The Drosophila formin DAAM regulates the tracheal cuticle pattern through organizing the actin cytoskeleton

Proteins of the formin family play key roles in the regulation of the cytoskeleton. Although the connection between formins and microtubules is less well understood, formins are implicated in a large number of actin-based processes, including cell polarization, division, movement, stress fiber formation and vesicular trafficking (reviewed by Evangelista et al., 2003; Wallar and Alberts, 2003). Recent work suggested that formins catalyze the assembly of unbranched actin structures and nucleate actin filaments directly(Evangelista et al., 1997; Evangelista et al., 2002; Pruyne et al., 2002; Sagot et al., 2002a; Sagot et al., 2002b).

Although all members of this highly conserved family are defined by the presence of the formin homology domain 2 (FH2), which is both necessary and sufficient to nucleate actin in vitro(Pring et al., 2003; Pruyne et al., 2002; Sagot et al., 2002b), formins are multi-domain proteins that contain several other conserved sequences as well. The FH2 domain is usually flanked on the N-terminal side by a proline-rich FH1 domain that can serve as a docking site for the G-actin-binding protein profilin (Chang et al., 1997; Evangelista et al.,1997; Imamura et al.,1997; Watanabe et al.,1997). In addition, the FH1 domain also binds WW domains and SH3 domains, including those of the Src family(Chan et al., 1996; Kamei et al., 1998; Tominaga et al., 2000; Vallen et al., 2000). A distinct formin subfamily, the DRF group (Diaphanous related formins) has the ability to interact with an activated Rho GTPase through an N-terminal GTPase-binding domain (GBD) (Evangelista et al., 1997; Imamura et al.,1997; Kohno et al.,1996; Watanabe et al.,1997). This binding alleviates the autoinhibitory interaction between the GBD and the short C-terminal DAD domain (Diaphanous autoregulatory domain) (Alberts, 2001). Crystallographic analysis of the N-terminal part of mouse Dia1 provided a structural basis of this regulation and identified several other functional domains within this part of the protein, including a dimerization domain (DD),(Fig. 1B)(Otomo et al., 2005a; Rose et al., 2005). In addition, the crystal structure of C-terminal formin domains reveals that the FH2 domain also forms a dimer required for actin nucleation and processive filament capping (Otomo et al.,2005b; Shimada et al.,2004; Xu et al.,2004), strongly suggesting that native formins act in dimeric forms.

Recently, a novel formin subtype, DAAM (Dishevelled-associated activator of morphogenesis), has been identified and implicated in planar cell polarity signaling during Xenopus gastrulation(Habas et al., 2001). The polarized orientation of cells within the plane of a tissue, or planar cell polarization (PCP), is an important aspect of cellular differentiation and is often necessary to the formation of functional organs. Genes controlling PCP have been extensively studied in Drosophila, revealing a crucial role for frizzled (fz) signaling through dishevelled(dsh) during the course of PCP establishment. The downstream components of the Fz/PCP pathway include RhoA, Drok, the JNK cascade and several other genes, most of which act tissue specifically(Adler, 2002; Mlodzik, 2002). Recent work has demonstrated that many of the genes involved in PCP signaling are also required for polarized morphogenetic cell movements such as convergent extension during early vertebrate embryogenesis(Fanto and McNeill, 2004; Mlodzik, 2002; Strutt, 2003; Wallingford et al., 2002). Habas et al. (Habas et al.,2001) suggested that the novel FH2 protein Daam1 is required for convergent extension is Xenopus embryos and that Daam1 might function as a bridging factor between Dsh and RhoA. Moreover, Wnt/Fz-mediated activation of RhoA appeared to depend on Dvl (a Dsh homolog) and Daam1(Habas et al., 2001). However,in contrast to this model, previous work has provided evidence that formins act as Rho effectors downstream of the Rho GTPases (reviewed by Wasserman, 1998).

To gain further insights into the function of this novel class of FH2 proteins, we analyzed the single Drosophila member of the DAAM family(DAAM). Phenotypic analysis of DAAM mutants showed that it plays either no role or possibly a redundant role in PCP establishment in Drosophila. However, we found evidence that DAAM is involved in the regulation of the actin cytoskeleton in several different tissues,including the tracheal system. The Drosophila tracheal network is one of the best characterized model systems of branching morphogenesis(Affolter and Shilo, 2000; Uv et al., 2003). During the first phase of tracheal development, the primordial cells invaginate in each embryonic hemisegment from the epidermis, migrate and undergo cell shape changes to form the primary branches(Affolter and Shilo, 2000;Manning, 1993; Uv et al.,2003). Subsequently, some tracheal branches fuse with an adjacent branch to build up a continuous tubular network. This process is mediated by fusion cells located at the branch tips, which recognize each other in the adjacent hemisegments and become doughnut shaped, forming a lumen that connects the two branches (Samakovlis et al., 1996). Before the end of embryogenesis, tracheal cells secrete a cuticle on their apical, luminal surface that protects the larvae from dehydration and infection. The tracheal cuticle is distinguished from the epidermal cuticle by the presence of cuticle ridges, often called taenidial folds that are thought to prevent the collapse of tracheal tubes while allowing them to expand and contract along their length.

