A wave of EGFR signaling determines cell alignment and intercalation in the Drosophila tracheal placode

The invagination of epithelia allows the segregation of organ primordia during morphogenetic processes in development, such as gastrulation and the formation of neural tubes, gill slits and sensory organs. It involves the specification of organ placodes and the ordered internalization of epithelial sheets, followed by the epiboly-like movement of cells in neighboring regions that fill the remaining space. Apical constriction of individual cells often precedes the invagination and is thought to play important roles in gastrulation in Drosophila (Costa et al., 1994; Leptin and Grunewald, 1990; Oda and Tsukita, 2001; Parks and Wieschaus, 1991) and in the invagination of other organ placodes(Llimargas and Casanova, 1999; Myat and Andrew, 2000b; Simoes et al., 2006).

Cell intercalation is thought to be the major force driving the global transformation of epithelial shape, such as in embryonic body axis elongation(Irvine and Wieschaus, 1994; Keller, 2002) and gastrulation(Ettensohn, 1985; Hardin and Cheng, 1985). In Drosophila, the orientation of intercalating individual cells or multicellular rosettes of cells accompanies germ band extension under control of the anterior-posterior (AP) patterning system(Bertet et al., 2004; Blankenship et al., 2006; Zallen and Wieschaus, 2004). Oriented cell intercalation is also observed in elongating tracheal branches(Ribeiro et al., 2004). In addition, cell intercalation and apical constriction depend on non-muscle myosin, which is recruited to the cell-cell junctions and provides the contractile force (Bertet et al.,2004; Kiehart et al.,2000).

The tracheal system of Drosophila is formed by the invagination and branching of ectodermal epithelia(Manning and Krasnow, 1993; Samakovlis et al., 1996). The tracheal fate is specified at stage 10, when the HLH-PAS protein Trachealess(TRH) begins to be highly expressed in the tracheal placode(Isaac and Andrew, 1996; Wilk et al., 1996)(Fig. 1A). Apical constriction is observed in the dorsomedial region of the placode, where cells start to invaginate until all the TRH-positive cells are internalized(Fig. 1B). EGFR signaling is activated within the tracheal placode under the control of TRH(Gabay et al., 1997b). In Egfr mutant embryos, the prospective tracheal cells fail to concentrate F-actin at the constriction site(Brodu and Casanova, 2006), and the invagination is partially defective(Llimargas and Casanova,1999). It is not clear, however, to what extent apical constriction and cell intercalation account for the highly coordinated process of tracheal invagination or how EGFR signaling regulates these processes. Progress in elucidating these events has been slow because of the lack of precise knowledge about the cellular movements during invagination.

Here we used time-lapse imaging to trace the dynamic movements of cells and cell interfaces before and during tracheal invagination. We show that changes in the distribution of non-muscle myosin accompanies the rearrangement of cells in the tracheal placode, which encircle the invagination site by aligning in short rows arranged in arcs (a process called `boundary smoothing') and undergoing cell intercalation. These processes are controlled by EGFR, which is activated in an expanding circular pattern, thereby providing spatiotemporal information to cells in the tracheal placode. In the absence of EGFR, tracheal cells were internalized individually, suggesting that the key role of EGFR signaling is to coordinate the intrinsic ingression activity of the cells into the ordered process of invagination.

The following fly strains were used: sqh-GFP-Moe(Kiehart et al., 2000), trh-lacZ (Kassis et al.,1992), MRLC-GFP (myosin-GFP)(Royou et al., 2004), btl-GFP-Moe (Kato et al.,2004), rho-lacZ (a gift from Drs Kenji Matsuno and Yukio Nakamura, Tokyo University of Science, Japan). Information on the following stocks can be found in FlyBase(http://flybase.net/): rho^del1, Egfr^f24, Dsor1^Gp158,pnt^delta88, UAS-sspi, prd-Gal4, ovo^D1 FRT101;hs-FLP³⁸, TM3, Kr-gal4 UAS-GFP.

