Multiple requirements for the receptor serine/threonine kinase thick veins reveal novel functions of TGFβ homologs during Drosophila embryogenesis

A major goal of embryological studies is to understand the mechanisms by which the spatial organization of an animal emerges from a fertilized egg. A particularly important role in this process has been attributed to mechanisms whereby a signal generated by a cell (or a group of cells) controls the fate of neighboring cells, a phenomena referred to as induction (Spemann and Mangold, 1924; reviewed in Gurdon, 1992).

In vertebrates, the formation of most adult organs seems to depend on inductive interactions between mesenchymal and epithelial cells, i.e. between mesoderm and ectoderm/endoderm. Many of the early developmental decisions also involve induction. In amphibians, for example, the mesodermal germ layer is established via a system of intercellular signalling originating from the endoderm (Nieuwkoop, 1969). In invertebrates suitable for genetic analysis, the difficulty in manipulating embryos has been a major limitation to studies on induction. However, recent cell ablation and molecular genetic approaches have revealed that this mechanism also occurs in various developmental pathways in invertebrates. Examples include the formation of the vulva and the anterior pharynx in Caenorhabditis elegans (Priess and Thomson, 1987; Sulston and White, 1980), as well as cell fate induction in the developing compound eye (Dickson and Hafen, 1994; Greenwald and Rubin, 1992), embryonic segmentation (Ingham and Martinez-Arias, 1992) and gene regulatory processes across germ layers in Drosophila melanogaster (Bienz, 1994; Crabtree et al., 1992).

Biochemical and genetic studies have led to the identifica-tion of several secreted signalling molecules which mediate inductive interactions. In Drosophila, one such molecule is encoded by the gene decapentaplegic (dpp) (Ferguson and Anderson, 1992; Irish and Gelbart, 1987; Padgett et al., 1987). dpp encodes a member of the TGFβ superfamily of secreted signalling molecules, to which other known growth and differ-entiation factors belong (i.e. TGFβ, activins, Müllerian inhibit-ing substance (see Kingsley, 1994)). Numerous studies in invertebrate and vertebrate organisms have revealed a large number of diverse biological activities of TGFβ superfamily members, but it is largely unknown how they mediate com-munication between cells.

Recent studies in mammalian systems revealed that members of the TGFβ superfamily interact with transmem-brane proteins that have been classified according to their ligand-binding properties as type I or type II receptor serine threonine kinases (Attisano et al., 1993; Ebner et al., 1993; Franzen et al., 1993; Lin et al., 1992; Massagué, 1992; Mathews and Vale, 1991; ten Dijke et al., 1994). Whereas the type II receptors can bind their ligand independently, type I receptors appear to require the association of a type II receptor for ligand binding. Neither type I nor type II receptors appear to be able to signal alone. These proteins are therefore consid-ered to form heteromeric signalling receptor complexes composed of a type I and a type II receptor (Wrana et al., 1992). How binding of these receptors to their ligands triggers the diverse biological activities of the signalling complexes is still unknown.

To initiate a genetic dissection of signalling processes involving members of the TGFβ superfamily, we have previously isolated two members of the type I receptor family in Drosophila and demonstrated that they are encoded by the genes saxophone (sax) and thick veins (tkv) (Nellen et al., 1994). Mutations in sax and tkv interact genetically with dpp. Both genes are required maternally for dorsoventral patterning of the early Drosophila embryo (Nellen et al., 1994), suggesting that tkv might act as a receptor for dpp. Indeed, tkv protein has been shown to bind dpp protein and BMP-2 with high affinity (Penton et al., 1994).

Although null mutations in dpp result in a strong ventralization of the embryo, dpp is also expressed, and presumably required, in various restricted sites later during embryonic development. However, due to the severe consequences of aberrant dorsoventral patterning of the early dpp mutant embryo, the function of dpp in these later processes could not be analyzed. The only two exceptions are dpp’s involvement in patterning the visceral mesoderm and the imaginal discs (Bienz, 1994, Crabtree et al., 1992, Spencer et al., 1982, Posakony et al., 1990). Fortuitously however, for both of these functions of dpp there are alleles available that affect the specific dpp expression pattern.