Here, we show that DAAM is required to organize an array of parallel running actin cables beneath the apical surface of the tracheal cells that define the taenidial fold pattern of the cuticle. We found that the actin ring pattern corresponds exactly to that of the taenidial fold pattern, and we propose that the actin rings organized by DAAM define the position of taenidial fold formation. Our genetic interaction and epistasis data are consistent with a model that DAAM activity is regulated by RhoA. In addition, we show that DAAM works together with the non-receptor tyrosine kinases Src42A and Tec29 to regulate the actin cytoskeleton of the Drosophila tracheal system.

Excision alleles of DAAM were generated by remobilization of two P-element insertions, EP(1)1336 and EP(1)1542. 300-300 dysgenic males were crossed individually to y,w,l(1)/y,w,FM6 females to recover excisions, and establish stocks subsequently. Eye clones were induced by crossing y,w,FRT19A, arm-lacZ/FM6; ey-Flp males to DAAM^Ex68, w, FRT19A/y,w,FM6 females. In phenotypic rescue experiments, Act-GAL4<UAS-FL-DAAM^13.59/CyO,Act-GFP; tub-GAL4<UAS-FL-DAAM^13.59/CyO,Act-GFP; or btl-GAL4<UAS-FL-DAAM^13.59/CyO,Act-GFP males were crossed to females of different DAAM alleles. Rescue was examined in the male progeny, DAAM^Ex68 was rescued to adult viability in 25.8% of the appropriate males, while DAAM^Ex65 was rescued in 93.6% of the males.

For gain-of-function and epistasis analysis, UAS-C-DAAM/TM3,Sb,Kr-GFP;Src42A^E1/CyO,Act-GFP; UAS-C-DAAM/TM3,Ser,Act-GFP;Tec29⁵⁶¹⁰<FRT40A/CyO,Act-GFP; UAS-C-DAAM/TM3, Ser,Act-GFP;RhoA^72F/CyO,Act-GFP; UAS-C-DAAM/TM3,Ser,Act-GFP stocks were established and crossed to btl-GAL4 (on II) homozygous females. In co-expression experiments btl-GAL4<UAS-FL-DAAM^13.59/CyO,Act-GFP was crossed to UAS-C-DAAM/TM3,Sb,Kr-GFP. The appropriate genotypes were selected by the loss of GFP expression. For live imaging, embryos were collected from the progeny of a btl-GAL4 to UAS-actin::GFP/CyO cross.

In addition, we used the following mutant alleles: w¹¹¹⁸,Oregon-R, w,Df(1)AD11/FM7c (BL#1098), w,Df(1)AC7/FM7c (BL#1144), Tp(1;3)sta sta¹:ss^sta/FM3 (BL#844), and y¹,l(1)EFd⁵/Dp(1;Y)Sz280, y²/FM6,l(1)11¹¹ (BL#6021)

Breakpoints of the excisions were mapped by standard Southern blot and PCR analysis. For overexpression and rescue experiments the UAS-FL-DAAM construct has been created by inserting the full-length RE67944 cDNA into the pUASp vector, while the UAS-C-DAAM construct carries a PCR amplified C-terminal region of the cDNA in the same pUASp vector encoding amino acids 518-1153. The polyclonal anti-DAAM antibody was raised in rabbit injected with a bacterially produced and purified His-tagged C-DAAM protein. After four boosts, the crude serum was used for immunostaining.

For RNA in situ hybridization, probe was prepared by the DIG RNA Labelling Kit (Roche). Fixation and whole-mount embryo hybridization was carried out according to standard protocols. For immunofluorescence staining, the following antibodies were used: rat anti-DCAD2 1:10 (DSHB), rabbit anti-DAAM 1:4000, rat anti-Elav 1:300, mouse anti-β-gal 1:1000 (Promega), mouse anti-GFP 1:1000, mouse anti-2A12 1:5 (DSHB) and Rhodamine phalloidin at 1:100(Molecular Probes). Secondary antibodies were: anti-rabbit-Alexa546, 1:500;anti-rabbit-Cy5, 1:300; anti-mouse-Alexa488, 1:500; anti-rat-Cy5, 1:300;anti-rat-Alexa488, 1:500; anti-mouse-Alexa488, 1:500 (all from Molecular Probes); and anti-mouse-biotinylated IgM 1:300 (Vector Laboratories). Embryos and larval tracheas were mounted in glycerol/PBS 4:1. Fluorescent and DIC images were collected using Zeiss LSM 510 and Olympus FV500 confocal microscopes. Bright-field images were collected using Zeiss Axiocam on Zeiss Axioskop MOT2. Images were assembled in Adobe Photoshop 7.0, GIMP 2.2.6 and ImageJ 1.32.