Imaging was performed as previously described(Kato et al., 2004). Ten 1-μm-thick Z stacks were taken every 2 minutes over a period of up to 160 minutes. Embryos at this stage often rotated during imaging. Only images of a few successful recordings that covered the entire invagination process were used for cell tracing. The number of successfully imaged invagination events and embryos (in parentheses) was: for GFP-Moesin-labeled embryos, control 9(4), rho^del1 5 (2), Egfr^f24 9 (4), pnt^delta88 7 (3), and for myosin-GFP-labeled control embryos, 9 (3). Intercalating cells were marked and traced from still images,and their centroid position was determined using the measurement function of ImageJ (National Institutes of Health, USA). The angle and distance of cell displacement was calculated. The invagination site was defined as the position of the first cell to internalize. For the Egfr^f24 embryos,only cases in which the invagination took place only at one position were used for analysis. The orientation of mitosis was determined in a similar manner,except that the position of invagination was defined as the centroid of the tracheal pit. Mutant embryos were identified by the absence of green balancer. Non-randomness of the orientation of cell displacement and the cell division axis were assessed by the Kolmogorov-Smirnov test.

The primary antibodies were against the following proteins: DLG, E-cadherin(Developmental Studies Hybridoma Bank), anti-double phosphorylated ERK(dp-ERK; Sigma), KNI (Kosman et al.,1998), GFP (rabbit, MBL; chick, Chemicon), myosin heavy chain(MHC: a gift from Fumio Matsuzaki, RIKEN, Center for Developmental Biology,Kobe, Japan) and β-galactosidase (Cappel). The dp-ERK signal was amplified with a TSA indirect kit (Perkin Elmer Life Sciences). Signal intensity of MHC-stained cell boundaries were quantified by the use of ImageJ and the statistical significance was tested by Student's t-test(two-sample assuming equal variances, two-tail).

To study the shape changes, movement, and proliferation of tracheal primordial cells, we used time-lapse imaging to observe embryos labeled with the F-actin marker GFP-Moesin (Kiehart et al., 2000) or myosin marker MRLC-GFP(Royou et al., 2004)(hereafter called myosin-GFP; Fig. 2, control, see Movies 1, 2 in the supplementary material) and compared the results with the images of fixed preparations(Fig. 1A,B)(Brodu and Casanova, 2006; Isaac and Andrew, 1996; Llimargas and Casanova, 1999; Manning and Krasnow, 1993; Wilk et al., 1996). The use of GFP markers allowed us to compare the dynamic distribution of F-actin and myosin during invagination.

About 50-70 minutes prior to the start of invagination, at cycle 15 of embryogenesis, the dorsal ectodermal cells entered mitotic quiescence(Fig. 2A and see Movie 1 in the supplementary material). Then, about 10 cells in the dorsal-medial part of the placode started to show apical constriction(Fig. 1Ba; Fig. 2B, yellow; defined as a reduction of apical cell surface below 5 μm²) and shifted basally about 3 μm (Fig. 1Bb, defined as time 0'). The cells with the constricted apical surfaces then became internalized, leaving a cleft called the tracheal pit(Fig. 1Bc, Fig. 2B, pink). Subsequently,cell internalization continued without apparent apical constriction(Fig. 1Bd), and mitotic activities resumed around the tracheal pit(Fig. 1Bc,d, Fig. 2B, asterisks). Time course analyses of apical constriction showed that those cells that internalized with sharply constricted apices (<1 μm²) were clustered at the center of the placode(Fig. 1C, red outline). We also mapped the timing of cell internalization within the tracheal placode and found that cells that internalize early were clustered at the center,surrounded by later internalizing cells in a stepwise manner, forming concentric circles (Fig. 1C). The cell size and temporal analyses of cell internalization demonstrated that the cells around a prospective tracheal pit internalized first with apical constriction (phase 1), followed by those surrounding them, with less extensive apical constriction (Fig. 1C). The timetable of the tracheal invagination events is shown in Fig. 1E.

