A corset function of exoskeletal ECM promotes body elongation in Drosophila

Body elongation is a general feature of development. Postembryonically, the body needs to be framed and protected by extracellular materials, such as the skeleton, the skin and the shell, which have greater strength than cells. Thus, body elongation after embryogenesis must be reconciled with those rigid extracellular materials. Here we show that the exoskeleton (cuticle) coating the Drosophila larval body has a mechanical property to expand less efficiently along the body circumference than along the anteroposterior axis. This “corset” property of the cuticle directs a change in body shape during body growth from a relatively round shape to an elongated one. Furthermore, the corset property depends on the functions of Cuticular protein 11 A and Tubby, protein components of a sub-surface layer of the larval cuticle. Thus, constructing a stretchable cuticle and supplying it with components that confer circumferential stiffness is the fly’s strategy for executing postembryonic body elongation. Tajiri et al. describe how the cuticle coating the Drosophila larval body expands less efficiently along the body circumference than along the anteroposterior axis to drive body elongation. This “corset” property depends on cuticular proteins Cpr11A and Tubby, which may work together to change larval body shape.

T ransformation of a relatively round-shaped body into an elongated one is a general feature of development. Embryonic axis elongation in a wide range of animals has been commonly explained by active and coordinated cell behaviors 1 , such as oriented cell division 2,3 , cell intercalation 4-8 and cell migration 9,10 . Postembryonic body elongation requires a different consideration: because materials that frame the body during postembryonic life, such as the skeleton, the skin and the shell, are mainly shaped by extracellular matrix (ECM), body shape change must involve shape change of those materials outside of cells.
The insect body is coated by cuticle, a seamless sheet of ECM composed of proteins and chitin (a linear polymer of N-acetyl-beta-D-glucosamine (GlcNAc)) that are deposited by the epidermis 11 . The cuticle is continuously required throughout postembryonic life as an external skeleton that gives the insect its shape, supports movement, and protects against injury and desiccation 11,12 . In the developmental context, the textbook view of the cuticle is that it restricts changes in body shape and size, and that it needs to be replaced through molting to allow for those changes 11,13,14 . Nonetheless, stretching of cuticle between molts has been shown to accompany body size increase in some insects [15][16][17][18] . Furthermore, ECM dynamics has emerged as a general mechanism of insect morphogenesis 19 . For example, deformation of the cuticle covering the larval body of Drosophila drives body shape change without molting 20 , and elongation of wings and legs in Drosophila is controlled by ECM remodeling 21,22 . These findings suggest that body shaping during postembryonic development in insects depends not only on cuticle replacement but also on dynamic properties of the existing cuticle.
Here we demonstrate a "corset" property of the Drosophila larval cuticle that permits less efficient expansion of the cuticle along the body circumference than along the anteroposterior axis. During larval growth, this cuticle property directs a progressive change from a relatively round-shaped body to an elongated one. We further show that the corset property depends on the functions of two proteins that localize to a sub-surface layer of the larval cuticle, Cuticular protein 11 A (Cpr11A) and Tubby (Tb). The stretchable cuticle on which circumferential stiffness is conferred by its components allows the fly to execute postembryonic body elongation.

Results
Abnormal shaping of the larval cuticle in Cpr11A mutants. We found that the pupae of MB00532 homozygotes or hemizygotes, which harbor a transposon insertion immediately upstream of Cpr11A (Fig. 1a), appeared laterally wider and anteroposteriorly shorter than the wild-type pupae (Fig. 1c-f). The abnormal shapes of MB00532 pupae were represented by their body axial ratios (length/width ratios), which were significantly smaller than those of the wild-type ( Fig. 1k; compare + /+ with MB/MB and + /Y with MB/Y). Df(1)Exel6244, a deficiency uncovering Cpr11A along with six neighboring genes (Fig. 1a), failed to compliment the tubby phenotype of MB00532 (Fig. 1g, k; compare MB/Df with MB/MB). Df(1)Exel6244 hemizygotes were viable and exhibited even tubbier pupal shapes (represented by smaller axial ratios) compared with the MB00532 hemizygotes (Fig. 1h, k; compare Df/Y with MB/Y). Df(1)Exel6244 homozygous females were larval lethal. The tubby phenotype was recapitulated by whole-body expression of dsRNA against Cpr11A in the wild-type background ( Fig. 1k; compare Act-GAL4 with Act > Cpr11A dsRNA). Conversely, the tubby phenotype of MB00532 or Df(1) Exel6244 was rescued by inducing whole-body expression of Cpr11A cDNA (Fig. 1i, k; MB/Y; Act > Cpr11A), or by introduction of Cpr11A:EGFP, a rescue construct containing the genomic sequence of Cpr11A into which an EGFP-coding sequence was inserted immediately upstream of the Cpr11A stop codon (Fig. 1a, j, k; Df/Y; Cpr11A:EGFP/ + ). These results indicated that loss of Cpr11A function was responsible for the tubby body shape phenotype. Df(1)Exel6244 hemizygotes are hereafter referred to as Cpr11A Df .