To investigate whether dpp plays additional roles during embryogenesis, we made use of mutations that inactivate the dpp receptor tkv. In contrast to dpp, which is expressed exclusively zygotically, tkv product required for the early dorsoventral patterning is provided maternally. Here we report the dynamic and spatially complex embryonic expression pattern of tkv and focus on the analysis of the zygotic functions of tkv. We found that tkv is required at multiple distinct steps during embryonic development. The earliest zygotic requirement of tkv activity appears to corre-spond to the process of dorsal closure of the epidermis. Later, tkv activity is crucial for visceral mesoderm and endoderm patterning, a function it seems to carry out primarily via the establishment of dpp expression in the visceral mesoderm. Interestingly, tkv mutant embryos fail to develop a functional tracheal system possibly due to a partial failure in cell migration. Our results assign novel functions to dpp or a closely related member of the TGFβ superfamily of signalling molecules in the process of dorsal closure, in the patterning of the visceral mesoderm and in trachea formation.

The two embryonic lethal alleles tkv^slater(tkv^str-I and tkv^str-II) were isolated by Nüsslein-Volhard et al. (1984). tkv^str-II has a stop codon at amino acid position 144 resulting in a predicted protein that terminates immediately N-terminal of the conserved cysteine cluster in the extracellular domain of the encoded receptor serine/threonine kinase protein; this allele presumably represents a null mutation of tkv (Nellen et al., 1994). We therefore used this particular allele for our mutant analysis. However, all phenotypes reported also occur in the tkv^str-I mutant, which changes a conserved glutamate residue 528 into a lysine residue at the C terminus of the kinase domain. The deficiency stock of tkv (Df(2L)tkv^sz-2) was obtained from J. Szidonya (Szidonya and Reuter, 1988). The tracheal system was visualized using the P-element-containing chromosome from the strain 1-eve-1 described in Perrimon et al. (1991).

In general, mutant embryos were identified using CyO chromosomes carrying β-galactosidase-expressing P-element constructs (see Affolter et al., 1993). To identify homozygous tkv-deficient embryos at the blastoderm stage, we used a tkv cDNA probe and performed whole-mount hybridizations on embryo collections from the tkv defi-ciency carrying flies.

In situ hybridization to whole-mount embryos was performed as described by Tautz and Pfeifle (1989) with minor modifications (Affolter et al., 1993). The entire plasmid carrying the tkv cDNA insert was used as a probe (Nellen et al., 1994). For wg, we have used a 2.9 kb cDNA insert (Rijsewijk et al., 1987). For dpp, a cDNA clone containing an approximately 4.5 kb insert covering the coding sequence was labelled (St. Johnston et al., 1990). For lab, the entire insert of the cDNA clone c241 was used (Mlodzik et al., 1988). For pdm-1, a bluescript vector containing a 1.7 kb EcoRI fragment encom-passing the entire coding sequence was labelled (Affolter et al., 1993). For pnr and zen, a cDNA clone containing the entire coding sequence was labelled (Ramain et al., 1993; Rushlow et al., 1987).

Embryos were immunostained according to standard procedures (Ashburner, 1989), followed by the addition of a secondary antibody conjugated with biotin or with alkaline phosphatase. The distribution of the second antibody was revealed either by using the horseradish peroxidase ABC kit (Vectastain) or by staining for alkaline phos-phatase activity. In double immunostainings, alkaline phosphatase staining was performed before the horseradish peroxidase reaction. The rabbit lab antiserum was kindly provided by Tom Kaufman, the Ubx monoclonal antibody and the rat abdA antiserum by Gines Morata, the rabbit Scr antiserum by Peter LeMotte and Walter Gehring, the monoclonal Antp antibody by Dan Brower and the rabbit abdA antiserum by François Karch. The monoclonal crumbs antibody was a generous gift of Elisabeth Knust (Tepass et al., 1990). To visualize β-galactosidase expression, a monoclonal antibody (Promega) was used. For light microscopy, immunostained embryos were viewed in a Zeiss Axiophot compound microscope using dif-ferential interference contrast. For documentation, images were pho-tographed on EKTAR 100 (Kodak) film.