To examine how the concentric groups of cells formed, we analyzed the pattern of cell intercalation in the 30-minute period prior to invagination,by tracing the relative positions of neighboring cells(Fig. 3A, blue and red arrows). The direction of cell displacement was quantified by measuring the angle of the line connecting the centroids of the two displaced cells(Fig. 3B). We tested two alternative hypotheses, (1) that cell intercalation follows a polar coordinate system with the center situated at the point of invagination, or (2) that it follows a Cartesian coordinate system defined by the AP and DV(dorsal-ventral) embryonic axes. The displacement angle relative to the polar coordinate (|θ|) was 24.4±3.4° [average and standard error (s.e.m.), n=21]. This value greatly deviates from the average of 45° that would be expected if displacement occurred randomly(Kolmogorov-Smirnov test, P<0.05). However, the average angle relative to the AP axis (|θ′|) was 41.8±6.1° (n=18), which was judged to be random. Therefore, the result favors the polar coordinate model.

After the initiation of invagination, the tracheal cells entered a final wave of mitosis. The orientation of the cell-division axis of the placodal cells was significantly biased toward the center of the tracheal pit(|θ|=14.6±2.78°, n=22, P<0.01; Fig. 1D). A radial orientation of cell division might help direct cells to flow into the site of invagination. Taken together, these observations indicate that the orientation of the cell displacement prior to invagination (phase 1), and the cell division axis in phase 2 of invagination in the tracheal placode are both aligned toward the site of invagination, suggesting that the planar cell polarity in the tracheal placode is polarized toward the invagination site.

In the neighborhood of the tracheal placode cells undergoing intercalation and we frequently noted groups of four to six cells forming arc-like rows that collectively surrounded the future invagination site(Fig. 3A, time 2'). By tracing the contours of the cells in these rows, we found that these smooth arcs of well-aligned cells arose from rows of cells whose contours formed zig-zagging boundaries (Fig. 3A; green and orange lines). We call this process `cell-boundary smoothing' and characterized its cellular properties by monitoring the accumulation pattern of myosin-GFP (Royou et al.,2002) as a marker for contractility of the cell-cell junctions(Bertet et al., 2004; Kiehart et al., 2000) (see Fig. 6 for quantitative assessment of this process). Myosin-GFP was highly concentrated at the dorsal border of the ectoderm that had begun to form the contractile supracellular actomyosin purse string (Kiehart et al.,2000) (Fig. 3C),suggesting that myosin-GFP accumulation is a hallmark of contractile activity at cell junctions. When dorsal ectodermal cells entered mitotic quiescence before invagination, the myosin-GFP level in apical cell interface was generally low except for a high accumulation between the straight rows of cells abutting the segment boundary (see Movie 2 in the supplementary material, time -40'). The prominent dot-like signals in each cell appeared to be an artifact of the GFP fusion protein, because they were not detected with anti-myosin heavy chain (MHC) antibody staining(Fig. 6B) and were not considered further. In the next 20 minutes, myosin-GFP accumulated at the apical cell junctions of other ectodermal cells in non-uniform and rapidly changing patterns. The signal was especially upregulated in shrinking cell junctions undergoing cell intercalation(Fig. 3E)(Bertet et al., 2004). When the arc-like rows of cells started to appear in the tracheal placode, the myosin-GFP accumulation was often elevated in the smoothed boundaries of the cellular arcs surrounding the invagination site(Fig. 3C and see Movie 2 in the supplementary material). We also noted a number of cell intercalation events associated with smoothing boundaries (Fig. 3A).

The enrichment of myosin-GFP in the arc-like row boundaries was transient,and took place in multiple waves: first in cells that invaginated early, and then in later-invaginating cells (Fig. 3D, boundary sets indicated with blue, green, and magenta lines;see Movie 2 in the supplementary material). Tracing the time-lapse images revealed that these arc-like boundaries were formed as initially separate cells, or ones that were newly juxtaposed, aligned themselves into continuous rows (Fig. 3D). Thus, the formation of a smooth row contour from an initially discontinuous and zig-zagging row boundary was accompanied by an elevated accumulation of myosin-GFP. As the rows of cells moved closer to the invagination site, their boundaries became discontinuous again (see the boundary marked in blue, Fig. 3D, time -2'). Therefore,the observed boundary smoothing was transient. Increase in myosin-GFP along cell boundaries was also detected in other parts of the ectoderm, but it appeared to be random and did not correlate with cell movement. For example,the tracing of cell contours in the dorsal ectoderm revealed that although these boundaries were occasionally broken by the insertion of new cells, their positions and mutual relationship remained relatively constant(Fig. 3D, orange lines). These dynamic patterns of myosin-GFP accumulation were different from the distribution of F-actin revealed by GFP-Moesin(Fig. 3A), which was more uniform and constant, suggesting that the selective recruitment of myosin to cell junctions might trigger localized contractile activities. We concluded that many cells in the tracheal placode are rearranged into transient arc-like rows through the processes of cell intercalation and smoothing of arc-like cell boundaries.