Cpr11A encodes a protein harboring an N-terminal signal peptide and a Rebers and Riddiford (R&R) domain, a chitinbinding domain conserved in numerous arthropod cuticular proteins 23 (Fig. 1b). This suggested that Cpr11A might affect shaping of the cuticle. The third instar cuticle undergoes anteroposterior contraction and circumferential expansion at the onset of metamorphosis to directly become the puparium (pupal case) that covers the pupal body 20 . To test whether the pupal shape differences between the wild-type and the Cpr11A mutants reflect the effect of Cpr11A on larval cuticle shaping or on the metamorphic shape change, we isolated and made flat preparations of cuticles of wild-type, Cpr11A Df and MB00532 third instar larvae, and measured their axial ratios. We confirmed that the differences in pupal axial ratios mirrored the differences in the axial ratios of the larval cuticle per se (Fig. 2aVII, bVII, f, Supplementary Fig. 1). These results suggest that Cpr11A is required for normal shaping of the larval cuticle.
Cpr11A-dependent anisotropy in cuticle expansion during larval growth. The Drosophila larval period is divided into three instars by two molts, and lasts for approximately four days in total. During this period, the larva undergoes rapid and continuous increase in body weight (approximately 100-fold) [24][25][26] , and the surface area of the cuticle increases in a continuous manner both between and across molts 18 . To investigate the origin of the cuticle shape abnormality of Cpr11A mutants, we analyzed the cuticle shapes of Cpr11A Df larvae of various stages and body sizes. Siblings carrying the FM7i, Act-GFP balancer were used as controls.
The first instar larval cuticle is formed de novo during late embryogenesis, and is completed near the end of embryogenesis (stage 17d according to ref. 27 .). The shapes of cuticles isolated at this stage were comparable between the control and the Cpr11A Df mutant (Fig. 2aI, bI, dI), as represented by comparable axial ratios of the cuticle (Fig. 2e, f). After hatching, the cuticle widths and lengths increased in different proportions in the control and the mutant (Fig. 2d). Notably, the cuticle of the control larvae expanded anisotropically: it expanded at a higher rate along the length (body anteroposterior axis) than along the width (body circumference) (Fig. 2aI-VII), as represented by the progressive increase in the cuticle axial ratio (Fig. 2e, f). In contrast, cuticle expanded in a relatively isotropic manner in Cpr11A Df larvae ( Fig. 2bI-VII): the cuticle axial ratio remained fairly constant while the cuticle length increased (Fig. 2e), and changes in the cuticle axial ratio between stages were insignificant except for a small increase from the first instar to the second (Fig. 2e, f). The difference in anisotropy of cuticle expansion between the control and the mutant ultimately resulted in the marked difference in the body shapes of third instar larvae (Fig. 2aVII, bVII) and pupae (Fig. 1). These results indicate that, while Cpr11A does not visibly affect the initial shaping of larval cuticle during embryogenesis, it is required for the anisotropy of cuticle expansion during larval growth. Cpr11A may limit expansion of cuticle along the body circumference, and/or promote cuticle expansion along the anteroposterior axis, during body size increase.
