Regulation of wingless transcription in the Drososphila embryo

The specification of positional identity within each developing parasegment of the Drosophila embryo is a highly complex process requiring the activity of a number of genes, identified by the segment polarity mutations (reviewed in DiNardo and Heemskerk, 1990; Ingham and Nakano, 1990; Ingham, 1991; Peifer and Bejsovec, 1992). The development of each parasegment is initiated at the blastoderm stage of embryogenesis when transcription of the engrailed (en) and wingless (wg) genes is activated at the presumptive boundaries of each parasegment. There-after, the expression of these two genes plays a crucial role in the elaboration of positional identity, the two cell populations defined by their expression domains acting as signalling centres that instruct neighbouring cells (Ingham and Martinez-Arias, 1992). The role of the wg gene product in this process has been particularly well characterised: wg protein is secreted by cells that express the gene and is taken up by neighbouring cells either side of its expression domain (van den Heuvel et al., 1989; Gonzalez et al., 1991). Posteriorly, the activity of wg is important for the maintenance of en expression since, in its absence, en transcript and protein rapidly disappear after gastrulation (DiNardo et al., 1988; Martinez-Arias et al., 1988; Bejsovec and Martinez-Arias, 1991). Anteriorly, the secreted wg protein seems to have a number of consequences at different developmental stages: during gastrulation, cells respond to wg protein by accumulating Armadillo protein at their membranes (Riggleman et al., 1990); later, wg activity seems to be required to suppress the differentiation of denticles by epidermal cells in the ventral part of each segment. (Baker, 1987; Bejsovec and Martinez-Arias, 1991). Since the response of cells to wg protein differs according to their location along the anteroposterior axis, it has been suggested that different cells have differing competences to respond to wg (Ingham et al., 1991). Thus only cells in the anterior half of each parasegment appear predisposed to express en in response to wg, a property that is revealed when wg is ectopically expressed through mutation of the regulatory gene patched (ptc) (DiNardo et al., 1988; Martinez-Arias et al., 1988; Ingham et al., 1991).

There are a number of striking similarities between the activity of wg and that of its vertebrate orthologue Wnt-1. Mutation of the latter results in the loss of engrailed expression in the mesencephalon of the mouse (McMahon et al., 1992), suggesting a functional relationship between the two genes paralleling that in Drosophila. Moreover, the results of grafting experiments in the chick provide circumstantial evidence that different regions of the vertebrate brain have differing competences to respond to the Wnt-1 signal (Martinez et al., 1991; Bally-Cuif et al., 1992). In addition, RNA injection experiments in Xenopus have shown that murine Wnt-1 and wg can have identical biological effects, their ectopic expression in the early embryo resulting in the duplication of the main body axis (McMahon and Moon, 1989; Chakrabarti et al., 1992), apparently through the induction of a second Spemann’s organiser (see also Smith and Harland, 1991; Sokol et al., 1991). The penetrance of this effect varies with the location of the injected mRNAs, implying that their protein products act over only a limited distance, consistent with the known properties of both proteins in tissue culture assays (Jue et al., 1992; N. Ramarkrishna and A. Brown, personal communication).

Thus, in both invertebrates and vertebrates, the Wg pro-tein and its mammalian counterpart appear to act as shortrange signalling molecules, the effects of which depend upon two parameters: the differing competences of responding cells and the restriction of the source of the signal (see Ingham, 1991). In this study, we analyse a number of the factors that control the spatial regulation of wg transcription in the Drosophila embryo.

prd^2.45, ptc^IN, eve^1.27, wg^IIS and hh^IJ are all null or near amorphic alleles of their respective loci; Df(3R)Scb deletes the ftz locus (Jürgens et al., 1984). All of these strains were obtained from C. Nüsslein-Volhard in Tübingen, Germany. ptc^S2 is an amorphic allele and ptc^G12 is a deletion of the ptc coding region (Nakano et al., 1989). The recombinant chromosomes: prd^2.45ptc^S2, ptc^S2eve^1.27, wg^IISptc^G12 and ftz^9H34hh^IJ were constructed using conventional genetic techniques. The triple mutant combination ptc^G12; ftz^9H34hh^IJ was generated by crossing parents of the genotype: ptc^G12/+; ftz^9H34hh^IJ/TM3. e recognised by their failure to express ptc and their reduced number of parasegments, due to the ftz mutation; since the ftz^9H34hh^IJ chromosome was balanced, all ftz embryos were necessarily also homozygous for hh^IJ.