A recently published phylogenetic analysis (based on sequence comparisons of 101 FH2 domains from 26 eukaryotic species) has concluded that metazoan formins can be grouped into seven major subclasses(Higgs and Peterson, 2005). This analysis suggested that the DAAM subfamily clearly represents a distinct class. This is consistent with the previous finding that the DAAM subfamily exhibits extensive sequence similarity both within and outside of their highly conserved FH1 and FH2 domains (Habas et al., 2001). At present, however, very little is known about the in vivo function of the DAAM subfamily. Sequence analysis revealed the presence of a single DAAM ortholog in the Drosophila genome corresponding to the annotated gene CG14622 (Fig. 1A). Although the integrated Drosophila database(FlyBase) predicts that this gene might code for several different transcript classes, we focused our work on the predicted CG14622-RB transcript that appears to be encoded by the RE67944 EST clone(Fig. 1A). Sequencing of this full-length cDNA clone indicated a protein consisting of 1153 amino acids. Further analysis revealed that the ORF contains several conserved domains including an FH1, FH2, GBD and a putative DAD domain characteristic of formins. Based on the overall homology level and the phylogenetic analysis of FH2 domains (Higgs and Peterson,2005), the DAAM subfamily appears to be most related to the Dia subfamily, raising the possibility that similar to the DRFs, DAAM formins are also regulated by an autoinhibitory mechanism that can be relieved upon RhoGTP binding.

To begin the functional analysis of DAAM, we generated loss-of-function mutant alleles by using two transposon insertions, EP(1)1336 and EP(1)1542, in the immediate vicinity of the DAAM locus (Fig. 1A). Remobilization of these insertions allowed us to isolate a set of deletion alleles, two of which were mapped to the 5′ region of DAAM, and five of them were mapped to the 3′ region. The large 5′ deletions affect only 5′ UTR exons and intronic sequences(Fig. 1A), while the 3′alleles carry smaller deletions (965-2538 bp), but affect the DAAMORF, leading to C-terminal truncations that (depending on the allele) remove up to 457 amino acids of the predicted protein(Fig. 1B). All 5′ and 3′ deletion alleles are lethal as hemi- or homozygotes, with the exception of DAAM^Ex1, the smallest C-terminal deletion allele which is semiviable and fertile (viability is 17% when compared with wild type). The lethal 5′ and 3′ deletion alleles do not complement each other and the deficiencies uncovering this region[Df(1)AD11, Df(1)AC7 and Df(1)sta]. They are, however,complemented by Dp(1;3)sta and Dp(1;Y)Sz280, two transpositional duplications that carry the 1F2-3 cytological region,including DAAM. All lethal alleles are viable over Daam^Ex1, the weakest and homozygous viable allele, with the exception of the two largest C-terminal deficiency alleles(Daam^Ex68 and Daam^Ex36). To demonstrate that the lethality associated with these alleles is solely due to the lack of DAAM, transgenic flies carrying UAS-FL-DAAM(containing the full length RE67944 cDNA clone) were generated and tested in rescue experiments. When UAS-FL-DAAM was expressed under the control of Act-Gal4 or tub-Gal4 in a DAAM mutant background lethality was rescued to adult viability, even in case of the largest deletion allele Daam^Ex68. Taken together, these results demonstrate that all lethal alleles represent DAAM alleles and these mutations do not affect other essential genetic elements.

Based on biochemical assays and tissue culture experiments it was suggested that a human DAAM family member, DAAM1, is required for non-canonical Wnt/Fz signaling, and might function as a bridging factor between Dsh and RhoA. Xenopus embryo experiments supported this view, as the Xenopus ortholog xDaam1 has been implicated in PCP signaling during the early steps of gastrulation (Habas et al., 2001). To test whether the Drosophila DAAM is involved in PCP signaling, we carried out a loss-of-function analysis of the gene. We have thus induced DAAM mutant clones in the wing and compound eye, two PCP organization model tissues. Although several alleles were tested, none of them exhibited PCP defects in the eye or wing (not shown). The same was observed in adult escapers of the DAAM^Ex1/DAAM^Ex249 allelic combination.

