The Drosophila Extradenticle and Homothorax selector proteins control branchless/FGF expression in mesodermal bridge-cells

Abstract

The stereotyped outgrowth of tubular branches of the Drosophila tracheal system is orchestrated by the local and highly dynamic expression profile of branchless (bnl), which encodes a secreted fibroblast growth factor (FGF)-like molecule. Despite the importance of the spatial and temporal bnl regulation, little is known about the upstream mechanisms that establish its complex expression pattern. Here, we show that the Extradenticle and Homothorax selector proteins control bnl transcription in a single cell per segment, the mesodermal bridge-cell. In addition, we observed that a key determinant of bridge-cell specification, the transcription factor Hunchback, is also required for bnl expression. Therefore, we propose that one of the functions of the bridge-cell is to synthesize and secrete the chemoattractant Bnl. These findings provide a hitherto unknown and interesting link between combinatorial inputs of transcription factors, cell-specific ligand expression and organ morphogenesis.

Introduction

Many epithelial organs consist of highly branched tubular networks that transport gases or fluids. Both in the Drosophila tracheal system and in the mammalian lung, members of the fibroblast growth factor (FGF) family of secreted signalling molecules have a key role in the branching process, and expression of the corresponding genes is, at the same time, highly complex, yet rather stereotyped (Metzger & Krasnow, 1999; Affolter et al, 2003). In the tracheal system, the Branchless (Bnl) ligand (Sutherland et al, 1996) binds to and activates the receptor tyrosine kinase Breathless (Btl)/FGFR (FGF receptor) in nearby tracheal cells (Klambt et al, 1992), which in turn triggers dynamic changes in the cytoskeleton (Ribeiro et al, 2002) and results in migratory behaviour of responding cells towards the ligand source. bnl expression is arrayed around the tracheal placode in a segmentally repeated fashion, and each expression domain is located to a specific anteroposterior (AP) and dorsoventral (DV) position within the segment (Sutherland et al, 1996). This suggests that the global AP and DV patterning systems of the Drosophila embryo specify these positions. Indeed, the bnl spot necessary for the outgrowth of the dorsal tracheal branch has been shown to be regulated by the BMP-like signalling molecule Decapentaplegic (Dpp), which is expressed in the dorsal-most epidermal cells (Vincent et al, 1997). However, no transcriptional regulators of bnl expression along the AP axis have been described so far.

We show that the Extradenticle (Exd), Homothorax (Hth) and Hunchback (Hb) transcription factors are required at stage 12 for the lateral expression of bnl in one specific cell per segment, and we identify the corresponding cell as the mesodermal bridge-cell. Bridge-cells have been proposed to serve as Bnl-independent guidance cues for elongation of the main structure of the tracheal tree, the dorsal trunk (DT), but their molecular mechanism of action has since remained elusive (Wolf & Schuh, 2000; Wolf et al, 2002). Here, we find that an important function of the bridge-cell is to synthesize and secrete the chemo-attractant Bnl. We also observed that Hox genes of the bithorax complex do not contribute to bnl activation in trunk segments of the developing embryo, which suggests a Hox-independent function of Exd and Hth in bridge-cells.

Results And Discussion

The exd and hth tracheal defects are similar

We have discovered that in embryos homozygous for a null allele of the homothorax (hth) gene, the tracheal branches elongating along the AP axis failed to form correctly, as observed for the dorsal and lateral trunks, or the visceral branches (Fig 1C,D). hth encodes a homeodomain-containing transcription factor of the TALE (three-amino-acid loop extension) family that is necessary for the nuclear translocation of another TALE homeoprotein encoded by the extradenticle (exd) locus (Peifer & Wieschaus, 1990; Rauskolb et al, 1993; Rieckhof et al, 1997). In addition, Hth cooperates with Exd to directly control target-gene regulation (Ryoo et al, 1999). To test whether exd was also required for tracheal development, we generated embryos lacking both maternal and zygotic exd contributions. The tracheal phenotype of exd-null embryos was similar to that of hth mutants, with selective defects in all tracheal branches elongating along the AP axis (Fig 1E,F). To assess whether the hth tracheal phenotype could indirectly result from impaired nuclear localization of Exd, we ubiquitously expressed a nuclear form of Exd in hth mutant embryos. In such conditions, we did not observe a rescue of the hth tracheal phenotype (Fig 1G), which is consistent with the proposed role of Exd and Hth in transcription regulation as a heterodimeric complex. The Exd/Hth complex could be required in tracheal cells or, alternatively, in cells surrounding the developing trachea. To distinguish between these two possibilities, we expressed the Hth protein in hth mutant backgrounds, either ubiquitously in the embryo (Fig 1H) or specifically in the tracheal system with the Btl-Gal4 driver (not shown). Significant rescue of hth DT defects was only observed when Hth was expressed in the entire embryo (Fig 5O), indicating that the two selector genes have an essential role in non-tracheal cells.