Embryos were collected at the appropriate times, and fixed and processed essentially as previously described (Tautz and Pfeifle, 1989) except that probes were single-stranded RNA. DIG-labelled antisense wg RNA was prepared from a full-length wg cDNA (Rijesiwik et al., 1987) subcloned in Bluescript. DIG-labelled antisense hh RNA was prepared from a partial hh cDNA (Mohler and Vani, 1992) subcloned in Bluescript. Staging of embryos was according to Campos-Ortega and Hartenstein (1985). Embryos were mounted in glycerol and photographed using differential interference contrast optics on a Zeiss Axioplan microscope.

Transcription of wg is activated at the blastoderm stage of Drosophila embryogenesis in three broad domains at either pole of the embryo and then in single-cell-wide stripes (Baker, 1987) that define the posterior margins of each parasegmental primordium (Fig. 1A); these correspond to the cells that accumulate little or no Ftz and Eve protein as the blastoderm cellularises (Lawrence and Johnston, 1989). The parasegmental activation of wg is controlled by both positive and negative regulatory effects of various pair-rule gene products. In the odd-numbered parasegments, wg is under the negative control of eve, since in the absence of eve activity, all the cells of each primordium express wg (Ingham et al., 1988). Similarly, in ftz mutants, wg is activated in all the cells of the even-numbered parasegments (Fig. 1B), indicating that the ftz product normally acts to repress wg expression (Ingham et al., 1988). By contrast, in the absence of prd activity, there is no activation of wg in the cells at the margins of each even-numbered parasegment (Fig. 1C) implying that prd acts as a positive regulator of wg expression.

During normal development, cells at the posterior margin of each parasegment continue to express wg as the germ band extends; this maintenance of expression is not an autonomous property of these cells but rather is a response to a signal(s) from their posterior en-expressing neighbours (Martinez-Arias et al., 1988) and depends upon the activity of the hedgehog (hh) gene (Hidalgo and Ingham, 1990) that is expressed in en-expressing cells (Mohler and Vani, 1992; see also Fig. 3C). To define more precisely the relationship between hh function and wg expression, we analysed the pattern of wg transcription from the onset of gastrulation in embryos homozygous for the strong loss of function allele, hh^IJ. At blastoderm, no differences in the activation of wg transcription can be detected between wild-type and hh^IJ homozygotes (data not shown). Reductions in the level of transcript can first be unequivocally distinguished at stage 8 in hh mutant embryos; by the end of stage 9, expression is normal in the so-called ‘head blob’ and labrum and in the foregut and hindgut primoria (Fig. 2A,B), but has disappeared from all other ectodermal cells with the exception of six neuroblasts in each parasegment (Fig. 2C-F). As embryogenesis proceeds, this neuroblast expression fades as in wild type. At stage 11, wg transcript appears in two patches in the labial segment of hh mutants (Fig. 2H); these correspond to the salivary gland primordia that also initiate wg expression in wild-type embryos (Fig. 2G). Expression of wg in the visceral mesoderm, that is initiated in wild-type embryos at stage 13 (Fig. 2I; van den Heuvel et al., 1989) does not, however, occur in hh mutants (Fig. 2J).

The ectopic expression of wg seen in ftz and eve mutants is not stable but fades away rapidly after gastrulation. In the case of ftz mutants, expression of wg persists only in a single-cell-wide stripe of cells at the posterior boundary of each of the abnormally large parasegments that becomes visible in the extended germ band (Martinez-Arias et al., 1988; Fig. 3H). In ftz hh double mutants, expression of wg is initiated throughout each even-numbered parasegment at blastoderm but disappears completely from the presumptive epidermis after gastrulation (Fig. 3I). Thus the persistence of wg in narrow stripes of cells at the posterior margins of the enlarged parasegments that form in ftz embryos depends upon the activity of hh; that only these cells maintain wg expression suggests that hh can act over only very short distances, i.e. on adjacent cells, to maintain wg transcription activated at blastoderm.

In the absence of eve activity, expression of wg com-pletely disappears from the epidermal cells following gastrulation (Fig. 3E). Since hh is required to maintain wg, one explanation for this disappearance may be that hh is not activated in eve mutants; activation of en in the same domains at the anterior of each parasegment is known to depend absolutely upon eve activity (Macdonald et al., 1986). To test this possibility, embryos lacking all eve activity were hybridised with hh probe; as predicted, there is no parasegmental expression of hh in these embryos (Fig. 3F).