We have also tested whether the PCP gain-of-function phenotypes induced by overexpression of PCP proteins can be modified by DAAM mutant alleles. To achieve this, we have compared the effects of Fz and Dsh overexpression in the wing or eye in wild-type and DAAM^Ex68 heterozygous mutant backgrounds. No significant difference between the wild-type and the DAAM mutant backgrounds was observed (not shown). Thus, the loss-of-function analysis of DAAMargues that the Drosophila DAAM ortholog is either not required for PCP signaling or at best plays a redundant role during PCP establishment.

As the DAAM alleles are lethal, it encodes a gene with essential function(s). To identify these, we examined DAAM mutant larvae, and noticed that these animals display severe trachea defects. These include the collapse and flattening of the tracheal tubes, in both the main airways and the side branches (Fig. 2B,C). Wild-type tracheal cells secrete a cuticle on their apical (luminal) surface,which is continuous with the epidermal cuticle. Nevertheless, it is easily distinguished by the presence of cuticle ridges called taenidial folds that project into the lumen (Fig. 2A,B). These ridges form annular rings or run a helical course around the lumen of the tubes. Tracheal tubes of DAAM mutant larvae fail to secrete such a highly ordered cuticle. Instead, although some local order is visible, we observed short and curvy apical ridges that rarely run perpendicular to the tube axis (Fig. 2C,D). These striking cuticle pattern defects are detected throughout the tracheal system independent of tube types, with the exception of the fusion cells that secrete a different type of cuticle characterized by a dotty pattern instead of taenidial folds (see below, Fig. 5C). To confirm that the trachea phenotype is caused by DAAM loss of function, we expressed a UAS-FL-DAAM construct under the control of a trachea specific btl-Gal4 driver. This significantly rescued the trachea disruptions(Fig. 2E), just as a general driver (e.g. Act-Gal4, not shown).

Next, we examined the expression of DAAM and confirmed that the mRNA is expressed in the developing embryonic tracheal system, starting from stage 13 of embryogenesis onwards (Fig. 3A). Similarly, immunostaining experiments (using a polyclonal anti-DAAM serum raised against the C-terminal half of the protein; see Materials and methods for details) revealed the presence of the DAAM protein in the developing embryonic tracheal cells in stage 13 and older embryos(Fig. 3B). In DAAM^Ex68 embryos, the antibody failed to detect protein in the trachea (Fig. 3D). Similarly, detectable protein was absent from mutant clonal tissue in eye-antennal imaginal discs, where DAAM is expressed during later stages of development (Fig. 3E-H). This confirmed that the antibody is specific to DAAM, and that DAAM^Ex68 is likely to be a protein null. Consistently, the trachea phenotype of DAAM^Ex68 homo- or hemizygous larvae is very similar to that of DAAM^Ex68/Df(1)AD11 larvae(Fig. 2E).

As it is well established that formins regulate the actin cytoskeleton, we investigated actin organization in wild-type and DAAM mutant tracheal cells. This revealed that in wild-type tracheal cells apically localized actin is organized into parallel bundles running perpendicular to the axis of the tubes (Fig. 4A,C,E), strongly resembling the organization of teanidial folds in the cuticle. Significantly,the number of actin rings and taenidial folds in a given tracheal region is equal to each other (Fig. 4A-B). To determine precisely the time when such actin rings or spirals become apparent, we have expressed an Actin-GFP (green fluorescent protein) fusion protein in the trachea of living embryos (using the btl-Gal4 tracheal driver line). This demonstrated that actin rings are first visible in late stage 15 embryos, just before the onset of cuticle secretion (Fig. 4N).

The embryonic tracheal cells are rather small, and thus it was difficult to study the formation of these actin rings at the subcellular level. Nevertheless, it is clear that these actin rings are not only present when cuticle secretion begins, but are also detected through the third larval instar stage. In contrast to the remarkable actin organization found in wild-type tracheal cells, in DAAM^Ex68 mutants this organization is completely abolished. Whereas the absence of DAAMdoes not appear to reduce the apical F-actin level of the tracheal cells,actin cables formed are much shorter and thinner than in wild type(Fig. 4L,M). Additionally,actin bundles often appear to be crosslinked to each other instead of running parallel (Fig. 4L). Taken together, in DAAM^Ex68 the actin bundles in the tracheal cells display abnormal morphology and fail to be organized into parallel running actin rings under the apical surface of the cells.