Figure 1.

hth and exd mutant embryos show similar tracheal defects. (A,C,E,G,H) Whole-mount staining of stage-14 embryos with the 2A12 antibody. (B,D,F) Whole-mount antibody staining with anti-GFP (green fluorescent protein) of stage-14 embryos bearing the recombined btl-Gal4 and UAS–actin–GFP constructs on the second chromosome. (A,B) Wild-type (WT) embryos. (C,D) hth^P2 mutant embryos. (E,F) Maternal and zygotic exd^Y012 mutant embryos. (G,H) Embryos triplestained (G) or double-stained (H) with anti-Exd (Extradenticle) and anti-Hth (Homothorax) antibodies to recognize the genetic backgrounds (not shown). (G) hth^P2 mutant embryos bearing arm-Gal4 and UAS-nls-exd transgenes. The constitutively nuclear-localized form of Exd cannot rescue hth^P2 dorsal trunk (DT) defects. (H) hth^P2 mutant embryos bearing arm-Gal4 and UAS-hth transgenes show a rescue (rarely complete; see Fig 5O) of DT defects.

Figure 5.

Constitutive activation of the Branchless (Bnl) signalling pathway can partially rescue the hb and hth dorsal trunk (DT) phenotype. Bnl signalling was induced with the arm-Gal4 and UAS-bnl transgenes (B,E,H,K), or with the recombinant Btl-Gal4/UAS–actin–GFP (green fluorescent protein) and UAS-Btl^tor transgenes (C,F,I,L). Embryos are stained with anti-2A12 (green) and anti-Blistered/DSRF (red; A,B,D,E,G,H,J,K), or with 2A12 (green) and anti-GFP (C,F,I,L). Embryos were triple stained with anti-β-galactosidase to identify homozygous mutants by the absence of the blue balancer. (A–C) Wild-type (WT) embryos. (D–F) Homozygous bnl^P1 mutant embryos. (G–I) Homozygous hb^FB mutant embryos. (J–L) Homozygous hth^P2 mutant embryos. (M–O) Graphical representation of the number of DT metameres observed (abscissa) per embryo (as a percentage; Y-axis), either in the mutant (red curves) or in the rescue (green and blue curves) context. In each situation, the total number (n) of embryos scored with the corresponding genotype is indicated. (M) Statistical representation of the bnl^P1 DT phenotype compared with the two rescue situations. (N) Statistical representation of the hb^FB DT phenotype compared with the two rescue situations. (O) Statistical representation of the hth^P2 DT phenotype compared with the two rescue situations. A supplementary curve (in black) illustrates the efficiency of hth^P2 DT rescue obtained with the arm-Gal4 and UAS-hth transgenes.

Exd and Hth control bnl expression in bridge-cells

During the past few years, a large number of genes with essential roles in diverse aspects of tracheal development have been identified. Several mutations have been reported to affect, rather selectively, the development of the DT, and we therefore tested whether we could link the hth and exd phenotypes to any of these previously reported genes. We first tested the expression profile of knirps (kni), which leads to a lack of DT formation when ectopically expressed in DT cells (Chen et al, 1998), and spalt (sal), which contributes to the unique morphological aspect of the DT (Kuhnlein & Schuh, 1996; Chihara & Hayashi, 2000; Ribeiro et al, 2004). In hth mutant embryos, both Kni (Fig 2A,A′) and Sal (Fig 2B,B′) proteins were correctly expressed in the tracheal system, and we thus concluded that the hth DT phenotype did not result from kni or sal deregulation. A third gene that has been reported to be essential for DT development is hunchback (hb), which is expressed in the mesodermal bridge-cell (Wolf & Schuh, 2000). Although hb, hth and exd mutations lead to similar DT phenotypes, we did not observe a loss of Hb localization in bridge-cells of hth mutant embryos (Fig 2C,C′). Therefore, we tested whether lack of the DT in hth and exd mutants could result from a loss of bnl expression, as this ligand is required for the migration of all tracheal branches. At stage 11, bnl is expressed in six stereotyped ectodermal sites around the tracheal placode. In hth or exd mutant embryos, bnl transcription in these sites seemed normal (data not shown). However, at later stages, in situ hybridization experiments clearly showed the loss of bnl expression between the DT anterior and posterior outgrowths, in both hth (Fig 2E) and exd (Fig 2F) mutant embryos.

