential role in scale elongation and ridge position in early development 25 . Recent findings show F-actin is also important for the morphogenesis of swallowtail butterfly wing scale nanostructures 26 . However, the involvement of CPs per se in the formation of the intricately layered cuticular microstructure with other associated proteins remains unclear 14,27,28 . Although an earlier proteomic analysis tried to identify the proteins in mature scales from Bombyx mori adults, only seven CPs were detected due to their insolubility 29 .
In this study, we succeeded in identifying many CPs and other classes of proteins by collecting scales from developing wings before crosslinking occurred and dissolving them in a buffer containing urea, a well-established method for studying cuticle CPs but a breakthrough approach for the molecular study of wing scales. LC-MS/MS of the soluble scale proteins allowed us to identify the CPs and enzymes related to cuticular sclerotization, which revealed a possible mechanism of scale formation. RNAi experiments conducted with histidine-rich (His-rich) CPs with different His-contents suggested that each His-rich CP identified in scales might play a distinct, but vital role in scale microstructure. Films produced in vitro by crosslinking reactions of the highest His-rich CP (45%; BmorCPR152) suggested a possible avenue to industrial application of CPs as new biomaterials.

Results and discussion
The development of scale microstructures. Morphological observation of B. mori wing scales revealed vivid and dynamic development during the pupal and pharate adult stages (Fig. 1a). As reported for Precis coenia, scales emerged from their sockets by 72 h after pupation 12 . In B. mori (upper panel of Fig. 1a), tiny round scales were clearly visible at the third day after pupation (P-3), and then grew to an elongated baseball bat-like shape in P-4. The size and shape of scales in P-5 became similar to the mature ones; however, they remained soft and even dissolved partially in phosphate buffered saline (PBS). Scales in P-6 retained their shape very well in PBS although they still were soft. High magnification images of scales by scanning electron microscopy (SEM) showed that the ridge and cross rib framework was already present at P-5 (lower panel of Fig. 1a). Lamellae showed clearly and the ridges became more complicated and three-dimensional at P-6. All the observations indicated that P-6 scales, which have the same microstructures as mature scales, could be dissolved in urea containing buffer, since the stabilization of the microstructure of the scales began at P-6 or later stages.
The dissolving of developing scales in buffer. The developing wing scales were collected from the sixth and eighth days after pupation, followed by dissolving in 8 M urea-containing PBS and SDT buffer (4% (w/v) SDS, 100 mM Tris/HCl, PH~7.6) as described in Fig. 1b and Supplementary Fig. 1. The SDS-PAGE gels of proteins extracted from solubilized P-6 scales presented clear and discrete bands, whereas very faint bands appeared in P-8 (Fig. 1c). These results showed that scales of P-8 became much more difficult to dissolve compared to P-6 scales, which was consistent with our morphological observations indicating that the scales started to stabilize around P-6. As the proteins and other components were stabilized, the number and amount of proteins extractable from scales were also expected to be reduced.
Identification of proteins involved in scales. We carried out liquid chromatograph-tandem mass spectrometry (LC-MS/MS) analysis of extracts from P-6 and P-8 wing scales which detected peptides (Supplementary Data 1-4) originating from 1257 and 1090 proteins. The proteins classified as unreliable are indicated by "U" in column G of Supplementary Data 1-4. Only the reliable proteins were used for the analysis in this report; however, the unreliable proteins may also contain useful information.
Our analyses also provided quantitative data as iBAQ (intensity-based absolute quantification) scores for all of the identified proteins. Table 1 shows the top 25 scores for the two replicates of each time point; complete data are listed in Supplementary Data 1-4. What is immediately obvious is that CPs were among the most abundant proteins identified. In the two P-6 samples, they were 68% of the top 25, while at P-8 they were 64% and 56% in replicate samples. These most abundant proteins belonged to the CPG family (CPs with glycine-rich repeats), CPLCP (CPs with low complexity proline-rich amino acids, and also characterized by three conserved sequence motifs 30,31 ), CPCFC (CPs with well-conserved cysteine residues), CPFL (containing a 44 amino acid motif-like domain in the conserved C-terminal region), CPAP3 and CPR family (CPs with the Rebers and Riddiford Consensus, which has been subdivided into three groups, RR-1, RR-2 and RR-3). There were also three different CPHs (a collection of unclassified CPs). Obviously, scale formation involves a complex selection of CPs to build the exquisite scale structure. This finding is reminiscent of an earlier LC-MS/MS study that identified members of several CP families in the corneal lens and other structures from the mosquito Anopheles gambiae 32 .
