Egfr is essential for maintaining epithelial integrity during tracheal remodelling in Drosophila

During morphogenesis, cells forming epithelial tissues show an amazing capacity to remodel and to rearrange while remaining attached to one another. To ensure proper organ formation, a correct balance must exist between the forces that trigger remodelling (whether internal or external), which promote cell movement, and the forces that maintain cell cohesiveness, which preserve tissue integrity. Much attention has been given to the cellular and molecular basis of tissue remodelling (Lecuit,2005; Pilot and Lecuit,2005); however, the signals and mechanisms that ensure sufficient cell adhesiveness during this process remain poorly understood. Loss of tissue integrity occurs during epithelial-mesenchymal transition (EMT), a process by which epithelial stationary cells acquire a migratory phenotype. EMT occurs during normal development (e.g. gastrulation or neural tube formation), and importantly, it is a key step in malignant transformation(Kang and Massague, 2004; Shook and Keller, 2003; Thiery, 2003). Given the importance of tissue integrity maintenance, both during normal development and during invasive malignancies, it is crucial to understand the molecular mechanisms and the signalling pathways that ensure it. Here, by studying Drosophila tracheal development, we unveil a new role for the Egfr pathway in maintaining epithelial integrity.

The tracheal (respiratory) system of the fly has become one of the most amenable and best-documented models for tubulogenesis (reviewed by Affolter et al., 2003; Hogan and Kolodziej, 2002; Lubarsky and Krasnow, 2003). The tracheal tree consists of a network of epithelial tubules that oxygenate the tissues. It arises from clusters of cells in the ectoderm, known as tracheal placodes, which are specified at mid-embryogenesis and invaginate. Subsequently, and by processes of directed cell migration, changes in cell shape and rearrangement of the cells within the tissue, tracheal cells undergo stereotyped processes of branching and branch fusion that generate the mature structure by the end of embryogenesis. Convergent efforts have identified the tracheal requirements of several genes and signalling pathways (reviewed by Ghabrial et al., 2003). However, the links between the genetic pathways that govern tracheal development and the cellular responses remain elusive.

As tracheal development occurs in the absence of cell proliferation and cell death (Samakovlis et al.,1996), tracheal cells undergo tremendous rearrangements to give rise to fully extended branches without compromising tube continuity. Tracheal development is, therefore, an ideal model with which to tackle the analysis of tissue remodelling and tissue integrity maintenance in vivo. To date, few elements have been shown to regulate tracheal epithelial integrity. The nuclear zinc-finger protein Hindsight (Hnt; Pebbled-FlyBase)(Wilk et al., 2000) and a component of the luminal extracellular matrix(Jazwinska et al., 2003),Piopio (Pio), are required for maintaining tracheal integrity. Other epithelial integrity regulators impinge on the cellular junctions that attach epithelial cells to one another and that establish their apico-basal polarity. Mutants for Lachesin (Lac) and other septate junction (SJ)components produce luminal breaks that result from an insufficient adhesion between tracheal cells (Llimargas et al.,2004; Wu and Beitel,2004). The small GTPase Rac has been shown to regulate epithelial integrity by modulating the assembly and disassembly of E-Cadherin (DE-cad) at the adherens junctions (AJs) (Chihara et al., 2003). Regulation of the activity of AJs lies at the basis of many of the coordinated cellular changes that occur during extensive tissue remodelling (Lecuit, 2005; Pilot and Lecuit, 2005). Indeed, the remodelling of AJs is required for proper tubulogenesis and intercalation in the developing trachea(Ribeiro et al., 2004).

Here, we unveil a new role for Egfr through the ERK-type MAPK pathway (hereafter referred to as the MAPK pathway) in maintaining epithelial integrity. We show that the downregulation of this pathway results in a loss of tissue continuity as branches extend. Conversely, upregulation of the pathway produces defects in branch extension consistent with excess stiffness. Remarkably, Egfr- promoted epithelial integrity does not require the nuclear transcription factor Pointed (Pnt), which otherwise mediates most of the Egfr pathway outcomes. In contrast to Egfr, MAPK pathway regulation by Breathless (Btl, another tyrosine kinase receptor known to trigger the pathway in the trachea) seems not to affect epithelial integrity. The defects we observe in tracheal integrity correlate with subtle differences in the accumulation of DE-cad and of an apical actin belt. Consistently, mutants impairing AJs or actin cytoskeleton assembly [shotgun (shg),encoding DE-cad, and crossveinless c (cv-c), encoding a RhoGAP involved in actin dynamics, respectively] are also required for integrity of the tissue.