The distribution of tkv transcripts during embryonic development was analyzed using in situ hybridization to whole mount embryos (Tautz and Pfeifle, 1989). High, uniform levels of tkv transcripts are detected in unfertilized eggs (Fig. 1A). During cellularization after nuclear cycle 13 (stage 5: all stages according to Campos-Ortega and Hartenstein (1985)), the pattern of tkv transcript distribution is extremely dynamic. Before the maternal contribution of tkv RNA disappears completely, tkv transcripts accumulate at the dorsal side of the embryo (Fig. 1C,D). With the exception of two small gaps, transcripts build up dorsally along the entire anteroposterior axis. During the late phase of cellularization, tkv transcripts start to accumulate on the ventral side of the embryo in a pattern reminiscent of the twist gene product (Fig. 1F,G). During the process of mesoderm invagination, tkv transcripts are detectable at high levels in all mesodermal cells (Fig. 1H,I). During germ band extension, tkv is still expressed in the mesoderm (Fig. 1J) and transcript levels only decline in this germ layer during the last phase of extension (Fig. 1L and data not shown). As transcripts fade away in the mesoderm, tkv transcription is activated in cells located in the medial ectodermal region, the neurectoderm, with lower levels detected towards the midline (Fig. 1K,L). Until full extension of the germ band is reached, the expression in the neurectodermal region refines into a narrow stripe in the ventrolateral ectoderm on either side of the midline (Fig. 1M,N). Transcripts are not detectable in the anterior and the posterior midgut, neither at this stage nor at later stages (see Fig. 1J,L). In stage 11 embryos (full germ band extension), tkv is expressed in the tracheal placodes before and during the invagination of the epidermal tracheal cells (Fig. 1O-R). Shortly after, the continuous stripe of expression along the anteroposterior axis in the ventrolateral ectoderm is split in a segmentally repeated manner (Fig. 1Q,R). In stage 13 embryos, during the last phase of germ band retraction, tkv transcripts are readily detected in parts of the visceral mesoderm (Fig. 1U,V). The regions of the visceral mesoderm that express tkv in stage 15 embryos are located just anterior and posterior to the developing gastric caeca and in the posterior part of the midgut (Fig. 1W). No expression is observed in the endodermally derived midgut cells (Fig. 1J,W,X). Expression is still detectable in the ventrolateral and dorsolateral epidermis in one to two rows of cells flanking, both anteriorly and posteriorly, the segment grooves (Fig. 1S,T and data not shown). tkv expression is also detected in a complex pattern in the head region from early develop-mental stages on (Fig. 1). This has not been analyzed in detail.

In a previous study, we have shown that the maternal expression of tkv (Fig. 1A) is required for patterning the entire domain of the presumptive ectoderm normally specified by dpp (Nellen et al., 1994). In order to learn more about the functional significance of the zygotic expression pattern of tkv, we have undertaken a detailed analysis of tkv mutant embryos using available molecular markers and have tried to correlate the observed phenotypes with the dynamic expression of tkv. We have identified three stages during embryonic development at which tkv activity (and therefore presumably also the activity of a TGFβ family member) is crucial for the proper specifica-tion of distinct cell types. Below, we will describe these tkv requirements separately and then discuss the possibility that dpp acts as the signalling molecule triggering these biological activities of tkv.

The asymmetric accumulation of tkv transcripts on the dorsal side of the embryo at a time when cells are assigned to specific positions along the two body axis (i.e. during cellularization) prompted us to examine in detail the cuticular phenotypes of tkv mutant embryos. In particular, we wanted to determine whether mutant embryos displayed defects along the dorsoven-tral axis. Thus we scored cuticles of mutant first instar larvae for the presence or absence of cuticular structures that are derived from dorsal regions of the blastoderm map (Jürgens, 1987; Jürgens et al., 1986; Lohs-Schardin et al., 1979).