Interestingly, 23 of the identified CPs from P-6 and 20 from P-8 were His-rich, with histidine contents amounting to 6-45%. More peptides with His-residues were detected in P-6 compared to P-8 ( Supplementary Figs. 2 and 3), and His-rich RR-2s had obvious predominance in both P-6 and P-8 scales (Fig. 1d) 34 . Compared to other his-rich RR-2 proteins, BmorCPR57 (15% His-content) was far more abundant in the recovered proteins of both P-6 and P-8 ( Table 1). The highest and second highest His-content proteins BmorCPR152 (45%) and BmorCPR82 (27%) were also found in the top 25 (Table 1).
Among the identified proteins, we reclassified three glycinerich CPs (CPG) that were initially classified by Futahashi 35 as CPLCPs. Based on BlastP searches against CPLCPs of An. gambiae, we renamed BmorCPG12, 13, and 24 as BmorCPLCP12, 11, and 3. Those were also His-rich (6-8%) and in high amounts in scales (Table 1 and Supplementary Data 1-4). Fu et al. previously reported 29 that BmorCPLCP12 is found in the mature body scales of B. mori. Moreover, BmorCPLCP11 and 12 are homologs of the An. gambiae CPs AgamCPLCP11 and 12, which previous proteomics analysis found to be present in larval head capsule and cast pupal cuticles 36,37 . Corman and Willis 38 found that the mosquito CPLCP family is similar to articulins, which are intracellular proteins of protists that form ordered, filamentous structures and have been suggested to provide support, patterning, and some elasticity to the cytoskeleton 39 . Accordingly, CPLCPs may also play an important roles in sculpting the basic structure of lepidopteran scales.
Gene 011414 was also among the most abundant proteins recovered, especially at P-8. PFAM classifies this protein as a retinin. The retinin domain was identified in the CPCLA family of An. gambiae, but the B. mori protein did not score as a CPLCA. Nonetheless, based on its abundance, it is probably another CP, underscoring the complexity of scale construction. Also among the 25 most abundant proteins recovered were apolipoporins, lipoproteins, two different sensory proteins, and one with the von  Willebrand factor domain, previously reported in the extracellular matrix (matrisome) of Drosophila melanogaster 40 .
As previously reported, F-actin and chitin play important roles in the development of scales in butterflies 25,26,41 . We identified actins in the scales, and at both P-6 and P-8 they were among the 25 most abundant proteins (Table 1). Some chitinases, which can cleave internal bonds in chitin polymers 42 , were also found in B. mori scales. This is consistent with observations in butterflies that F-actin bundles are degraded once the ridges have grown at approximately 60% of pupal development 26,41 . In addition, a recent report of scale development in a swallowtail butterfly shows that Arp2/3 appears to reorganize the degraded F-actin to a network after the development of ridges is complete. This reorganization contributes to the morphogenesis of wing scale nanostructures 26 . The detection of Arp2/3 (Gene009685) at P-6 (Supplementary Data 1-2) in our study suggested re-organization of F-actin may also occur in B. mori scales. However, we still cannot draw a conclusion for the functions of F-actin and chitin in forming the microstructure of scales, suggesting that how CPs, F-actin and chitin interact to contribute to the scales of Lepidoptera is an important study for the future.
In sum, this comprehensive analysis of the proteins that make up scales provides crucial information for an understanding of the process of scale formation. Notably, it reveals the existence of a critical previously unsuspected role for His-rich cuticle proteins, which are fundamental to achieving the complex structure of lepidopteran scales.
Identification of cuticular enzymes in scales associated with melanization and sclerotization. Previously discovered enzymes in the melanin pathway that regulate the intricate morphology of butterfly wing scales 17,22-24 led us to look for evidence of those enzymes in silk moth wing scales. We detected peptides for tyrosine hydroxylase (TH: Gene000277) 43 , dopa decarboxylase 44 (DDC: Gene002084), and dopamine N-acetyltransferase 45 (aaNAT: Gene010727) in developing P-8 scales. These findings suggested that N-acetyl-dopamine (NADA) 46 , a sclerotizing precursor molecule along with N-β-alanyl-dopamine (NBAD) [47][48][49] , can be produced from tyrosine (Fig. 1e). However, ebony (NM_001145321.1) 50,51 , which is an essential enzyme for the synthesis of NBAD from dopamine, was absent from silk moth scales, which indicated NBAD could not be produced during scale formation.