The following stocks are described in FlyBase(http://flybase.bio.indiana.edu):rl^10a, UAS-rl^sem, pnt^Δ88,pnt⁷³⁷, Egfr^f2, UAS-Egfr^DN, UAS-λtop,btl^LG19, btl^H82Δ11, UAS-btl^DN,UAS-βGal, Df(3L)H99, shg², Mkp3^5J4,UAS-Ras^N17, UAS-Raf^DN, rho^PΔ5,UAS-tauGFP, UAS-DE-cadGFP and EP3142. The following Gal4 drivers were used:btlGal4 to drive expression in all tracheal cells from invagination onwards,ptcGal4, armGal4 and 69B. Mkp3^M76-R2b and Mkp3^M76 have been described previously (Gomez et al.,2005), as has cv-c⁷(Denholm et al., 2005). UAS-p35 was obtained from S. Campuzano (CBM, Madrid), btlGFP and hsflp;btlenhancer>y+>Gal4, UAS-GFPactin; btlenhancer-mRFP-moe from M. Affolter (Biozentrum, Basel), and GS10283 from the Drosophila Gene Search Project (DGSP). To recognise the chromosomes carrying the desired mutations,we used second or third blue balancers.

To induce the expression of UAS-Egfr^DN or UAS-λtop in small groups of tracheal cells, we crossed the lines to hsflp;btlenhancer>y+>Gal4, UAS-GFPactin;btlenhancer-mRFP-moe (Ribeiro et al.,2004). Embryos were collected at 25°C for 5-6 hours, heat shocked for 35 minutes at 36°C, and transferred to 25°C before fixation. Raising the embryos at 29°C (with or without heat shock) induced btlGal4 in virtually all tracheal cells. Therefore, embryos could not be raised at 29°C after the heat shock to achieve the maximal efficiency of the Gal4/UAS system, and therefore the strongest and more penetrant phenotypes with the transgenes.

The GS element in line 801 was mapped by inverse PCR techniques following standard protocols (Berkeley Drosophila Genome Project, please contact authors for details).

Embryos were staged according to Campos-Ortega and Hartenstein(Campos-Ortega and Hartenstein,1985) and stained following standard protocols. Immunostainings were performed on embryos fixed in 4% formaldehyde for 20 to 30 minutes,except for DCAD2 stainings, for which embryos were fixed for 10 minutes. The following antibodies were used: anti-GFP (Molecular Probes and Roche), mAb2A12(Developmental Studies Hybridoma Bank, DSHB), anti-Pio (from M. Affolter),anti-Hnt (from R. Wilk, University of Toronto, Toronto), anticleaved caspase-3(Cell Signaling Technologies), anti-DSRF (2-161 from Cold Spring Harbor Laboratories), anti-Kni (developed by J. Reivitz and provided by M. Ruiz-Gomez, CBM, Madrid), anti-DE-cad (DCAD2, Developmental Studies Hybridoma Bank), anti-actin (MP Biomedicals), anti-Trh (made by N. Martín in J. Casanova's Laboratory, IBMB, Barcelona) and anti-β-Gal (Cappel and Promega). Biotinylated or Cy3-, Cy2- and Cy5-conjugated secondary antibodies(Jackson ImmunoResearch) were used at a dilution of 1/300. For HRP histochemistry, the signal was amplified with the Vectastain-ABC kit. For fluorescent stainings, the signal was amplified with TSA (NEN Life Sciences)when required. Photographs were taken in a Nikon Eclipse 80i microscope. Confocal images were obtained with a Leica TCS-SP1- or TCS-SP2-AOBS system,Leica DM IRE2 microscope and LCS software.

Unless otherwise stated, in all panels labelled `GFP' the embryos carried btlGal4 UAS-tauGFP, and these were stained with anti-GFP to highlight the shape of tracheal cells. btlGal4 also drove the expression of the indicated UAS constructs. We used mAb2A12 or CBP to visualise the lumen.

In situ hybridisation was performed according to standard protocols with digoxigenin-labelled RNA probes prepared from the Mkp3 cDNA clone LD02618.

Fifty embryos of the selected genotypes were ground up in 100 ml lysis buffer [125 mM Tris (pH 6.8), 21% Glycerol, 5% SDS, 10% bromophenol blue andβ-mercaptoethanol]. For better protein extraction, samples were boiled five times at 91°C for 1 minute, and sonicated for 1 second. Proteins were separated by SDS-PAGE (10% polyacrylamide) and transferred onto a nitrocellulose membrane (Schleicher and Schuell). The membrane was blocked in PBT-5% milk overnight at 4°C, incubated with primary antibody (anti-GFP 1/750, or anti-α-Tubulin 1/500, Sigma) and, subsequently, with HRP-linked secondary antibody (1/10000, Amersham). Proteins were visualized by an enhanced chemiluminescence (ECL) detection system (Amersham). The blots were reprobed sequentially with both antibodies to obtain the specific proteins levels of each sample. Intensities of the bands were quantified by densitometric scanning of the film exposed to chemiluminescence using the QuantityOne program (BioRad).