Comparison with the Tubby 1 mutant phenotype. The Cpr11A mutant phenotype was reminiscent of Tubby 1 (Tb 1 ) and Twee-dleD 1 (TwdlD 1 ) mutations, which individually cause tubby larval and pupal shapes in a dominant manner. Tb 1 and TwdlD 1 are amino acid deletions in Tb/TweedleA and TwdlD, respectively, which are two of the Tweedle family of cuticular proteins 28 . The axial ratios of Tb 1 homozygotes (referred to as Tb 1 hereafter) cuticle isolated at embryonic stage 17d were comparable to those of the control and Cpr11A Df cuticle at the same stage (Fig. 2aI, bI, cI, d-f). During the first instar, the cuticle length-width plot of Tb 1 overlapped with that of Cpr11A Df , and the cuticle axial ratios of Tb 1 and Cpr11A Df were comparable ( Fig. 2aII-III, bII-III, cII-III, d-f). Then, the axial ratio of Tb 1 cuticle decreased significantly from the first instar to the second (Fig. 2e, f), representing a higher rate of cuticle expansion along the body circumference relative to expansion along the anteroposterior axis. The ratio further decreased from the second instar to the third (Fig. 2e, f). Thus, the Tb 1 mutation reduces or reverses the anisotropy of cuticle expansion during larval growth.
To test whether the Cpr11A and Tb mutations affected the growth of the larva, the ultimate body sizes of wild-type, Cpr11A Df, MB00532 and Tb 1 larvae were assessed by weighing male post-feeding (wandering) larvae. No statistically significant difference was found among them ( Supplementary Fig. 2), indicating that the mutations did not affect the absolute body size.
Direct measurement of cuticle stiffness. Stretching of the larval cuticle due to rapid and continuous growth of the body volume has been proposed as one mechanism for cuticle expansion 18 . To test whether the difference in anisotropy of cuticle expansion between the control and the Cpr11A Df or Tb 1 mutants might reflect difference in how the cuticle responded to mechanical stretch, we directly applied circumferential or anteroposterior pulling force to the cuticle of third instar larvae through metal pads glued to the cuticle and examined how the cuticle extended in response (e.g. Figure 3a, see Methods). Figure 3b and d show how the cuticle strain, defined as increase in cuticle width/length divided by the initial width/length, changed in response to stepwise increasing stress, defined as pulling force applied per unit length/width of the cuticle. The strain-stress plots were roughly The full-length Cpr11A protein fused at its C-terminus with EGFP is expressed from the Cpr11A:EGFP transgene. Nonsecretory DsRed is expressed from the Cpr11A-DsRed transgene. b Domain structure of Cpr11A. The signal peptide predicted by SignalP5.0 46 , the R&R domain (PF00379) by Pfam 49 , the low complexity region and the disorder regions by MobiDB 50 are shown. c-j Pupae of the indicated genotypes, viewed from the dorsal side. Anterior is to the left. The length (L) and the width (W) of a pupa were measured as indicated by arrows. MB, MB005322; Df, Df(1) Exel6244; Act > Cpr11A, Act-GAL4/UAS-Cpr11A; Act > Cpr11A dsRNA, Act-GAL4/UAS-Cpr11A dsRNA. Bar, 1 mm. k Mean ± S.D. of pupal axial ratios (L/W) of individual genotypes. n, the number of pupae measured for each genotype. Significance was assessed using one-way ANOVA (p < 0.0001) and Tukey's multiple comparisons test. *p < 0.05; ****p < 0.0001; broken brackets, not significant. and Tb 1 (c) at embryonic stage 17d (I), early/late first instar (II/III), early/late second instar (IV/V), and early/late third instar (VI/VII). In (I-III), magnified images are shown on the right. Cuticles are viewed from the ventral sides. Anterior is to the left. L, cuticle length; W, cuticle width. Bars, 500 μm. d Plot of cuticle length (L) versus cuticle width (W). 1st, first instar; 2nd, second instar; 3rd, third instar. e Plot of cuticle axial ratio (L/W) versus cuticle length (L), calculated from the data in d. f Mean ± S.D. of axial ratios for each developmental stage, calculated from the data in (d). n, the number of embryos or larvae measured for each genotype for each stage. Significance was assessed using two-way ANOVA (p < 0.0001) and Tukey's multiple comparisons test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; broken brackets, not significant. linear in the area where stress was larger than 1 N/m, as indicated by high strain-stress correlation coefficients in the area (Supplementary Data). Stiffness was defined as the slope derived from linear regression of stress against strain in the area. (See Methods for details.) While we did not find statistically significant differences in the anteroposterior stiffness (Fig. 3d, e), the circumferential cuticle stiffness of Cpr11A Df was significantly smaller than that of wildtype control cuticle (Fig. 3b, c). These results indicated that Cpr11A normally functions to limit the circumferential extension of cuticle. It should be noted that, although force application and measurement of extension were done on the cuticle, potential contributions from internal tissues such as the epidermis and the musculature were not eliminated, because the cuticles were not isolated in these measurements. Cpr11A may restrict circumferential cuticle expansion either directly, or indirectly through its effect on physical forces that the internal tissues exert upon the cuticle. The "corset" property conferred by Cpr11A, either directly or indirectly, is consistent with the Cpr11A-dependent anisotropy of cuticle expansion during larval growth; in the absence of Cpr11A function, the cuticle extends more easily along the circumference in response to internal tissue growth.