In embryos lacking the activity of ptc, the domain of wg expression becomes expanded anteriorly so that by the beginning of stage 10 it occupies about one third of each parasegment (Fig. 4A,B). Prior to gastrulation, the distribution of wg transcripts is the same as in wild type, but subsequently broadening of the wg domains is detected with the beginning of germ band extension (stage 8/9). The broad wg domains are maintained throughout embryogenesis (Fig. 4C,D), though expression is still subject to the same selective repression along the dorsoventral axis as in wild type (data not shown; but see Fig. 4E,F).

To determine whether this repressive activity of ptc is involved in the suppression of ectopic wg expression that we observe in eve and ftz mutant embryos following gastrulation, we constructed ptc; ftz and ptc eve double mutant embryos. In both cases, expression of wg is activated throughout the even-numbered or odd-numbered parasegments respectively at blastoderm but instead of fading from most or all of these cells, it persists throughout the extended germ band stage and during germ band retraction (Fig. 4E,F). Thus absence of ptc activity results in the maintenance of wg transcription in the progeny of all cells in which it is initiated at blastoderm, with the exception of lateral cells in which transcription is repressed, as in wild type, during mid stage 10.

Since hh is not expressed in eve mutants (see above), the persistence of wg expression in ptc eve double mutants suggests that the derepression of wg by removing ptc activity is independent of hh activity. To test this proposal directly, we investigated whether the persistent expression of wg observed in ptc; ftz double mutants is hh-independent by constructing embryos mutant for ptc, ftz and hh (see Materials and Methods for experimental details). Expression of wg persists throughout germ band extension in these embryos, the transcription pattern being indistinguishable from that of ptc; ftz double mutants (data not shown); thus derepression of wg transcription in the absence of ptc activity is independent of hh function.

While ptc activity appears to repress wg transcription in cells in which it is aberrantly activated at blastoderm, we wondered whether it might also act to suppress wg expression in other cells. In particular, we wished to investigate the possibility that some cells become ‘competent’ to express wg at blastoderm, even if they do not activate transcription at this stage. We therefore constructed a recombinant chromosome carrying null alleles of both prd and ptc and analysed wg expression in embryos homozygous for this chromosome. At blastoderm, wg is activated in these embryos only in the odd-numbered parasegments, as in embryos homozygous for the prd allele alone (see Fig. 1C). At stage 9/10, these odd wg domains broaden as in ptc mutants; in addition, new stripes of wg-expressing cells appear in the even-numbered parsegments (Fig. 4G,H). These stripes are initially narrower than their odd-numbered counterparts but subsequently become broader. The position of the stripes, located mid way between the odd-numbered ones, implies that they are composed of cells that would normally have activated wg transcription in response to prd activity.

In tissue-culture systems, the murine orthologue of wg, Wnt-1, can act in both a paracrine and an autocrine fashion to transform cells morphologically (Jue et al., 1992). We therefore wondered whether wg might itself be required for the maintenance of its own expression Embryos homozygous for the allele wg^IIS exhibit a typical wg null cuticular phenotype but still produce a transcript. Transcription of wg in these embryos is initiated normally at blastoderm, but as in hh embryos disappears from the ectoderm during germ band extension. By stage 10, the ectodermal stripes of wg transcript that characterise the trunk region of the embryo are missing (Fig. 5A) though expression persists in the same groups of neuroblasts seen in wild-type and hh mutant embryos (Fig. 5B; compare with Fig. 2C,D). In contrast to hh mutants, however, the transcript also disappears from the stomodeum and hindgut primordium (Fig. 5B,C; compare with Fig. 2H) and is much reduced in the head blob by stage 11 (Fig. 5C); at the same time, new expression of wg is initiated in ectodermal cells at the dorsal margins of each segment (Fig. 5C). The similarity between the patterns of wg transcript in hh and wg mutants suggests that wg may exert its effects indirectly, maintenance of hh expression depending upon wg activity (Mohler and Vani, 1992). To test this possibility, we constructed the double mutant wg ptc in which any effect of the loss of hh activity (due to the wg mutation) on wg transcription should be overriden by the removal of ptc. In these double mutant embryos, wg transcript persists in the most ventral part of the ectoderm throughout germ band extension (Fig. 5D); this indicates that wg function is not essential for wg transcription at the parasegment boundaries and suggests that the effect of the wg mutation is indeed mediated through its control of hh expression. By contrast, expression of wg still disappears from the stomodeum and hindgut (Fig. 5D) indicating that this requirement for wg activity cannot be compensated for by removal of ptc activity.

At the blastoderm stage of embryogenesis the initiation of wg transcription in single-cell-wide stripes at the posterior boundary of each parasegmental primordium depends upon opposing activities of various pair-rule gene products. The expression of prd is required in alternate primordia for the activation of wg transcription whereas both the ftz and eve proteins act to repress this activation.