Embryos carrying btlGal4 UAS-tauGFP;801 were collected at 29°C and dechorionated for 2 minutes with sodium hypochlorite diluted 1/100. They were glued to a coverslip and mounted in 10S Voltalef oil with the hanging drop method in an oxygen-permeable chamber. Images were collected from stage 14 at 21°C on a Leica TCS-SP2-AOBS system, Leica DM IRE2 microscope and LCS software. The 488 nm emission line of an Argon laser was used for excitation and sections were recorded every 6 minutes over a 3-hour period. TIFF projection images were processed into 3D and 4D LCS software, and the movie was assembled using ImageJ.

For DCAD2 stainings, the mutant and control populations were collected,fixed and immunostained together. Mutant and control sibling populations were obtained by crossing btlGal4 UAS-tauGFP/CyO to UAS-λtop, UAS-Egfr^DN, or line 801(identifying the mutant and control population by the presence or absence of GFP), or from a rho^PΔ5/TM3 ftzlacZ stock (identifying the mutant and control population by the presence or absence of anti-β-Gal staining). We compared the levels of DCAD2 staining between the two populations within each experiment in a high number of embryos (n>20). We note that there was variability in the accumulation of DCAD2 in each experiment, both within the mutant population and within the internal control population, but the differences we report in the Results between controls and mutants, although subtle, were consistently and reproducibly seen. In Fig. 5, we show representative, although extreme, examples from the different genotypes.

To quantify DE-cad levels, eight to 11 embryos of each genotype were scanned with a 63× objective zoomed 4×, using the same laser power for each experiment. Projections of 0.5 μm confocal sections of the posterior region of the tracheal tree were analysed. We quantified the pixel intensity of staining by taking four measurements along the dorsal branches(DBs) and six measurements of the contour of DT cells in each embryo with a freeline tool in ImageJ 1.25 (Rasband, http://rsb.info.nih.gov/ij). We calculated the average of the four measurements of the DBs and six measurements of the DT for each embryo, and compared the mutant and control populations. The levels for each experiment depended on the staining and the laser power used, and therefore different experiments cannot be directly compared. However, to facilitate the visualisation of the differences, the average levels of the control population of each single experiment were normalised to 100%, and the level of staining of the mutant population is expressed as the percentage of the level of the control population.

To identify new genes involved in tracheal development, we used the Gene Search (GS) system (Toba et al.,1999) to generate an original collection of lines (C.C. and M.L.,unpublished). We selected a line containing a GS element, termed 801,on the basis of its embryonic tracheal phenotype when crossed to the btlGal4 driver, which is expressed in all tracheal cells from invagination onwards. Tracheal development of such embryos proceeded normally up to stage 13. At stage 14, we detected defects in branch formation and outgrowth, as several dorsal branches (DBs) and ganglionic branches (GBs)never formed, or contained only one or two cells(Fig. 1B).

In addition to these defects in primary branching, we detected many branch breaks when luminal markers were used (not shown). This phenotype was highly penetrant (100% of embryos, n=98) and commonly observed in thin unicellular branches, such as the most ventral (lateral trunk posterior,LTp-GB) and dorsal branches (Fig. 3J). The branches could be broken at any position, giving rise to cells completely isolated from the rest of the tracheal tree, or to cells abnormally separated and only attached to the stalk by thin cytoplasmic extensions or bridges (Fig. 1C,D). Branch breaks increased with time and they were very conspicuous at late stages of embryogenesis, when sometimes almost no branches, not even the dorsal trunk (DT), appeared continuous(Fig. 1B,C). In vivo imaging showed that tracheal cells start to pull apart, establishing long cytoplasmic extensions that eventually can completely break (see Movie in the supplementary material). We interpret these phenotypes as resulting from a loss of tissue integrity that is likely to be needed to counteract the pulling forces driving branch extension. In addition, the tracheal cells appeared more rounded or cuboidal when compared with the wild type(Fig. 1C,D).

This phenotype was also observed when we used Gal4 drivers that induce a more general and earlier expression, such as armGal4, 69B or ptcGal4 (data not shown). We reasoned that the over- or misexpression of the gene(s) near the GS insertion caused the loss of epithelial integrity that maintains tracheal cell attachments to one another.

We mapped the GS element insertion of line 801 to 75F6. The two closest genes are MAP Kinase Phosphatase 3 (Mkp3; positioned 21 bp 5′ of the GS element) and Misexpression Suppressor of Ras 6 (MESR6; at 6.8 kb 3′ of the element; Fig. 2A). Two independent GS lines, GS10283 [Drosophila Gene Search Project (DGSP)] and Mkp3^M76 (Gomez et al.,2005), located near our insertion, produced the same phenotype when crossed to btlGal4 (Fig. 2B,C), whereas the EP3142 line that drives MESR6expression did not produce a consistent phenotype (data not shown). In addition, a null mutation in Mkp3, Mkp3^M76-R2b(Gomez et al., 2005), reverted the integrity defects of Mkp3^M76 tracheal overexpression(Fig. 2D). Altogether these results point to a major contribution of over- or misexpression of Mkp3 in the tracheal cells to the phenotype.