As a candidate signal for coordinating the boundary smoothing and cell intercalation, we studied the expression pattern of EGFR signaling. Active ERK MAP kinase was visualized with an anti-double phosphorylated ERK (dp-ERK)antibody (Gabay et al.,1997a). dp-ERK was first detected in both the cytoplasm and nuclei of the cells in the tracheal placode before they entered mitotic quiescence and started to constrict apices (Fig. 4A), in a pattern roughly overlapping with the expression of the EGFR activator rhomboid (rho; Fig. 4F; detected by the expression of rho-lacZ), within the cells in the tracheal placode expressing trh-lacZ and KNI (Fig. 4F,G). The dp-ERK expression then expanded to cover the dorsal half of the placode (Fig. 4B). A y-z sectional view of the tracheal placode doubly labeled for dp-ERK and nuclei (Fig. 4Bb) showed that the dp-ERK in the central domain was concentrated in the apical cell cortex, and in the peripheral domain it accumulated in both the cytoplasm and nuclei. Thus, cells with nuclear dp-ERK signals formed a ring in a horizontal optical section at the basolateral level(Fig. 4Bc). We found no sign of apical constriction and cell boundary smoothing was partial at this stage(Fig. 4Bd,e). During the next stage, cell boundary smoothing proceeded and the central cells with an apical accumulation of dp-ERK showed constricted apices and invaginated, and then rapidly lost the dp-ERK signal during phase 2 of invagination, when mitotic activity resumed (Fig. 4C and see Fig. S1 in the supplementary material). dp-ERK was reduced in rhomutants (data not shown), and lost completely in Egfr mutants(Fig. 4D,E). We concluded that the intracellular location of the dp-ERK changed from nuclear and cytoplasmic to the apical cortex in about a 30-minute period prior to invagination, and dp-ERK was subsequently eliminated after the onset of invagination. In the tracheal placode, the pattern of nuclear dp-ERK expression was spatiotemporally regulated by EGFR: it originated from a central spot and expanded into a ring that encircled the prospective site of the apical constriction and invagination.

EGFR signaling is required for tracheal cell invagination(Llimargas and Casanova, 1999; Wappner et al., 1997). In mutants of rho, a positive regulator of EGFR signaling, some tracheal precursor cells fail to invaginate and remain at the epithelial surface(Llimargas and Casanova, 1999)(Fig. 5B). A similar phenotype is observed in mutants of EGFR and the ETS domain transcription factor Pointed(PNT), a key target of ERK signaling(Brunner et al., 1994; O'Neill et al., 1994)(Fig. 5C,D), suggesting that EGFR signaling and its nuclear transduction are essential for proper invagination.