Our analysis of wg in hh embryos shows that hh activity is required for the maintenance of wg transcription within one hour of its activation in response to these pair-rule cues. Interestingly, this requirement is limited to those regions of the embryo where wg is expressed in narrow bands of cells; thus, in the head, the stripe of cells in the antennal primordium requires hh activity whereas expression in the adjacent broad patch of cells, the ‘head blob’ is independent of hh function. This implies a separate mechanism for the maintenance of wg expression in large patches of cells such as those of the head blob and of the foregut and hindgut primordia that are also independent of hh function; the initiation of expression in these regions is known to be under different control, the head blob expression depending upon activity of the head gap gene orthodenticle (Cohen and Jurgens, 1990). Within each parasegment, six neuroblasts also continue to express wg in hh mutant embryos after the expression in the overlying ectoderm has disappeared. Presumably transcription in these cells becomes fixed as they delaminate from the ectoderm and is maintained autonomously until the neuroblasts divide, when expression ceases both in the mutant and in wild-type embryos.

The misexpression of wg that ensues in the absence of either ftz or eve, is rapidly compensated for by a second mechanism that is able to repress wg transcription following its aberrant activation. Here we have shown that this repressive mechanism depends upon the activity of ptc; in the ptc eve and ptc; ftz double mutant combinations that we analysed, the ectopic expression of wg along the anteroposterior axis in each parasegment persists throughout embryogenesis, although the gene is still subject to regula-tion along the dorsovental axis. Thus ptc activity is capable of repressing wg expression in any cell in which it has been activated, irrespective of its location within the parasegment, but this negative regulatory activity of ptc is not the only mechanism by which wg can be repressed.

In normal development, however, the activation of wg is restricted from the outset to cells at the posterior margin of each parasegment; thus the ability of ptc to repress wg in other regions of the parasegment might seem gratuitous. However, when ptc is eliminated from an otherwise genetically wild-type embryo, there is a broadening of the wg expression domain anteriorly. This ectopic expression of wg is limited to cells in the posterior part of each parasegment, even though ptc is itself expressed (and potentially functional) in a much broader domain (Nakano et al., 1989; Hooper and Scott, 1989). One way of interpreting these findings is to suppose that the normal function of ptc is to repress wg transcription in a subset of the polyclone that initiates wg expression at blastoderm. It has previously been proposed that ptc acts constitutvely to repress wg transcription and that hh antagonises this repressive effect of ptc, thereby positively regulating wg (Ingham et al., 1991; Hidalgo, 1991). Accordingly, following the cellular rearrangements and divisions that occur during germ band extension, wg expression would be maintained in just those cells that are within range of the putative hh signal, a range that from our observations seems to be limited to adjacent cells. Recent cell lineage studies have suggested that an analogous mechanism acts to restrict the en domain fol-lowing gastrulation, some cells that initially express en at blastoderm subsequently switching off expression if they become displaced more than a certain distance from the neighbouring wg-expressing cells (Vincent and O’Farrell, 1992). The difference between the respective widths of the en and wg domains presumably reflects the differing ranges of the putative signals: wg, like its mammalian orthologue, Wnt-1, seems able to act across more than one cell diameter (Gonzalez et al., 1991; Bejsovec and Martinez-Arias, 1991; Jue et al., 1992) whereas hh appears to be restricted to acting only on its immediate neighbours. This latter finding fits well with the molecular structure of the Hh protein which, from sequence analysis is predicted to contain a single trans-membrane domain (Mohler and Vani, 1992); thus the activity of the protein might be expected to be limited to cells that come into direct contact with those expressing hh.

Whilst our data are consistent with such a mechanism, we consider it likely that the activity of ptc is not exclusively restricted to repressing wg expression within the polyclone in which it is activated at blastoderm, but also to suppressing its expression in other cells that have the poten tial to activate wg transcription. This proposal is based partly upon consideration of the extent to which the wg expression domain is expanded in ptc mutant embryos; wg- expressing cells occupy about one third of each parasegment, a rather larger proportion than would be expected to be contributed by the polyclone that initiates wg expression at blastoderm. Thus, in the absence of ptc, it seems that cells anterior to the wg polyclone have ectopically activated wg expression following gastrulation.

