A Matrix Metalloproteinase Mediates Airway Remodeling in Drosophila

Abstract

Organ size typically increases dramatically during juvenile growth. This growth presents a fundamental tension, as organs need resiliency to resist stresses while still maintaining plasticity to accommodate growth. Extracellular matrix (ECM) is central to providing resiliency, but how ECM is remodeled to accommodate growth is poorly understood. We investigated remodeling of Drosophila respiratory tubes (tracheae) that elongate continually during larval growth, despite being lined with a rigid cuticular ECM. Cuticle is initially deposited with a characteristic pattern of repeating ridges and valleys known as taenidia. We find that for tubes to elongate, the extracellular protease Mmp1 is required for expansion of ECM between the taenidial ridges during each inter-molt period. Mmp1 protein localizes in periodically-spaced puncta that are in register with the taenidial spacing. Mmp1 also degrades old cuticle at molts, promotes apical membrane expansion in larval tracheae, and promotes tube elongation in embryonic tracheae. Whereas work in other developmental systems has demonstrated that MMPs are required for axial elongation occurring in localized growth zones, this study demonstrates that MMPs can also mediate interstitial matrix remodeling during growth of an organ system.

Introduction

During embryonic morphogenesis, organisms and their tissues adopt characteristic forms. These juvenile forms are maintained in the face of external pressures from the environment and internal growth processes. Resilience is conferred mainly by the extracellular matrix (ECM), a network of proteins and proteoglycans that can vary from rigid (e.g. in bone or exoskeleton) to flexible (e.g. in basement membrane). However, despite ECM and tissue resilience, tissues and organisms do not remain static. Developmental transitions from the juvenile to the adult form require tissue remodeling, and even maintaining homeostasis in response to growth or wound healing requires plasticity. This tension between maintaining and altering form is a fundamental feature of multicellularity.

In Drosophila melanogaster, larvae secrete a chitinous exoskeleton called the cuticle, which covers the epidermis and also lines the tubes of the internal respiratory organ, called the tracheal system. The rigid nature of cuticle limits the ability of the organism to grow, yet despite the cuticle, the larva elongates about 8-fold before undergoing metamorphosis to take on adult form. Two larval molts occur during which animals shed and replace their exoskeletal cuticle, but significant axial elongation occurs in the inter-molt period, or instar. This instar growth raises questions about how the exoskeletal cuticle maintains its integrity while permitting elongation and growth.

Within the Drosophila larva, the tracheal organ presents an opportunity to examine the mechanisms of ECM maintenance and growth. The tracheae are a system of ramifying epithelial tubes that supply oxygen to the cells throughout the animal. The tubes are lined with a barrier of apical (lumenal) chitinous ECM continuous with the exoskeletal cuticle of the animal. When a larva sheds its exoskeleton at molting, the attached tracheal cuticular lining is dragged out through the spiracles, openings that connect the tracheal tubes with the atmosphere, as the animal crawls out of its old cuticle (reviewed in Manning and Krasnow, 1993). Before this old tracheal cuticle is shed, a new tracheal cuticle is deposited along the apical surface of the epithelial tube so that there is continuous protection of the epithelium. Tracheal tubes expand their diameters only at molts when the cuticle is replaced; in contrast, they elongate continuously throughout larval growth (Beitel and Krasnow, 2000). Tracheal growth occurs in the absence of cell division, with tracheal cells growing by increasing cell size and ploidy without increasing cell number. As tracheal elongation occurs in the continuous presence of cuticle, it would seem to require cuticle remodeling.

We previously found that a Drosophila matrix metalloproteinase, Mmp1, was required for normal tracheal growth, as Mmp1 mutant tracheae become stretched and broken as the larva grows (Page-McCaw et al., 2003). Matrix metalloproteinases are a conserved family of extracellular proteases that are upregulated in cancer and inflammation (reviewed in Sternlicht and Werb, 2001); analysis of mouse and fly MMP mutants indicates that they are broadly required for tissue remodeling (Page-McCaw, 2008; Page-McCaw et al., 2007). Although little is known about fly Mmp1 substrates, it is known that the vertebrate MMP family is capable of cleaving many components of the extracellular matrix as well as signaling molecules (reviewed in Egeblad and Werb, 2002; see Lu et al., 2009 for a recent example). MMP activity can be regulated at many levels: proteolytic activation of the latent zymogen, interactions with endogenous inhibitors, and localization. In cultured cell lines there are reports of MMPs being regulated by secretion, which can serve as an additional mechanism of regulating MMP spatial activity (Sbai et al., 2008; Tanaka et al., 2007).

In this study we determined how tubes remodel their apical ECM between molts to permit elongation. In every instar, a cuticle forms in the tracheae that contains ridges, called taenidia, which are regularly spaced thickenings circumscribing the inner surface of the tubes. Taenidia appear as rings or spirals (see Fig. 1A), which are patterned by the underlying actin cytoskeleton (Matusek et al., 2006). Taenidia are believed to allow the tubes to be flexible while simultaneously providing strength against collapse, similar to the ridges on a vacuum-cleaner hose (Manning and Krasnow, 1993). We find that in wild-type larvae, taenidial spacing expands as the larval tubes elongate, so that the final density is reduced almost two-fold by the end of the inter-molt period. Mmp1 is required for both taenidial expansion and tube elongation, and in a gain-of-function experiment, Mmp1 promotes tracheal tube elongation. In wild-type larvae, Mmp1 is localized in puncta along the taenidia during tube elongation. We also find that Mmp1 is required for the degradation of cuticle into manageable pieces for shedding at the molt. Thus a matrix metalloproteinase mediates the major ECM remodeling events required for this organ system to grow.

Figure 1. Taenidial spacing expands during third instar.

A. Schematic of the tracheal dorsal trunks; cells are outlined and nuclei shaded gray. Taenidia are not to scale as fewer are shown for clarity; intertaenidial distance is depicted accurately in next panels. B–C. Taenidia in an early and late third instar (L3) tracheal trunk, viewed by DIC optics. 20 µm bars were used to calculate intertaenidial distance. D. Schematic showing how taenidial expansion relates to tube elongation. E–F. Average intertaenidial distance for wild-type 3^rd instar larvae (E) and Mmp1² 3^rd instar larvae (F), shown by stage and anterior-posterior position. The number of trunks examined is at the base of each column. Intertaenidial distance in late third instar wild-type animals is significantly different between anterior and posterior (p=0.0157), asterisk. Intertaenidial distance is nearly constant in Mmp1 mutants throughout 3^rd instar (F). G. Mmp1 mutant animals transheterozygous for independently derived null alleles of Mmp1 fail to increase intertaenidial distance (p<10⁻¹⁴), demonstrating the specificity of this phenotype forMmp1 mutants.