To investigate the role of EGFR signaling in the tracheal placode invagination events, time-lapse imaging of rho, Egfr and pntmutants was performed, and their phenotypes were compared with that of control embryos. As a staging reference, we used the cell division in the dorsal epidermis that normally resumes about 60 minutes after the onset of invagination (Fig. 2B-E). We found that defects in the Egfr mutants were already apparent at the onset of invagination. In the Egfr mutants apical constriction was nearly absent (Fig. 2C,compared with the yellow-colored cells in 2B), and the remaining apical constriction event did not correlate with the position of invagination. These defects in apical constriction are consistent with a previous report implicating EGFR signaling in the initiation of local F-actin assembly(Brodu and Casanova, 2006). Furthermore, we found that the onset of invagination was delayed in the Egfr mutants, and some cells resumed mitosis before the start of invagination (Fig. 2C, 46 minutes, asterisk), suggesting that the temporal order of invagination and mitosis was miscoordinated. The delayed invagination was confirmed by examining fixed mutant embryos. In control embryos, tracheal precursor cells started to constrict their apices and shift basally when the groove between the maxilla and labium ingressed deeply, but no such morphological changes were detected in the tracheal placodes of Egfr mutants at the equivalent stage (data not shown). We also imaged rho and pnt mutant embryos and found that invagination was delayed and apical constriction was less extensive than in control embryos(Fig. 2D,E and see Movies 4, 5 in the supplementary material). In addition, we monitored cell boundaries in pnt mutants and found some degree of smoothing (see Fig. S2 in the supplementary material). The phenotypes of rho and pntmutants are weaker version of the Egfr mutant phenotype. These data suggest that EGFR signaling is required for the apical constriction and specification of the timing of invagination.

We next assessed the position of the tracheal pits in control and mutant embryos. In the control embryos, a single tracheal pit opened at the dorsal part of the tracheal placode (at about 25% of the length of the tracheal placode from the dorsal margin; Fig. 5H). In Egfr mutants, the tracheal placode expanded ventrally to include more cells at the expense of reduced ventral ectoderm(Fig. 5F)(Raz and Shilo, 1993). In the expanded tracheal placode of the Egfr mutants, the position of the tracheal pit was more variable and tended to shift ventrally, and occasionally two tracheal pit-like indentations were observed(Fig. 5G,I). To determine whether ERK signaling was involved in this process, we studied the phenotype of maternal-zygotic mutants of Dsor1, which completely lack ERK kinase activity (Hou et al.,1995; Tsuda et al.,1993). These mutants had misplaced and duplicated tracheal pits(Fig. 5J). Time course analyses of the Egfr mutants revealed multiple incidences of one or two cells being internalized prior to the full invagination of the larger group of cells forming the visible tracheal pit (Fig. 2C and see Movie 3 in the supplementary material). Most of the cells undergoing these individual internalization events were adjacent to mitotic cells.

Although frequent mitosis appeared to interrupt the cell rearrangement process, we were able to detect some arc-like cellular rows in the Egfr mutant placode (Fig. 2C, time 60'; see Movie 3 in the supplementary material). However,the pattern of these arc-like rows did not correlate with the position of the invagination site (Fig. 2C). We also detected a limited number of cell intercalation events, and measurement of their orientation revealed a |θ| that was judged to be non-biased (43.9±7.2°, P>0.05). Taken together, these results suggest that EGFR signaling is required for the proper patterning of cell rearrangement and to restrict the site of tracheal invagination to a single focus at the dorsal region in tracheal primordia.

We next investigated the role of EGFR signaling in the regulation of myosin recruitment to cell-row boundaries. We found that the outer boundaries of cells with high nuclear dp-ERK often correlated with the arcs of cells with a high accumulation of myosin-GFP (Fig. 6Aa and see Fig. S1 in the supplementary material). To determine whether EGFR was required for this myosin accumulation, we stained embryos with an antibody against MHC. The staining revealed the smoothed boundaries of cells aligned in arc-like rows in the tracheal placode of control embryos(Fig. 6B). In order to quantify the occurrence of arc-like rows, we classified tricellular junctions in the tracheal placode into two classes (Fig. 6D). Class T junctions contain two of the cell interfaces forming an angle of 180±30°. Other junctions were classified as class Y. Cells were judged to be forming an arc if `horizontal' class T junctions(Fig. 6D) were found in adjacent cells (less than one cell boundary apart). In control tracheal placodes (Egfr/+, n=3), an average of 11.0 class T junctions were observed per placode (T+Y=83 per placode), and 7.3 (66%) of them were judged to be in arcs. No such arc-like rows were apparent in Egfrmutants (Fig. 6C). The average number of class T junctions was reduced to 5.2 per placode despite 49%increase in placode size (T+Y=124 per placode). We found one case where three class T junctions were in a row. It is likely that those cells were not participating in invagination (Fig. 2C and see Movie 3 in the supplementary material).

