Fascin links Btl/FGFR signalling to the actin cytoskeleton during Drosophila tracheal morphogenesis

The development of the Drosophila tracheal system is a well-established in vivo model for tubulogenesis and cell migration (Affolter et al., 2003; Hogan and Kolodziej, 2002; Lubarsky and Krasnow, 2003). The Breathless (Btl)/FGFR pathway is a key player in tracheal formation and regulates both branch migration and cell diversification (Ghabrial et al., 2003). Whereas Btl targets for cell specification have been identified, much less is known about how Btl signalling is transmitted to direct migration. Btl receptor is activated at the tips of the primary branches by its ligand Branchless (Bnl)/FGF (Klämbt et al., 1992; Ohshiro et al., 2002; Sutherland et al., 1996). Previous reports indicated that Btl activation stimulates filopodia formation, which may promote motility (Ribeiro et al., 2002). However, how Btl triggers filopodia formation and how these filopodia contribute to branch migration remain unclear.

Filopodia are dynamic actin-based projections with sensing capacity, playing important roles in adhesion and migration. Filopodia formation requires efficient bundling of actin filaments (Faix et al., 2009; Khurana and George, 2011). Fascins belong to an evolutionary conserved family of proteins that cross-link actin filaments into tightly packed bundles, providing stiffness to different types of actin-based structures, such as microspikes, filopodia, stress fibres or microvilli (Hashimoto et al., 2011; Khurana and George, 2011; Vignjevic et al., 2006). Importantly, fascins have recently received a lot of attention in the biomedical field for their involvement in cancer and metastasis and they are at present tumoral biomarkers emerging as potential therapeutic targets (Hashimoto et al., 2011). A wider understanding of fascin function and regulation, required to design such therapeutic strategies, may come from the analysis of fascin activity in normal conditions in in vivo systems. In Drosophila there is one single fascin protein, named Singed (Sn). sn is required for different actin-based events during fly development, such as blood cell migration, oogenesis or bristle formation (Cant et al., 1998; Tilney et al., 1998; Zanet et al., 2009).

In this study we find that sn is transcriptionally upregulated at the tip cells of the tracheal branches in response to the Btl activation, through the effectors Pointed (Pnt) and Anterior open (Aop). We find that this regulation is functionally relevant, as sn mediates, at least in part, Btl activity. Our functional analysis indicates that sn is required at the tips of the branches, precisely where the Btl pathway is more active, to properly organise the actin cytoskeleton. At the cellular level we find clear defects in filopodia morphology and in the shape of the tip cells in sn mutants, and we propose that these defects underlie the defects we observe at the tissue level, including delayed branch migration and terminal cell extension. We find that tracheal formation requires the actin-bundling activity of non-phosphorylated Sn, and that an unphosphorylatable (active) form of Sn is able to promote tip cell features. Finally, we report that during tracheal formation sn acts in concert with forked [f; another actin cross-linker with homology to Espin (Bartles et al., 1996)]. A role for f in the organisation of long-lived actin structures such as bristles was previously documented (Tilney et al., 1995); here we unveil for the first time a role for f in dynamic filopodia. Taken together, our results provide new insights into how the Btl pathway regulates the actin network and into how this finely tuned bundled actin network regulates tracheal morphogenesis.

sn was known to be expressed in different embryonic tissues (Zanet et al., 2009), and Berkeley Drosophila Genome Project reported expression in the tracheal system, suggesting a tracheal requirement.

Our analysis of Sn protein accumulation indicated a spatially regulated pattern in the trachea. Sn accumulated highly at the tips of the primary branches from stage 12 (Fig. 1B). At stages 13 and 14 Sn accumulated strongly in the tip cells of the dorsal trunk (DT), dorsal branches (DB), lateral trunk (LT) and most cells of the visceral branch (VB) (Fig. 1C,D,F). By stage 15 high Sn levels were restricted to the one or two cells nearest the tip in each branch, and this pattern was maintained until late stage 16 (Fig. 1E,G). Besides this enhanced accumulation, Sn also accumulated at lower or basal levels in all tracheal cells throughout development. At the subcellular level we observed a cytoplasmic accumulation of Sn, including the specialised actin-rich filopodia (Fig. 1F,G).

A PlacZ reporter of sn expression (sn^P1) showed comparable patterns of expression and protein accumulation (supplementary material Fig. S1), indicating that sn is transcriptionally regulated during tracheal development.