Materials and Methods

Staging larvae

Flies were maintained at 25°C. Mmp1 mutant stocks were maintained over a CyO, arm-GFP balancer chromosome, and homozygous mutant 1^st instar animals were selected by the absence of GFP expression. To obtain staged larvae, embryos were collected 3–5 hours and aged for ~21 h in order to collect 1^st instar animals just after hatching. For second instars, plates containing new 1^st instar larvae were aged for ~26 hours, then cleared of animals that had already completed the first molt and checked every hour for newly molted 2^nd instars; these were moved onto a new plate. These 2^nd instars were either dissected as early 2^nd instars or aged for ~ 24 h and dissected as late 2^nd instars. To obtain 3^rd instar animals, animals were selected as synchronized 1^st instars and aged for ~ 48 h. The plate was cleared of any animals that had already completed the 2^nd molt; the plate was then checked every hour and any animals that had molted to third instar within that hour were transferred to a new plate. These animals were dissected as early 3^rd instars, or aged for ~24 h and dissected as mid 3^rd instars, or aged ~48 h and dissected as late 3^rd instars.

Dissection, fixation, and antibody staining for light microscopy

To observe tracheal dorsal trunks, animals were placed in a drop of PBS, anterior tissue was removed, and the animals were turned inside out leaving the dorsal trunks connected to the body wall. Tissue was fixed in 4% formaldehyde in PBS 20 minutes, washed, permeabilized, blocked, incubated with 1° antibodies overnight at 4°C, and incubated with 2° antibodies 3 hours at RT. Dorsal trunks were removed from the body wall with forceps, placed in Vectashield mounting media (Vector Labs) and imaged with a Zeiss Axio Imager Z1 with Apotome for optical sectioning. Mouse anti-Mmp1 monoclonals 3B8, 5H7, and 23G1, which we previously raised against the catalytic domain of Drosophila Mmp1 (Page-McCaw et al., 2003), were all obtained from the Developmental Studies Hybridoma bank and mixed as a 1:1:1 cocktail and diluted 1:100. Each monoclonal is specific for Mmp1 (Fig. S2, see also Glasheen et al., 2009). Other antibodies used were mouse anti-Fasciclin III diluted 1:5 (DSHB), guinea pig anti-Ninjurin A diluted 1:500 (Zhang et al., 2006), mouse anti-integrin antibodies αPS1, αPS2, and βPS1 diluted 1:10 (DSHB), mouse anti-Armadillo diluted 1:50 (DSHB), and rabbit anti-Piopio diluted 1:100 (gift from M. Affolter). Secondary antibodies were Cy3 labeled goat anti-mouse (Jackson) diluted 1:1500, FITC-labeled donkey anti-guinea pig (Jackson) diluted 1:500, and FITC-labeled donkey anti-rabbit (Jackson) diluted 1:300.

Cryosectioning

Wandering 3^rd instars of w¹¹¹⁸ were placed into a drop of PBS and dissecting scissors were used to open the side of the animal. Animals were fixed as above, and transferred to 1.0 ml of 30% sucrose in PBS overnight, 4°C. The bottom of a small plastic mold (15 mm × 15 mm × 5 mm) was covered with a base of O.C.T. (Tissue Tek) and frozen at −16°C. Intact 3^rd instar larvae were placed onto the frozen blocks, covered completely with O.C.T. and frozen at −18°C. A Microm HM550 cryostat was used to make 20 or 30 µm sections of the entire animal. Sections were dried onto glass microscope slides (Fisherbrand Superfrost Plus) for at least 30 min at RT. Slide-mounted sections were stained with Mmp1 antibodies as described above beginning with permeabilization. Samples were mounted in Vectashield containing DAPI and examined on the Zeiss Z1 as above or on a Zeiss LSM510 META Confocal microscope. Tracheal cuticle prepared by whole mount or cryosectioning display intense autofluorescence in the FITC channel. We exploited autofluorescence as a marker for epicuticle in confocal microscopy, exciting at 488nm and collecting emission using a 505LP filter.

Measurement of taenidial folds by light microscopy

Tracheal samples were dissected, fixed, mounted, and imaged under a 40X objective with DIC optics. Zeiss Axiovision software was used to label a 20 µm section of the dorsal trunk on the resulting DIC image as shown in Fig. 1B,C. The number of taenidia in each 20 µm section was counted manually, and from these an intertaenidial distance was calculated as, (# taenidia)/20µm. This was repeated on 6–10 dorsal trunks (as noted on each figure) and then an average intertaenidial distance among trunks was determined: sum of (calculated intertaenidial distance in each dorsal trunk)/(# dorsal trunks). Standard error of the mean is shown on bar graphs. Statistical significance was determined by the Student’s two-tailed unpaired t-test with equal variance.

Measurement of body length, tracheal segment length, and tracheal angles

Late 3^rd instar larvae (obtained as described above) were heat-killed by placing larvae on a coverslip on a 95°C heat block for a few seconds until movement ceased, then placed in drop of 65% glycerol in 1X PBS. Lateral images were taken with a Zeiss Stereo Lumar V12, and the length from the transverse connectives of Tr8 to Tr9 was measured. The angles made by the intersection of the dorsal trunk and transverse connective of Tr5 were measured. Body length was measured on dorsal images. Zeiss Axiovision software was used for these measurements.

Overexpression of Mmp1 was accomplished by crossing UAS-Mmp1.f2/CyO dfd GMR YFP (Le et al., 2006) to the tracheal-specific driver btl-GAL4 (Shiga et al., 1996). Embryos were fixed and stained as previously described (Samakovlis et al., 1996). Primary antibodies used were mouse anti-2A12 sera (1:1, DSHB) and rabbit anti-GFP (1:250, Molecular Probes). Secondary antibodies were conjugated with Cy5 (1:125, Jackson) or Alexa 488 (1:250, Molecular Probes). Images were obtained with a Leica SP2 Confocal Laser Scanning System. Tracheal length was assessed in staged embryos as the distance along the center of the dorsal trunk between dorsal branches 5 – 9. As reported in Fig. 9, this led to 115% increase in tube length in embryos overexpressing Mmp1. When the length was normalized using the straight distance between individual branch junctions 5 through 9 to control for strain differences, the Mmp1 overexpressing tracheae were 121% longer than controls (p<0.002). Normalization of tracheal length/straight distance in stage 14 overexpressors before cuticle deposition resulted in no significant difference in Mmp1 overexpressors vs. controls (p>0.092). Measurements were performed on 3-D reconstructions of Z-stacks using PerkinElmer Volocity software. Statistical analysis was determined by one or two sample Student’s two-tailed unpaired t-test assuming equal variance.

Figure 9. Mmp1 promotes tube elongation.