The expression pattern of Mkp3 is consistent with a role during normal tracheal development. The gene showed a dynamic pattern during embryogenesis (data not shown) (Gomez et al., 2005) [BDGP in situ(http://www.fruitfly.org/cgi-bin/ex/insitu.pl)]. Mkp3 accumulated in the tracheal pits at stage 11(Fig. 2E), and we also observed expression in tracheal branches at later stages(Fig. 2F). As expected, Mkp3 was overexpressed in the tracheal cells of btlGal4-801embryos (Fig. 2G).

We studied the tracheal requirement for Mkp3. The null mutations Mkp3^M76-R2b and Mkp3^5J4 exhibited a mild delay in branching, as if the loss of Mkp3 caused increased tissue stiffness. To quantify the delay, we visualised the tip cells with a DSRF antibody (Affolter et al.,1994) and analysed the number of DBs between metameres 4 to 8 that have reached the dorsal midline at stage 14-15 (not shown). We found a delay in 12% of DBs (n=190) in Mkp3^5J4 mutants compared with 0.3% in wild type (n=50). Increased stiffness may produce defects in cell intercalation. To test this hypothesis, we analysed DBs of late embryos stained with a Trachealess (Trh) antibody to determine whether tracheal cells were properly positioned end-to-end, and with an AJ marker(DCAD2 antibody) to reveal the cell intercalation state(Ribeiro et al., 2004). We detected defects in cell intercalation with both markers(Fig. 3K,N) in Mkp3^5J4 mutants. The lack of a stronger tracheal phenotype could indicate that Mkp3 is not absolutely required during embryonic stages, or that it shares redundant functions with other MAP Kinase Phosphatases (Mkps). Alternatively, the zygotic requirement might be rescued by the maternal contribution. Homozygous females for the null Mkp3^M76-R2b allele are sterile and only lay unfertilised eggs, precluding us from determining the tracheal requirement of the maternally provided protein.

Mkp3 has been shown to act as a specific negative regulator of the ERK-type MAPK (Kim et al., 2002; Rintelen et al., 2003). ERK is a central element of the Ras/MAPK pathway whose activity is regulated by phosphorylation (Roux and Blenis,2004). Upon receptor activation, the signal is transduced through activation of the small GTPase Ras and several serine-threonine protein kinases. We therefore investigated whether this pathway is required to maintain tracheal epithelial integrity. Indeed, we found that tracheal expression of dominant-negative versions of Ras and Raf(Fig. 3A; not shown) produced branch integrity defects in most embryos (95%, n=20, and 100%, n=27, respectively) and a similar expressivity(Fig. 3J). Similarly, the zygotic absence of ERK, encoded by rolled (rl),gave rise to defects in branch continuity(Fig. 3B), and to branch patterning defects. Conversely, the constitutive activation of rl(UAS-rl^sem) in the tracheal cells produced a branching extension delay in 27% of DBs (n=90), and incomplete or impaired cell intercalation (Fig. 3L,O). These defects are consistent with an excess of tissue stiffness as opposed to a loss of tissue integrity upon MAPK pathway downregulation. These results indicate that the MAPK pathway regulates epithelial integrity, unveiling a new role of this pathway during tracheal development.

We next tested for genetic interactions between line 801 and rl. We found that the tracheal overexpression of line 801and a simultaneous reduction of rl in heterozygous rl^10a embryos, increased the branch integrity defects of line 801 (not shown). Conversely, tracheal co-expression of rl^sem and line 801 rescued the phenotype caused by the overexpression of Mkp3(Fig. 3C), demonstrating that rl acts downstream of Mkp3 in tracheal epithelial integrity.

The ETS protein Pnt acts as a nuclear positive effector of the MAPK pathway in many developmental contexts (Rebay,2002). We therefore investigated whether pnt is also required to maintain tracheal integrity. As previously described, null or hypomorphic mutations in pnt,pnt^Δ88 or pnt⁷³⁷, show an absence of secondary branching and migration defects in most branches (Myat et al., 2005; Samakovlis et al., 1996). However, we did not detect breaks in the branches, as we did when the MAPK pathway is downregulated(Fig. 3D). This result indicates that pnt is not involved in branch integrity maintenance and suggests that the MAPK pathway regulates this aspect directly by modulating cytoplasmic targets, or through other nuclear effectors.