As reported by Nüsslein-Volhard et al. (1984), most of the cuticles of tkv mutant embryos display a prominent dorsal hole in the trunk region (Fig. 2A). In addition to this defect in dorsal closure, tkv cuticles lack parts of the dorsal hypoderm in the trunk region (Fig. 2A,B). Keilin’s organs and both the ventral and lateral (T2 and T3) or dorsal (T1) black dots, respectively, are present in the thoracic region (data not shown). Outside the trunk region, all dorsally derived structures at the posterior end, such as the Filzkörper, the spiracular hairs, as well as the anal plates are retained in tkv mutants (Fig. 2B). In the head, mouth hooks and cirri are present, and both the antennal and the maxillary sense organs are formed. The cephalopharyngeal skeleton is severely disrupted or absent, and its remains are forced out of the body cavity.

The thoracic and abdominal regions of the dorsal epidermis do not contain scorable cuticular markers that define different dorsolateral positions. In the wild-type embryo, the dorsal-most blastoderm cells in the trunk region give rise to the extraembryonic amnioserosa, which does not contribute to the final cuticular structure of the larva, but appears to be necessary for the proper morphogenetic movements during gastrulation (Lohs-Schardin et al., 1979). We have used the expression patterns of zerknüllt (zen) and Krüppel (Kr) as markers for proper amnioserosa specification and differentiation, respec-tively. We found that zen and Kr are expressed in zygotic tkv mutant embryos in patterns indistinguishable from those observed in wild-type embryos (Fig. 3A,B). In addition, the Krüppel (Kr) expressing amnioserosa cells in tkv mutant embryos contain large and flattened nuclei which are charac-teristic of the amnioserosa (Fig. 3B). This indicates that the dorsal-most cells in the trunk region of the cellular blastoderm embryo differentiate properly in the absence of zygotic tkv activity.

Our analysis of tkv mutant embryos and larvae demonstrates that most epidermal structures that derive from the dorsal region of the cellularizing embryo are present and that no expansion of ventral or lateral pattern elements occurs in zygotically mutant tkv cuticles. Furthermore, germ band extension and retraction occur normally in tkv embryos (see Figs 3, 4); mutants for genes that are known to affect the early patterning of the dorsal side show severe defects in these processes (Arora and Nüsslein-Volhard, 1992; Wharton et al., 1993). In conclusion, we find that zygotic tkv expression in dorsally located nuclei during cellularization is not required to specify epidermal cells with respect to their position along the dorsoventral axis. However, in the absense of maternal tkv gene product, zygotic tkv activity can rescue some dorsolateral pattern elements and is thus able to play a role in early dorsoventral patterning (Nellen et al., 1994).

The absence of dorsal closure is the only major defect of tkv mutants (aside from certain head defects), which might correlate with the lack of tkv expression on the dorsal side of mutant embryos during cellularization. A number of mutations have been isolated by Nüsslein-Volhard and col-leagues, which fail to close up the hypoderm along the dorsal side (anterior open, canoe, kayak, pannier, punt, schnurri, slater (tkv), yurt; see Tearle and Nüsslein-Volhard, 1987). Beside tkv (this study), only pannier (pnr) has been analyzed at the molecular level. pnr encodes a protein containing two zinc fingers with high homology to those of the GATA-1 protein, a vertebrate transcription factor required for temporal regulation of the globin and other erythroid-specific genes (Evans and Felsenfeld, 1989; Ramain et al., 1993; Winick et al., 1993). pnr transcripts first appear in the dorsal portion of the embryos just prior to cellularization. As development proceeds, pnr RNA persists at high levels in the dorsal epidermis but is excluded from the region of the amnioserosa and is not detected elsewhere in the embryo (Winick et al., 1993).