A,B,D,E: Dorsal trunks from embryos stained with monoclonal antibody 2A12. At stage 14, control embryos (A) and embryos overexpressing Mmp1 in the tracheal system (B) have dorsal trunks with similar morphology. At state 16, after cuticle is deposited in the dorsal trunks, Mmp1 overexpression in the tracheal system promotes excess tube elongation (E) when compared to controls (D). C: Quantification of the dorsal trunk lengths in control and btl>Mmp1 embryos show no significant difference (p=0.11). F: Quantification of the differences in trunk length show that Mmp1 overexpressing tracheae are 15% longer than controls. Scale bar in E is 10 µm, applies to A,B,D,E.

Electron microscopy and ultrastructural measurements

Mmp1² flies were maintained over a CyO dfd GMR YFP balancer chromosome. Homozygous mutant 1^st instar larvae were selected by absence of YFP and staged as described above. Larvae were heat-killed on a coverslip on a 70°C heat block for 20 seconds. Animals were filleted using surgical scissors, then fixed in Karnovsky’s (50 mM phosphate buffer pH 7.1, 2% glutaraldehyde, 2% paraformaldehyde). Animals were washed, incubated 2.5 hours in 4% osmium tetroxide at RT, incubated in 50 mM phosphate buffer overnight at 4°C, washed, and dehydrated through an ethanol series. EMbed 812 resin (Electron Microscopy Sciences) was prepared according to manufacturer’s instructions. Larvae were infiltrated with resin in four steps: 25% resin/75% ethanol; 50% resin/50% ethanol; 75% resin/25% ethanol; 100% resin. Larvae were embedded into molds, resin was cured at 60°C for 24 h, and larvae were sectioned and imaged with a JEOL 1230 Transmission Electron Microscope. Dorsal trunk segments from the posterior third of the larva were imaged. Technical limitations prevented exact determination of segmental identity. Three larvae were examined for each of the four conditions.

Taenidial core width, interstitial cuticle width, and apical membrane length were measured from TEM micrographs using Gatan DigitalMicrograph software. To measure the apical membrane length, first a 10 µm straight line was drawn along the axis of the tube, then the apical membrane was traced and measured along that length. Ratio of apical membrane length to tube length was calculated by dividing apical membrane length by 10 µm. Three wild-type animals and two Mmp1 mutant animals were measured, and three measurements were taken for every image. Variation noted in the text and in Fig. 8 is the standard deviation.

Figure 8. Ultrastructure of tube elongation in wild-type and Mmp1 tracheae.

A,B,D,E: Transmission electron (TEM) micrographs of 3rd instar dorsal trunks (longitudinal sections, as in Fig 1A inset). C,C’: Schematic of panels A and B respectively, showing structures. Cell layer is shown dark gray, cuticle layer light gray. Both wild-type and Mmp1 early 3rd instar larvae display apical membrane protrusions (arrows in A,C,D) positioned to deposit cuticular material between the cores, although there was significant variability within each genotype. In wild type, the space between taenidial cores expands between early and late 3rd instar (A and B, respectively). D,E: Mmp1 mutant cuticle appears not to be remodeled between early (D) and late 3rd instar (E). Diamond in panel E shows a gap between the cuticle and cell layers in Mmp1 late 3rd instar. In some regions of Mmp1 mutants, this gap was a complete delamination of the membrane from the cuticle. F: Table quantifying mean taenidial spacing, as shown by the lines in C and C’, and the ratio of apical membrane length to overall tube length, in late L3 wild-type and mutant larvae, (** p-value<0.001). Scale bar in E is 0.5 µm, applies to A,B,D,E.

We calculated the expected ratio of apical membrane length to tube length in Mmp1 mutants as follows: in wild-type animals, the ratio of apical membrane length to tube length was found to be 1.22; a wild-type tracheal segment (tube) expanded 1.5-fold more than an Mmp1 mutant segment (Table 1), so if the mutant cells expanded their membrane to the wild-type extent the expected ratio would be 1.22 × 1.5 = 1.83 in the mutants.

Table 1.

Tracheal Elongation Is Coordinated with Body Elongation

Genotype	3rd Instar Stage	Inter-Taenidial Distancea,b in µm (n)c	Taenidial Distance Fold-Increase	Tracheal Segmentd Length in µm (n)c	Tracheal Segment Fold-Increase	Whole Body Length in µm (n)	Body Length Fold-Increase
w1118	early	0.78±0.01 (18)	1.9x	231±8 (14)	1.8x	2298±49 (16)	1.7x
	late	1.51±0.05 (20)		409±12 (8)		3950±106 (8)
Mmp12	early	0.82±0.02 (17)	1.0x	92±5 (4)	1.2x	1770±158 (4)	1.5x
	late	0.86±0.02 (20)		107±6 (12)		2594±174 (24)

^aas measured in Figures 1 and 2 for posterior segments Tr8-Tr10

^berror margins represent standard error

^cnumber of dorsal trunks measured is given in parentheses

^dmeasured from transverse connective of Tr8 to transverse connective of Tr9

Results

Tracheal cuticle expands by increasing intertaenidial distance

The larval tracheal system is lined with cuticle that is shed at each molt. However, tracheal tubes lengthen continuously throughout the intermolt periods, or instars, despite the continuing presence of cuticle (Beitel and Krasnow, 2000). To investigate how this matrix remodeling is accomplished, we took advantage of the distinctive structures called taenidia, visible ridges of cuticle that run in helical or annular rings perpendicular to the axis of tracheal tubes (Fig. 1A). We asked whether the taenidia retained their spacing and density as the tube elongated, or whether the taenidial spacing increased and density decreased. We dissected dorsal trunks from third instar larvae at three developmental periods – beginning of third instar (immediately after the second molt), mid-third instar (~24 hours after molting), and late third instar (~48 hours after molting) – and counted the number of taenidia observed in a 20 micron segment (Fig 1B,C). We converted these data into intertaenidial distance to determine the size of the repeating unit (taenidial ridge plus valley). Comparing animals that had just entered third instar (early L3) with those that had nearly completed third instar (late L3), we found that intertaenidial distance increased about two-fold as the larva aged and the dorsal trunk elongated (Fig. 1B–E, Table 1). In some mutants embryonic tracheal tube elongation is not constant along the anterior-posterior axis (Beitel and Krasnow, 2000), so we examined taenidial expansion in three different anterior-posterior positions along the dorsal trunk, corresponding to tracheal segments Tr2–4 for anterior, Tr5–7 for middle, and Tr8–10 for posterior (Fig. 1E). The intertaenidial distance expanded significantly more in the posterior segments than the anterior segments (p=0.0157). Thus, apical matrix is remodeled by region-specific expansion of intertaenidial distance.