The Btl pathway is activated by its ligand Bnl/FGF at the tips of tracheal branches (Klämbt et al., 1992; Ohshiro et al., 2002; Sutherland et al., 1996). The correlation between Sn pattern and Btl activation suggested that sn is a target of Btl. Consistent with this we observed that loss of bnl prevented high accumulation of Sn in tracheal tip cells, whereas the pattern elsewhere (e.g. in plasmatocytes) was unaffected (Fig. 2A). Conversely, bnl tracheal misexpression led to high levels of Sn in all tracheal cells (Fig. 2B).

Btl signals through the Ras-MAP kinase pathway and controls transcription through the transcription factor Pnt (Cabernard and Affolter, 2005; Samakovlis et al., 1996a). However, it was also shown that Btl can signal in the trachea in a Pnt-independent manner (Ribeiro et al., 2002). pnt is expressed in the trachea in a pattern comparable to that of sn (Fig. 2C,D), suggesting that sn is a target of pnt. The pattern of Sn in the absence or generalised expression of pnt (Fig. 2E,F) confirmed this hypothesis.

In contrast to Pnt, the transcription factor Aop is inactivated by the Btl pathway and negatively regulates Btl activity by competing with Pnt for ETS binding sites (Ohshiro et al., 2002). Accordingly, we found that aop negatively controls sn expression (Fig. 2G,H).

Collectively our results show that sn expression is regulated by the Btl pathway through the nuclear effectors Pnt and Aop. The use of the reporter allele showed that this regulation is transcriptional (supplementary material Fig. S1).

Our results showed that the Btl pathway regulates sn expression. To confirm the functional relevance of this regulation we carried out genetic interactions. bnl is a haploinsufficient locus with two-thirds of the heterozygotes showing branching defects (Sutherland et al., 1996). We took advantage of this sensitised background and asked whether we could modify the phenotype by modulating sn activity. In sn mutants, as in the wild type, 20 ganglionic branches (GBs) entered the ventral nerve cord (VNC), while in bnl heterozygotes only 6.9 GBs out of 20 GBs entered (n=51 embryos). These GB defects of bnl heterozygotes were significantly increased when sn activity was downregulated, and only 3.6 GBs entered the VNC (Fig. 2I-M) (n=51 embryos). This result indicates that bnl and sn interact, suggesting that sn modulates bnl-mediated cell migration.

We characterised sn tracheal requirements in DBs. Because sn is required during oogenesis (Cant et al., 1994), we could not analyse the complete lack of Sn (maternal and zygotic). Alternatively, we analysed maternal and zygotic hypomorphic mutants (sn^P1, viable and fertile), zygotic amorphic mutants (sn²⁸, viable and female sterile), and homozygous hypomorphic females crossed to amorphic males (sn²⁸/sn^P1). Loss of sn activity in these combinations was confirmed by absence of Sn (supplementary material Fig. S2 and not shown). sn downregulation produced defects at the tips of the branches (see following subchapters), where sn levels are high and Btl activity is maximal.

sn mutants displayed defects in branch pathfinding as many DBs were misguided and often contacted the adjacent ones (Fig. 3A-C; supplementary material Movie 2). In addition, we found cases of branch outgrowth delays. To investigate a requirement for sn in branch extension we performed time-lapse analyses to measure the approaching of contralateral wild-type DBs in dorsal views until they meet at the dorsal midline (Fig. 3D; supplementary material Movie 1). We plotted the distance separating each pair of contralateral DBs independently with respect to time and extracted the speed of approaching as a measure of branch extension. We found that the average speed ranged from 0.66 to 0.81 μm/minute between DB4 and DB7 (Fig. 3F shows DB4 results). A comparable analysis of sn mutants indicated that the absence of sn slows down the speed of migration (from 0.54 to 0.61 μm/minute between DB4 and DB7) (Fig. 3E,F; supplementary material Movie 2), indicating defects in DB extension.

Terminal branches, which arise from terminal cells expressing Dsrf (Bs - FlyBase) (Guillemin et al., 1996), form at the tips and require Btl activity (Samakovlis et al., 1996a). In sn mutants we detected three different types of terminal branching defects: extra terminal branches, defects and delays in terminal branch extension, and misguidance of terminal branch lumina (Fig. 4A-J).