The absence of pnr activity results in a phenotype that mimics aspects of the tkv mutant phenotype (i.e. incomplete dorsal closure). We therefore investigated whether pnr expression is affected in tkv mutants and whether this could explain the lack of dorsal closure in tkv. We found that pnr transcripts are distributed in tkv mutants at normal levels and following the spatial restriction observed in wild-type embryos up to full germ band extension (Fig. 3C-E). However, after the initiation of germ band retraction, there is a significant reduction not only in the level of pnr transcripts, but also in the number of dorsal epidermal cells along the dorsoventral axis which express pnr (Fig. 3F,H; compare Fig. 3H to G).

What aspect of the tkv expression pattern is required for the maintenance of pnr transcription in the dorsal epidermis? The late reduction of pnr transcription observed in tkv mutants suggests that the lack of the dorsal expression of tkv during cel-lularization is not responsible for the decay of pnr expression during germ band retraction. Indeed, tkv expression is also observed at low levels in the dorsal epidermal cells that accu-mulate pnr transcripts at the extended germ band stage in wild-type embryos (Fig. 3J). These observations suggest that it is this late aspect of the tkv expression pattern that, upon disrup-tion in tkv mutants, leads to a reduction in pnr expression and to the failure to close up the hypodermis on the dorsal side (see Discussion).

Both the analysis of tkv cuticle preparations as well as molecular markers indicate that, due to the maternal rescue, dorsoventral patterning is largely normal in zygotic tkv mutants. We conclude that zygotic expression of tkv along the dorsal surface of the cellularizing embryo is not required to determine dorsoventral cell fates. However tkv is required zygotically for proper dorsal closure and appears to contribute to the maintenance of pnr activity in dorsal epidermal cells during germ band retraction.

After the tkv RNA levels on the dorsal side of the embryo have declined during the late phase of cellularization, tkv transcripts appear in all ventral cells that will form the mesodermal cell layer (Fig. 1). Transcripts remain detectable in the mesoderm up to the stage of full germ band extension (Fig. 1J).

To elucidate the function of tkv during mesoderm differentiation, we have used various available antibodies and/or DNA probes to analyse different aspects of mesoderm differentiation in tkv mutants. We could not find any major defects in the early phases of cell specification in the somatic mesoderm. For example, both the early and the late expression patterns of twist (see Bate et al., 1991) are normal in tkv mutants. Furthermore, the segregation of mesodermally derived pericardial precursor cells is largely normal in the absence of tkv activity, as evidenced by the expression of the homeobox gene even-skipped (data not shown; see Azpiazu and Frasch, 1993). In addition, the subdivision of the mesoderm into somatic and visceral mesoderm appears to occur quite normally in tkv mutants (see below). However, distinct alterations in gene expression patterns were found in the visceral mesoderm (and in the adjacent germ layer, the endoderm). As we demonstrate in the following paragraphs, most of the changes that we identified can be explained by the failure to establish expression of the gene dpp in the visceral mesoderm of tkv mutants.

After the process of gastrulation, the activity of dpp is required for midgut morphogenesis, presumably by controlling the spatially restricted expression of several genes in the visceral mesoderm (e.g. Scr, Ubx, wg). dpp is expressed in the developing visceral mesoderm in regions overlapping paraseg-ment 3 and 7 and loss of this expression causes the lack of gastric caeca and second midgut constriction, respectively (Immerglück et al., 1990; Panganiban et al., 1990). In addition to its morphological function, dpp has been proposed directly to mediate the transfer of positional information from the visceral mesoderm to the endoderm, resulting in the restricted expression of the homeotic gene labial (lab) in endodermal cells of the central portion of the midgut (Immerglück et al., 1990; Panganiban et al., 1990).

tkv mutant embryos completely lack dpp transcripts in both the anterior domain (around parasegment 3) and the central domain (parasegment 7) (Fig. 4B). dpp expression in the pharynx and the oesophagus is normal, but transcripts also fail to accumulate in the clypeolabrum (Fig. 4B and data not shown). Consistent with the midgut phenotypes observed in dpp^shv mutants (which lack cis-regulatory DNA elements responsible for dpp expression in the visceral mesoderm), tkv mutants fail to establish wg transcription in parasegment 8 (Fig. 4D) and lack induction of lab and repression of pdm-1 (Affolter et al., 1993) in the endodermal cells adjacent to parasegment 7 (Fig. 4F,H). With respect to morphological criteria, tkv mutant embryos fail to form the second midgut constriction and, during later stages, to build the gastric caeca (data not shown). These phenotypes have previously been described for dpp mutants (Panganiban et al., 1990; Immer-glück et al., 1990).