To understand if the expansion in intertaenidial distance was sufficient to account for tube elongation, we measured the change in length of one segment of the dorsal trunk, measured from transverse connective of tracheal segment Tr8 to the transverse connective to Tr9. We found this segment elongated by 1.8-fold, similar to the 1.9-fold expansion of the intertaenidial distance (Table 1). We also measured the overall change in length of the animal during third instar, as elongation of the tracheal trunks would be expected to correspond to the change in animal length. We found that larvae elongated 1.7-fold during third instar. Thus, the change in intertaenidial distance is sufficient to account for the expansion of the cuticle required for elongation of both the tracheae and the larva.

The extracellular protease Mmp1 is required for tracheal cuticle expansion

Previously, we found that Mmp1 mutants have defective larval tracheae (Page-McCaw et al., 2003). Mmp1 encodes a matrix metalloproteinase, one of a conserved family of extracellular proteases that cleave ECM components in vitro. In Mmp1 mutants, the tracheal system appears to develop normally during embryogenesis, but as larvae grow the tracheal tubes become stretched, frequently displaying one or more breaks in the dorsal trunks by second or third instar (Page-McCaw et al., 2003). This phenotype is tissue-autonomous, as the tracheal-specific knock-down of Mmp1 (btl-GAL4, UAS-Mmp1-IR) reproduces the Mmp1 tracheal phenotypes (Uhlirova and Bohmann, 2006). The tracheal defects in Mmp1 larvae cause secondary defects in growth, possibly because of hypoxia, resulting in smaller larvae, imaginal discs and brains (Beaucher et al., 2007).

By examining intertaenidial distance in dissected tubes from Mmp1 null third instar larvae, we found that at the beginning of the instar, mutants had taenidial spacing nearly identical to wild-type (compare Fig. 1E to 1F). However, the intertaenidial distance did not increase as the mutants progressed through third instar (Fig. 1E; Fig. 2H p=9.2×10⁻⁸, comparing wild-type and Mmp1 late L3 intertaenidial distances). Accordingly, in intact mutants, the tracheal segments were able to elongate only 1.2-fold, probably from mechanical stretching of the ECM, while body length increased 1.5-fold (Table 1). This overall elongation of animals more than their tracheal tubes explained why the Mmp1 tubes were pulled tight or broken (Fig. 2C). To demonstrate the specificity of this phenotype for Mmp1, we examined transheterozygous animals carrying two independently-derived null alleles in different genetic backgrounds; these animals also failed to increase their intertaenidial distance during third instar (Fig. 1G). We recently showed that Mmp1^Q273* homozygotes do not exhibit tracheal breaks even though they are missing the entire Mmp1 hemopexin domain, believed to be involved in protein-protein interactions and substrate recognition (Glasheen et al., 2009). We found that Mmp1^Q273* mutants were able to partially increase intertaenidial distance (Fig. S1), indicating that the hemopexin domain is not strictly required for taenidial expansion. The hypomorphic taenidial expansion phenotype is consistent with the reduction in protein levels in this mutant (Glasheen et al., 2009).

Figure 2. Tb partially suppresses Mmp1.

A–E. Third instar larval dorsal trunks; images are displayed at different magnifications (see scale bars). Insets in A, B, D, E are magnifications of the angles formed at the joint of the transverse connective (tc) and the dorsal trunk (dt) in tracheal segment 5 (Tr5) as outlined in panels. A. Dorsal trunks of wild-type larvae have some slack, as demonstrated by the angle in Tr5. B. Mmp1 null tracheae are under tension, demonstrated by the reduced angle in Tr5. C. Mmp1 dorsal trunks break frequently in 3rd instar. Arrows show broken ends. D. Tb/+ larvae have shortened bodies with relatively longer tracheal trunks, providing extra tracheal slack as demonstrated by the increased angle in Tr5. E. Mmp1; Tb/+ tracheae are under less tension than Mmp1 tracheae, demonstrated by a greater angle in Tr5. F. Quantification of the angle between the dorsal trunk and transverse connective at Tr5. The number of trunks examined is shown at the base of each column. G. Table showing Tb/+ suppression of two other Mmp1 phenotypes: larval lethality as assessed by the ability to pupariate, and dorsal trunk breaks in 2^nd instar (L2) and 3^rd instar (L3) larvae. H. Mmp1; Tb/+ mutants do not increase intertaenidial distance, even though Tb/+ partially suppresses Mmp1 tracheal phenotypes. Statistical analysis of Mmp1 in taenidial expansion (T-test): wild-type and Mmp1 late L3 intertaenidial distances are highly significantly different (p=9.2×10⁻⁸). Mmp1; Tb/+ late L3 is highly significantly different from Tb/+ late L3 (p=1.5×10⁻¹⁰).

Tubby partially suppresses Mmp1 tracheal phenotypes

Larvae heterozygous for a dominant mutation in Tubby (Tb, also known as TwdlA, Guan et al., 2006) have shortened bodies with relatively long tracheal tubes, producing extra tracheal slack in these animals (Fig. 2D). We asked whether Tb could partially suppress the tracheal breaks and lethality of Mmp1 mutants, as less tracheal elongation would be required in the Tb background. We assessed the tension on the tracheal tubes of late 3^rd instar larvae in two ways. First, we measured the angle formed by the transverse connective joining the dorsal trunk in Tr5 (Fig. 2A,B,D–F), and second we counted the number of animals with tracheal breaks (Fig. 2C, G). By both measures, we observed that Tb was able to compensate for the Mmp1 tracheal elongation defects in double mutants (Mmp1²; Tb/+). The double mutant tracheae were not under as much tension, as evidenced by changes in the angle between the dorsal trunk and the transverse connective: in wild-type animals, this angle was 46.1° ± 3.6° (Fig. 2A, F); in Mmp1 mutants, whose tracheae were under tension because they cannot elongate, this angle was reduced to 13.6° ± 2.4° (Fig. 2B,F). In a double mutant background, this angle was relaxed to 27.0° ± 2.8° (Fig. 2E,F). Consistent with these observations, only 55% of double mutants had tracheal breaks whereas nearly all (93%) the Mmp1 third instar larvae developed breaks in their tracheal trunks (Fig. 2C,G). Interestingly, Tb also partially suppressed the larval lethality of Mmp1 null mutants. Whereas as all Mmp1² larvae died before pupariation, 29% of double mutant larvae survived to pupariation (Fig. 2G). Thus, Tb partially suppressed the effects of Mmp1 mutations.

Tb encodes a cuticle protein in the Tweedle family, Tweedle A (Guan et al., 2006). This finding raised the possibility that Tb partially suppressed Mmp1 because of defects in Tb mutant cuticles that removed the requirement for Mmp1 in tracheal cuticle remodeling – for example, if the Tb cuticle stretched more easily than wild-type. We examined taenidial spacing in early and late third instar Mmp1; Tb/+ double mutants (Fig. 2H). We found that in the double mutants, as with Mmp1 single mutants, the intertaenidial distance did not increase during third instar. In the single mutant Tb/+, however, the intertaenidial distance increased similarly to wild-type, indicating that tension is not the trigger for expansion of taenidial ridges. The requirement for Mmp1 in intertaenidial expansion is not specifically suppressed by Tb at the level of cuticle structure, but rather Tb relieved the requirement for tracheal elongation by shortening the body axis.