The extra terminal branches arose from extra Dsrf-expressing cells (Fig. 4B), indicating that sn is directly or indirectly involved in tracheal identity specification. To better characterise the two other terminal branching defects, we resorted to in vivo analyses. In wild type, DB terminal cells typically extend a cytoplasmic projection directed ventrally during the approaching of contralateral branches. This growing projection accommodates and directs the formation of a new intracellular lumen (Gervais and Casanova, 2010; Kato et al., 2004; Oshima et al., 2006) (Fig. 4F,H; supplementary material Fig. S3A, Movie 3). In sn mutants, the ventral cytoplasmic projection was often delayed and extended only during or after contact of contralateral branches. In addition, the lumen was often misguided, generating bifurcated or turned terminal branches (Fig. 4G,I; supplementary material Fig. S3B, Movie 4). These results indicated that sn participates in the cellular events required to extend and direct terminal branches.

Branch fusions that connect the tracheal network are mediated by the specialised fusion cells (Gervais et al., 2012; Lee et al., 2003; Lee and Kolodziej, 2002; Samakovlis et al., 1996b; Tanaka-Matakatsu et al., 1996), which also accumulate high levels of Sn (Fig. 1G). In sn mutants, both the penetrance and expressivity of DB fusion defects were increased when compared with wild type (Fig. 4N). These defects were not due to an incorrect specification of fusion cells, as the fusion marker Dysfusion (Dys) (Jiang and Crews, 2003) was normally expressed (Fig. 4K,L). In wild-type embryos fusion cells first emit filopodia towards each other (Gervais et al., 2012; Samakovlis et al., 1996b; Tanaka-Matakatsu et al., 1996) and then extend a polarised cytoplasmic process that initiates a broader cell contact (supplementary material Fig. S3C, Movie 3). Later, a new structure organised by E-cadherin and the actin cytoskeleton forms at the contact point and assists the formation of the penetrating lumen (Lee et al., 2003; Lee and Kolodziej, 2002). We detected different defects in sn mutants that result in a lack of DB fusion: branch misguidance (as already described); formation of an irregular cytoplasmic process, which seemed inefficient to promote and/or maintain a full contact of fusion cells (supplementary material Fig. S3D, Movie 4); formation of a faint track of actin (Fig. 4M). In summary, sn downregulation leads to branch fusion defects derived from different abnormalities, which can be attributed to defects in guided migration and cell contact.

Two sets of experiments confirmed an autonomous requirement of sn in the trachea: (1) adding back sn in the tracheal system rescued branch fusion, terminal lumen guidance or branch extension (Fig. 3F, Fig. 4E,J,N, Fig. 6C,J); (2) the tracheal expression of sn RNAi (which decreased Sn in the trachea, supplementary material Fig. S2) produced tracheal defects comparable to those observed in sn mutants (Fig. 4E,J,N) and increased the defects in DB fusion in a sn mutant background (Fig. 4N).

We aimed to understand the cellular basis of the aforementioned tissue defects of sn mutants. Fascins organise actin-based structures, such as filopodial extensions (Hashimoto et al., 2011). Because filopodia form at the tip of tracheal branches in response to Btl (Ribeiro et al., 2002), precisely where Sn levels are higher, we investigated whether sn regulates this filopodial activity. High-resolution time-lapse analysis (10 seconds/frame) in the wild type indicated that these filopodia are very dynamic (Fig. 5A; supplementary material Movie 5). Morphologically, these filopodia are straight, perpendicular to the membrane and display a stiff appearance. Interestingly, we also noticed that the dynamics of these filopodia correlated with the advance of lamellipodia-like cell fronts, so that dense filopodial activity anticipates the formation of organised cytoplasmic processes. For instance, the angle of a prominent pool of filopodia directed dorsoanteriorly correlates with the direction of migration of the branch (see Materials and Methods), suggesting a role in directed migration. By contrast, a dorsally directed pool correlates with branch fusion. Finally, a ventrally directed pool anticipates the formation of the cell extension that will generate the terminal branch (supplementary material Movies 3, 5).