MAPK pathway is activated by tyrosine kinase receptors(Rebay, 2002). During tracheal development, two tyrosine kinase receptors have been shown to activate the pathway, Egfr and Btl (reviewed by Ghabrial et al., 2003). We investigated whether either or both receptors are responsible for maintaining branch integrity.

Null mutants for Egfr produced embryos with a strong morphological phenotype, nevertheless, we could also distinguish clear branch integrity defects (Fig. 3E). Consistently, 96% of embryos (n=37) expressing a dominant-negative version of the receptor, Egfr^DN, displayed defects such as branch breaks or cells abnormally separated and attached by cytoplasmic bridges (Fig. 3G,J), as in line 801 overexpression. But in contrast to btlGal4 801 embryos,the general pattern and outgrowth of DBs and GBs was not grossly affected in btlGal4 UAS-Egfr^DN. Accordingly, we detected a normal pattern of kni expression (Fig. 4F) and the correct number of cells in the DBs, indicating that primary branching proceeded normally in these embryos. Conversely, a constitutively active form of the receptor, λtop, gave rise to delayed tracheal extension in 35% of DBs (n=120; Fig. 3M). In addition, we also detected incomplete cell rearrangements and confirmed the lack of cell intercalation in some DBs by the absence of autocellular AJs(Fig. 3P, Fig. 5E). These phenotypes are consistent with an excess of tissue stiffness.

rhomboid (rho) encodes a peptidase involved in the secretion of the Egf ligand (Urban et al.,2001). Apart from other tracheal defects associated with invagination, null rho^PΔ5 mutants showed a similar phenotype to that of the overexpression of line 801(Fig. 3F). Furthermore, the haploinsuficiency of Egfr^f2 or rho^PΔ5 in btlGal4 801 embryos significantly increased the branch integrity defects(Fig. 3J; data not shown). Altogether, the data assign a new role for Egfr signalling in maintaining the epithelial integrity of tracheal branches through the activation of the MAPK pathway.

In contrast to Egfr mutants, null mutants for btl(btl^LG19) did not visibly display tissue integrity defects(not shown), suggesting that btl is not required to maintain branch integrity. However, a btl integrity requirement could be masked by the absence of extended branches in null mutants. Therefore, we used other btl alleles. A dominant-negative form of the receptor(Reichman-Fried and Shilo,1995), UAS-btl^DN, expressed in tracheal cells did not produce branch integrity defects, although it produced mild defects in fusion and terminal branching (Fig. 3H). Additionally, the hypomorphic mutation btl^H82Δ11, which allows some branch outgrowth, did not cause a branch integrity phenotype(Fig. 3I). Moreover, halving btl dose in btlGal4 801 embryos did not significantly affect the penetrance of the branch integrity phenotype, but resulted in a significant increase of branches that do not form(Fig. 3J). These results indicate that btl does not play an essential role in maintaining integrity of the tracheal tissue, although it is essential for proper branching pattern, as has been already shown(Klambt et al., 1992). Hence,we propose that the branching defects caused by MAPK pathway downregulation might be mainly attributed to btl, whereas the branch integrity defects might depend on Egfr.

Tracheal development proceeds in the absence of cell death and cell proliferation (Samakovlis et al.,1996). However, the Egfr pathway has been shown to promote cell survival in several tissues(Cabernard and Affolter, 2005; Kurada and White, 1999). We therefore investigated whether cell death precedes and accounts for the tracheal defects of Egfr downregulation. A caspase-3 antibody revealed scattered cell death throughout the tracheal tree of btlGal4 801 embryos (Fig. 4A), and accordingly we detected few apoptotic cells, and the presence of macrophages(see Movie in the supplementary material). This suggests that, although it could contribute in part to it, cell death does not play a major role in tissue integrity. Consistently, protecting the tracheal cells of btlGal4 801 embryos from dying by expressing at the same time the viral antiapoptotic p35 protein (Hay et al., 1994) still produced branch integrity and primary branching defects (Fig. 4B). In addition,tracheal cells, even those present in fragmented branches, showed a normal(although more rounded) shape, instead of the characteristic condensed shape of apoptotic cells (Fig. 1). Similarly, preventing cell death with p35 or Df(3L)H99, a deficiency that suppresses reaper-dependent apoptosis(White et al., 1994), in btlGal4 UAS-Egfr^DN or Egfr^f2 embryos,respectively, did not rescue the branch integrity phenotype(Fig. 4C,D). These results indicate that cell death is not the primary cause of the branch integrity phenotype.

An aberrant acquisition of tracheal identity, causing a deficient integration of the cells into the tracheal tree, could account for the branch integrity defects. Arguing against this hypothesis we found that most tracheal cells of btlGal4 801, btlGal4 UAS-Egfr^DN or rho^PΔ5 embryos expressed general markers, such as btl(Fig. 4A,B,D,E), the lumen marker 2A12 (Fig. 4C), trh (Fig. 5B,G), hnt, pio and others (not shown), spatially restricted markers, such as kni (Fig. 4F), and later markers, like DSRF (Fig. 4E). Additionally, most branch fusions occurred (not shown),indicating that cells underwent normal differentiation. These results rule out that misspecification of tracheal fates is the cause of the branch integrity phenotype.

