A dual role for the protein kinase shaggy in the repression of achaete- scute

Sensory bristles of the peripheral nervous system arise in a precise pattern in adult Drosophila Hies. On each hemitho- rax, small bristles or microchaetes arc uniformly spaced over the scutum and eleven large bristles or macrochactcs occupy stereotyped positions (Fig. 1 A). On the wing blade adjacent bristles arc arranged al the margin in two or three rows. Bristle development is dependent on achaete (ac) and scute (sc). two genes of the achaete-scute complex (AS-C) (Garcia-Bellido, 1979; Ghysen and Dambly-Chaudiere. 1988) and in their absence sensory organs are missing. Both genes encode transcriptional regulators of the neural fate and contain DNA-binding/dimerization domains of the basic helix-loop-helix type (b-HI.H) (Villares and Cabrera. 1987; Murre et al., 1989a,b).

The wing and thorax arise from a single imaginal disc. In the thoracic epithelium ac and sc are expressed in a complex and dynamic pattern in “proneural clusters” of 20 to 30 cells at the sites of the future sensory organs (Romani et al. 1989; Cubas et al., 1991; Skealh and Carroll. 1991). Within each cluster a group of four, live or six cells, the “proneural held”, accumulate ac and sc products to a higher level (Cubas and Modolell, 1992). All cells of the proneural field arc ncurally competent but a single sensory organ precursor (SOP) is selected and continues to accumulate ac/sc proteins to a high level, whereas in the remaining cells ac/sc expression ceases (Cubas and Modolell, 1992). From some proneural clusters two or three SOPs arise sequentially. The selection of a single SOP from the group ol equivalent cells (Simpson and Carteret. 1990), and hence the spacing of bristles, is achieved by cell interactions during a process known as lateral inhibition (Simpson. 1990). All of the macrochaetes arise from these proneural clusters. After pupariation ac and sc arc re-expressed to allow development of the microchacle precursors. In the wing epithelium ac and sc are expressed in two rows of cells along the prospective wing margin where the bristles arise and along the prospective third vein al the sites of other sensory organs (Cubas et al.,. 1991). Unlike those of the thorax, bristles along the wing margin arc not spaced out from one another.

There are thus two separate issues relating to the devel- opment of the bristle pattern. I low is the expression of ac/sc at specific sites controlled? What is the mechanism allowing a single cell to be chosen from a group of equivalent ones?

There is evidence for the existence of a complex array of c/.s-regulalory sequences within the AS-C that mediate expression of each gene at specific sites of the epithelium. Site-specific sequences responsible for ac and sc expression have been inferred from the study of mutants causing absence of one or a subset of sensory organs: they arc mostly associated with chromosomal breakpoints in the vicinity of the genes (Ruiz-Gomez and Modolell. 1987: Romani el al.,. 1989; Ruiz-Gomez and Ghysen. 1993). achaete and so arc activated indcpcndantly such that ac is expressed at certain sites on the notum and sc al others (Martinez and Modolell. 1991). Subsequently, however, they cross-activate one another, so that both arc expressed at the locations of all future sensory organs (Martinez and Modolell. 1991; Skcath and Carroll. 1991). I lowcver both proteins appear to display equivalent properties.

Two genes, extramacrochaetae (emc) and hairy (h) act as negative transregulators of ac and .so. In mutant einc flies supernumerary macrochactes appear in new. ectopic, locations (Fig. 1B) whereas in Hies mutant for/z. additional microchaetes are observed in ectopic locations (Moscoso del Prado and Garcia-Bellido. 1984a,b: Garcia-Alonso and Garcia-Bellido, 1988). The ectopic bristles result from new. ectopic accumulations of ac and .so proteins, presumably caused by a lack of repression (Moscoso del Prado and Garcia-Bellido. 1984a: Cubas et al.,. 1991; l992;Skeath and Caroll, 1991. 1992; Blair et al.,. 1992). extramacrochaetae and h. like ac and sc, encode an HLH motif (Ellis el al.,. 1990: Garrell and Modolell. 1990). Il has been shown that cine competes with daitghterless/ac and daughlerless/sc hel- crodimers and thus interferes with their DNA-binding prop- erties in a manner similar to the mammalian homologue /</ that associates with Myo I) (Van Doren cl al.,.1991: 1992; Benezra el al.,. 1990: Duncan el al.,. 1992). These observa- tions reveal that einc may not repress the transcription of ac and sc directly but instead interferes with auto- and cross- regulation (Mariincz and Modolell. 1991) and prevents local accumulation of ac and sc proteins (Van Doren cl al., 1992). extramacrochaetae transcripts arc heterogeneously expressed throughout the imaginal epithelium and high levels preferentially coincide with low levels of ac and sc and vice versa (Cubas and Modolell. 1992). Since the expression of einc is indepcndant of ac and sc. einc may thus play a role in reducing ac and sc accumulation al non- proneural sites.