Not only are NADA and NBAD the final products of the melanin pathway in insects 24 , they are also considered the most common precursor molecules in the sclerotization process 48,49 . Andersen suggested a single highly reactive dehydro-NADA or dehydro-NBAD could react with two amino acid residues in different CPs, thereby forming stable crosslinks 48,49,52,53 . That peptides from laccase2A (Gene005466) 54 , tyrosinase (Gene009702), and prophenoloxidase (Gene009293) 55 were also identified in scales suggested that dehydro-NADA can be oxidized from NADA, which will lead to NADA sclerotin in scales. The absence of ebony indicates that NBAD sclerotin might not occur in scales. Silk moth scales are mainly white, consistent with the fact that NADA sclerotin is often colorless or very light straw-colored, whereas NBAD products of sclerotization contribute to a yellow color 24,49,53 .
His residues, which contain the strongest nucleophilic imidazole groups, have been confirmed to be involved in the sclerotization process in the mealworm (beetle) Tenebrio molitor and the desert locus Schistocerca gregaria 56 . The identification of CPs with a range of His residues together with a series of cuticular enzymes involved in sclerotization suggest the possibility that His-rich CPs have specific roles in the construction of scale microstructure. DDC and aaNAT were only detected in P-8 scales indicating that the pigmentation and sclerotization of wing scales occurred after P-6 or during P-8, before adult eclosion. These observations are consistent with our finding that scales were more soluble in P-6 than P-8. These findings are also consistent with the model proposed by Matsuoka and Monteiro for delayed cross-linking of melanin pathway gene products after the initial Table 1 Rank order of the most abundant proteins recovered based on iBAQ scores (data from Supplementary Data 1-4). Reduction of the mRNAs of the four His-rich CPs with different His-content (45, 27, 15, and 10%) and Bmlac2A led to abnormal microstructures in over half of the examined scales (details in Fig. 2b and Supplementary Data 5). Compared to normal (Fig. 3a, b), reduction of the mRNA for BmorCPR152, which encodes a 45% His-rich CP, produced distorted and bent ridges (magenta arrows in Fig. 3b) in the scale. Reducing mRNA for the 27% His-rich CP BmorCPR82 rendered a lack of microribs (orange arrow head in Fig. 3b) and disordered crossribs (yellow arrow head in Fig. 3b). After depletion of transcripts for BmorCPR57 (15% His) and BmorCPR129 (10% His), the junctions between ridges and crossribs (white arrows in Fig. 3b) were disordered or destroyed and the spacing of ridges became irregular; this damage was especially obvious with BmorCPR129. Reduction of Bmlac2A mRNA beginning in the pupal stage caused many abnormal microstructures. The cross ribs were intertwined at the window region (yellow arrow heads in Fig. 3b), which resulted in the interruption of the connection between adjacent ridges, and many unexpected striations intersected the microribs (white arrow head in Fig. 3b). The ridges also had abnormal structures (magenta arrows in Fig. 3b); and extra windows were found in the microrib area (blue arrow head in Fig. 3b). All the abnormal phenotypes that were observed from RNAi treatment of individual His-rich CP genes were found after depletion of transcripts of Bmlac2A. This similarity in the knockdown results of His-rich CP genes and Bmlac2A indicated that the processes of cuticular sclerotization played critical roles in making the microstructure of scales.
In addition, the overall development of the wing was affected by reduction of mRNAs of the His-rich CPs and Bmlac2A compared to control (Fig. 2c, Supplementary Fig. 4, Supplementary Data 5). After siRNA injection of the 15 and 10% His-rich CPs, wings were not fully expanded after the moth emerged ( Supplementary Fig. 4, rows 2 and 3). Reduction of Bmlac2A mRNA produced a more pronounced effect, in that moths had rather small wings ( Supplementary Fig. 4, row 4), whereas there was no apparent effect on wing structure from knockdown of the 45% and 27% His-rich CPs (Fig. 2c).
Staining with antibodies against BmorCPR152 (45% His-rich) and BmorCPR82 (27% His-rich) showed they were mainly distributed in the scales (yellow arrows in Fig. 4 rows 1 and 2), whereas BmorCPR57 (15% His-rich) was localized not only in scales (yellow arrows in Fig. 4 row 3), but also in the wing blades (white arrows in Fig. 4 row 3). This is consistent with our finding that RNAi with BmorCPR152 and BmorCPR82 caused severe damage only to the structure of the scales, whereas the depression of BmorCPR57 resulted in more widespread general wing abnormalities despite its lower His-content ( Supplementary Fig. 4, row 2).