To quantify the differential distribution of myosin, we classified cell boundaries of class T junctions into horizontal and vertical(Fig. 6D), and their intensity was compared. Intensity of horizontal class T boundaries was 1.52±0.11(mean±s.e.m., arbitrary unit, n=33), that was significantly increased compared to 1.00±0.11 in vertical boundaries(n=37, P<0.002, Student's t-test). Such enrichment of myosin to horizontal boundaries was observed in class T junctions both in arcs or not. Since two-thirds of class T junctions formed arcs, we conclude that myosin is preferentially enriched in cell junctions consisting of arcs.

To test whether EGFR was capable of inducing this cortical myosin accumulation, we ectopically activated EGFR in otherwise flat epithelia by expressing the secreted (activated) form of an EGFR ligand, Spitz (sSPI)(Schweitzer et al., 1995). The expression of sSPI by the paired enhancer(Brand and Perrimon, 1993)activated ERK kinase in broad bands that covered the even-numbered parasegments. At the interface of cells expressing ectopic dp-ERK and their neighbors, we often observed an elevated myosin accumulation(Fig. 6E, arrowheads). Comparison of myosin accumulation in this type of border (DV borders, Fig. 6F) versus intersecting borders (AP borders) demonstrated that dp-ERK activation specifically enriched myosin accumulation in the interface with low dp-ERK-expressing cells(Fig. 6F). The result suggests that the juxtaposition of cells with high and low levels of EGFR activity can trigger myosin accumulation at the cell interface. Furthermore, within the region of high dp-ERK activity, we noted the formation of deep epidermal depressions at the sites where segmental furrows would form(Fig. 6E, asterisk): the depressed region had abundant cells with constricted apices and became continuous with the expanded tracheal pit. The odd-numbered parasegments remained flat (arrow). These results suggest that a high level of EGFR signaling promotes the apical constriction of epidermal cells and precocious epidermal depression.

The mechanical basis of invagination has long been a subject of debate. Cell-shape change, cell migration and cell intercalation are some of the proposed mechanisms (reviewed by Keller et al., 2003). Apical constriction converts cells into a bottle-like shape and is thought to be an important process in invagination(Myat and Andrew, 2000a). However, the observation that cells can be internalized without undergoing apical constriction during normal invagination of the Drosophilamesoderm (Oda and Tsukita,2001) and trachea (this work), and as a result of mutations that disturb apical constriction but allow the invagination of a majority of placodal cells (Llimargas and Casanova,1999; Oda and Tsukita,2001; Sweeton et al.,1991) suggest that the critical role of apical constriction is not in invagination per se, but is likely to be modulatory. We have shown here that invagination of the tracheal placode involves, prior to the start of invagination, the rearrangement of cells to form multicellular arcs with smooth boundaries. We have also shown here that polarized cell intercalation and boundary smoothing take place simultaneously in multiple waves, and together with apical constriction, coordinate the cells' intrinsic internalization activities to allow precisely regulated invagination. Our findings suggest a model in which EGFR signaling regulates this process(Fig. 6G).

We have shown that the accumulation of myosin-GFP was elevated in the boundaries of cells that were arranged in arc-like rows. The linkage of a cortical actomyosin network through such a chain of cells via a cell-cell adhesion complex would result in the formation of supracellular actomyosin cables similar to those observed at the leading edge of the dorsal epidermis(Hutson et al., 2003; Kiehart et al., 2000). Although the cables formed were transient and weak compared with their counterparts in the epidermal leading edge, their contractile force in convex cell contours would help smooth the boundary and compress cells toward the center, thereby accounting, at least in part, for the dense cell packing in the prospective tracheal pit. In addition, polarized cell intercalation contributes to cell packing by pushing cells toward the center and connecting cell boundaries to form larger arcs. These packed, apically constricted cells were first to internalize (Fig. 6G).