Epithelial cells are attached to one another by several types of junctions composed of different protein complexes that occupy the most apical region of the lateral membrane. As a consequence, the cells show a marked apico-basal polarity (Knust and Bossinger,2002; Tepass et al.,2001). The rounded shape of the tracheal cells and the loss of epithelial integrity upon Egfr pathway downregulation are reminiscent of non-polarised cells. We analysed markers for different junctional complexes and did not detect defects in the localisation of most of them (not shown; see below), indicating that tracheal cells did not loose their general apico-basal polarity.

DE-cad, encoded by the shg gene, is a classical cadherin that represents a major constituent of AJs(Tepass et al., 2001). When analysing the accumulation of DE-cad in rho^PΔ5, btlGal4 UAS-Egfr^DN or btlGal4 801 embryos, we detected a reproducible, but mild, decrease in the levels, and a loss in the sharpness of staining when compared with the internal control embryos(Fig. 5B-D; see also Fig. S1 in the supplementary material). The effects were more conspicuous in thin unicellular branches, such as the DBs, where the staining that appears as a line lining the lumen in the wild type(Fig. 5A) was sometimes lost or very reduced. Nevertheless, the decrease was detected in all branches, even in the DT, where the characteristic mesh-like staining that reflects the apical outline of the cells was, in extreme examples, lost or became spotty around some cells (Fig. 5A-D). Conversely, high levels of DE-cad were observed in embryos expressing a constitutively activated Egfr (Fig. 5E; Fig. S1 in the supplementary material), and occasionally we observed abnormally straight junctions compared with the wavy wild type ones,as if the cells were subjected to higher tension.

To further prove the modulation of DE-cad levels upon modulation of the Egfr pathway, we downregulated or overactivated Egfr signalling in small groups of tracheal cells by inducing the expression of Egfr^DN (Fig. 6A-D) or λtop(Fig. 6E-G). In spite of the fact that we had to perform these experiments under conditions of moderate activation of these transgenes (see Materials and methods), we could detect,respectively, a mild decrease or increase in DE-cad levels in several examples, validating the above results.

The relationship and interdependence of AJs and the actin cytoskeleton have been extensively documented (Bershadsky,2004; Carthew,2005; Gates and Peifer,2005; Goodwin and Yap,2004; Zhang et al.,2005). For this reason, we analysed the actin cytoskeleton in our mutant conditions. In the wild type, we observed a prominent cortical actin bundle in the tracheal tubes by late embryogenesis(Fig. 5F). In loss-of-function conditions of the Egfr pathway, we detected thinner accumulation of cortical actin in most embryos when compared with the internal controls(Fig. 5G-I). Conversely, when the Egfr is constitutively activated in tracheal cells, we observed an enrichment of apical actin (Fig. 5J). Altogether, our results establish a correlation between the levels of DE-cad and cortical actin, and the activity of the Egfr pathway during tracheal development.

shg is maternally provided and this contribution allows the development of shg mutants to proceed until late embryogenesis(Uemura et al., 1996). The antibody used to detect DE-cad recognises the endogenous protein, including the maternally provided protein. To analyse the effect of the Egfr pathway on the newly synthesized protein, we made use of a functional DE-cadGFP(Oda and Tsukita, 1999)expressed under the control of heterologous promoters. In otherwise wild-type conditions, this protein is incorporated into the cell junctions(Chihara et al., 2003). We first used btlGal4 to express DE-cadGFP in the tracheal system, and found lower levels of GFP signal, when compared with wild type,under several conditions of Egfr pathway downregulation(Fig. 7A-C; not shown). We confirmed these results in western blot experiments(Fig. 7F). Similarly, when we used widely expressed drivers, such as 69B or ptcGal4, we also detected lower levels of DE-cadGFP in several tissues, such as the trachea(Fig. 7D,E), the salivary glands (Fig. 7G,H) or hindgut(not shown) upon downregulation of Egfr pathway. These results show that the Egfr pathway can regulate DE-cad levels posttranscriptionally in various contexts.

If the branch integrity phenotype of Egfr pathway downregulation is due to a decrease in cell-cell adhesion, we would expect a similar phenotype when the integrity of AJs is compromised. Mutants for shg had been described as being defective in tracheal branch fusion and they display interruptions in the tubes (Uemura et al.,1996). In addition to these phenotypes, we observed that tracheal cells show a rounded shape and that branches tend to break, leaving cells isolated from the rest of the tree, as occurs when the Egfr pathway is downregulated (Fig. 8A). Furthermore, the levels of cortical actin decreased in shg mutants(Fig. 8C).