We also examined the scales from the body and legs after RNA knockdown, which showed abnormalities similar to the microstructure of wing scales (Supplementary Figs. 5 and 6 and Supplementary Data 6). We found severe damage to the compound eyes and mandibular bristles after interference in the expression of the His-rich CPs and Bmlac2A ( Supplementary  Fig. 7). Especially, the surface of the compound eyes was shrunken by reduction of the mRNAs of BmorCPR152 and Bmlac2A (Supplementary Fig. 7a). In addition, some mandibular bristles were missing after RNAi treatment with BmorCPR152   c Percent of abnormal wings after RNAi (n = 3). Blue, GFP control; magenta, BmorCPR152; green, BmorCPR82; purple, BmorCPR57, orange, BmorCPR129, black, Bmlac2A. Significance difference: *P value < 0.05, **P value < 0.01, ***P value < 0.001, and ****P value < 0.0001.
Altogether these results suggested the His-rich CPs identified in wing scales also play important roles in the formation of cuticle structures in other organs.
Crosslinking with a His-rich CP in vitro. Crosslinking experiments of BmorCPR152, the CP with the highest His-content found in our study, were carried out in vitro to examine whether it can be cross-linked by NADA and NBAD derivatives with laccase2. The crosslinking reaction was performed according to previous reports 57, 58 and the reaction products were resolved on SDS-PAGE gels (Fig. 5a). The reaction of BmorCPR152 with NBAD/NADA and Bmlac2A formed dimer, trimer and tetramer at 0.5 h after the reaction started, and the higher polymer bands appeared at 1 h. However, using NADA as substrate presented weaker oligomer bands and rare polymer bands compared with NBAD, and importantly, monomers were deleted faster and far weaker at 1 h and almost used up at 2 h with NADA (Fig. 5a). This suggested that polymer formation with NADA was more rapid than with NBAD and led to polymers too large to penetrate into the running gel. That NADA showed very rare high polymer bands compared with NBAD may be because both samples were centrifuged at 12,000 rpm for 5 min after the reaction, which removed large suspended protein complexes from the NADA supernatant before loading onto the SDS-PAGE gel.
The amino acid analysis of BmorCPR152 showed that HHHH/ XHHH/XHH (X = H/D/G/S) repeats were evenly spread throughout the His-region ( Supplementary Fig. 8). Since the imidazole group of a histidine residue is highly nucleophilic, it is a candidate for binding to α, β-dehydro-NADA/NBAD-paraquinone methide 49 (Fig. 5b). Although other amino acids like lysine (Lys), alanine (Ala), and tyrosine (Tyr) could also be involved in crosslinking, the content of those amino acids in BmorCPR152 was very low (Lys: 0.43%, Ala: 2.5%, Tyr: 2.5%). It would also be worthwhile to study the relationship between crosslinking rate and His-content of different CPs.
Films synthesized by a His-rich CP. Unexpectedly, films appeared on the surface of the solution after eight hours of crosslinking using the 45% His-rich CP BmorCPR152 with Bmlac2A and NADA/NBAD. The films were easily crumpled and partly folded (yellow arrows in Fig. 5c, d) when hit with droplets of PBS from a pipette. Films synthesized using NADA as the substrate (Fig. 5c) appeared white after transfer to a Petri dish for observation, whereas the films synthesized with NBAD (Fig. 5d) were golden with the occurrence of light interference in the winkled area (white arrow in Fig. 5d). As Andersen suggested, a CP will link to the α, β carbon of a dehydro-NADA methide side chain, resulting in the formation of a colorless product, which we found in films synthesized using NADA; however, dehydro-NBAD methide is suggested to form links to CPs via the aromatic ring positions and β-positions of side chains, which lead to yellow or dark brown cuticle 45,49,59 , as we observed using NBAD. Furthermore, no film was synthesized in the absence of Bmlac2A or NADA or NBAD ( Supplementary Fig. 9), again consistent with previous reports 45,49,59 . Further study is needed to clarify the binding sites between the histidine imidazole groups of CPs and dehydro-NADA/NBAD-para-quinone methide. The film produced by BmorCPR152 with NBAD or NADA and Bmlac2A creased easily when removed from the reaction buffer with forceps, whereas it unfolded when returned to the buffer, suggesting that it was very soft. The elastic modulus value was estimated to be 0.17-0.20 GPa by AFM measurements (Supplementary Fig. 10). These values fall into the range of 1 MPa-10 GPa for soft rubbers and polymers 60 . Although more detailed measurement of the physical properties of such cuticular films is needed, the fact that they can be formed readily in vitro and in the absence of chitin points to a potential industrial application of cuticular proteins as new biomaterials.