Although it was previously noted that EGFR plays an important role in the accumulation of myosin in cells undergoing apical constriction(Brodu and Casanova, 2006), it was not clear whether EGFR influences cells surrounding the invagination site. Here we showed that the nuclear transduction of EGFR signaling monitored by dp-ERK formed a transient wave that swept from the center to the periphery of the placode. Since PNT, the major nuclear target of EGFR-ERK signaling, was required in this process, nuclear dp-ERK must represent the major EGFR signaling output required for invagination. Our observation that nuclear dp-ERK was already extinguished at the time of invagination suggests that the effect of EGFR in invagination might be indirect. Since the spreading front of the dp-ERK often correlated with high myosin accumulation, we speculate that the circular propagation of EGFR signaling in the peripheral region helps to specify the cell-cell interface where myosin accumulates.

A good candidate for the link between nuclear dp-ERK and myosin regulation is the RHO GAP gene cv-c, which is transcriptionally upregulated by EGFR (Brodu and Casanova,2006). Since RHO GAP is linked to the regulation of myosin contractility through the RHO-RHO kinase-myosin regulatory light-chain pathway(Amano et al., 1997; Winter et al., 2001), the transcriptional activation of cv-c within a region of high EGFR activity should downregulate myosin activity. Although the mechanism is not known, a difference in myosin activity between neighboring cells might trigger contractile activity. It follows that the circular spread of EGFR activity would cause the sequential modulation of myosin contractility throughout the tracheal placode. The requirement for pnt for proper invagination indicates transcriptional regulation is essential. However, the relatively weaker invagination phenotype in pnt mutants compared to Egfr mutants suggests that another nuclear target, or a non-nuclear pathway of EGFR signaling through apically accumulated dp-ERK regulates invagination.

In Egfr mutants, the onset of invagination was delayed and took place at variable positions, and phase 1 of invagination was apparently skipped. These phenotypes can be explained by the loss of the putative centripetal force driven by the myosin-dependent contraction of the arc-like rows of cells. One striking observation in the Egfr mutants was the internalization of multiple individual cells prior to invagination. This observation indicates that EGFR is dispensable for cell internalization per se, but is essential for focusing the cell internalization activity, thus specifying the time and place of the initial invagination. This result also indicates that cells in the tracheal placode possess an EGFR-independent cell-ingression activity. Although the origin of this activity is not understood, it may be under the control of genes specifying the tracheal placode, such as trh and vvl. It is interesting to note that the cell internalization events observed in the Egfr mutants were often associated with mitosis, which might alter the local surface tension in the placode to permit neighboring cells to ingress. Thus, EGFR seems to restrict the invagination site to one place through two mechanisms: first by coordinating the circular cell rearrangement, and second by suppressing cell division to stabilize the surface tension of the placode.

The activation of EGFR signaling started from one spot and spread out in a ring that encircled the prospective invagination site. Since the pattern of rho transcription was not sharply focused in the tracheal placode, it is likely that the secreted Spitz ligand is broadly distributed. The circular spread of EGFR activation from this crude distribution of EGFR ligand might require feedback regulation of EGFR signaling(Shilo, 2005), as has been shown in other cases (Brodu et al.,2004; Wasserman and Freeman,1998).

In this study, the precise mapping of dp-ERK and cell behaviors allowed us to demonstrate the role of EGFR signaling in cell rearrangement. Drosophila leg development bears several similarities to tracheal invagination: it develops by evagination, a process fundamentally similar to invagination, and requires graded EGFR activity(Campbell, 2002; Galindo et al., 2002) and TRH(Tajiri et al., 2006). FGF signaling is known to have a role in the invagination of the lens and otic placode in the chick and mouse (Faber et al., 2001; Ladher et al.,2005). The concept of cell rearrangement at the signal propagation front may be useful for understanding the cellular mechanisms underlying the shape transformation of other organ placodes under the control of receptor tyrosine kinases.

We are grateful to Kagayaki Kato (RIKEN, CDB) for help with the image analysis. We thank Jordi Cassanova, Dan Kiehart, Fumio Matsuzaki, Kenji Matsuno, Yukio Nakamura, the Bloomington Stock Center and the Developmental Study Hybridoma Bank, for Drosophila strains and antibodies, and members of the Hayashi laboratory for helpful comments on the manuscript. This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas(C) `Genome Science' from MEXT Japan.