As an alternative approach to analyse the contribution of the AJs and the actin cytoskeleton to tissue integrity, we compromised the proper assembly of cortical actin. For this, we made use of cv-c mutants. cv-cencodes a RhoGAP involved in promoting the coordinated assembly of the actin cytoskeleton (Denholm et al.,2005) and, accordingly, we observed that it is required for the proper accumulation of cortical actin in tracheal tubes (not shown). cv-c mutants showed a mild branch integrity phenotype that consists of cells abnormally separated and sometimes attached by cytoplasmic bridges(Fig. 8B), and we detected a mild decrease in DE-cad accumulation (Fig. 8D). These results indicate that compromising AJs or the actin cytoskeleton, which can impinge on one another, results in tracheal integrity defects.

In this study, we show that the MAPK pathway, in response to Egfr,but not to btl, ensures the maintenance of tissue integrity during tracheal branching in a pnt-independent manner. The effects of Egfr pathway modulation on tissue integrity correlate with differences in the accumulation of DE-cad and cortical actin. Accordingly, the impairment of proper AJ or actin cytoskeleton assembly gives rise to similar defects to those produced by downregulation of the Egfr pathway. Together, these results suggest that the Egfr pathway regulates the maintenance of tracheal epithelial integrity by modulating, at least in part, and most likely posttranscriptionally, cadherin-based cell adhesion.

The requirement for Egfr signalling in the developing tracheae has been already studied by us and by others. The pathway plays a pivotal role during tracheal invagination and also in primary branching(Bradley and Andrew, 2001; Llimargas and Casanova, 1997; Llimargas and Casanova, 1999; Wappner et al., 1997). In fact, it was suggested that the defects observed in branching could be a consequence of an abnormal invagination of the cells(Bradley and Andrew, 2001; Llimargas and Casanova, 1999). This or these requirements seem to depend on a peak of Egfr activity observed prior to invagination, and correlate with a peak of ERK phosphorylation, as visualised with a dpERK antibody(Gabay et al., 1997).

Here, we document a new role for the Egfr pathway in the regulation of tissue integrity. This new requirement could depend on the described early peak of Egfr activity (Gabay et al., 1997), which would be sufficient to prevent defects at later stages. However, we propose that Egfr-promoted epithelial integrity depends on a later, or continuous but lower, or basal activity of the pathway that does not correlate with detectable ERK phosphorylation. Consistent with this hypothesis, we find that downregulation of the pathway by overexpressing 801 or UAS-Egfr^DN with btlGal4, which is expressed after the early peak of ERK phosphorylation, produces a conspicuous branch integrity phenotype. In any case, tissue integrity defects are mainly observed in the most dorsal and ventral tracheal branches, which are subjected to stronger pulling forces as development proceeds, and, therefore, it is precisely at late stages when defects in tissue integrity are expected.

AJs connecting epithelial cells dynamically disassemble and reassemble,thereby allowing tissue remodelling. Tracheal tissue remodelling might require the fine-tuning of cell adhesion properties, as tracheal cells need to be able to change their relative position (probably by loosening cell adhesion) while maintaining epithelial continuity. Our data indicates that the Egfr pathway is a modulator of this balance, not only in the tracheal system, but also in other tissues undergoing extensive remodelling, such as the salivary glands,where we find a similar regulation of DE-cad and actin levels upon modulation of Egfr signalling (C.C. and M.L., unpublished). Conversely, we do not find such a regulation in more static tissues, like the ectoderm (C.C. and M.L.,unpublished), whose maintenance was proposed to depend on the maternally provided DE-cad protein (Uemura et al.,1996). We suggest that the Egfr pathway plays a role in the modulation of cell adhesion in tissues that undergo dramatic morphogenetic events, which might need the zygotic DE-cad contribution and a more dynamic regulation of cell adhesion. Our results indicating a modulation of junctional complexes and/or the actin cytoskeleton by the Egfr pathway establish a link between a developmental pathway required for many biological events and cell biology in terms of cell adhesiveness and cell shape.

Our results show that downregulation of several intracellular elements of the MAPK pathway produce defects in branch integrity, whereas a constitutively activated form of rl (rl^sem) rescues the phenotype of btlGal4 801 embryos. This suggests that the conserved MAPK cassette is required to maintain branch integrity.

Two tyrosine kinase receptors, Egfr and Btl, activate the MAPK pathway during embryonic tracheal development. However, we find that the two receptors, acting through the same intracellular cascade, elicit different responses. The MAPK pathway requirement in primary branching is likely to depend on input by btl, whereas the tissue integrity requirement is likely to depend on input by Egfr. How does the same MAPK pathway trigger distinct outcomes depending on the receptor that activates it? A temporal and/or spatial differential activation of the MAPK pathway could account for the different outcome. In addition, differences in the composition of the intracellular cascade due to specific transducers for one type of receptor, such as downstream of FGFR (dof; stumps-FlyBase) (Petit et al.,2004), could contribute. Finally, quantitative and/or qualitative differences in the activation of the intracellular transducers by the different receptors could also underlie the outcome diversity.