Although the global actin organization is severely impaired, local order can often be detected in small patches within a cell. The cuticle pattern of DAAM^Ex68 mutants displays a similar phenotype, indicating that actin cables direct the run of taenidial folds, even in the DAAMmutant situation. Thus, it appears that the major function of DAAM in Drosophila tracheal cells is to organize an array of actin rings that directs cuticle patterning by specifying the site of taenidial fold formation. Despite the fact that the apical actin level is not significantly reduced in DAAM mutants, we cannot exclude the possibility that in the wild-type situation, DAAM also contributes to actin assembly, a well characterized formin function. Consistent with this possibility, DAAM largely colocalizes with the apically enriched actin bundles in the embryonic trachea cells (Fig. 4E-J). Nevertheless, the potential involvement in actin polymerization appears be a redundant function. The recently characterized form3 is the best candidate to be functionally redundant with DAAM for actin assembly,because it is the only other Drosophila formin known to be expressed in the tracheal system (Tanaka et al.,2004).

Truncated formins consisting only of the C-terminal FH1 and FH2 domains behave as constitutively active forms in many contexts(Evangelista et al., 1997; Habas et al., 2001; Tominaga et al., 2000; Watanabe et al., 1999). To test whether DAAM activity is regulated in a similar manner, UAS-C-DAAM (carrying a truncated DAAM encoding the C-terminal 637 amino acids including the FH1 and FH2 domains; Fig. 1B) transgenic flies were generated. Strikingly, C-DAAM expressed in the trachea (with the btl-Gal4 driver) caused distinct phenotypic effects in the fusion cells and the regular tracheal cells.

After the cuticle is laid down in the tracheal system, fusion cells are easily distinguished from others because these narrow, doughnut-shaped cells have a characteristic granulated cuticle pattern that is very distinct from the parallel running apical ridges (Fig. 5C). The same difference can be found at the level of actin organization, as apical actin is also concentrated into a granular pattern(Fig. 5D). Significantly,anti-DAAM staining in wild type revealed that fusion cells, unlike the normal tracheal cells, do not express DAAM (Fig. 4F,G). When C-DAAM is expressed in fusion cells it induces a dramatic change in the actin organization and cuticle pattern, causing strong actin accumulation at the apical surface. However, most of the actin is found in largely unorganized bundles (Fig. 5F), although the typical granular pattern is barely visible. In addition, such fusion cells are often wider than their wild-type counterparts,and their granular cuticle pattern is transformed towards a stripy pattern,partly resembling the taenidial folds of normal tracheal cells(Fig. 5E). However, when full-length DAAM was expressed in the fusion cells (or elsewhere in the tracheal system) we could not detect any phenotypic effects (not shown).

Several conclusions can be drawn from these data. First, C-DAAM behaves as an activated formin, consistent with its domain structure and showing that the activity of DAAM is regulated by N-terminal sequences like the DRF subfamily. Second, although actin cables polymerized by C-DAAM failed to organize into such perfect rings as found in wild-type tracheal cells, C-DAAM plays an instructive role in cuticle secretion by polymerizing actin cables that are sufficient to change the cuticle pattern in a DAAM non-expressing cell towards the pattern characteristic of DAAM-expressing cells. Third, the mere presence of DAAM is not sufficient to change the cuticle pattern of fusion cells, only the activated form can do that, suggesting that fusion cells lack the activator(s) of this formin.

Overexpression of C-DAAM in normal tracheal cells (where the endogenous wild-type DAAM is expressed) resulted in severe impairment of taenidial fold formation and actin organization (Fig. 5A,B) in a qualitatively similar manner to the absence of DAAM. Recent results indicated that formins normally act in dimeric or multimeric forms (Li and Higgs,2003; Otomo et al.,2005a; Otomo et al.,2005b; Rose et al.,2005; Takeya and Sumimoto,2003; Xu et al.,2004; Zigmond et al.,2003). It has also been demonstrated that homotypic protein interaction of the mammalian formin, Fhos, is mediated by the FH2 domain, and an FH1-FH2 (C-DAAM equivalent) Fhos fragment is able to bind the full-length protein (Takeya and Sumimoto,2003). Therefore, it was possible that the btl-Gal4/UAS-C-DAAM phenotype is, at least partly, caused by an antimorphic effect of C-DAAM by interfering with the wild-type protein through dimerization. In that case, it is expected that increasing wild-type DAAM levels will suppress the btl-Gal4/UAS-C-DAAM phenotype.