It has been observed (Panganiban et al., 1990; Hursh et al., 1993) that the lack of dpp expression in paraseg-ment 7 results in a reduction of the level of Ubx expression in the same parasegment of the visceral mesoderm. In contrast to dpp mutants, tkv mutants completely fail to activate Ubx (Fig. 5F). However, the homeotic genes Scr, Antp and abd-A are expressed in the visceral mesoderm indicating that patterning in this germ layer is not disrupted all together (Fig. 5B,D,H; see Tremml and Bienz (1989) for a detailed analysis of homeotic gene expression in the visceral mesoderm).

We wanted to find out whether the lack of Ubx expression in the visceral mesoderm of tkv mutants reflects the absence of parasegment 7 information in this cell layer or whether the cells normally forming parasegment 7 were respecified to a different identity. In Ultrabithorax (Ubx) mutants, which lack paraseg-ment 7, the domain of the anteriorly expressed Antennapedia (Antp) gene is extended by one parasegment, bringing Antp expression adjacent to abdominal-A (abd-A) expression in the visceral mesoderm (Tremml and Bienz, 1989). Using double immunostainings, we observed the same phenomenon: the Antp and abd-A expression domains abut each other in tkv mutants (Fig. 5J). We therefore conclude that tkv is crucial for the establishment of visceral mesoderm parasegment 7 and rep-

Upon full extension of the germ band, tkv transcripts start to accumulate in segmentally repeated clusters of cells on the lateral side of the epidermal cell layer. Slightly later, the tracheal pits invaginate in the middle of these cell clusters (Fig. 1 and data not shown). tkv expression therefore marks the tracheal placodes slightly before the tracheal cells invaginate into the underlying mesoderm. After the invagination, the complex branching pattern of the tracheal system is established, without cell division, via cell migration and cell extension (for a detailed description of the development of the trachea, see Campos-Ortega and Hartenstein, 1985; Manning and Krasnow, 1993).

To find out whether tkv is essential for the establishment of the trachea, we have visualized the developing tracheal system in wild-type or mutant embryos using the anti-crumbs mono-clonal antibody (Tepass et al., 1990). The branching pattern of a wild-type stage 14 embryo is shown in Fig. 6A (see Figure legend and Manning and Krasnow (1993) for nomencla-ture used). tkv mutants reveal striking defects in the tracheal system (Fig. 6B). Whereas the lumina of the dorsal trunk and the visceral branches were virtually intact, there is a complete absence of the dorsally directed dorsal branches and the remainder of the tracheal lumen is only established in a rudimen-tary fashion (Fig. 6B).

To investigate in more detail whether and how tracheal cell migration is affected in tkv mutants, we have introduced into this mutant background a chromosome (1-eve-1) that provides early and persistent accumulation of β-galactosidase immunoreactivity in the cytoplasm of all tracheal cells (see Materials and Methods). This allows the migration pattern of tracheal cells during development in mutant embryos to be followed and to compare it to the wild-type situation.

Consistent with the observations made with the lumen-specific crumbs antibody, the distribution of β-galactosidase protein demonstrates that tkv mutant embryos do not develop dorsal branches and lack both the ganglionic branches and the lateral trunk (Fig. 6D). The dorsal trunk and the visceral branches that develop along the anteroposterior axis (cells migrate from posterior towards anterior) are properly formed and the dorsal trunk derivatives of the ten trachomeres subse-quently fuse, as they do in wild-type embryos (compare Fig. 6D to C). In the head region, most of the tracheal branches appear largely intact (Fig. 6F). For example, the pharyngeal branch that connects to the pharyngeal muscle and originates in the first thoracic segment reaches the pharynx during stage 15. In addition and as found in wild-type embryos, the tracheal cells that are in direct contact with the pharyngeal muscle express the DSRF gene product (data not shown; see Affolter et al., 1994). Therefore neither pathfinding nor migration of tracheal branches in the head region are affected in tkv mutant embryos. Defects are thus largely confined to those branches that in the early phase of cell migration grow in either dorsal or ventral direction, i.e. dorsal branch, lateral trunk and ganglionic branch.