A uniform expression of .se. obtained through the use of a heal shock construct, resulted in the development of bristles at the correct sites in the absence of endogenous ac/sc activity (Rodrigues et al.,. 1990). This experiment suggests that the epithelium is differentially sensitive to the ncuralising effects of these proteins. These differences could be due to differing amounts of cine protein and perhaps other proteins. Thus both spatial restriction of ac/sc expression and local differences between cells are important.

The selection of a single SOP from a competent group requires cell interactions (Doc and Goodman. 1985) and it is thought that the emerging SOP produces an inhibitory signal preventing the other cells of the group from realising their neural potential (Stern. 1954; Wigglesworth. 1940; Simpson. 1990). The products of the Notch (N) and Delta (DI} genes (Lehmann et al.,. 1983; Campos-Ortega and Knusl. 1990) may function as receptor and ligand respec- tively (Fehon el al.,. 1990; Rebay et al.,. 1991; lleitzler and Simpson. 1991). during transmission of the inhibitory signal.,

shaggy (sgg) encodes several cytosoluble protein kinases with predicted serine/threonine specificity (Bourouis cl al.,. 1990, Ruel et al.,, 1993a: Siegfried et al.,. 1990) that are required fora number of different developmental processes (Bourouis et al.,. 1989; Pcrrimon and Smouse. 1989). Here we discuss the role of sgg in the regulation of ac and sc. Mutant sgg clones show a wild-type morphology on the wing margin and. in fact, this is the only place on the entire fly body where sgg is not required (unpublished observa- tions). Elsewhere over the wing blade, mutant clones cause the appearance of adjacent ectopic bristles (Simpson et al.,. 1988; Fig. 1F) and it has been shown that ac is derepressed in these clones (Blair. 1992). On the thorax, absence of sgg does not lead to the development of ectopic bristles: bristles only arise al the usual wild-type positions, but there are more of them (Simpson and Carteret. 1989; Fig. 1D.E). Thus al the site of each macrochaclc a group of about three macrochactes develops, and microchaetes are more numerous and often adjacent. Hence, on the wing, sgg is required outside of the domains of ac/sc expression whereas on the thorax it is not. Furthermore, on the thorax, sgg is required within the domains of ac/sc expression whereas on the wing it is not.

In this paper we shall present arguments in favour of a dual role for sgg in the regulation of ac/sc. First, the gene may be part of a general repression mechanism preventing ac/sc expression. This is effective. for example, on the wing blade and explains the derepression of ac/sc there in the absence of sgg (Blair. 1992). We have been able to demon- strate that sgg also represses ac/sc in a similar fashion on the thorax, however, outside of the proneural sites this effect is masked by other, additional repression mechanisms. Al the proneural sites, .s.e.e represses both ac and sc. Since ac and sc arc nevertheless expressed al high levels at these sites a mechanism must exist to overcome or antagonize the effects of sgg. Second, genetic analyses reveal that sgg is also required downstream of N for lateral inhibition. In this case, when sgg is missing more than one cell per group adopts the neural fate, hence a cluster of macrochaetes develop at each site. This results from a failure to transduce the inhibitory signal., On the wing margin where the bristles are adjacent, lateral inhibition does not take place and hence there is no requirement there for sgg. Finally, it has been shown that sgg is a functional homologue of glycogen synthase kinase-3 (GSK-3) (Woodgett, 1991; Siegfried et al.,, 1992; de Groot et al.,, 1992; Ruel et al.,, 1993b). GSK-3 itself has been found to be phosphorylated and is active in resting cells, whereas a non-phosphorylated form of the protein is inactive (Hughes et al.,, 1993). Therefore it is possible that activity of the sgg protein in Drosophila may be regulated by phosphorylation or dephosphorylalion at different times. For ease of description we shall refer to a “constitutive” role for sgg in the general repression of ac and sc, and to a second “induced” role after activation of N during signal transduction. We shall first discuss the “induced” role.