In sn mutants we observed a similar pattern to that of the wild type in terms of length and direction of filopodia extensions, but a decrease in the number (Fig. 7F,G). We also noticed a fully penetrant phenotype in sn mutants: filopodia were curved or wavy (the presence of curved filopodia was five times higher than in controls; see Materials and Methods), often parallel to the membrane and with a flaccid appearance. These consistent filopodia defects were accompanied by the irregular advancement of cytoplasmic processes or cells fronts that resulted in abnormal tip cell shapes (Fig. 5B; supplementary material Movie 6). Our results show that Sn controls the morphology of Btl-regulated actin-based filopodia at tracheal tips and suggest that these filopodia may contribute to sustaining fully organised lamellipodia-like cytoplasmic processes.

To analyse the organisation of the actin cytoskeleton, we visualised F-actin in fast time-lapse movies by expressing Utrophin-Green fluorescent protein (UtrGFP) in the tracheal cells. UtrGFP revealed a well-organised cortical actin network formed by long and persistent actin ribs or microspikes probably formed by actin bundles. These actin microspikes arise inside the cytoplasmic processes of the tip cells (which we therefore define as a lamellae) and protrude beyond the cell front forming the filopodia. In addition, a dense, thick and persistent actin mesh organises at the front of the tip cell, which can be defined as lamellipodia (Fig. 5D; supplementary material Movie 7). By contrast, in sn mutants, the actin network was disorganised: (1) instead of long actin ribs from the lamellae we found actin accumulated in punctae or short filaments that hardly extend through the lamellipodium or filopodial bases; (2) the lamellipodial actin meshwork was weakly and transiently organised (Fig. 5E; supplementary material Movie 8). Similar results were obtained when Moesin-GFP (MoeGFP) was used to visualise the actin network (supplementary material Fig. S4). Thus, our results show an abnormal organisation of the actin network in sn mutants consistent with a defective bundling of actin filaments that would lead to inefficient migratory lamellipodia as well as to flaccid and bent filopodia.

Sn bundling activity is controlled post-transcriptionally by phosphorylation on Serine 52 (in flies)/39 (in vertebrates) (Tilney et al., 2000; Vignjevic et al., 2006; Yamakita et al., 1996; Zanet et al., 2009). Accordingly, a phosphomimetic form of Sn (Sn^S52E) cannot provide the sn-bundling activity required for bristle formation in Drosophila, whereas the non-phosphorylatable form (Sn^S52A) can perform such activity, indicating that S52 phosphorylation inactivates Sn. However, in other tissues (plasmatocytes and nurse cells), S52 phosphorylation does not inactivate Sn (Zanet et al., 2009). We asked whether phosphorylation on S52 is important for tracheal formation.

We expressed sn^S52E and sn^S52A in wild-type tracheal cells. Overexpression of sn^S52A produced defects in branch fusion, and, remarkably, ameliorated unspecific defects in DB pathfinding and terminal branching observed in control embryos (Fig. 6A,B and not shown). By contrast, the phenotypes of sn^S52E were in line with loss of sn function and produced defects in terminal branching and fusion (Fig. 6A,B). This result indicates that Sn^S52E acts as a dominant negative. It has been shown that fascin active and inactive conformations exist in equilibrium, regulated by interactions with actin (Yang et al., 2013). The overexpression of Sn^S52E must somehow displace this equilibrium towards the inactive form.

To more clearly assess the activity of each phosphovariant, we tested their capacity to rescue sn endogenous downregulation. Consistent with dominant-negative activity, Sn^S52E exacerbated several sn tracheal defects, such as branch fusion failure (Fig. 6A) or the speed of branch extension (Fig. 6C). At the cellular level we observed stronger defects than in sn mutants, such as very abnormal cytoplasmic extensions at tip cells (Fig. 6K). Conversely, sn^S52A rescued most aspects of sn phenotype at the tissue level, including branch extension, fusion and terminal branching defects (Fig. 6A-C), as wild-type sn does (Fig. 4J,N, Fig. 6C,J,). At the cellular level, sn^S52A also rescued the formation of straight filopodia, further confirming the tracheal autonomous requirement of sn (Fig. 5C). This correlated with an increased accumulation of actin detected along the length of filopodia, as observed in time-lapse imaging (supplementary material Fig. S4C). In addition, tip cells formed prominent and thick cytoplasmic extensions that often arose slightly earlier than in wild-type embryos (Fig. 6G,J). Strikingly, tip cells appeared very often clearly clustered in an advanced position connected to the rest of the stalk cells by a narrow bottleneck, and these clusters sometimes contained more cells (arrows in Fig. 6G; supplementary material Movie 9). These results unveil a dominant effect of Sn^S52A, and indicate that it can promote migratory and protrusive tip cell features.