All aspects of signalling by TGFβ1 and activin in tissue culture cells can be accounted for by the presence and activity of their type I and type II transmembrane serine/threonine kinase receptors (Wrana et al., 1992; Attisano et al., 1993). There is no evidence so far that cells mutant for one of these receptors would be affected in any other aspect than by the lack of responsiveness to factors of the TGFβ superfamily. The zygotic defects that we observed in tkv mutant embryos are thus most likely due to the lack of responsiveness of certain cells to a TGFβ homolog.

Since dpp has been shown, both by genetic and biochemi-cal means (Nellen et al., 1994; Penton et al., 1994), to constitute a ligand for the tkv receptor, we will discuss the zygotic defects in tkv mutants in the light of the possibility that they are caused by a failure to respond to dpp activity. These additional functions of dpp might have been masked due to the severity of the early dorsoventral defects in dpp null mutants. The tkv receptor required for dpp-mediated dorsoventral patterning is provided maternally (Nellen et al., 1994), which accounts for the largely normal patterning along the dorsoventral axis in tkv mutant embryos derived from heterozygous mothers. However, we cannot exclude the possibility that tkv functions additionally as a receptor for another as yet unidentified TGFβ homolog.

The most obvious defect observed in tkv mutants is the lack of dorsal closure. To our knowledge, pnr is the only gene (besides tkv) that has been analysed molecularly and that is specifically required for the process of closing up the dorsal hypoderm of the embryo. We found that the expression of pnr, which encodes a putative transcription factor, is not maintained in tkv mutants at high levels and declines during germ band retrac-tion. Interestingly, both tkv and dpp are expressed in the dorsal epidermal cells during the last phase of germ band extension (see Fig. 3I,J; Jackson and Hoffmann, 1994; St. Johnston and Gelbart, 1987). It is therefore possible that maintenance of pnr expression requires dpp activity and the receptor serine/threonine kinase encoded by tkv. To date, no mutations have been described in the dpp locus that remove selectively the expression in the dorsal epidermal cells. It is interesting to note, however, that the early expression of pnr during cellu-larization requires dpp activity (Winick et al., 1993), which in turn is mediated by tkv (Nellen et al., 1994).

It has been proposed that secreted dpp protein regulates gut morphogenesis, in part, by regulating homeotic gene expression in the visceral mesoderm and endoderm of the developing midgut (Immerglück et al., 1990; Panganiban et al., 1990; Bienz, 1994). Interestingly, most of the midgut defects that we observe in tkv mutants also occur in dpp mutants lacking dpp expression exclusively in the visceral mesoderm: (1) tkv mutants lack wg and have reduced Ubx expression in the visceral mesoderm, (2) lab induction as well as pdm-1 repression no longer occur in the adjacent endodermal cells, and (3) tkv mutants lack the second midgut constriction and do not develop the gastric caeca. Not only are the dpp-mediated effects on midgut gene regulation and morphology disrupted in tkv mutants, but dpp expression in the visceral mesoderm is missing all together. This would place the putative receptor (tkv) upstream of its possible ligand (dpp) and would suggest that tkv activity is required for the establishment of dpp expression rather than for mediating its signalling capacities during midgut development. Consistent with this idea, it has been shown that dpp expression in the visceral mesoderm in both parasegments 3 and 7 is autoregulated; DNA fragments from the dpp locus which generate parasegment 3 and 7 expression patterns, strictly require endogenous dpp activity in order to be expressed in these two parasegments (Hursh et al., 1993). It is thus possible that tkv expression during the early development of the mesodermal cell layer is involved in dpp-mediated autoregulation in visceral mesoderm parasegments 3 and 7. Due to the failure of dpp to autoregulate its expression, tkv mutations result in a lack of dpp transcription.