Figure 2.

The transcription factors Extradenticle (Exd), Homothorax (Hth) and Hunchback (Hb) are specifically required for the dorsolateral bnl expression sites. All embryos are in stages 12–13 and carry the recombined btl-Gal4 and UAS–actin–GFP (green fluorescent protein) transgenes on the second chromosome. The tracheal system is visualized with anti-GFP by fluorescent (in red in A,A′,B,B′,C,C′) or enzymatic (in brown in D–G) immunostaining. (A) The expression of Knirps (Kni, in green) is unchanged in dorsal and ventral parts of tracheal metameres (white arrows in the higher magnification, A′) of hth^P2 mutant embryos. (B) The expression of Spalt (Sal, in green) is unchanged in the dorsal trunk (DT; a white arrow in the higher magnification, B′) of hth^P2 mutant embryos. (C) The expression of Hb (in green) is unchanged in bridge-cells (a white arrow in the higher magnification, C′) of hth^P2 mutant embryos. (D–G) The expression of bnl is visualized in blue by in situ hybridization. The dorsolateral bnl expression sites necessary for DT elongation are shown (dotted circles). Compared with wild-type (WT) embryos (D), these bnl expression sites are lost in hb^FB (E), exd^Y012 (F) and hth^P2 (G) mutant embryos. Note that in these mutant embryos, the dorsal bnl expression sites are also diminished but still present, whereas the ventral bnl expression sites are not affected.

The lack of specific markers for cells expressing and secreting the Bnl/FGF ligand has thus far hindered the identification and further characterization of most of these cell clusters. However, as the bridge-cell and the dorsolateral Bnl spot, which is under the control of hth and exd, seemed to be in a similar position (compare Fig 3A,A′ and B,B′; see also Fig 1 in Wolf et al, 2002), we asked whether the bridge-cell was transcribing the bnl gene. Enzymatic revelation of bnl messenger RNA by in situ hybridization, and anti-Hb by immunostaining procedures, indicated that bnl was indeed transcribed by the cell expressing nuclear Hb protein (Fig 3C). This result was confirmed by high-resolution colocalization studies using confocal microscopy, which showed the nuclear protein Hb to be surrounded by the cytoplasmic bnl mRNA (Fig 3D,D′). Moreover, the observation of a specific loss of the dorsolateral bnl expression sites in hb mutants (Fig 2G), which results in apoptosis of bridge-cells (Wolf & Schuh, 2000), shows that bnl is indeed transcribed in these cells (in contrast to the findings of Wolf & Schuh, 2000). Loss of bnl expression in hth and hb mutant embryos was also inferred from the absence of filopodia formation in DT branches of living embryos (supplementary Fig S1B and C online, respectively).

Figure 3.

Expression of bnl in mesodermal bridge-cells. (A,A′,B,B′) Whole-mount antibody staining with anti-GFP (green fluorescent protein; red) of stage 12–13 embryos bearing btl-Gal4 and UAS–actin–GFP constructs. (A,A′) Hunchback (Hb; green) is expressed in the nucleus of the bridge-cell (white arrowhead). (B,B′) Fluorescent in situ hybridization against bnl RNA (green). The spot necessary for DT elongation is indicated (white arrowhead). (C) Enzymatic costaining of Hb (immunostaining in brown) and bnl (in situ hybridization in blue) shows that bnl messenger RNA is produced by the bridge-cell (black arrowhead). (D) Fluorescent triple staining of btl-Gal4/UAS–actin–GFP embryos with anti-GFP (blue), anti-Hb (red) and bnl mRNA (green). (D′) Highlight of the region boxed in (D). The nucleus of the bridge-cell (white arrowhead) is surrounded by the cytoplasmic bnl mRNA. WT, wild type.

In many instances, the Exd and Hth proteins act as cofactors for homeotic proteins (Mann & Morata, 2000). Therefore, we tested whether Hox proteins also participate in bridge-cell specification. As we observed that both bnl and Hb expressions, as well as DT elongation, were normal in Hox mutant embryos (supplementary Figs S1D and S2 online), we concluded that exd and hth carry out Hox-independent functions in the bridge-cell.