New tracheal cuticle is not templated by a previous cuticle

We were surprised that the Mmp1 taenidial ridges were spaced like wild-type at the beginning of third instar. Second instar Mmp1 tracheae displayed both taut and sometimes broken tracheae occasionally leading to death – so how could mutant larvae display appropriate taenidial spacing later, at the start of third instar? To address this question, we compared intertaenidial expansion in wild-type and Mmp1 mutant animals during second and third instar (Fig. 3). Strikingly, at the onset of second instar, wild-type intertaenidial distance was 0.8 microns, the same spacing observed at the onset of third instar; this distance expanded about two-fold during second instar before being replaced at the molt with a new cuticle whose taenidia were again spaced at 0.8 microns (Fig. 3A–E). In Mmp1 mutants, larvae began second instar with intertaenidial distance like wild-type, but the cuticle did not expand as the instar continued and the animal elongated (Fig. 3E). Thus Mmp1 mutants are able to secrete appropriately patterned cuticle but not remodel it. The normal spacing of Mmp1 third instar cuticle demonstrates that the density of taenidia in newly synthesized cuticle does not depend on the density in the preceding cuticle, suggesting that taenidial patterning is driven by intracellular processes rather than being templated by extracellular cues in the apical ECM.

Figure 3. Tracheal cuticles of each instar are formed with a fixed intertaenidial distance and undergo similar expansion during wild-type larval growth.

A–D. Wild-type dorsal trunks dissected from early and late 2^nd instar larvae (L2, A and B) and early and late 3^rd instar larvae (L3, C and D). E. Wild-type and Mmp1² intertaenidial distances at early and late 2^nd and 3^rd instar, calculated as in Fig. 2. Wild-type early 2^nd and early 3^rd instars have indistinguishable intertaenidial distances, and both expand about two-fold during the instar. Mmp1 early 2^nd and 3^rd instars have intertaenidial distances similar to wild-type, but the taenidial spacing does not expand during either instar. The number of trunks examined is at the base of each column. Scale bar on A represents 10 µm; A–D are at same magnification.

Shedding cuticle at molts requires Mmp1

In wild-type larvae, tracheal cuticle is remodeled in two ways: it expands during instars as we found above, and it is also replaced at molts. We previously observed that Mmp1 mutants sometimes had difficulty releasing external molted cuticle (Page-McCaw et al., 2003). To determine whether Mmp1 tracheae were able to replace tracheal cuticle at molts, we examined third instar dorsal trunks. Wild-type mid-instar trunks all had a single cuticle layer (N=24 trunks; Fig. 4A). In Mmp1 mutants, however, 17/23 dorsal trunks contained two visible cuticles, indicating that the second-instar cuticle had not been shed. In one case, we found all three larval cuticles present inside an Mmp1 third instar dorsal trunk (Fig. 4B). To understand the wild-type progression of replacing cuticle at a molt, we examined wild-type dorsal trunks at closely spaced intervals throughout the second molt (Fig. 5). The first observable sign of molting was dilation of the tracheal epithelium, as cells released the cuticle and increased the diameter of the dorsal trunk; this step is known as apolysis (Fig. 5B). Next, new cuticle was deposited at the expanded inner surface of the tube; two cuticles were clearly visible, each with taenidial ridges (Fig. 5C). Next, the inner cuticle broke in the vicinity of the fusion cells, a pair of distinctive toroidal cells located at the fusion site of each tracheal metamere (Fig. 5D). Cuticle breakage at fusion cells has been previously observed, and it has been assumed that cuticle must be broken into metameric fragments to be discarded efficiently (Whitten, 1957). Finally, each metameric piece of inner cuticle crumpled, presumably as it was dragged out the lateral spiracle (Fig. 5E). These cuticle fragments are sloughed off with the exocuticle as the animal crawls out of its old cuticle at the end of the molt (Whitten, 1957), leaving behind an intact third instar tube lined with a new cuticle (Fig. 5F).

Figure 4. Mmp1 mutants retain cuticle from previous instars.

A. Wild-type third instar dorsal trunk with single cuticle. B. Mmp1² dorsal trunk with three cuticles, from the first (L1), second (L2), and third (L3) instars.

Figure 5. Tracheal molting in wild-type animals.

A–F. Wild-type molting dorsal trunks. A. Second instar trunk before the onset of the 2^nd molt. B. Second instar trunk beginning the molt. Cells have pulled back from the cuticle (black arrows), expanding the tube diameter. C. Late 2^nd instar trunk. The expanded cellular tube has deposited a new cuticle layer (white arrow) complete with taenidial ridges (box and inset). The old 2^nd instar cuticle is visible in the center of the tube (black arrow). D. Dorsal trunk mid-2^nd-molt with broken cuticle (bracket) at fusion cells. E. Dorsal trunk discarding a segment of 2^nd instar cuticle. F. Third instar tube with a single cuticle lining, cleared of all previous instar cuticle. Scale bar on A represents 20 µm; all panels at same magnification.

The Mmp1 phenotype of extra cuticles suggested that either Mmp1 was required to release the old cuticle from the epithelium (as in Fig. 5B) or to degrade the released cuticle into manageable pieces (as in Fig. 5D). We examined Mmp1 mutants for the localization and expression levels of cell-surface adhesion molecules including the zona pellucida protein Piopio (Bokel et al., 2005; Jazwinska et al., 2003), the integrins αPS1, αPS2, βPS1 (Dominguez-Gimenez et al., 2007), the adhesion regulator Ninjurin A (Zhang et al., 2006), and β-catenin/Armadillo (Riggleman et al., 1990). All were unchanged in mutant dorsal trunks compared to wild type (data not shown). We next examined cuticle degradation in mutants. Fusion cells were identified by their location, their distinctive cuticle patterning which lacks taenidia (Matusek et al., 2006), and their ring-shape morphology. Of five wild-type dorsal trunks found in the brief stage where they contain a broken inner cuticle, all had breaks that disrupted the fusion-cell pattern of the inner (broken) cuticle; these breaks were in the immediate vicinity of the fusion cells themselves (Fig. 6A,C,E,G). In Mmp1 mutants, of five trunks with inner cuticle breaks, the fusion-cell cuticle pattern was intact on the inner cuticle in all samples, indicating the breaks were ectopic; additionally, 4/5 breaks were not located at the fusion cells of the tube (Fig. 6B,D,F,H). These ectopic breaks were presumably caused by pulling forces as the attached exocuticle was sloughed off; as such they are expected to be rare and insufficient to break cuticle into manageable pieces. Thus Mmp1 is required for the regular degradation of cuticles at fusion cells during molts, and in the absence of Mmp1 tracheal cuticle cannot be effectively removed from tracheal tubes.