The circumferential cuticle stiffness of Tb 1 was significantly smaller than that of the normal cuticle and was comparable to that of Cpr11A Df (Fig. 3b, c). We did not find significant difference between Cpr11A Df and Tb 1 in cuticle stiffness, either along the body circumference or along the anteroposterior axis ( Fig. 3b-e) (see Discussion).
Expression and protein localization of Cpr11A. Figure 4a shows the temporal profile of Cpr11A mRNA expression extracted from modENCODE developmental transcriptome data 29 . Cpr11A expression was up-regulated in the latter half of embryogenesis (when the synthesis of the first instar larval cuticle initiated 27 ), persisted through first, second and early third instar larval stages, and decreased concomitantly with feeding cessation in the late third instar. Thus, high levels of Cpr11A expression occurred coincidently with larval cuticle synthesis and larval body growth.
According to in situ hybridization data of the Berkeley Drosophila Genome Project 30,31 , Cpr11A mRNA was expressed in the epidermis in a segmentally striped pattern in late embryos. To efficiently monitor gene expression and protein localization in the larva, we established a transgenic strain harboring two reporters, Cpr11A:EGFP and Cpr11A-DsRed (Fig. 1a). Functionality of the former in rescuing the body shape abnormality of the Cpr11A Df mutant as described above (Fig. 1i, j) suggested that the expression and the localization of the EGFP-tagged Cpr11A protein recapitulated those of the endogenous Cpr11A protein.
The Cpr11A-DsRed reporter consisted of the Cpr11A gene region in which the 5' part of the Cpr11A coding sequence was replaced by a DsRed-encoding sequence followed by a stop codon and a polyadenylation signal, so that DsRed is not secreted and cells expressing the gene would be labeled by nonsecretory DsRed (Fig. 1a, see Methods). In first, second and third instar larvae, Cpr11A-DsRed and Cpr11A:GFP were expressed in a segmentally reiterated graded pattern, with high-level expression in the middle of every segment (Fig. 4b, c). In close-up views, DsRed signals were detected in the epidermal cells while the Cpr11A: EGFP protein was localized in the cuticle (Fig. 4d, cross-sections). When viewed from the surface, Cpr11A:EGFP was distributed throughout the cuticle and was enriched in a meshwork that transcended cell borders and continued over cells with much lower levels of Cpr11A-DsRed expression (Fig. 4d). The protein was also enriched in muscle attachment sites, despite lack of Cpr11A-DsRed expression in the underlying tendon cells (Fig. 4d). These observations suggest that the Cpr11A protein spreads and forms meshwork in the extracellular space.
Protein interactions in a sub-surface region of the cuticle. The insect cuticle is histologically subdivided into three layers: the envelop (the most apical (external) layer), the epicuticle (the medial layer consisting of proteins) and the procuticle (the basal layer consisting of proteins and chitin) 32 . The fluorescently tagged Tweedle proteins Tb:GFP 33 and TwdlD:DsRed 28 have been reported to mark an apical region of the larval cuticle. In our observation, the two proteins colocalized in a layer apical to the chitin-containing procuticle (Fig. 5a), and thus probably marked the epicuticle. Cpr11A:EGFP formed a layer that spanned the basal portion of the TwdlD layer and the most apical portion of the procuticle (Fig. 5b-e), indicating that Cpr11A:EGFP was localized around the epicuticle-procuticle boundary. The apical side of Cpr11A:EGFP signals were flat while the basal side had  fine protrusions, which corresponded to the meshwork of Cpr11A:EGFP enrichment observed in surface views (Fig. 4d). Restricted localization of Cpr11A to the sub-surface region of the cuticle that is not in direct contact with the epidermis is suggestive of the protein's direct effect on cuticle mechanical property.