Conclusion
Scales are a common attribute of Lepidoptera; however, no comprehensive information about the association of specific cuticle protein sequences with scale structure has been available. The experiments reported here indicate the intrinsic properties of the previously unstudied scale-forming CPs may play a central role in the final formation of the exquisite microstructure of lepidopteran scales.
In our present work, we were successful in dissolving the developing scales in pupal stage and pharate adult stages before they were cross-linked; this enabled us to identify a distinctive class of His-rich CPs in lepidopteran scales. Functional studies using RNA interference on four His-rich CP genes and Bmlac2A gave direct evidence that His-rich CPs and cuticle sclerotization play crucial and precise roles in the formation of the ridges, crossribs, and microribs of the scales.
Previous reports discussed above present details about the role of chitin and actin in scale development 25,26,41 . Importantly, none of these papers mentioned cuticular proteins. We have now shown that reduction of specific CPs can produce specific modifications of scale microstructure similar to those achieved by alteration of actins and chitin.
Importantly, our experiments established that His-rich CPs could be crosslinked with Bmlac2A and NADA/NBAD, yielding a His-rich CP film even without the involvement of other fibrilforming materials like F-actin or chitin.
Our results also represent a breakthrough for the future application of CPs as a new industrially useful biomaterial. Thus, it may be possible to design a diverse array of films and fibers made of CPs by controlling the His content, thickness, stiffness, strength, elasticity, and color during synthesis in vivo or in vitro.

Methods
Insects and sample preparation. B. mori, strain Dazao, was provided by the State Key Laboratory of Silkworm Genome Biology, Southwest University, China. The larvae were reared on fresh mulberry leaves at 25°C (day/night, 12/12 h). The wings were easily separated from the pupal cuticle at the seventh day after pupation (P-7), with adult eclosion occurring at day 8 or 9.
Wings were dissected out at the sixth and eighth days after pupation. After washing with PBS (pH~7.4) scales were carefully pulled off from the wings in PBS using sharpened forceps under a ZEISS Stemi2000 digital microscope (ZEISS, Germany). The PBS with scales was centrifuged (2000 rpm for 10 min) precipitating small fragments of wing and fatbody. These were discarded. The supernatant was then centrifuged at 12,000 rpm for 10 min to precipitate scales and leave contaminating hemolymph in the supernatant.
The purified scales were dissolved by shaking overnight in 8 M Urea/100 mM DTT in PBS at 4°C, and the supernatant was collected after centrifugation at 12,000 rpm for 30 min. The pellet was washed with 25 mM NH 4 HCO 3 , suspended in 50 µl SDT buffer, and boiled for 15 min, followed by treatment with an ultrasonic Processor (Power, 80 W; working time, 10 s; interval, 15 s; cycles, 10×). After centrifugation at 12,000 rpm for 30 min, the supernatant was collected. Both supernatants were combined for LC-MS/MS analysis.Thus there were four samples for LC-MS-MS, two replicates from P-6 and two from P-8. A flow chart of this procedure is in Supplementary Fig. 1.
Since we could only see a negligibly small amount of pellet after this treatment, we determined that scales from P-6 pupal wings were almost completely dissolved in the buffer. Scale samples from P-8 were incompletely dissolved after the same treatment. We could not collect scales before P-6, since the wing broke easily and scales disintegrated in the PBS buffer.
Microscopic analyses. The development of scale shape and size was observed using an OLYMPUS TH-200 digital microscope with accompanying software, Cellsens standard (OLYMPUS, Japan). Images of the whole wing and His-rich films were taken with a ZEISS Stemi2000 digital microscope (ZEISS, Germany). Detection of the fluorescence signal was carried out with a Zeiss LSM880 confocal microscope (ZEISS, Germany). Two days after moths emerged from siRNAinjected pupae, their dried wings, head, compound eye, mandibles and leg were removed and coated with Au (MCI000, Japan) for SEM using a Hitachi SU3500 SEM (Japan). For each repeat, we randomly picked three individuals from ten eclosed moths per siRNA gene treatment for phenotypic observation. Twenty scales from each wing and ten scales from each body and leg were checked. The compound eyes and mandibles of each individual were also checked.