Finally, we looked at the nuclear localization of Exd and Hth in the mesodermal bridge-cell. Although hb expression starts at stage 11 in a posterior lateral margin of each tracheal metamere (Wolf & Schuh, 2000), we did not observe nuclear colocalization of Exd in this first hb-positive epidermal cell (Fig 4A,A′,A″). Nuclear localization of Exd was only observed later in the daughter bridge-cell (Fig 4B,B′,B″,B′″), concomitantly with Hth (Fig 4C,C′,C″). We also confirmed that nuclear colocalization of Exd and Hb occurred in mesodermal cells (Fig 4D,D′,D″). The nuclear translocation of the two selector proteins at stage 12 coincides with the appearance of the bridge-cell and its guidance role for the outgrowth of DT branches. This suggests that local expression of bnl in the bridge-cell could result from the simultaneous presence of the three transcription factors Hb, Exd and Hth. However, the fact that the same transcription factors are also present in the ventral daughter cell of the bridge-cell precursor (Fig 4B,B′,B″,B′″,C,C′,C″), which does not express bnl, indicates that more factors are required to determine the fate of the particular mesodermal bridge-cell.

Figure 4.

Dynamic expressions of Extradenticle (Exd) and Homothorax (Hth) in the nuclei of bridge-cells. All embryos bear the recombined btl-Gal4/UAS-nls-GFP (green fluorescent protein; A,A′,A″) or btl-Gal4/UAS–actin–GFP (B,B′,C,C′,C″) transgenes, and are stained with anti-GFP (blue) and anti-Hb (Hunchback; green). The red staining shows either the Exd (A,A′,A″,B,B′,B″,B′″) or the Hth (C,C′,C″) proteins. White arrows indicate bridge-cells. At stage 11, the Exd protein (red) is not localized in the nucleus of the bridge-cell (A,A′,A″). Nuclear translocation of Exd (red) is observed at stage 12 (B,B′,B″,B′″), concomitantly with the nuclear expression of Hth (red) in bridge-cell s (C,C′,C″). (D,D′,D″) Fluorescent immunostaining of embryos carrying the mesodermal twist-24B-Gal4 driver and the UAS-lacZ reporter gene, with anti-β-galactosidase (βGal; blue), anti-Exd (red) and anti-Hb (green). WT, wild type.

Bridge-cells have bnl-dependent functions

To verify the importance of bnl expression in bridge-cells at the functional level, we analysed whether the DT defects in hb or hth mutant embryos could be rescued by expressing Bnl ubiquitously, or by expressing a constitutively active form of the Btl receptor in the tracheal system (Dossenbach et al, 2001). As a control, DT formation was also followed in wild-type or bnl^P1 mutant embryos under the same conditions. To take into account the variability of the DT phenotypes in each mutant situation, we reported, for each genetic background, the proportion of embryos with the corresponding number of fused DT branches formed along the AP axis.

As shown earlier (Sutherland et al, 1996), activation of Bnl signalling in all tracheal cells leads to a significant rescue of the DT in bnl mutants (Fig 5D–F,M). In hb mutant backgrounds, most of the embryos (89%) have from zero to three fused DT metameres (Fig 5G), but this number can increase to four or five (for 10% and 1% of the population, respectively; Fig 5N). Activation of Bnl signalling increases the proportion of embryos with four or five fused DT metameres (around 70%; Fig 5H,I,N). As a maximum of six fused DT metameres were formed in such embryos, because of the lack of segmental anlagen (Tautz et al, 1987), we conclude that a significant rescue of the hb DT phenotype occurs in the presence of Bnl signalling. In hth mutant backgrounds, most of the embryos (90%) have from zero to three DT metameres (Fig 5J,O). Activation of Bnl signalling leads to a significant, although not complete, DT rescue; a large proportion of hth mutant embryos (65%) showed four or five fused DT metameres, and this number varied up to six or seven (for 10% of embryos), a frequency never observed in hth mutants without added Bnl signalling (Fig 5K,L,O).