In the Tb 1 mutant background, although Cpr11A:EGFP localization was normal during the first instar (Fig. 5f), a fraction of Cpr11A:EGFP formed bubble-like structures that contained TwdlD:DsRed and bulged basally during the second and the third instars (Fig. 5g, h).The remaining fraction localized to a subsurface layer of the cuticle (Fig. 5g, h) but did not form a meshwork pattern (Fig. 5i, compare with Fig. 4d). Cpr11A:EGFP bubbles were also induced by the presence of the Tb 1 :DsRed transgene, which expressed the DsRed-tagged mutant form of Tb and reproduced the Tb 1 body shape phenotype 34 . The Cpr11A: EGFP bubbles contained the Tb 1 :DsRed protein (Fig. 5o). The basal expansion of the epicuticle marker TwdlD:DsRed in the Tb 1 background was reminiscent of abnormal epicuticle formation in the Tb 1 Chubby/Kugel v double mutant, which was associated with accumulation of epicuticle materials in the procuticle 18 . We speculate that the Tb 1 protein singly interfered with normal epicuticle formation, which in turn affected Cpr11A localization.
In contrast, the localization of Tb:GFP was not visibly affected in the Cpr11A Df mutant (Fig. 5p). Consistently, the cuticle ultrastructure of the Cpr11A Df mutant, including the epicuticleprocuticle boundary, appeared normal (Fig. 6). Moreover, the localization of Tb:GFP and TwdlD:DsRed to the cuticle apical layer was invariable over areas with different levels of endogenous Cpr11A expression (Supplementary Fig. 3). Thus, the Cpr11A protein is likely to confer circumferential stiffness on the cuticle without affecting its overall organization.
To address whether the Cpr11A mislocalization was responsible for the abnormal cuticle shape of the Tb 1 mutant, we tested whether the Tb 1 mutation affected pupal shapes in the absence of Cpr11A. Interestingly, the Tb 1 mutation caused significant difference in pupal axial ratios even in the absence of Cpr11A (Fig. 5j, k, n), suggesting that some corset property was retained even without Cpr11A. Loss of Cpr11A function in the Tb 1 mutant background caused significant decrease in pupal axial ratios relative to the Tb 1 single mutant (Fig. 5k-n), indicating that, despite partial mislocalization and disruption of the meshwork pattern, Cpr11A was still functional in the Tb 1 mutant background.
Extension of the epidermis. We have shown that Cpr11A enables anisotropic cuticle extension through direct regulation of the cuticle mechanical property. Meanwhile, the epidermis continues to adhere to the cuticle within each instar, which necessitates anisotropic extension of the epidermis that matches the cuticle extension. Tissue extension events have been accounted for by coordinated cell behaviors such as cell rearrangement and oriented cell division. On the other hand, the Drosophila larval epidermal cells are known to undergo size increase by endoreplication throughout the larval period 35 . To examine cell dynamics in the epidermis during larval growth, we examined how the epidermal cell outlines (labeled with Neuroglian (Nrg): GFP 36 ) changed during growth, either across a molt or within a single instar, in live larvae of the wild-type, Cpr11A Df or Tb 1 . In all three genotypes, the initial cell outlines were closely matched to those after growth by simply enlarging the initial images by different percentages along the body length and width (Fig. 7,  Supplementary Fig. 4). This result indicates that active cell behaviors such as rearrangement and oriented division occur rarely, if at all, in the epidermis during larval growth. Anisotropic extension of the epidermis could be either mechanically induced through adherence to the anisotropically extending cuticle, or oriented cell growth, or both. Within each genotype, the lengthto-width ratios of epidermis extension within the third instar were not significantly different from those from the second instar to the third ( Supplementary Fig. 4m, n). The continuous extension of the epidermis is in line with the linear increase in cuticle lengths and widths, both between and across molts, within each genotype (Fig. 2d). The epidermal cells in the Tb 1 mutant appeared highly anisometric (elongated along the body circumference; Fig. 7g, h, Supplementary Fig. 4i-l). It is consistent with the constantly circumference-biased extension of the epidermis in this mutant (Supplementary Fig. 4m).