In contrast to dpp mutants, tkv mutants completely lack Ubx expression in the visceral mesoderm and Antp is expressed in the cells that are devoid of Ubx. This represents a homeotic transformation of visceral mesoderm parasegment 7 into parasegment 6 caused by the absence of the tkv receptor kinase. Therefore, tkv appears to be tightly linked to the segmentation process of the visceral mesoderm and might be involved in the initial setting up of the parasegment 7 restricted expression of the Ubx gene. It has been shown that the dpp ligand mediates Ubx autoregulation through a defined dpp response element within upstream sequences of Ubx (Thüringer and Bienz, 1993; Thüringer et al., 1993). Further studies of tkv and its ligand(s) should provide insight into the establishment of parasegment 7 identity in the visceral mesoderm.

Although tkv may be involved in mediating dpp signalling in the visceral mesoderm, tkv is not expressed in the embryonic endoderm. It thus appears that dpp-mediated induction of the lab gene across germ layers must involve an additional type I receptor(s) expressed in the endodermal portion of the midgut. lab induction is significantly reduced in mutants of saxophone (sax) which encodes another type I receptor serine threonine kinase (Nellen et al., 1994). sax is expressed in the visceral mesoderm and in the endoderm and could thus be involved in either of the two cell layers in establishing lab expression in the endoderm. Alternatively, it is possible that lab induction results as an indirect consequence of dpp expression in the apposing visceral mesoderm and is mediated by a different sig-nalling molecule.

Analysis of tkv mutant embryos revealed striking alterations in the development of the tracheal system. Although some aspects of tracheal development appear to be unaffected by the lack of tkv activity (i.e. the dorsal trunk and the visceral branches are properly established), other aspects such as the formation of the ventral trunk and the ganglionic and dorsal branches are strongly perturbed. tkv is expressed in the tracheal placodes slightly before the tracheal cells invaginate into the underlying mesoderm and start to migrate. dpp is not expressed within the tracheal placodes, but its transcript appears immediately prior to germ band shortening in a segmentally repeated pattern in the lateral ectoderm aligned with the ventral margin of the tracheal pits (St. Johnston and Gelbart, 1987; Jackson and Hoffman, 1994). Double labelling experiments revealed that these lateral ectodermal domains of dpp expression do not overlap with, but are adjacent to, the tracheal cells in the placodes (Nicole Grieder and M. A., unpublished observation). This is also the case for the dpp expression pattern in the dorsal-most cells which are located just dorsal of the tracheal placodes. The spatial apposition of cells that express tkv or dpp around the tracheal placodes suggests that the dpp signalling molecule might trigger the activity of the tkv kinase in adjacent cells and contribute thereby to the migration behaviour of tracheal cells. In this respect, it is interesting to note that most of the tracheal structures established along the dorsoventral axis (i.e. dorsal branch, ganglionic branch) are absent in tkv mutant embryos, whereas the structures extending along the anteroposterior axis (i.e. dorsal trunk, visceral branch) are not affected at all. In contrast to tkv, the Drosophila FGF receptor encoded by the breathless gene is required for migration of all tracheal cells (Klämbt et al., 1992, Reichman-Fried et al., 1994). To confirm a functional requirement of dpp in tracheal development, mutants that specifically remove cis-regulatory sequences responsible for expression in the lateral and dorsal epidermal region of germ band extended embryos will have to be constructed; such mutants are not available at present (see Jackson and Hoffmann, 1994).

The labial antibody was kindly provided by T. Kaufman (HHMI, Indiana, USA). We thank P. Simpson and C. Rushlow for DNA probes, D. Brower, W. Gehring, F. Karch and E. Knust for antibodies and the Tübingen stock center, G. Reuter and J. Szidonya for fly stocks. We also thank R. Burke, B. Dickson, N. Grieder, E. Hafen and E. Ruberte for critically reading the manuscript. This work was supported by the Kantons Basel Stadt and Zürich and grants from the Swiss National Science Foundation.