To test this hypothesis, we examined btl-Gal4 driven co-expression of FL-DAAM and C-DAAM. This caused a strong suppression of the btl-Gal4/UAS-C-DAAM phenotype in regular tracheal cells(Fig. 7E), but not in fusion cells (not shown). These data indicate that C-DAAM indeed behaves as an antimorph in regular tracheal cells. In fusion cells, however, the lack of suppression by expressing FL-DAAM is consistent with DAAM having no endogenous function in these cells. Because the UAS-Gal4 system usually induces a robust expression, C-DAAM is likely to be in excess when compared with endogenous DAAM in btl-Gal4/UAS-C-DAAM tracheal cells. Thus, the tracheal phenotype in the regular tracheal cells is possibly the sum of two effects: an antimorphic effect through interference with the endogenous protein and a neomorphic effect through an excess of activated C-DAAM, which increases actin polymerization rate but fails to organize actin bundles properly. An alternative hypothesis to explain that C-DAAM overexpression leads to similar phenotypic effects to the loss of DAAM, could be that the amount of F-actin produced by C-DAAM exceeds the actin organizing capacity of endogenous DAAM. As we could not detect higher levels of apical actin in these tracheal cells, we favor the first hypothesis.

The precise mechanism by which C-DAAM impairs actin organization and cuticle patterning in tracheal cells is unclear. It is remarkable that although activated DAAM is lacking most of the formin regulatory domains, it was specific and only disrupted cuticle patterning without affecting other biological processes in the trachea (Fig. 5G). By comparison, expression of a constitutively activated form of dia (Dia^CA), the most closely related Drosophila formin to DAAM, destroys the normal branching pattern of the tracheal tubes (Fig. 5H). In addition, dorsal trunk fusion is often blocked and tracheal cuticle is not secreted (Fig. 5H). Thus, it appears that Dia^CA has an earlier effect on trachea development than C-DAAM. Together, these results suggest that DAAM carries a highly specific trachea function that is restricted to the assembly and organization of apical actin cables, which direct cuticle patterning, with most of the functional specificity coming from the C-terminal FH1-FH2 domains.

To gain insights into the regulation of DAAM and to identify potential partners that work together with DAAM to organize apical actin in tracheal cells, we carried out a genetic interaction test with a panel of cytoskeleton regulators known to be expressed in the Drosophila respiratory system. We used the weak hypomorphic DAAM^Ex1 allele that exhibits a mild phenotype (Fig. 6A), and thus appears to be the most appropriate for dominant genetic interaction assays. This approach identified RhoA, and two Src family non-receptor tyrosine kinases, Src42A and Tec29,as strong enhancers of the DAAM^Ex1 phenotype(Fig. 6B-D). To test whether these interactors have tracheal cuticle patterning phenotypes on their own, we examined the trachea of their homozygous mutant embryos and larvae. In agreement with the observation that btl-Gal4 driven expression of dominant-negative RhoA (RhoA^N19) blocks lumen formation (Lee and Kolodziej,2002), we found that absence of RhoA arrests tracheal development before dorsal trunk fusion takes place (not shown). Therefore, it was not possible to assess directly the effect of RhoA single mutants on apical actin organization and/or cuticle structure. However, although some of the very few hatching Src42A and Tec29 mutant larvae displayed a tracheal necrosis phenotype, others displayed a superficially normal-looking tracheal tree, where cuticle secretion has been completed. These larvae die as first instar, and, strikingly, exhibit an abnormal taenidial fold pattern similar to that of DAAM mutant larvae(Fig. 6E-H). Although the Tec29⁵⁶¹⁰ and Src42A^E1 tracheal phenotypes are weaker than that of the DAAM null allele, consistent with our previous conclusion that apical actin organization determines the taenidial pattern, these mutations also affect apical actin bundles in the tracheal system (Fig. 6I,J). Thus, these results indicate that Src42A and Tec29 are involved in taenidial fold patterning and during this process probably act together with DAAM. Significantly, the FH1 domain is known to bind SH3 proteins (Kamei et al.,1998; Tominaga et al.,2000; Vallen et al.,2000), and the Src family kinases have been implicated in several forming-related processes (Gasman et al.,2003; Tominaga et al.,2000). Additionally, we found that the RhoA, Src42A and Tec29 mutations not only enhance the tracheal cuticle defects caused by DAAM, but also decreased the viability of the semilethal DAAM^Ex1 allele (not shown), suggesting that these proteins may work together in other tissues as well. The observation that RhoAis an enhancer of DAAM^Ex1 further supports the model that DAAM family formins are regulated by Rho GTPases, similar to the regulation of the DRF subfamily. These data are also consistent with the previous report demonstrating that RhoA is a direct binding partner of human DAAM1(Habas et al., 2001). Surprisingly, however, DAAM^Ex1 does not show a genetic interaction with dsh, the Drosophila homologue of Dvl,another protein found to bind human DAAM1(Habas et al., 2001). Although this result does not exclude the possibility that DAAM activation in the tracheal cells normally requires Dsh binding, a detailed analysis of the trachea function of wg signaling, including dsh, failed to reveal a tracheal cuticle defect(Llimargas, 2000), suggesting that Dsh is not involved in DAAM regulation in the trachea.