In the absence of either N, DI or sgg neural hyperplasia occurs at the expense of epidermal cells (Lehmann et al.,, 1983; Bourouis et al.,, 1989). On the thorax of the adult fly this leads to the differentiation of a tuft of adjacent macrochaetes at the sites where in the wild type there is a single one, and a uniform field of microchaetes over the area where they are usually found in a spaced pattern (Simpson and Carteret, 1989; Heitzler and Simpson, 1991; see Fig. 1C). Supernumerary bristles are only found at wild-type positions, they do not arise at ectopic locations. The distri- bution of bristles corresponds to the areas of expression of the genes ac and sc (Simpson and Carteret, 1989, Cubas et al.,, 1991; Skeath and Carroll, 1991). On the notum, outside the domains of ac/sc expression, these mutants have no effect on the epithelium (Heitzler and Simpson, 1991).

Clones of cells triply mutant for ac, sc and sgg, or ac, sc and N, differentiate as epidermis (Simpson and Carteret, 1989; Heitzler and Simpson, 1991). This means that in the absence of ac and sc, cells do not have neural potential and the default state of the epithelium is then to develop as epidermis; this will happen whether or not N and sgg are present. Therefore N and sgg are not required for the differ- entiation of epidermal cells per se. On the contrary, cells expressing ac and sc do have neural potential and they will all develop as neural precursors in the absence of N and DI and many do so in the absence of sgg (Cabrera, 1990; Heitzler and Simpson, 1991; Skeath and Carroll, 1992). Therefore these genes are required to single out spaced pre- cursors and to prevent the other cells from realising their neural potential.,

Notch, DI, sgg and perhaps other genes of the neurogenic class may define a signalling pathway for lateral inhibition that would result in a repression of ac and sc expression. Cells mutant for N autonomously adopt the neural fate when adjacent to wild-type cells. They are thus insensitive to the inhibitory signal from their wild-type neighbours and this suggests that they are defective in the reception of the signal (Heitzler and Simpson, 1991). In contrast, cells mutant for DI can be rescued by adjacent wild-type cells and will form epidermis; this suggests that the defect in these cells may reside rather in the signal itself. Notch, a phylogenetically conserved molecule, and DI encode large transmembrane proteins with EGF-like motifs in the extracellular domains and N also carries a series of ankyrin repeats in the intra- cellular domain (Wharton et al.,, 1985; Kidd el al.,, 1986; Coffman et al.,, 1990; Ellisen et al.,, 1991; Weinmaster et al.,, 1991; Vassin et at, 1987; Kopczynski et at, 1988). A receptor-ligand relationship between these two gene products is consistent with the observation that >/-express- ing and ^-expressing cells in culture bind together (Fehon et al.,, 1990; Rebay el al.,, 1991).

Mosaic analyses showed that from a group of cells expressing ac and sc any cell can become the precursor and that the cells form an equivalence group (Heitzler and Simpson, 1991; Simpson and Carteret, 1990). Therefore before a neural precursor can inhibit its neighbours, it must first be singled out. N and DI are both expressed in all cells of the proneural clusters (Fehon et al.,, 1991; Kooh et al.,, 1993; Heitzler et al.,, 1993). We have shown that wild-type cells will adopt the epidermal fate if adjacent cells express a lower level of N activity than themselves, but produce neural precursors if adjacent cells express a higher level of N activity (Heitzler and Simpson, 1991). The opposite pertains for DI. This shows there is competition between the cells and that the N and DI proteins are required for the mechanism whereby cells choose between alternative fates. Cells with a reduced amount of N thus always inhibit their neighbours but in order to do so they require DI (Heitzler and Simpson, 1993). This suggests that reception by N is negatively coupled to the signalling capacity of cells, via DI, in a feedback loop (Fig. 2B). Thus a cell that finds itself with slightly less N product or slightly more DI product would gain an early advantage, which would be reinforced by the feedback and then greatly amplified with time.