Effects of Cpr11A Df and Tb 1 mutations on adult shapes. The Tb 1 mutation has been anecdotally reported to cause "tubby" body shape even in the adult 37 . To examine how Cpr11A Df and Tb 1 mutations affect the adult body shape, we measured the axial ratios in the abdomens and the thoraces (nota) of adult males. The Tb 1 mutant indeed showed lower axial ratios (meaning shorter and/or wider shape) in both the abdomen and the notum relative to the control ( Supplementary Fig. 5a, c, d, f, g, h). In the Cpr11A Df mutant, the axial ratios of the abdomen were significantly lower than those of the control, although the axial ratios of the notum did not show significant difference ( Supplementary  Fig. 5a, b, d, e, g, h). As neither Cpr11A nor Tb is expressed at a substantial level during the pupal stage, in which the adult cuticles are formed (Fig. 4a and Supplementary Fig. 5i), they are unlikely to function directly within the adult cuticles. We speculate that the tubby puparia of the mutants mechanically confine the shapes of the adult bodies that form inside them.

Discussion
We have shown that circumferential stiffness ("corset" property) of the Drosophila larval cuticle directs anisotropic cuticle expansion during body size increase by restricting circumferential cuticle stretching, which results in progressively more elongated shape of the whole body (Fig. 8a). Axis elongation has been commonly explained by cell behaviors such as oriented cell division, cell shape changes, cell intercalation and cell migration. Our finding demonstrates that the cuticle can play an instructive mechanical role in body elongation. The corset property depends on the function of protein constituents of the cuticle, Cpr11A and Tb, which show overlapping localization in a sub-surface region of the cuticle (Fig. 8a). Postembryonic survival in harsh environments necessitates animals to support and protect their bodies with materials with greater strength than cells. Coating the body with a stretchable cuticle and equipping it with components that confer circumferential stiffness appears to be a potent strategy for conciliating active morphogenesis with the protective function of the cuticle. In the nematode Caenorhabditis elegans, a number of mutations in cuticle collagen genes are known to result in the "dumpy" phenotype (short and fat body shape) 38 . Regulation of cuticle physical properties by cuticle constituents may be a general mechanism for body shaping in diverse organisms bearing exoskeleton.
Generally, the cuticle remains adhered to the epidermis during its synthesis. During a molt, separation of the epidermis from the cuticle, initiation of new cuticle synthesis, and shedding of the old cuticle occur sequentially. The normal shape of the first instar cuticle of the Cpr11A Df and Tb 1 mutants before hatching (Fig. 2aI, II, bI, II, cI, II, d-f) is likely to mirror normal morphogenesis of the epidermis during embryogenesis. Within each instar, cuticle length and width increase continuously (Fig. 2d). Contrarily to some insects whose cuticle is initially formed in folds which later unfold to accommodate body growth 39 , the fly larval cuticle does not show unfolding; it undergoes continuous expansion within an instar, both by stretching and by addition of new components 18 . Assuming that addition of new components to the cuticle contributes equivalently to anteroposterior and circumferential expansion of the cuticle, restriction of circumferential cuticle stretching by the Cpr11A-and Tb-dependent "corset" property would result in anisotropic cuticle expansion. Upon molting, cuticle lengths and widths, as well as differences in cuticle axial ratios between genotypes, are carried over from one instar to the next (Fig. 2d-f). Presumably, the shape of the old cuticle is transmitted to the new cuticle through the epidermis: when new cuticle starts to be synthesized, its shape conforms to the surface shape of the epidermis, which in turn mirrors the shape of the cuticle of the previous instar from which the epidermis has just separated.
Here let us simplify stretching of the cuticle within each instar as inflation of a tube-shaped balloon in response to increasing internal pressure. The thin-walled pressure vessel theory posits that circumferential stress acting on the tube wall (cuticle) is twice as large as longitudinal (anteroposterior) stress 40 . Consistently, a blown up tube-shaped rubber balloon expands more efficiently along its circumference than along its longitudinal axis (Fig. 8b). If the cuticle stiffness along the body circumference is equivalent to that along the anteroposterior axis, the cuticle would stretch at a higher rate along the circumference than along the anteroposterior axis, which may be the case for the Tb 1 mutant. Fig. 8 A corset function of the cuticle. a A summary of this study. b A rubber balloon before (upper) and after (lower) inflation. In this case, the balloon expanded 3.2-fold along its longitudinal axis and 5.4-fold along its circumference.