The expression of wg that we observe in prd ptc mutant embryos tends to support the idea that ptc suppresses a latency to express wg. In these embryos, expression of wg in the even-numbered parasegments is not initiated at blastoderm, due to the absence of prd activity, yet following gastrulation, these expression domains appear in the appropriate locations. This implies that, although wg transcription was not initiated, the cells in the posterior part of each parasegment are selectively specified to be competent to express wg, a competence that is revealed if the repressive activity of ptc is removed. In embryos mutant for prd alone, this competence to express wg is not manifest, presumably because the corresponding domains of en/hh-expressing cells are not established; thus even though the wg-competent cells are present, they do not activate transcription because the repressive activity of ptc cannot be antagonised by hh.

How such transcriptional competence is determined is at present uncertain, but one possibility is that the pair-rule genes themselves are instrumental in establishing the potential of the wg promoter to be activated. According to this view, low levels of Eve or Ftz protein at the blastoderm stage would render the wg transcription unit ‘open for business’ (see Peifer et al., 1987); transcription would only be initiated immediately in nuclei of cells expressing prd (and in the odd-numbered parasegments opa); however, the other nuclei would retain their ‘open for business’ conformation and hence be receptive to activity of a second level of transcriptional activation that commences as gastrulation proceeds. It is this second level of transcriptional activators (which might be ubiquitously distributed), necessary for the maintenance of wg expression after pair-rule activity has ceased, that would be modulated by ptc and hh activity. According to this interpretation, the levels of ftz and eve activity at blastoderm would play the key role in specifying the future propensities of the cells of each parasegment, subdividing them into wg-competent and en-competent equivalence groups. In embryos lacking a functional prd gene, alternative wg domains fail to be activated at blastoderm; however, the cells in the posterior half of every parasegment primordium should become ‘open for business’ with respect to wg, as in wild type, since the prd mutation has no effect on ftz or eve expression.

The significance of this hypothesis is that it predicts that the specification of cell fate at the blastoderm stage of embryogenesis is not restricted to the cells at the boundaries of each parasegment that activate transcription of en or wg, but in addition includes the intervening cells. Thus cells adjacent to those expressing wg or en do not remain developmentally naive, but rather acquire differing potentials to respond to inductive signals by being allocated to different ‘equivalence groups’ in response to varying levels of expression of ftz, eve and possibly other pair-rule genes. The notion of equivalence groups helps to explain why cells in different locations can respond to the same signal with different outcomes. For instance, cells anterior to the wg domain respond to wg activity by accumulating Arm protein at their surface (Riggleman et al., 1990), while those posterior to this domain respond to the same signal by expressing en. Similarly, cells anterior and posterior to the hh domain respond differentially to hh activity (Hidalgo and Ingham, 1990); in our terms, the differential responses reflect the differing competences of the responding cells.

The data that we have presented indicate that the spatial restriction of wg expression is of crucial importance to the patterning of each parasegment; although wg is initially activated under strict spatial control, subsequent cell-cell interactions are required to maintain its expression in the appropriate cells. We have shown that hh and ptc play a central role in this process; however, by analogy with its mammmalian orthologue, Wnt-1, it is possible that wg itself could act in an autocrine manner to maintain its own expression. The finding that wg expression depends upon wg activity might at first sight seem to support this view; however, our finding that this requirement is suppressed in wg mutants that also lack ptc activity suggests that the effect of wg mutants on its own expression is indirect. In the absence of ptc activity, the cells expressing wg can be considered physiologically equivalent to those that normally express wg in the wild-type embryo, assuming that the latter also effectively lack ptc activity due to the antagonistic effect of hh. Thus the persistence of wg transcription in mutants lacking ptc and wg activity suggests that there is no direct effect of Wg protein on the cells that transcribe wg; rather, the disappearence of wg transcript in wg mutants is likely to be due to the requirement for wg activity to maintain hh expression (Mohler and Vani, 1992). Thus we conclude that, at the parasegmental boundaries at least, wg acts exclusively in a paracrine fashion to specify cell fate. Although expression of wg persists for some time in some of these domains in the absence of wg activity, the expression in the stomodeum and hindgut primorium rapidly fades in the absence of wg activity. This may indicate an autocrine control of wg in these regions, although we cannot rule out the possibility of regulation by some other paracrine mechanism that is itself dependent upon wg activity.

We would like to dedicate this paper to the memory of Carlos Cabrera our friend and one time colleague who shared with us an interest in the role of wingless in the Drosophila embryo. We thank C. Nüsslein-Volhard and R. Whittle for mutant stocks, J. Mohler for hh DNA, A. Forbes, Y. Nakano and A. Taylor for many stimulating discussions, U. Strähle for comments on the manuscript and A. Taylor for technical assistance. This work was supported by the Imperial Cancer Research Fund.