Within proneural clusters that give rise to only a single SOP, the SOP generally arises from a cell positioned near the centre where cells tend to have a higher level of ac/sc (Cubas et al.,, 1992). Mosaic analyses showed that the quantity of ac and sc products themselves can influence the choice of fate: cells with more ac/sc product are more likely to become neural (Cubas et al.,, 1991). This effect is mediated by the neurogenic genes: we have found that the expression of DI and the signalling capacity of the cells is dependent on the genes of the AS-C (Heitzler et al.,, unpub- lished observations; Fig. 2B). Therefore the amount of ac/sc product may serve as an initial trigger and the role of N and DI would be to ensure that only a single cell will ever become a precursor.

A group of dominant alleles of N, called Abruptex (Ax) (Welshons, 1971; Foster, 1975; Portin, 1975) cause the opposite phenotype to that of loss of function alleles and flies make fewer bristles, the cells instead adopt the epidermal fate (Heitzlcr and Simpson. 1993). These alleles are associated with single amino acid changes in a cluster of EGF-like repeals in the extracellular domain of the protein (Kelley et al.,. 1987; Hartley et al.,. 1987). These altered proteins appear to have an enhanced affinity for the ligand, they arc suppressed when the amount of DI is reduced and double mutant Av DI cells differentiate as neural cells showing that, in order to take up the epidermal fate. Av cells require the ligand (Heitzlcr and Simpson, 1993). In culture, however, Ax-expressing cells continue to adhere to ^/-expressing cells but with reduced efficiency (Lieber et al.,, 1992).

As well as taking up the epidermal fate, and unlike N⁻ cells, Ax cells fail to inhibit their neighbours, so, in these cells with hyperactive N molecules the Dl molecules are not available to inhibit adjacent cells. This has led to the sug- gestion that the N and Dl proteins of the same cell may bind together and that this could be the molecular basis of the feedback between the two (Fig. 2B, Heitzler and Simpson, 1993). When expressed in the same cell after transfection, N and DI proteins co-localise suggesting that they can interact within the cell membrane (Fehon et at, 1990). Binding of a cell’s Dl molecules to its own TV molecules would reduce the availability of the Dl protein to interact with the TV molecules of neighbouring cells. The Ax molecules could be altered in such a way as to only poorly bind the DI protein of neighbouring cells (Lieber et at, 1992), but strongly bind to that of the same cell. The possi- bility of autocrine signalling in this manner remains to be tested. For a more complete review see Simpson et al., (1992).

The protein kinase nature of sgg is consistent with a role in the intracellular transduction of the inhibitory signal through a signalling cascade. Cells mutant for sgg, like N, autonomously adopt the neural fate (Heitzler and Simpson, 1991). This suggests that sgg is required for reception of the inhibitory signal and that perhaps it is part of a sig- nalling cascade downstream of N. The situation is compli- cated, however, by the fact that cells mutant for sgg are also impaired in their ability to inhibit their neighbours, that is to send the inhibitory signal via DI. Thus, in mosaics, mutant bristles can be found adjacent to wild-type ones showing that the mutant cell can neither recieve nor send the signal., Indeed this phenotype is similar to that of clones doubly mutant for N and DI (Heitzler and Simpson, 1993). However, sgg is not required upstream of DI since double mutant Nsgg cells continue to behave like N sgg⁺ cells and always signal inhibition to their neighbours. On the other hand, double mutant sgg Ax cells take up the neural fate, like sgg, and so these hyperactive N molecules are unable to transmit the inhibitory signal in the absence of sgg. Taken together, these results strongly suggest that sgg acts after N in the inhibitory pathway (Ruel et at, 1993b; Fig. 2B).