We analysed the subcellular accumulation of Sn^S52E and Sn^S52A proteins. Although both proteins accumulated in the cytoplasm and in projections, we detected an enrichment of Sn^S52A in filopodia protrusions, whereas Sn^S52E accumulated more diffusely in the cytoplasm (Fig. 6D,E; supplementary material Movies 10, 11). These results are consistent with a correlation between the activity, accumulation and phosphorylation state of Sn (Nagel et al., 2012; Zanet et al., 2012).

Most actin-based structures are assembled by more than one actin cross-linker (Khurana and George, 2011). In Drosophila, bristle formation requires both sn and f (Tilney et al., 2003; Tilney et al., 1998; Tilney et al., 1995). We investigated whether F cooperates with Sn in the trachea.

We detected additive effects in branch fusion in sn³f³⁶ double mutants (Fig. 7E). We also detected synergistic interactions, as a proportion of double mutants showed defects not observed in single mutants alone (Fig. 7A-D). In particular, we detected lack of DBs (in 71% of embryos, n=46, Fig. 7D) and lack of terminal branches (in 100% of embryos, n=14, Fig. 7C). Stainings for Dsrf showed lack of terminal cells in all embryos analysed (from two to six Dsrf cells missing per embryo, n=14; Fig. 7C). The absence of Dsrf-expressing cells partially explains the lack of terminal branches, but we also found examples in which the terminal cell was correctly specified and yet the terminal branch was shorter or defective (arrows in Fig. 7C).

Synergistic interactions between two genes usually reflect the fact that they are functionally related. In agreement with this hypothesis, we find that both f and sn mutants alone display bent and wavy filopodia. In addition, filopodia quantifications in double mutants showed significant differences with control embryos and with each mutant alone. In particular, whereas in f mutants filopodia were shorter and in sn mutants were decreased, in double mutants they were very strongly decreased and short (Fig. 7F,G). Time-lapse imaging confirmed that double mutants extended only few and short filopodia, which displayed a curved and flaccid aspect. In addition, the cytoplasmic processes and consequently the shape of tip cells were more affected than in single mutants (Fig. 7H; supplementary material Movie 12). Thus, our results reveal a previously undescribed role of F in embryogenesis, acting in concert with Sn to regulate the actin cytoskeleton in the tracheal tissue. Accordingly, we found that f is expressed in a general pattern, including tracheal cells (supplementary material Fig. S5).

To gain insight into the molecular mechanisms of sn and f interaction we assayed the ability of sn to rescue sn³f³⁶ defects. We found a rescue of filopodia morphology and of tip cell shape (Fig. 7I; supplementary material Movie 13), indicating that Sn can overcome the requirement of F when it is overexpressed.

The FGFR pathway is an essential regulator of many cellular activities (Brooks et al., 2012). In Drosophila, the Btl/FGFR pathway stimulates filopodia formation at the branch tips through unknown effectors (Ribeiro et al., 2002). How these filopodia contribute to migration was hampered by the fact that, in btl mutants, filopodia were not formed and, in addition, tip cells were not properly specified. In this study we identify sn as a target of Btl activity. In sn mutants, filopodia form but they are defective, allowing us to assess their relevance in branching and migration. Furthermore, our results unveil a dual regulation of filopodia by the Btl pathway, as it is both required for filopodia formation in a transcriptional-independent manner (Ribeiro et al., 2002) and for filopodia organisation in a transcriptional-dependent (Pnt-mediated) manner (this study).

Sn is specifically upregulated in branch tip cells in response to Btl activation. Similarly, Fascins are expressed in epithelial cells that require specific programmes for motility or epithelial to mesenchymal progression (Machesky and Li, 2010). The restricted expression of fascins is physiologically relevant because their misregulation can cause severe disorders. For instance, fascin-1 misexpression is observed in several carcinomas and correlates with a poor prognosis (Hashimoto et al., 2011; Machesky and Li, 2010). Fascin-1 stabilises actin bundles in invadopodia (Li et al., 2010), which are long filopodia-like protrusions in cancer cells thought to stimulate invasion and cancer metastasis (Sibony-Benyamini and Gil-Henn, 2012; Yamaguchi, 2012). But in spite of the importance of fascins both in physiological and pathological conditions, only a few regulatory transcriptional mechanisms have been described so far (Hashimoto et al., 2011). Therefore, identifying fascin regulators, as we do in this study, is relevant for the biomedical field. We suggest that the Btl pathway, acting through its canonical pathway, might be a general regulator of sn in migrating cells. Importantly, dysregulation of FGF signalling has also been observed in tumour formation and progression (Brooks et al., 2012), suggesting that dysregulation of fascin could mediate in part this malignant FGFR activity.