Quantitative reverse transcription PCR. Whole wings were dissected and the pupal cuticle removed on the 4th, 6th, and 8th day after pupation and placed into TRIzol (Invitrogen, USA) after washing with PBS three times. RNA was isolated from wings with TRIzol (Invitrogen, USA) following the manufacturer's instructions and treated with Recombinant DNase I (RNase-free) (Takara Bio, China) to   Fig. 11) and their specificity were confirmed by Western blotting using wing scales dissolved on P-6 and P-8 ( Supplementary Fig. 12). For tissue immunolocalization, pupal wings were dissected at P-6 as previously described, fixed in 4% paraformaldehyde for 4 h at 4°C, and the cell nuclei were stained with DAPI (Beyotime Biotechnology, China). The primary and second antibodies were used at a dilution of 1:1000, and tissues were washed with PBST (0.01 M PBS, 0.01%(V/V) Triton X-100) more than three times after each incubation.
RNAi on His-rich CP genes and Bmlac2A. siRNA sequences (Supplementary Table 2) were based on candidate genes annotated in KAIKObase using GENETYX (Software Development Co., Tokyo, Japan). Chemically synthesized doublestranded siRNAs were purchased from Takara Bio Inc. (Dalian, China) and stored at −80°C. Two siRNAs designed against different regions of each target gene were mixed equally before use 63,64 . Injection time was determined based on the expression profile of His-rich CP genes in the wing during the pupal stage and pharate adult stage ( Supplementary Fig. 13). Because transcripts for all genes examined for RNAi treatment were at a maximum level after P-4, in these experiments siRNAs were injected into pupae twice, once at P-3 and again at P-5. About 6.6 µg (20 µl × 0.33 µg/µl) of siRNAs were injected into each abdominal cavity with a glass capillary needle inserted into the third spiracle on the left side.
The RNAi experiments were repeated three times using 13 pupae for each set of injections of each gene. The injected pupae were put back into their original cocoons and incubated at 25°C. Twelve hours after the second siRNA injection, wings were dissected from three pupae that were randomly chosen from each replicate to determine the efficiency of RNAi. Three of the remaining ten injected pupae of each set were used for the observation of phenotypic difference after adult eclosion.
Cross linking experiments with His-rich CPs in vitro. Signal peptide coding sequences of BmorCPR152 were predicted by SignalP 5.0 (http://www.cbs.dtu.dk/ services/SignalP/) 65 and removed from all CDS before using them to construct recombinant prokaryotic expression vectors by insertion into a pET-SUMO vector (provided by Wuhan Genecreate Biological Engineering Co., China), followed by expression and purification using a Genecreate Kit (Wuhan Genecreate Biological Engineering Co., China). SUMO Protease (Solarbio, China) was used to remove the SUMO-tag from the recombinant BmorCPR152 protein.
NADA and NBAD were provided by Dr. Tsunaki Asano after preparation according to Jerhot et al. 66 and Yamasaki et al. 67 , respectively. Bmlac2A was purified from the abdominal cuticles of pharate pupae 54,68 . A trypsin concentration of 0.25μg/μl was used to activate laccase2 at 37°C for 1 h before use followed by inhibition of trypsin activity with a protease inhibitor cocktail (MCE, USA) 67 .
Measurement of physical properties of in vitro cross-linked films. The films synthesized in vitro with the 45% His-rich CPs (BmorCPR152) were removed with forceps and, after washing with ddH 2 O, were dried on a silicon slice (provided by Kawasaki laboratory of Hitachi High-Technologies, Japan). Their physical properties were measured with an Atomic Force Microscope (AFM5100N, Kawasaki laboratory of Hitachi High-Technologies, Japan).
Statistics and reproducibility. Graphpad Prism6 (Graph Pad software, New York, NY, USA) was used for analyzing and plotting the data. Statistical values are shown as means ± SEM. Mean values were compared using the Student's t-test with the following significance thresholds: *P < 0.05, **P < 0.01, and ***P < 0.001. ****P < 0.0001.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The LC-MS/MS analysis raw data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD024401, PXD024365, PXD024362, and PXD024341. All other data generated or analysed during this study are included in this published article and presented in Supplementary Information and Supplementary Data files. All relevant data are available from the corresponding author K. Mita upon request (to mitakazuei@gmail.com).