To further investigate the regulatory relationship between DAAMand its genetic interactors, we examined if the btl-Gal4/UAS-C-DAAMphenotype was sensitive to the gene dose of RhoA, Src42A and Tec29. This epistasis analysis indicated that while RhoA did not modify the effect of C-DAAM expression, Src42A and Tec29dominantly suppressed the respective cuticle defects(Fig. 7A-D). These data are consistent with the model that RhoA is an upstream regulator of DAAM, whereas the non-receptor tyrosine kinases act downstream of or in parallel to it. Hence, these results further support the view that the DAAM family formins are regulated by Rho GTPases. With respect to non-receptor tyrosine kinases, our data support similar conclusions to those of previous reports, demonstrating that mouse Dia1 and mouse Dia2 act upstream of Src in the regulation of actin dynamics (Tominaga et al., 2000), and that human DIA2C is required for Src activation during the process of endosome regulation(Gasman et al., 2003).

The basic structure of the insect tracheal system is a highly conserved tubular network in every species. The most important function of this network is to allow oxygen flow to target cells. Thus, tracheal tubes need to be both rigid enough, to ensure continuous air transport, and flexible enough along the axis of the tubes, to prevent the break down of the tube system when body parts or segments move relative to each other. These requirements are mainly ensured by the tracheal cuticle, which covers the luminal surface of the tubes and displays cuticle ridges (making the overall structure similar to the corrugated hose of a vacuum cleaner). Analysis of DAAM mutants provides the first direct evidence that this hypothesis is correct. Our data demonstrate that in the absence of DAAM the taenidial fold pattern is severely disrupted, often leading to the collapse of the tubes and to discontinuities in the tubular network. In addition, our analysis revealed that the remarkably ordered cuticle pattern, displayed in the wild-type trachea tubes, depends on DAAM-mediated apical actin organization. Apical actin is organized into parallel-running actin cables, much the same way teanidial folds run in the cuticle. Significantly, the formation of these actin bundles precedes the onset of cuticle secretion, and the number and phasing of the actin rings correspond exactly to that of the taenidial folds in the cuticle. Thus our studies revealed a novel aspect of apical actin organization in the tracheal cells that has not been appreciated before.

The DAAM gene encodes a novel member of the formin family of proteins, involved in actin nucleation and polymerization. Consistent with this, DAAM is colocalized with apical actin in the tracheal cells, and the activated form of DAAM is able to induce actin assembly when expressed in tracheal cells (this work) and in other cell types (T.M. and J.M.,unpublished). In DAAM mutant tracheal cells, apical actin is still detected, albeit at a somewhat lower level than in wild type, but the bundles formed in the mutant are much shorter and thinner than in wild type, and often appear to be crosslinked to each other. Most strikingly, global actin organization is almost completely lost, although some local order can still be detected. Remarkably, the cuticle pattern in mutant tracheal cells still follows the underlying aberrant actin pattern. Overall, in DAAMmutants, both the tracheal cuticle and the apical actin pattern resemble a mosaic of locally ordered patches that failed to be coordinated and aligned with each other and the axis of the tracheal tubes. We thus propose that the apical actin bundles play a key role in patterning the tracheal cuticle by defining the place of taenidial fold formation. Regarding the function of DAAM, our results suggest that the major role of this formin in the tracheal cells is to organize the actin filaments into parallel running actin rings or spirals instead of simply executing the well characterized formin function related to actin assembly. However, whether this is a direct effect on actin organization, and thus represents a novel formin function, needs to be further elucidated. An alternative model could be that DAAM is primarily required for actin polymerization but tightly coupled to an actin `organizing' protein. In such scenario, the polymerization activity should be a redundant requirement,whereas the link to the organizing protein would be a DAAM-specific function,thereby explaining the presence of unorganized actin bundles in DAAMmutant tracheal cells.