Since N⁺sgg⁻ cells display impaired signalling, but N⁻sgg⁻ cells signal constitutively, it follows that sgg modulates the signal but only in the presence of N molecules. Therefore, sgg is downstream of N for transduction of the received signat, but upstream of N for the capacity of the cell itself to send the signal (Fig. 2B). Hence it is possible that sgg modifies N which in turn will affect the ability of the cell to signal via DL Again, such an effect could result from a coupling of the N and DI proteins such that inhibited cells become locked into an inhibited state, mediated by an autocrine signal.,

Absence of sgg leads to a de repression of ac/sc and the development of macrochaetes outside of the proneural areas in the wing (Fig. IF). Like the wild-type bristles found on the margin, these ectopic bristles are adjacent to one another. Notch is not required for the epidermal versus neural decision of cells on the wing: clones of N null mutants make epidermis over the entire wing blade (with the exception of course of the wing margin) and the study of double mutant sgg N^ts1 clones show that sgg is epistatic over N in this area (unpublished observations). (Note that N is required in the precursors of all bristles at a later step for the differentiation of the four cells of the bristle organ, Hartenstein and Posakony, 1990). This suggests that on the wing (excluding the margin) sgg represses ac/sc via a mechanism that is N- independent and thus different from the mechanism of lateral inhibition described above.

On the thorax, clones mutant for sgg do not display ectopic macrochaetes. Nevertheless sgg also represses ac and sc on the notum as shown by the following series of experiments. In sgg mutant clones a group of macrochaetes develop at each wild-type site, whereas in ac/sc mutants no bristles develop (Santamaria and Garcia-Bellido, 1978). Triply mutant ac sc sgg clones are devoid of bristles showing that the sgg mutant phenotype requires the activity of ac and sc (Simpson and Carteret, 1989).

The two mutants ac¹ and sc¹ are caused by lesions in reg- ulatory sequences that result in a loss of expression of ac or sc, respectively, at specific sites in the epithelium resulting in the absence of a small subset of bristles on the notum (Modolell et al, 1983; Campuzano et al, 1985; Ruiz-Gomez and Modolell, 1987; Romani et al, 1989). These two mutants are epi static over sgg and, in each case the double mutant clones are devoid of bristles at the ac³ and sc¹ mutant sites (Simpson and Carteret, 1989; Fig. 2A). In the wild type, ac comes on at the ac-dependent sites and it then activates sc, and vice versa. Therefore, in this experiment, if ac is not expressed at the appropriate site then sc will not be activated regardless of the presence or absence of sgg. Similarly sc will be unable to activate ac.

The mutant Hw^49cR+5 produces a truncated non-func- tional sc protein due to a deletion in the coding sequence (Balcells et at, 1988). The ac³ mutation is associated with a chromosomal inversion that results in a complete loss of all ac expression rather than a site-specific loss as in ac¹ (Campuzano et at, 1985; Skeath and Carroll, 1991). In contrast to the previous results, sgg is epistatic over these two mutants and in ac³sgg and Hw^49cR+5sgg clones a cluster of bristles arises at each site (unpublished observa- tions; Fig. 2A). In this case then, sc has been switched on at ac sites in the absence of ac itself and ac has been switched on at sc sites in the absence of, vc itself. Thus, here, a loss of sgg has apparently resulted in a de novo expression of ac and sc. Therefore sgg must normally repress their expression.

As the regulatory mutants ac¹ and sc¹ are epistatic over sgg this suggests that the repression of ac and sc by sgg requires intact enhancer sequences and that sgg may act via an intermediate transcription factor(s).

These results suggest that the mechanism by which sgg represses ac and sc in the wing, functions also on the thorax, at least al the proneural sites. Therefore sgg represses ac and sc at many, if not all, sites in the epithelium, but at certain special proneural sites, ac and sc are expressed in an as yet unknown manner that overcomes this repression.

Mutant sgg embryos derived from mutant germ lines are composed exclusively of abnormal cells with neural char- acteristics (Bourouis et al.,, 1989). It is possible that this too reflects a requirement for sgg in the repression of the genes of the AS-C.