The isotropic cuticle extension seen in the Cpr11A Df mutant would necessitate that cuticle be twice as stiff along the circumference as along the anteroposterior axis. If so, the Cpr11A protein in the wild-type larva would function to make the cuticle even stiffer along the circumference, thereby directing the cuticle to extend more preferentially along the anteroposterior axis. Cpr11A may confer circumferential stiffness directly on the cuticle, although, as noted above, the indirect action of Cpr11A through mechanics of internal tissues such as the epidermis and the musculature cannot be ruled out. We have not been able to explain the divergence in cuticle shaping between Cpr11A Df and Tb 1 . The large variations in our measurement of the anteroposterior cuticle stiffness (Fig. 3e) might have masked potential difference between Cpr11A Df and Tb 1 . The large variations also made it difficult to compare circumferential and anteroposterior stiffness values within individual genotypes.
Larval body elongation assisted by anisotropic cuticle expansion is reminiscent of the Drosophila egg elongation assisted by the surrounding ECM. Although it was initially proposed that the ECM itself acts as a molecular corset 41,42 , it has recently been reported that F-actin within the egg acts as a molecular corset and that the ECM acts as a cue to polarize the F-actin 43 . In the case of larval body elongation, localization of Cpr11A and Tb to the subsurface region of the cuticle that is not in direct contact with the epidermis suggests that those proteins directly confer the "corset" property on the cuticle. Nonetheless, the possibility that the proteins affect biophysical properties of the internal tissues, for example by acting as "signals" to the epidermis or the musculature, is not excluded. Future studies are needed to specify sites of action of those proteins.
Overlapping localization of Cpr11A, Tb and TwdlD, together with Tb 1 -induced mislocalization of Cpr11A and TwdlD in or around the basal Tb 1 aggregates, suggests attractive interactions among those proteins. We speculate that those multiple interactions contribute additively to the corset property. The result that the Cpr11A Df and Tb 1 double mutant shows a more severe effect on cuticle shaping than either single mutant supports this model. Localization of Cpr11A and Tb to the sub-surface region of the cuticle is consistent with their effect on cuticle expansion and with the lack of effect on active cell behaviors in the epidermis. How are the protein localized to the cuticle region away from the cells from which they derive? One explanation would be that temporary deposition of those proteins by the epidermis prior to procuticle formation allows sequential stacking of the proteins and the procuticle. The temporal expression pattern of Cpr11A contradicts this scenario. For example, Cpr11A is expressed at high levels near the end of embryogenesis (18-24 h after egg laying (AEL), Fig. 4a), whereas the procuticle of the first instar cuticle is apparent in the embryo as early as 13-15 h AEL and has substantially thickened by 19 h AEL 27 . Likewise, Cpr11A expression is sustained at high levels for at least 12 h after the second molt (Fig. 4a), whereas the procuticle of the third instar cuticle is present before the molt and thickens continuously 18 . From the observation that the epicuticle thickness shows no decrease despite large increase in surface area during the second and the third instars, Kaznowski et al. inferred that the epidermis secretes material that moves through the procuticle to be added to the epicuticle 18 . Exclusive localization of the Cpr11A protein to the epicuticle-procuticle boundary despite its sustained production by the epidermis suggests that the protein penetrates through the procuticle to reach its destination. The R&R chitinbinding domain of Cpr11A may facilitate movement through the chitin-containing procuticle. Intracuticular movement of Cpr11A could also contribute to its planar spread into regions where the underlying cells do not express the Cpr11A gene (Fig. 4d). The attractive interaction between Cpr11A and the Twdl proteins discussed above, together with Cpr11A affinity to chitin, may stabilize Cpr11A around the epicuticle-procuticle boundary. We envision that cuticular protein interactions and dynamics that occur within the cuticle are of general significance in understanding cuticle organization and properties.

Methods
Fly stocks. Flies were maintained on standard yeast-cornmeal food at 25°C unless otherwise mentioned. Canton-S was used as wild-type. MB00532, Df(1)Exel6244, Tb 1 , FM7i Act-GFP, CyO 2xTb 1 -RFP (2xTb 1 :DsRed) and Actin5C-GAL4 (Act-GAL4) were obtained from the Bloomington Stock Center, and Nrg G00305 (Nrg: GFP) was from the Kyoto Stock Center. v103321 obtained from the Vienna Drosophila Resource Center was used as UAS-Cpr11A dsRNA. Tb:GFP was a gift from B. Moussian and TwdlD:DsRed was a gift from S. Wasserman. Tb 1 heterozygotes and homozygotes are morphologically inseparable 37 . To simplify mating schemes, we introduced transgenes and other mutations into the Tb 1 heterozygous background in Figs. 5, 7 and Supplementary Fig. 4.