We identified two types of defects in sn mutants at the cellular level: filopodia defects and cell shape defects. Filopodia defects were penetrant and consisted of curved and apparently flaccid filopodia that often orient parallel to the membrane, in contrast to the straight, stiff and perpendicular filopodia of control embryos. We propose that these defects are due to the abnormal actin organisation we describe: whereas in the wild type long and conspicuous actin ribs are detected and actin clearly accumulates at the filopodia base, in sn mutants only punctae or short bundles are detected in the lamellae and no clear actin accumulation is observed at the base of filopodia. Thus, our results are consistent with the described activity of sn, cross-linking actin filaments into bundles that provide stiffness to filopodia (Hashimoto et al., 2011; Vignjevic et al., 2006; Vignjevic and Montagnac, 2008). Besides filopodia, sn is required in other types of actin-based structures in flies, such as actin bundles in nurse cells (Cant et al., 1994), dendrite terminal branchlets in sensory neurons (Nagel et al., 2012), actin bundles in the bristles (Cant et al., 1994; Overton, 1967; Tilney et al., 1995) or lamellae microspikes in plasmatocytes (Zanet et al., 2009). In these structures sn promotes the formation of straight and stiff structures, whereas its absence leads to the formation of curved ones.

Besides filopodia defects, the morphology of tracheal tip cells was clearly irregular in sn mutants and the cytoplasmic processes of terminal and fusion cells were often delayed or abnormal. These cell shape defects may derive from the defective filopodia, which would be inefficient to deform the membrane or stabilise this deformation. Additionally, we propose that they also reflect a role for the sn-dependent microspikes and cortical actin meshwork at the lamellipodia that we describe, which would be able to generate membrane protrusions that shape the tip cells. The sn-dependent actin accumulation at the lamellipodia and its role in tracheal cell shape is consistent with evidence in other cell types showing that fascin forms densely packed, parallel actin bundles that provide rigidity to push out membrane tethers upon protrusion (Schäfer et al., 2011), as in vertebrate and Drosophila lamellipodia (Adams, 1997; Nemethova et al., 2008; Zanet et al., 2009) or in invadopodia (Li et al., 2010; Schoumacher et al., 2010).

Are these defects at the cellular level underlying the tracheal tissue defects in sn mutants? We propose that they are and that the defects in tracheal formation stem from the defective filopodia and cell morphologies (Fig. 7J). In the case of branch extension, we propose that the abnormal filopodia and lamellipodia of sn mutants are not sufficient to support the bnl-mediated migration, producing branch growth delays. This is in agreement with reported evidences indicating that cell migration requires filopodia to sense the environment and to adhere to the substrate to transmit the force to advance (Adams, 2004; Schäfer et al., 2011) and that lamellipodia are a crucial protrusive component of migrating cells. In the case of branch fusion and terminal branch extension defects, we propose that they are due to the defective advance of lamellipodia, which impair a timely and efficient branching and fusion, or to defective exploratory-recognition abilities of filopodia in branch fusion. Finally, in the case of terminal lumen guidance, it was shown that it requires an actin network that directs lumen pathfinding (Oshima et al., 2006). Our results provide new insights into this mechanism by showing that this actin network needs to be properly bundled.

Strikingly, we also find a role for sn in regulating tracheal cell specification, as we observed presence of extra terminal cells expressing Dsrf in sn mutant conditions. Previous reports may shed light into the possible mechanism of this phenotype, as it was described that actin dynamics can regulate Dsrf activity, which can regulate its own expression (Miralles et al., 2003; Somogyi and Rørth, 2004; Sotiropoulos et al., 1999). Alternatively, the misregulation of Dsrf expression may be indirectly caused by an abnormal branch extension producing altered reception of cell specification signals or altered tip cell interactions.