Clones mutant for sgg on the noium cause additional bristles to form but only al proneural sites, a consequence of the fact that several cells of the proneural field, instead of one. adopt the neural fate. In contrast clones mutant for einc cause ectopic macrochactes to form at new sites (Moscoso del Prado, 1984a,b; Garcia-Alonso and Garcia-Bellido, 1988; Fig. 1 B; 2A). The role of einc outside the proneural sites to limit ac/sc expression is well documented (Cubas et al.,, 1991; Skeath and Carroll. 1991; Van Doren et al., 1992). High levels of emc in these regions prevent accumulation of ac and sc (Cubas and Modolell. 1992). We have found that many of the ectopic macrochactes in emc mutants arise later than those at wild-type positions (Heitzler el al.,, unpublished data). The following observations show that the ectopic macrochactes do not arise from cells within the proneural fields. In clones mutant for Dl^9P39 a tuft of about six macrochactes develops at each site usually occupied by a single one (Heitzler and Simpson. 1991). This is the result of all cells of the proneural field becoming bristle precur- sors; consequently they stop dividing. Therefore if the later arising ectopic bristles in the case of emc were to derive from these same cells after division, they should be absent in double mutant emc Dl^9P39 clones. In fact, in such clones, bristles arc still found at ectopic sites in addition to the wild- type ones (unpublished observations).

Hence the additional bristles in sgg mutant clones are derived from cells of the proneural field, while those of emc are derived from cells outside the proneural fields. Also, we have found that in emc mutants, bristles arc normally spaced, suggesting that emc is not required for lateral inhi- bition (unpublished observations). Nevertheless a synergism is found between emc and sgg: sgg/+; emc/+ double het- erozygotes display one additional macrochaete on the thorax (Simpson and Carteret. 1989). Furthermore, whereas sgg clones in wild-type Hies do not cause ectopic macrochactes on the thorax, if the amount of emc is reduced (for example in emc¹/+ flies that by themselves do not have extra macrochactes) then ectopic macrochactes do appear in sgg mutant clones (Simpson and Carteret. 1989). shaggy and emc thus act synergistically in spite of the fact that the two genes appear to affect different cell populations. However, it seems likely that the synergism reflects the other, “con- stitutive”. function of sgg in the general repression of ac and sc on the thorax. Both genes therefore act to repress ac and sc and the question arises as to whether they act together or within independent pathways (see below).

Study of the mutant phenotypes has thus led us to propose two roles for sgg: a “constitutive” role concerning the repression of ac/sc and an “induced” role (.luring lateral inhi- bition (Fig. 2). Here we discuss molecular and biochemical studies whose ultimate aim is an understanding of the molecular basis of these observations.

A family of protein kinases are encoded al the shaggy locus. Two transcription units give rise to ten transcripts and live different proteins (called SGG 10, SGG39, SGGY, SGG46. and SGGX) with a common kinase catalytic domain that predicts serine/threonine specificity, and over- lapping patterns of expression during development (Ruel et al.,. 1993a). Mutational analysis of sgg defines a single com- plementation group, lethality of which is associated with the loss of two of the major proteins (SGG 10 and SGG39). Phe- notypes of flics expressing individual sgg proteins revealed that although there is some redundancy between the different forms they do not all carry out identical functions in vivo (Ruel cl al.,. 1993a). Of the three proteins expressed in the epithelium of (he wing and thorax, one. SGG 10. carries out functions required for the normal segregation of neural precursors.

The sgg protein kinases show extended similarities to the rat GSK-3 enzymes (Woodgett, 1991; Fig. 3). The highest level of amino acid conservation is found in a region encom- passing the kinase catalytic domain, but also extends to either side (Fig. 3). Homology is lower in a region immedi- ately adjacent to the catalytic domain and is insignificant over the remaining 130 C-terminal amino acids. The rat has two forms of GSK-3, GSK-3α and GSK-3β (Woodgett, 1990). more similar to each other than to the sgg kinases. GSK-3β appears slightly more similar to sgg than GSK-3β. Of the sgg proteins, SGG10 is the closest to GSK-3β. The sgg proteins differ between themselves at the C and N termini which have no counterpart in the rat homologues. There seem to be subtle differences between the different sgg enzymes and the two GSK-3 enzymes. The study of transgenic flies showed that GSK-3β, but not GSK-3α can substitute to some extent for sgg (Ruel et al.,, 1993b). Regions outside the kinase catalytic domain may therefore confer differences between the proteins. Although they can phosphorylate the same substrates in vitro, kinetic parame- ters of phosphorylation by GSK-3α. GSK-3β and SGG 10 differ (Plyte el al.,. 1992). Additionally subcellular localisa- tion or tissue distribution may distinguish them. Cellular compartmentalisation of SGG/GSK-3 is still under investi- gation.