Figure 6. Mmp1 tracheal cuticle does not break at the fusion cells.

A,C,E,G. Wild-type dorsal trunk in the 2^nd molt. White arrowheads show 2^nd instar cuticle break (A,E); white arrow marks fusion cells, revealed by FasIII staining (C,E). E. Merged image of A and C showing tracheal cuticle break at fusion cells. G. Schematic of the inner cuticle breaking at the fusion cells (black double rings). B,D,F,H. Mmp1 tracheal dorsal trunk in the 2^nd molt. White arrowheads show second instar cuticle break (B,F); white arrows mark two sets of fusion cells (C,F). F. Merged image of B and F showing cuticle break far from fusion cells. H. Schematic of the inner cuticle breaking far from the fusion cells (black double rings). Scale bar on B represents 20 µm; panels A–F are at the same scale.

Mmp1 protein localization at fusion cells and at taenidia

Our data indicate that in larval tracheae, Mmp1 functions at taenidial ridges for ECM expansion and at fusion cells for ECM degradation. To gain insight into these functions, we stained tracheal dorsal trunks with anti-Mmp1 antibodies to determine protein localization. In control Mmp1^null tracheae no antibody staining was observed (Figure S2). In wild-type animals, we unexpectedly observed several different expression patterns, often within the same tube: fusion cell staining, diffuse cell staining, cell-cell borders, and cytoplasmic granules (Fig.7). To investigate potential developmental regulation, we examined dorsal trunks from staged second and third instar larvae. Before the 2^nd molt, we frequently saw weak cell-cell border staining and stronger fusion cell staining (Fig. 7A). Right after the 2^nd molt, in addition to cell-cell border and fusion cell staining, we observed individual stochastic cells that appeared to be expressing high levels of Mmp1 (in 14/20 tubes, like Fig. 7B). By three hours into 3^rd instar, intensely staining cytoplasmic granules became evident (10/14 tubes; Fig 7C), suggesting that Mmp1 was trafficked in vesicles. Eighteen hours after the molt, cell-cell borders and fusions cells, and cytoplasmic puncta were all visible. Later in third instar, staining resolved to the cell-cell borders and fusion cells (32–40h after the molt, Fig. 7D). The finding that the Mmp1 protease is expressed at fusion cells and is required for old cuticle degradation provides a mechanism for the decades-old observation that molting cuticle breaks regularly at fusion cells (Whitten, 1957). Because the localization of Mmp1 to fusion cells is not restricted to the peri-molt period, it is likely that proteolytic activation, or loss of a proteolytic inhibitor, is required to activate Mmp1 for cuticle degradation during molts.

Figure 7. Mmp1 dynamically localizes to fusion cells, cell borders, and taenidial ridges.

A–G. Anti-Mmp1 antibody staining in dorsal trunks; Mmp1 is shown red and DAPI-stained nuclei blue. A. Dorsal trunk dissected immediately before the 2nd molt shows Mmp1 generally diffuse but highly expressed at fusion cells (arrows). B. Dorsal trunk dissected just after the 2nd molt, showing Mmp1 expression upregulated in individual cells. C. By three hours after the 2nd molt, Mmp1 is in cytoplasmic granules. D. Mid-3rd instar (32 hours after molt) Mmp1 is observed at cell-cell borders and fusion cells. E–G. Cryosections of wandering 3rd instar wild-type dorsal trunks. E. Transverse cryosection showing Mmp1 protein at high levels in the cytoplasm (arrowhead). F. Longitudinal cross-sections show Mmp1 localized in periodically spaced puncta at the apical cell surface, in addition to its localization in cytoplasmic granules. F’. Magnification of the boxed regions in F, showing alignment between Mmp1 apical deposits (yellow arrows) and taenidial ridges. G. Confocal micrograph showing Mmp1 and the taenidial ridges, visualized by cuticle autofluorescence in the FITC channel. Scale bar in panel D is 10 µm; panels A–D are at the same magnification. Scale bars in panels E and F are 10 µm. Scale bar in panel G is 1 µm.

For apical/basal resolution, we stained transverse cryosections of dorsal trunks for Mmp1; these clearly demonstrated the cytoplasmic nature of the granules but did not reveal taenidial localization (Fig. 7E). To image the apical surface of the tubes, we stained longitudinal cryosections of dorsal trunks for Mmp1 (7F–G). We found Mmp1 localized at the apical cell surface, in small repeating puncta aligned with the taenidia as viewed by DIC microscopy (Fig. 7F) or by cuticular autofluorescence (Fig. 7G), generally one puncta per taenidial ridge. The formation of these apical puncta appeared to be regulated, as they were present in some regions but not others, even within the same tube (see Fig. 7F). We speculate that the formation of these puncta is regulated by secretion of Mmp1 from cytoplasmic vesicles (appearing as granular staining) at the apical cell surface, and that secreted Mmp1 forms puncta in contact with the ECM. Thus Mmp1 is expressed at taenidia where it is poised to expand the cuticle intertaenidial distance.

Remodeling taenidial morphology requires Mmp1

To visualize taenidial remodeling, we used electron microscopy (EM) to examine wild-type and Mmp1 third instar tracheae at early and late stages (Fig. 8). In wild-type early third instar larvae, the cuticle had a regular toothed appearance with prominent “cores” at the center of each taenidial ridge (Fig. 8A,C,C’). Cellular apical projections extended through the matrix material toward the inter-core areas (arrow in Fig. 8A,C,D). During growth in wild-type larvae, the distance between taenidial cores increased, apparently from addition of matrix material between the cores (Fig. 8A–C). Accordingly, the average distance between two taenidia increased from 0.52 ± 0.12 µm to 1.33 ± 0.16 µm (Fig. 8F), in excellent agreement with the increased intertaenidial spacing observed by light microscopy. Although we observed variation in the intensity of the staining between similar samples (see Figure S3), the geometries of the taenidia and their expansion were consistent. Thus, electron micrographs show dramatic remodeling of the apical ECM in wild-type tracheal tubes during third instar growth.

In Mmp1 mutants, the variable staining of cuticular material precluded assessment of fine changes in matrix organization, but we did observe that late third instar Mmp1 taenidia appeared consistently different from wild-type. In the mutants, little remodeling of lumenal cuticle was evident: the average spacing between mutant taenidia in early third instar was 0.64 ± 0.19 µm, and this did not change significantly over the course of the instar as late third instar spacing was 0.82 ± 0.22 (Fig. 8F). In contrast, the cuticle deposited in Mmp1 mutants at the start of 3rd instar had a near-normal taenidial pattern (Fig. 8D). Thus the late third instar taenidia resembled the early third instar taenidia in the Mmp1 mutants but not the late third instar wild-type taenidia (Fig. 8D,E). We also observed striking and reproducible separations between the cell and cuticular layers in late 3^rd instar Mmp1 larvae (Fig 8E, diamond) that were not present in early 3^rd instar Mmp1 or wild-type larvae (see Discussion).