In the case of the main tracheal airways, which are multicellular along their periphery, it is striking that in wild type the run of the actin bundles is perfectly coordinated across cell boundaries. In addition, the run is always perpendicular to that of the tube axis. How does DAAM ensure the coordination of these two aspects of actin organization? Because the DAAM protein and the apical actin cables are both found at the level of the adherens junctions, it is possible that DAAM regulates the coordination of the actin cables at the cell boundaries through a direct interaction with junctional protein complexes. However, other explanations are also possible,and further experiments will be required to elucidate the molecular mechanism of this regulatory function. The fact that actin cables normally run perpendicular to the tube axis seems to suggest that tracheal cells are able to sense a global orientation cue and align their actin bundles accordingly. The nature and source of this cue is unknown, as is the mechanism by which DAAM is involved in the read-out of this signal. Nevertheless, it is interesting that in DAAM and btl-Gal4/UAS-C-DAAM mutant trachea, the main pattern of the cuticle phenotype is often changing from one segment to the other, suggesting that the effect of the `global' orientation cue is limited to metameric units.

Sequence comparisons of FH2 proteins suggest a close phylogenetic relationship between the DRF, FRL and DAAM subfamilies(Higgs and Peterson, 2005). Members of these three subfamilies have a high level of conservation in the FH2 domain, and importantly, also in the region of the GBD and DAD domains,suggesting that the FRL and DAAM family formins are also regulated by autoinhibition and RhoGTPases, like the DRFs. We present further evidence in support of this view. First, DAAM and RhoA display a strong genetic interaction. Second, C-DAAM (an N-terminally truncated form of DAAM)behaves like an activated form much the same way DRF family formins behave. Third, epistasis experiments with C-DAAM and RhoA suggest that RhoA acts upstream of DAAM. Thus, our data support the model in which DAAM, at least in the Drosophila tracheal system, is regulated by autoinhibition that can be relieved by RhoGTPases.

This conclusion, however, contradicts the observation that human DAAM1 is required for Wnt/Fz/Dvl dependent RhoA activation in cultured cells and that xDaam1 appears to mediate Wnt-11 dependent RhoA activation in Xenopusembryos (Habas et al., 2001). These results suggested that DAAM functions upstream of RhoA in non-canonical Wnt/Fz-PCP signaling. An explanation for these distinct conclusions might be related to the fact that DAAM, in contrast to xDaam1, does not appear to be required for Fz/Dsh-PCP signaling. Hence, it is possible that the Drosophila ortholog is regulated in the same way as the DRF formins,while the vertebrate family members can be regulated in a different way, once bound by Dsh/Dvl and recruited into PCP signaling complexes.

Genetic interactions with the hypomorphic DAAM^Ex1allele identified two non-receptor tyrosine kinases, Src42A and Tec29, as strong interacting partners. Although both of these kinases play multiple roles during embryogenesis, we demonstrate that single mutants for both affect the tracheal cuticle pattern in a similar way to DAAM. These results suggest that DAAM and the Src family kinases work together to regulate the actin cytoskeleton and cuticle pattern in tracheal cells. Although it is not known whether DAAM physically binds Src42A and/or Tec29, it has been established that the FH1 region of DRFs and other formins can bind SH3 domains, including those of the Src family kinases(Gasman et al., 2003; Tominaga et al., 2000; Uetz et al., 1996). In agreement with our data that DAAM acts upstream of Src42A and Tec29 in tracheal cells, cytoskeleton remodeling and SRF activation mediated by mouse Dia1 and mouse Dia2 requires Src activity(Tominaga et al., 2000). Moreover, a recent report suggests that RhoD and human DIA2C regulate endosome dynamics through Src activation, proposing that Src activity is stimulated via human DIA2C dependent recruitment to early endosomes(Gasman et al., 2003). Similarly, the Limb deformity protein (a formin) interacts with Src on the perinuclear membranes of primary mouse fibroblasts(Uetz et al., 1996). Based on these examples, we speculate that in Drosophila tracheal cells the RhoA/DAAM/Src module may not only be required to organize apical actin bundles, but additionally it might represent a link to secretory vesicles and to the regulation of exocytosis. Future studies will be required to test this hypothesis, and to unravel the mechanisms whereby DAAM family formins and Src family kinases contribute to cytoskeletal remodeling in the Drosophila tracheal system and in other tissues.

We thank M. Krasnow, P. Rorth, M. Affolter, the Developmental Studies Hybridoma Bank, the Bloomington Stock Center and the Szeged Stock Center for fly stocks and reagents. We are grateful to H. Gyurkovics, T. Török and J. Reynolds-Kenneally for critical reading and helpful comments on the manuscript. We are particularly grateful to Anna Rehák, Edit Gyányi and Edina Ördög for technical assistance. We thank L. Homolya and B. Sarkadi at the National Medical Center (Budapest) for their patience and generous help with microscopy facilities. This research was supported by NIH FIRCA grant TW05893 to M.M. and J.M., and by a NIH NIGMS RO1 grant GM62917 to M.M. J.M. is an EMBO/HHMI Scientist and a Bolyai János Research Scholar.