SGG10 and SGG39 display redundant activities, they only differ by a C-terminal extension present in SGG39. SGG46 is not redundant with other sgg proteins. In a mammalian cell transfection assay. GSK-3α, GSK-3β, SGG10 and SGG39, but not SGG46, can modulate the activity of the proto-oncogene c-JUN presumably as a result of direct phosphorylation (deGroot et al.,, 1992; Nikolakaki et al.,, 1993). indicating that sgg and GSK-3 can act on the same substrates. The in vivo targets of sgg will not, however, necessarily be the same. The phosphorylation sites of c-JUN. conserved in all members of the jun family in mammals, arc not conserved in Drosophila JUN. GSK-3 is implicated in the insulin response pathway where glycogen synthase is targeted by at least live protein kinases including GSK-3 which has inhibitory effects on the enzyme. Other substrates of GSK-3 are involved in the regulation of cell metabolism or growth control and include transcription factors and components of signal transduction pathways (Woodgett. 1991; Plyte et al.,, 1992).

In many cases GSK-3 appears to have an inhibitory effect maintaining its targets in a phosphorylated but inactive slate in resting cells (Woodgett, 1991). Hormonal stimulation leads to dcphosphorylation of the inhibitory GSK-3 sites either by induction of a phosphatase or through inhibition of GSK-3 activity (Plyte et al., 1992). Indeed GSK-3 activity was found contingent to phosphorylation of a tyrosine residue in resting cells (Hughes et al., 1993). These studies suggest that function of these enzymes is itself subject to modulation.

We have used a cell transfection assay (the transactivation properties of ac and sc together with daughterless (da) on a reporter construct in Drosophila S2 cells; Van Doren et al., 1992) in an attempt to identify possible downstream targets of the kinase. The effect of transfection of sgg protein kinases was measured in the presence or absence of factors thought to be regulators of ac/sc. such as emc, h. the E(spl) m-prolcins and other new putative transcriptional regulators isolated in our laboratory (unpublished observations). Amongst others, shaggy was found to modulate the effects of emc and E(spl) (unpublished observations). Our results provide evidence that sgg and emc probably do act together in the same pathway and that this is reflected in the synergism observed in the genetic studies.

Constructs expressing one of the basic-1 IIJI proteins of the E(spl) complex have also been shown to have negative effects in the same cell transfection assay (unpublished results in collaboration with E. Knusl). This is specifically enhanced by co-transfcction with sgg. The E(spl) genes are thought to be negative transcriptional regulators of the genes of the AS-C (Klämbt et al., 1989; Schrons et al., 1992; Kunst et al., 1992; Campos-Ortega. 1993), and clones simultane- ously mutant for m3. m5. m7 and m8 on the thorax display a N-like phenotype (unpublished observations), shaggy and E(spl) may perhaps function together during lateral inhibi- tion.

Experiments are now in progress to lest whether these results could be due to phosphorylation of emc and E(spl) by sgg. Clearly other explanations arc possible and further- more any demonstration of phosphorylation in vitro will have to be followed by more stringent molecular and genetics analysis of target phophorylation sites in vivo.