Mmp1 is required for expansion of tracheal cell apical membrane

There is substantial evidence that apical ECM remodeling in embryonic tracheae causally contributes to membrane elongation because all known mutations that disrupt tracheal lumenal matrix organization induce lengthening of the trachea, and therefore tracheal membrane elongation (Affolter and Caussinus, 2008; Andrew and Ewald, 2010; Wu and Beitel, 2004). Furthermore, there is also evidence that membrane expansion can induce matrix remodeling since mutations or transgene-expression that increases activity of the apical polarity gene crb also increase tracheal length (Laprise et al., 2010). How growth of the apical ECM and apical cell membranes is coordinated is not known. The lack of cuticular remodeling in Mmp1 mutants provides a tool for investigating this coordination. We asked whether the apical cell surface of Mmp1 mutant trachea elongated despite the near-static apical ECM, or alternatively if the failure of apical ECM remodeling prevented apical membrane elongation. In wild-type late third instar larvae, the apical membrane was relatively straight indicating coordinated growth of cuticle and membrane. However, in Mmp1 mutants “waviness” was sometimes present in late third instar Mmp1 mutant apical membranes, suggesting that the tracheal cell membranes had elongated more than the matrix (data not shown). Quantifying the ratio of apical membrane length to tube length in EM micrographs of late third instars showed that wild-type larvae had a ratio of 1.22 ± 0.06 and that mutant larvae had a ratio of 1.44 ± 0.17, a significantly different ratio (p<0.001). Notably, this somewhat increased ratio in mutants was considerably less than the ratio of 1.83 (p<0.001) that would have been expected had Mmp1 tracheal cells expanded their apical membranes to the wild-type extent (see Materials and Methods for expected ratio calculation). Because the Mmp1 apical membrane does not expand to the wild-type extent, either Mmp1 or Mmp1-mediated cuticle elongation is required for normal membrane elongation. These results reveal a previously undocumented coordination of growth of the apical ECM and apical cell membrane.

Mmp1 promotes embryonic tube elongation but blocks tube dilation

The finding that Mmp1 is required for apical membrane expansion of tracheal cells raises the question of whether Mmp1 has a permissive or instructive role in apical membrane expansion. To test the idea that Mmp1 might promote membrane elongation in an instructive manner, we ectopically expressed Mmp1 in the embryonic tracheal system. Although Mmp1 is endogenously expressed in embryonic tracheae, the system does not require Mmp1 as embryonic tracheae are not affected by Mmp1 null mutations (Page-McCaw et al., 2003). Expression of Mmp1 with the tracheal specific btl-GAL4 driver caused early larval lethality, and thus we examined the dorsal trunks of embryos before the lethal phase. In stage 16 embryos, Mmp1 expression increased embryonic tracheal length by 15% compared to wild-type (Fig. 9D–F). The dorsal trunk followed a convoluted rather than straight path between segments (Fig. 9E), a phenotype reminiscent of that caused by mutations in verm or serp, which encode putative ECM modifying proteins (Luschnig et al., 2006; Wang et al., 2006). Notably, as for verm and serp mutations, increased tracheal length was not accompanied by increased diameter. On the contrary, Mmp1 expression blocked the genetically programmed diameter increase that normally occurs during stage 14 (compare Fig. 9A and D). These results show that Mmp1 can profoundly influence the growth of apical cell membrane in a spatially restricted manner. Interestingly, we did not observe this tube elongation phenotype in stage 14 embryos, before the cuticle lining the tubes is deposited (Fig. 9A–C). This restriction of cellular elongation to stages after cuticle deposition suggests that the presence of cuticle is necessary for Mmp1-mediated tube elongation.

Discussion

In this study we find that larval tracheal tubes elongate their apical matrix at discrete sites along the long axis of the tube. Immediately after a molt in wild-type animals, the new apical matrix (cuticle) is constructed with taenidial ridges at a fixed interval of ~0.8 µm. During intermolt tube elongation, this matrix expands the taenidial interval to ~1.6 µm. At molting, this fully expanded matrix is discarded and replaced with a new matrix with a taenidial interval again at 0.8 µm, which will again expand about two-fold. This precise remodeling of the taenidia is accomplished by the extracellular protease Mmp1, which is localized in discrete apical puncta, each associated with an individual taenidium. In Mmp1 mutants, the taenidial ridges do not expand, the taenidial interval remains fixed, and larvae cannot elongate their tracheae as their bodies elongate, causing stretched tubes that eventually break. In normal tube elongation, ECM expansion is coupled with cellular apical membrane expansion, and Mmp1 is required for the coordination of ECM and cellular expansion; Mmp1 is able to promote both aspects of tube expansion when overepressed in larvae. Mmp1 is also required for the important tracheal cuticle remodeling event in larvae: degrading cuticle into pieces that can be discarded at each molt. Thus Mmp1 tracheal tubes cannot elongate because they cannot remodel apical extracellular matrix either to expand it or to discard it.

On first inspection, it is difficult to account for the progressively deteriorating phenotype of Mmp1 mutants. Although the tracheal system appears morphologically normal at hatching in Mmp1 null mutants, within several hours they develop taut stretched tracheal systems. As we have previously reported, this phenotype worsens throughout larval life: by second instar, many animals have broken dorsal trunks and some death occurs; by third instar nearly all animals have tracheal breaks and all eventually die (Page-McCaw et al., 2003). This presents an apparent paradox as early third instar mutants, with shortened tracheal systems, have normal spacing of their taenidia. This paradox is resolved, however, when one considers that in each instar taenidial expansion is a requirement for tube elongation. Thus at the start of second instar, although the taenidia are correctly spaced, they comprise a tube that is already shorter than wild-type and is virtually unexpandable, and thus tubes frequently break as the animal grows. By third instar, although the taenidia are again deposited with normal spacing, the collective failures of elongation in the previous instars make the entire tracheal system very short and highly abnormal. Taken alone, our taenidial expansion data might suggest that the Mmp1 tubes fail to elongate at all. If Mmp1 mutant tubes were unable to achieve any elongation, then the tracheal system of third instar larvae would be ~1/8 the length of wild-type controls; we observe that an Mmp1 tracheal segment is about ¼ the length of wild-type (Table 1). Thus despite the lack of taenidial expansion, there appears to be some kind of other elongation at work in these mutant tubes. Possibilities include an aberrant brute-force stretching of unremodeled cuticle, or a burst of tube elongation during molts when the tube releases cuticle.