Similar to our observations, air sac development in Drosophila has been recently reported to require both Btl and Egfr, and each receptor seems to elicit different responses. Furthermore, as we find during embryonic tracheal development, an uncoupling of the MAPK cassette and pnt has been observed during air sac development(Cabernard and Affolter, 2005). These parallels suggest a common mechanism for generating different responses from the same intracellular transduction pathway.

The loss of tissue continuity and cell detachment observed in Egfrdownregulation conditions may be due, at least in part, to a decrease in cell adhesion. Accordingly, we detect a mild, but reproducible, decrease in the accumulation of DE-cad and cortical actin. As inferred from the phenotypes,such a mild decrease could cause a loss of cell adhesion during tracheal remodelling, while not grossly affecting other processes requiring DE-cad-based cell adhesion, such as branch fusion(Tanaka-Matakatsu et al.,1996; Uemura et al.,1996). As expected, we find that compromising AJ assembly or the actin cytoskeleton also gives rise to defects in tracheal tissue integrity.

Cadherins have been shown to support cell cohesion and participate in morphogenetic events. The actin cytoskeleton also plays an important role in shaping the cell architecture and in many morphogenetic processes. AJs and the actin cytoskeleton are intimately coupled, and their formation and maintenance is interdependent (Bershadsky,2004; Carthew,2005; Gates and Peifer,2005; Goodwin and Yap,2004; Zhang et al.,2005). We also observe such interdependence in the tracheal system.

Cadherin-based cell-cell adhesion can be regulated at transcriptional and posttranscriptional levels. The modulation of a DE-cadGFP chimaera driven by heterologous promoters shows that, in our case, DE-cad regulation is posttranscriptional. Several posttranscriptional mechanisms of DE-cad regulation have been proposed(D'Souza-Schorey, 2005), and we can envisage a role for the Egfr pathway in each of them. A first mechanism is at the level of DE-cad endocytic trafficking. In this context, the Egfr pathway could modulate the balance between recycling to the plasma membrane of internalised DE-cad or lysosomal targetting and degradation. A second mechanism of cell-cell adhesion regulation is posttranslational modifications of AJ components, such as phosphorylation or ubiquitination. Finally, another possible mechanism of regulation is through the cytoskeleton. The Rho family of small GTPases plays a key role in actin cytoskeleton regulation, and growth factor receptors such as Egfr have been reported to regulate their activity(Burridge and Wennerberg,2004). Remarkably, the Egfr pathway has been recently shown to regulate the expression of the rhoGAP cv-c in the tracheal placodes(Brodu and Casanova, 2006), and we find that cv-c mutants display tracheal integrity defects,although they are milder than those seen upon downregulation of the Egfr signal. We therefore propose that cv-c is at least one of the effectors of Egfr-mediated modulation of DE-cad levels and tracheal tissue integrity. Further analysis will be needed to disentangle the exact molecular mechanisms and to find other possible mediators of the Egfr signal.

The decrease of cadherin activity upon activation of the Egfr pathway has been extensively reported in the literature(Comoglio et al., 2003; Dumstrei et al., 2002; Lilien and Balsamo, 2005). Here, we report the opposite: that Egfr pathway downregulation correlates with a decrease of cadherin-based cell adhesion. Although this is not the first example of such a relationship (Brown and Freeman, 2003), it illustrates the versatility and complexity of the interactions occurring between signalling pathways and adhesion molecules,and establishes another model with which to analyse how cell adhesion is modulated.

We are grateful to J. Casanova, M. Furriols and other members of J. Casanova's lab for discussions, and to M. Domínguez, V. Brodu, D. Shaye, S. Araújo, M. Furriols and J. Casanova for critical comments and improvement of the manuscript. We thank R. Mendez Castro and N. Martín for excellent technical assistance, N. Ninov for help with time-lapse experiments, and ML. Espinàs for help with western blots. We also thank D. Andrews, B.-Z. Shilo, E. Martín-Blanco, M. D. Martín-Bermudo,E. Hafen, H. Oda, S. Hayashi, M. Affolter, J. F. de Celis, the Bloomington Stock Centre, the Drosophila Genetic Resource Center, and the Developmental Studies Hybridoma Bank for providing flies and reagents. C.C. is supported by a fellowship from the Ministerio de Educación y Ciencia and M.L. by a contract of the `Ramón y Cajal' program. This work was supported by the Ministerio de Ciencia y Tecnología (BMC2002-00359).