We have presented arguments for a dual role of sgg in the regulation of ac/sc. shaggy is required to repress ac/sc expression on the wing blade, it functions in a similar fashion on the thorax and perhaps even in the embryo. This repression mechanism may be mediated by a transcription factor(s) and is N-indcpcndcnl. By analogy to the biological effects of GSK-3, it seems likely that this “constitutive” function is the result of the activity of a phosphorylated form of the sgg protein kinase. It is known that some of the targets of GSK-3 are transcription factors, so sgg could maintain potential transcriptional activators of ac/sc in an inactive state or potential repressors in an active one. At special proneural sites ac and sc are expressed in spite of the presence of veg by an unknown mechanism that overcomes or antagonizes sgg activity, shaggy is subsequently required downstream of N during lateral inhibition and we postulate that this may be an “induced” function following binding of N to D1. Molecular studies of GSK-3, as well as sgg, provide evidence for modulation of activity of this enzyme by tyrosine phosphorylation (Hughes et al.,. 1993). Therefore, shaggy could be regulated by this means at different steps. It is possible that some of the same target proteins may be involved in both aspects of sgg function: if, for example, its activity were to be antagonized at the proneural sites, but then “re-activated” after N signalling.

Both emc and sgg repress ac/sc on the thorax. High levels of emc, usually occurring at non-proncural sites, prevent accumulation of ac and sc by binding to these proteins and preventing auto and cross-regulation. This phenomenon lues not absolutely require sgg: derepression of ac/sc is not bserved in sgg mutant clones, repression by cine suffices. Shaggy, on the other hand, may repress ac/sc through a mechanism that requires a transcription factor(s) since intact cis-regulalory sequences of ac and sc are necessary for repression. The observed synergism between emc and sgg could simply reflect the fact that, in the double heterozy- cotes, both repression mechanisms are less effective and this can cause sufficient derepression of ac/sc for an extra bristle to appear. Our cell transfection assay has revealed, however, that the activity of emc is modulated by sgg and so it is possible that the two genes act together in the same pathway.

It is not known, of course, how a signal is generated after binding of N to D1, nor how sgg acts in the signalling cascade. Our results arc in favour of a role of both sgg and E(spl) in the same pathway in this process. The E(spl) proteins may be transcriptional repressors of ac and sc. It is also probable that there is more than one pathway leading to the downregulation of ac/sc during lateral inhibition since ones mutant for sgg display partial penetrance (Fig. 2B).

The physiological significance of the signalling pathway implicating the protein kinases of the SGG/GSK-3 family will undoubtedly be the subject of further investigation. Of major importance will be the demonstration of in vivo mod- ulation of kinase activity following extracellular signaling, shaggy and GSK-3 contain a conserved site for potential phospho-tyrosine regulation and phosphorylation of this site correlates with catalytic activity. Tyrosine (and also threonine) phosphorylation al the equivalent position of the MAP protein kinase family is responsible for their induction during stimulation by various mitogens (Sanghera, 1992). This results from activity of an upstream kinase (the MAP kinase kinase) that exhibits a dual speci- ficity for serine/threonine and tyrosine residues and which is itself regulated by scrinc/thrconine phosphorylation. Both the MAP kinase family and the MAP kinase kinase are highly conserved in several organisms including Drosophila (Biggs and Zipursky. 1992; Tsuikt et al.,. 1993). Unravelling the entire regulatory cascade could be difficult since it has been documented in at least one case that activation of a GSK-3 target was obtained by the stimulated dephosphory- lation of the GSK-3 phospho-sites by a phosphatase (Dent et al.,. 1990) and furthermore that GSK-3 can itself be inac- tivated by phosphorylation by PKC and thus is potentially antagonized by the phosphatidyl inositol stimulated pathways (Goode et al.,. 1992). Genetic dissection together with in vitro studies arc important tools that will help to further unravel this complexity.

The technical help ol Cathie Carteret. Claudine Ackerman and Véronique Pantcsco has been greatly appreciated throughout the years. We thank our colleagues at the LGME for fruitful discus- sions. Our work is supported by the Institut National de la Santé et de la Recherche Médicate, the Centre National de la Recherche Scientifique, the Centre Hospitalier Universitaire Régional, the Association pour la Recherche sur le Cancer and the Pondation pour la Recherche Médicale. I.. R. is supported by a grant from the Minislére de la Recherche et de la Technologic and P. 11, by a grant from the Association pour la Recherche sur le Cancer.