In electron micrographs of late Mmp1 mutants, the cuticle appears to separate from the epithelial layer in late Mmp1 samples, but not in wild-type; the explanation of artefactual sample fracturing is unlikely, as cell membranes appeared intact in TEMs from both genotypes. We can envision two models to explain this separation. One possibility is that matrix components are still secreted by the mutant cells, but they require matrix remodeling in order to become incorporated into cuticle, and so they accumulate between the cells and the unremodeled cuticle, creating a gap. They would have to be electron-transparent to be consistent with our images, and there are reports of electron-transparent cuticle layers in insects (Whitten, 1972); indeed the taenidial cores of our TEMs can appear electron-transparent. Alternatively, it is possible that Mmp1 is required for processing adhesion molecules that hold the cuticle to the cell layer, so that adhesion is lost in the absence of the Mmp1. We favor the first model as it accounts for the fate of the matrix components that should have been deposited in the cuticle in the absence of remodeling.

Specific but dynamic regulation of MMP localization

An important question arising from this study is how MMP localization and activation is controlled to produce uniform tube elongation. One simple model is that the cells can sense tension and respond by secreting or activating Mmp1. However, this model is not supported by our observation that Tb mutants, with slack in the tracheal tubes, still increase the intertaenidial distance during the intermolt period. Thus we favor the hypothesis that tube elongation is under developmental regulation. The Tb mutant phenotype indicates that such a tube elongation program is independent of body elongation. One possible mechanism is that the developmental program could trigger Mmp1 secretion from cytoplasmic vesicles to the apical ECM; how MMP secretion is controlled is an open question. The protease appears to be localized in extracellular puncta precisely coordinated with the taenidia. This localization pattern may be ultimately directed by the actin cytoskeleton, which appears to pattern the taenidia during new cuticle secretion (Matusek et al., 2006). Consistent with the possibility of actin localizing Mmp1, the actin cytoskeleton generally regulates apical secretion in the embryonic tracheal system (Massarwa et al., 2009); and in cultured neurons, it has been observed that MMP-containing vesicles traffic along microfilaments (Sbai et al., 2008). Mmp1 mutants that lack a hemopexin domain, or dominant-negative mutants that interfere specifically with the Mmp1 hemopexin domain, are still able to elongate their tubes (Fig. S1, and Glasheen et al., 2009), indicating that the hemopexin domain is not required for secretion, localization, or activation of Mmp1 in these discrete puncta.

Mmp1 mediates interstitial axial elongation

The interstitial remodeling of tracheal cuticle stands in contrast to other developmental mechanisms of ECM deposition and remodeling that occur during axial growth of rigid or load-bearing structures. During vertebrate long-bone growth, elongation takes place at the growth plates near the ends of the bones, concentrating ECM remodeling and deposition distally. In plants, axial elongation takes place at the meristem regions, which are located distally like growth plates, again confining ECM deposition to distal regions. These cases of spatially restricted remodeling occur in contexts where maintaining the integrity of a rigid ECM presents a structural hurdle to simultaneous remodeling. Interestingly, MMPs appear to be required for both these kinds of growth. Mouse MMP13 mutants cannot remodel the cartilage at the growth plate to make bone (Inada et al., 2004; Stickens et al., 2004), and MMP9 is also required for normal long-bone growth (Vu et al., 1998). In Arabidopsis, the MMP mutant At2-mmp1 cannot extend shoots (Golldack et al., 2002), and in the Loblolly pine MMP expression is correlated with embryonic root (radicle) protrusion, which is hindered by an MMP inhibitor (Ratnaparkhe et al., 2009). In Drosophila larval tracheae the cuticle is not fully sclerotized and so retains some plasticity (Chapman, 1998), which might be expected as the cuticle does not provide rigidity along the axis but instead protects circumferentially from crushing forces. This different set of structural requirements probably explains why in tracheae deposition of matrix is not confined to distal regions, nor even to one region per segment, but rather is dispersed across the length of the tissue. The requirement for tube rigidity in the circumferential axis, but not along the body axis, also addresses why tracheal elongation can occur continually, modifying an existing cuticle. In contrast, tube circumferential expansion cannot occur continually but is limited to the molt, when the cuticle is completely replaced. The importance of MMPs in axial growth is underscored by the common requirement for MMPs during all these cases, despite the different structural contexts for matrix remodeling in bone elongation, plant growth, and tracheal elongation.

A cuticle-derived signal could coordinate cell elongation with ECM expansion

In Mmp1 mutants, where the cuticle remains unelongated, it is striking that elongation of the underlying cells and their apical membranes is also impaired. As described in the results, although there is some excess of apical membrane in Mmp1 mutants, the excess is much less than expected had the cells underlying the cuticle expanded their membranes to the wild-type extent. These results suggest that either Mmp1 activity or cuticle elongation is required for cells and apical membranes to elongate normally.

How are Mmp1 activity and/or cuticle remodeling coordinated with membrane expansion? Although it is possible that matrix remodeling and membrane expansion represent distinct activities of Mmp1, the simplicity of having a direct causal relationship between matrix remodeling and membrane expansion is appealing. One example of a combined model is that extracellular matrix elongation, mediated by Mmp1, places cells under tension, and cells respond by elongating apical membrane. Another model is that Mmp1 proteolytic activity, which remodels cuticle, may simultaneously generate an inductive signal (or inactivate a negative regulator) that causes the underlying cells to elongate, thus coordinating the elongation of both the rigid cuticle and the underlying cells. Production of such a signal would be analogous to mammalian MMPs cleaving laminin-5 or collagen IV to unmask cryptic signaling sites that promote cell migration (Giannelli et al., 1997; Xu et al., 2001); this model would also be consistent with recent findings that release of a collagen IV domain by MT2-MMP is required for branching morphogenesis of the submandibular gland in mice (Rebustini et al., 2009). Consistent with Mmp1 activity generating a cuticle-derived signal, misexpression of Mmp1 in embryonic tracheae causes tube elongation only when cuticle is present (Fig. 9A–F, cuticle deposition begins during stage 15).

Concluding Remarks

Our results show that extracellular matrix remodeling is a critical aspect of tube elongation in larval tracheae. Mmp1 mutants cannot remodel cuticle and cannot properly elongate their tubes or degrade unnecessary matrix material to be discarded at molts. These remodeling events are regulated separately from the initial deposition and patterning of cuticle, as Mmp1 mutants are able to secrete normally patterned cuticles. Additionally, Mmp1 appears to regulate cell elongation, perhaps directly by processing a signaling molecule, or indirectly by regulating the ECM. Hence a matrix metalloproteinase appears to act as a critical co-regulator of matrix and cellular growth. Finally, it is significant that tracheal remodeling but not initial tracheal morphogenesis requires a matrix metalloproteinase. Our analysis of this remodeling phenotype reinforces the notion that matrix metalloproteinases are specialists for remodeling existing tissues, rather than forming tissues, likely because of the need to alter existing ECM that limits plasticity.