Our current knowledge of fascins unveils an enormous complexity. Fascins act as actin-bundling proteins (Khurana and George, 2011) and this activity is regulated by phosphorylation on S52 in flies/S39 in vertebrates (Tilney et al., 2000; Vignjevic et al., 2006; Yamakita et al., 1996; Zanet et al., 2009). However, in in vivo systems such as Drosophila, phosphorylation on S52 inactivates sn activity in certain conditions but not in others (Zanet et al., 2009). Furthermore, although it is widely accepted that fascins act as actin-bundling proteins, a recent report proposed a bundling-independent activity of fascins that promote filopodia formation (Zanet et al., 2012).

Our work in an in vivo system sheds light into sn and actin-based mechanisms. Our results are consistent with a requirement of the actin-bundling activity of sn, regulated by phosphorylation on S52, for proper tracheal formation. This conclusion is based on the observation that the phosphomimetic form Sn^S52E cannot rescue sn tracheal defects, whereas the non-phosphorylatable form Sn^S52A, as well as wild-type Sn, rescue these defects. This correlates with the subcellular accumulation of these phosphovariants: Sn^S52A is clearly enriched in filopodia protrusions whereas Sn^S52E accumulation is more cytoplasmic. Our results are in line with results in bristles (Zanet et al., 2009) and sensory neurons (Nagel et al., 2012) but in contrast with observations in plasmatocytes and nurse cells (Zanet et al., 2009). Remarkably, we find that Sn regulation in plasmatocytes and tracheal cells is different not only at the post-transcriptional level, but also at the transcriptional one; although Btl regulates sn expression in the tracheal tissue, it does not regulate it in plasmatocytes. Altogether this indicates that sn can employ different molecular mechanisms that are cell-type and context specific.

The parallels for sn during tracheal and bristle formation extend beyond the requirement of sn-bundling activity, as in both cell types f interacts with sn. In bristles, F and Sn bind and act sequentially to cross-link actin filaments [reviewed by Tilney and DeRosier (Tilney and DeRosier, 2005)]. F appears at early stages of bristle formation and organises previously formed (by an unidentified cross-linker) actin bundlets into loosely ordered modules. Sn appears later, displacing F and tightly organising the bundles into hexagonally packed structures, thereby increasing rigidity. Hence, in the absence of f the number of actin filaments per bundle are decreased but they are regularly packed, in the absence of sn more actin filaments per bundle are present but these are not hexagonally packed, and in the absence of both almost no actin filaments arrange into bundles (Tilney et al., 2003; Tilney et al., 1995). We find that a similar mechanism may be operating in tracheal filopodia: whereas absence of each cross-linker affects filopodia morphology, length (f mutants) or number (sn mutants), the absence of both proteins strongly affects all features. Thus, our results show, for the first time, a role for F in transient actin structures, as it is the case in tracheal filopodia. Furthermore, our experiments provide new insights into the mechanisms of actin bundling. While we show that F and Sn provide distinct functions to filopodia organisation indicating that in normal conditions both cross-linkers are required, our rescue experiments also indicate that Sn can overcome F requirements and restore filopodia organisation in overexpression conditions, indicating that their activities are partially redundant.

Intriguingly, in fsn mutants many terminal cells are not differentiated. Terminal cell specification requires Btl signal (Guillemin et al., 1996; Samakovlis et al., 1996a). The fact that in fsn double mutants there are few and short filopodia and that the cell fronts are incorrectly organised, leads us to speculate that Bnl signal might not be properly received by tracheal tip cells. Indeed, besides the recognised role of filopodia in environment sensing acting as mechanosensitive organelles, filopodia are also proposed to play roles in responding to external guiding cues, here acting as chemosensitive organelles (Khurana and George, 2011).

y¹w¹¹⁸ (wild type), sn^P1, sn²⁸, f^36a, sn³f^36a, bnl⁰⁰⁸⁵⁷ (bnl^PI), pnt^Δ88 and aop² (aop^11S) are described in FlyBase. ETS-12 (lacZ line) revealed pnt expression. Mutant chromosome were balanced over marked balancers. For overexpression in the trachea we used btlGal4. The transgenes used are described in FlyBase: UAS-GFPsn, UAS-GFPsn^S52E, UAS-GFPsn^S52A (kindly provided by S. Plaza), UAS-bnl, btl-MoeGFP, UAS-pnt, UAS-aop^act.2, UAS-UtrGFP (kindly provided by T. Lecuit) and UAS-snRNAi (from VDRC and Bloomington Stock Center). In panels labelled ‘tracheal cells’ the embryos were stained for Trachealess (Trh, nuclear staining) or for GFP in embryos carrying btlGal4 UAS-srcGFP (cell membrane staining).