The partial rescue of the hb and hth DT phenotypes can be explained by three hypotheses. First, these genes could also be required in cells other than the bridge-cell, and/or for later events during tracheal system morphogenesis, such as those implicating Ubx and abdA. Second, our rescue experiments did not rely on specific activation of Bnl signalling in bridge-cells. To bypass this limitation, we tested several EP insertions as potential Gal4 drivers (listed in the Methods), but none of them was exclusively expressed in bridge-cells (data not shown). However, as our genetic approach allowed a significant rescue of the DT phenotype of bnl mutant embryos, we think that this hypothesis is unlikely to be correct. Third, the bridge-cell could provide other functions as a substrate for migration and/or as a regulator of the fusion process. If these bnl-independent processes are also controlled by hb, hth and exd, constitutive Bnl signalling alone would not be sufficient to completely rescue the DT trunk phenotype in the absence of these transcriptional regulators.

In conclusion, we identify the selector genes exd and hth as important factors controlling the bnl/FGF expression site necessary for DT elongation. We identify this bnl expression site as the previously described mesodermal bridge-cell, which expresses a third transcription factor, Hb. These results link the cell-specific expression of three transcriptional factors to the three-dimensional architecture of an organ system, the trachea. Future studies will have to identify such factors involved in the determination of bridge-cell fate, and to determine whether the regulation of bnl by Exd, Hth and Hb is direct.

Methods

Fly stocks. The following fly strains and alleles were used in this study: hb^FB (Wolf & Schuh, 2000), bnl^P1 (Sutherland et al, 1996) and hth^P2 (Casares & Mann, 1998); the exd^Y012 allele was used to generate female germline mosaics, as described previously (Rauskolb et al, 1993). To drive Gal4 either ubiquitously in the embryo or specifically in the tracheal system, we used arm-Gal4 and btl-Gal4 (recombined on second chromosome with UAS–actin–GFP (green fluorescent protein)) driver lines, respectively. UAS–actin–GFP (Verkhusha et al, 1999) was used to detect cytoskeletal dynamics in tracheal cells. We also used UAS-bnl (Sutherland et al, 1996), UAS-btl^tor (Dossenbach et al, 2001), UAS-nls-exd (Abu-Shaar et al, 1999) and UAS-hth (Casares & Mann, 1998). For the rescue experiments shown in Fig 5, the bnl^P1, hb^FB and hth^P2 mutant chromosomes were balanced over a blue-balancer, and the transgenes were either homozygous (UAS-bnl, UAS-Btl^tor, arm-Gal4) or heterozygous (the recombined Btl-Gal4/UAS–actin–GFP) on the second chromosome. The fly strains tested as potential bridge-cellspecific Gal4 drivers were l(3)PL00790, l(3)PL00759 and EP(3)3557.

Immunostainings and whole-mount in situ hybridization. The following primary antibodies were used: monoclonal antibody 2A12 (DSHB, Iowa City, IA, USA) diluted 1:5, monoclonal anti-Crumbs (kindly provided by E. Knust) diluted 1:20, guinea-pig anti-Hunchbach (kindly provided by R. Schuh) diluted 1:400, rabbit anti-Spalt (Kuhnlein et al, 1994) diluted 1:50, mouse anti-Knirps (Chen et al, 1998) diluted 1:20, mouse anti-DSRF (Guillemin et al, 1996) diluted 1:20, rabbit anti-Exd (kindly provided by I. Duncan) diluted 1:500, and mouse or rabbit anti-GFP and anti-β-galactosidase (Promega or Cappel, respectively) diluted 1:500. Embryos were fixed and immunostained according to standard procedures, with minor modifications. Whole-mount in situ hybridizations on embryos were performed with digoxigenin-labelled antisense RNA probes of bnl (Sutherland et al, 1996), as described (Tautz & Pfeifle, 1989).

Affinity-purified secondary antibodies were coupled to alkaline phosphatase, biotin or peroxidase (Jackson Immuno Research Laboratories, West Grove, PA, USA), or were Alexa-488, Alexa-594 or Alexa-Cy5 conjugated (Molecular Probes Inc., Eugene, OR, USA). For some fluorescent staining, the signal was amplified with the aid of a Tyramide Signal Amplification kit (NEN Life Sciences, Boston, MA, USA). The fluorescent-stained embryos were then mounted in Vectashield (Vector Laboratories, Burlingame, CA, USA) for further observation under a confocal microscope (Leica TCS SP2 or LSM 510 Zeiss). Images were processed with the Leica TSC NT 1.6, Zeiss LSM5 Image Browser, Imaris 5 and Adobe Photoshop 7.0 programs.

Supplementary information is available at EMBO reports online (http://www.emboreports.org).

Supplementary Material

Supplementary Figures