Transgenic flies. To generate the UAS-Cpr11A construct, we subcloned Cpr11A cDNA from RE57452 of Drosophila Gold Collection 44 into pUAST using EcoRI and KpnI. For generation of Cpr11A:EGFP and Cpr11A-DsRed constructs, pCaS-peR4-Cpr11A was established first by amplifying the 7.1 kb Cpr11A gene region flanked by the neighboring genes (Fig. 1a)  The two amplified fragments were inserted into the BglII-digested construct by In-Fusion, so that the protein to be expressed from the construct would consist of Cpr11A amino acids 1-9 (predicted by SignalP 46 to be nonfunctional as a signal peptide) fused in frame to DsRed. Transgenesis was done by BestGene. The transgenic strains are available from the corresponding authors upon request.
Microscopy. Larval cuticle and epidermis was fluorescently labeled and observed on the Olympus FV-1000 confocal laser microscope as described in 20 . Larval cuticle preparations and adult abdominal cuticle preparations were made according to 47 and 48 , respectively, and were observed on the Olympus BX51 microscope. For transmission electron microscopy, ultrathin sections were made and were viewed on the JEOL JEM-1400 transmission electron microscope as described in 20 .
Measurement of cuticle elasticity. For circumferential measurement, a planar metal pad (a half-bent staple) was glued to the right side of posterior cuticle of a third instar larva deeply anesthetized with diethyl ether vapor. The larva was placed on a rubber plate with its ventral side up, and the metal pad was fixed to the rubber plate with a pin. Another metal pad was glued to the left side of posterior cuticle, and was attached to one end of a tension spring (spring constant = 5.1 ×10 −4 kgf/mm; Goko Hatsujo Co., Ltd.). The other end of the spring was pulled away in a stepwise manner (<1 mm per step, from step 0 (before pulling) to step 10) (Fig. 3a). The images of the cuticle and the spring were captured at each step on a Leica M165 FC stereo microscope, and the cuticle width and the spring length at each step were measured on ImageJ. We limited the placement of the metal pads (and thus the pulling) to the posterior part of the larval body so that increase in internal pressure arising from cuticle pulling would be released to the anterior part of the body. Circumferential cuticle strain at step i was defined as w i /w 0 −1, where w i is the cuticle width at step i and w 0 is the cuticle width before pull (step 0). Force applied to the spring at step i (f i ), calculated from the spring extension, was normalized by the average length of the pad-cuticle adhesion (d) to yield stress per cuticle unit length (f i /d). As shown in Fig. 3b and d, the stress-strain plots were roughly linear in the area where stress >1 N/m, as indicated by high correlation coefficients (Supplementary Data). We defined circumferential stiffness of cuticle as the slope derived from linear regression of the stress-strain plot in this area. For anteroposterior elasticity measurement, metal pads were attached to anterior and posterior parts of cuticle on the ventral side, so that increased internal pressure would be released to the dorsal side of the body. Subsequent procedures and data analyses were carried out as done for the circumferential measurement.
Imaging of epidermal cell outlines. Second instar or early third instar larvae expressing Nrg:GFP were anesthetized with diethyl ether vapor and mounted with water on glass slides. The epidermis of the dorsal side of the abdominal segment 1 was imaged by confocal laser microscopy, and the larvae were individually returned to fly media in 1.5 ml tubes. After incubation for one or two days at 25°C, the epidermis of the same region was imaged again. In Fig. 7, Df(1)Exel6244, Nrg:GFP/Y are described as Cpr11A Df . The progenies of crosses between Nrg:GFP females and either Canton-S males or Tb 1 males are described as control and Tb 1 /+, respectively.
Weighing larvae. For each measurement, ten wandering larvae were collected and weighed on an electronic scale (Mettler Toledo AB104). 5-6 measurements were conducted per genotype.
Statistics and reproducibility. Data are presented as means ± S. D. The significance of differences between multiple groups were compared by one-way or two-way analysis of variance (ANOVA), followed by Tukey's multiple comparisons tests.