We followed standard protocols for immunostainings and in situ hybridisations. The following primary antibodies were used: mouse anti-Sn [Developmental Studies Hybridoma Bank (DSHB)], mouse mAb2A12 (to visualise the tracheal lumen; DSHB), mouse anti-mActin (MP Biomedicals), rat anti-Trh (J. Casanova’s lab), rabbit or goat anti-GFP (Molecular Probes and Roche), chicken anti-β-gal (Abcam), rabbit anti-Dys and mouse anti-Dsrf.

An f riboprobe was generated using the following primers: forward, 5′-AGATACAGATCGCCCAGCAG-3′; reverse, 5′-TGTGATTTGCCAGACATGGT-3′.

Confocal images of fixed embryos were obtained with a Leica TCS-SPE system. Images were imported into Fiji and Photoshop, and assembled using Illustrator.

The Fiji Freehand Line tool was used to measure the length of the terminal cell filopodia of fixed stage 16 embryos carrying btlGal4-UASsrcGFP in the indicated backgrounds.

Total number of embryos, DBs or terminal cells (n) are provided in text and figures. Error bars indicate standard error (s.e.). P-values were obtained with two-sample t-test using R Commander v1.7-0 software (http://cran.r-project.org/bin/macosx/) or with Fisher’s exact test (http://www.langsrud.com/fisher.htm#INTRO) for binomial distributions. ^*P<0.05, ^**P<0.01, ^***P<0.001.

Live imaging was performed on an inverted Leica TCS-SP5 or a Zeiss Lsm780 confocal. Dechorionated embryos were mounted on Menzel-Gläser coverslips with glue and covered with Oil 10-S Voltalef (VWR).

To measure DB approaching, GFP fluorescence was captured every 90 seconds over 26-58 μm z-stacks of stage 13-14 embryos mounted dorsally and recorded in a Leica TCS-SP5 during ∼3 hours. Each DB front was manually followed with the Fiji Manual Tracking, and the results were treated in Excel.

To follow tip cell filopodia and cell shape changes we visualised stage 14-15 btlGal4-UASsrcGFP embryos in the indicated genetic backgrounds. Embryos were imaged in a Leica TCS-SP5 at 10-second intervals with 16×0.67-0.86 μm z-stacks.

To analyse actin dynamics we used btlGal4-UASUtrGFP or btl-MoeGFP in different genetic backgrounds and acquired images of 0.82-0.85 μm z-stacks every 14 seconds in a Zeiss Lsm780 using a water 63× objective and 1.8 zoom.

To measure the correlation of the angle of dorsoanterior filopodia with respect to branch migration, we measured all dorsoanterior filopodial angles of tip cells of stage 14 control embryos (btlGal4-UASsrcGFP) every minute using the Straight Line Selection tool of Fiji software. We compared the average of these angles with the tip cell advance. Tip cell advance was estimated with the Manual Tracking tool of Fiji software by following a path obtained by analysing every minute the position of the terminal cell front (measured at three different positions in the front).

To compare the straightness of filopodia between stage 14 control (btlGal4-UASsrcGFP) and sn mutants (sn^P1; btlGal4-UASsrcGFP) we measured the number of new bent filopodia (with an angle higher than 135°) formed and normalized it with the number of filopodia at each time point.

Tiff original supplementary material movies are shown as Avi and compressed as PNG.

Measurements were imported and treated into Excel software, and R Commander software was used for statistics.

We thank N. Luque, E. Fuentes, E. Rebollo and L. Bardia for technical assistance. We are grateful to S. Plaza and J. Zanet for reagents and the Developmental Studies Hybridoma Bank, the Bloomington Stock Center and Vienna Drosophila RNAi Centre for fly lines and antibodies. We thank members of the Llimargas and Casanova labs for helpful discussions and J. Casanova, M. Furriols, A. Letizia, S. Araújo, S. Ricardo and G. Lebréton for critically reading the manuscript.