wingless expression mediates determination of peripheral nervous system elements in late stages of Drosophila wing disc development

The appearance of spatial patterns of cell differentiation in the epithelia of imaginal discs in Drosophila depends upon signalling mechanisms between adjacent cells (Haynie and Bryant, 1976; Wilcox and Smith, 1977; Mohler, 1988; Phillips et al., 1990). In particular, the specification and differentiation of the wing cuticular elements, including bristles in stereotyped distributions, presents a very accessible system for analysis (Campuzano and Modolell, 1992). The application of genetic methods, particularly molecular probes to gene products in situ in the disc tissue, has suggested that the patterning is a series of cell-specific gene activations, acting in a cascade to increase the precision of the pattern.

In a wing imaginal disc, each sensillum comprises four different cells, derived clonally from a single precursor cell (Hartenstein and Posakony, 1989; Huang et al., 1991). This precursor or sensory mother cell (SMC) in turn arises within a group of cells, called a proneural cluster, characterised by expression of the ‘proneural’ genes of the achaete-scute complex (Cubas et al., 1991; Skeath and Carroll, 1991; Romani et al., 1989). The achaete-scute genes are expressed in the proneural clusters, dependent upon cis-regulatory regions (Ruiz-Gómez and Modolell, 1987). Proneural expression is refined by a balance between activators and inhibitors of a positive feedback system (Cubas and Modolell, 1992; Martinez and Modolell, 1991). Inhibitory factors include gene products that influence the specification of an SMC (Cubas and Modolell, 1992; Blair, 1992) as well as components of a system of lateral inhibition, which act after SMC specification to inhibit formation of additional SMCs in adjacent cells (Heitzler and Simpson, 1991; Simpson, 1990). Subsequently, genes such as neuralised and cut, which may be specific for neural functions, maintain this state of determination (Jack et al., 1991; Jan and Jan, 1992).

We chose to examine the contribution of the wingless gene to these events because it appears to be a component in a cell-to-cell signal transduction pathway both in embryogenesis (Martinez-Arias et al., 1988; DiNardo et al., 1988) and during imaginal disc development. It is a member of the highly conserved Wnt family of genes, which code for putative intercellular signalling molecules (Nusse and Varmus, 1992). It is expressed in a complex pattern in third instar imaginal discs implying a role in local developmental events (Baker, 1988a) and mutations in it affect discderived structures, including the wing (Sharma and Chopra, 1976; Morata and Lawrence, 1977).

We describe the time-course and pattern of activation of wingless in late stages of the wing disc development using a wingless reporter genotype which is simultaneously a temperature-sensitive genotype for wingless. Its role may be to establish a link between proximodistal positional cues and specification of sense organ primordia. Our findings are consistent with its participation in post-transcriptional regulation of the achaete-scute-complex (AS-C), in antagonism to the action of the gene zeste-white 3-shaggy (zw3sgg). At the presumptive wing margin, wingless activity affects cell proliferation.

Flies were reared on standard maize medium at 25°C. For the temperature shifts of conditionally expressed genotypes, collections were made at 16°C for 12 hour periods and upshifts were to 28°C. The precise stage reached at any shift point was determined (a) by X-gal staining a sample of the larvae and matching that pattern with the time course for the reporter in a wild-type genetic background at 25°C and (b) by inspecting morphological features associated with particular developmental stages. Those we have used include embryonic stages judged by differential interference optics, identification of the mouth parts, which are instar-specific, and the presence of wandering stage larvae, white prepupae, brown prepupae and pupae. We have distinguished five periods in the 3rd instar; stages III₍₁₎, III_(1/2), stage III₍₂₎ and III₍₃₎ are separated by changes in their wingless reporter expression (see Fig. 2A-C), whilst stage III_(w) is the wandering stage.

Adult cuticles were softened in 10% potassium hydroxide, then mounted in Berlese fluid and examined under phase-contrast optics. The frequency per hemi-notum was determined for each notal bristle, and wing margins were scored for a complete set of costal, triple row and double row bristles and hairs per wing.

The temperature-sensitive wingless mutation used was wg_1L (Nüsslein-Volhard et al., 1984) and the wingless reporter was wglacZ (Kassis et al., 1992) which is simultaneously a recessive lethal mutation in wingless. For the temperature-shift experiments, eggs were collected from a cross between CyO/wg-lacZ females and en-lacZ/wg_1L males. Positive X-gal staining in a wing disc but without the typical engrailed reporter pattern (Hama et al., 1990) identifies it as coming from the genotype wg-lacZ/wg_1L. The development of SMCs and proneural expression was examined in temperature-shift experiments where eggs were collected from a cross between T(2;3) CyO/wg_1L;TM2/Ch Tb females and T(2;3) CyO/wg_CX4;neu-lacZ_B28/Ch Tb males. Larvae and pupae of the wg_1L/wg_CX4;neu-lacZ_B28/TM2 genotype can be distinguished from their ‘Tubby’ sibs. neu-lacZ_B28 is an enhancer trap allele of neu ralised (D. Clements and J. Merriam, unpublished observations) which shows an imaginal disc lacZ pattern identical to that of the A101 strain (Huang et al., 1991; Phillips et al., 1990; Clements and Phillips, unpublished observations). Except where otherwise referenced, genotypes are described in Lindsley and Zimm (1992).

In situ hybridization was performed by the digoxigenin method and used in combination with X-gal staining as described previously (Phillips et al., 1990). The template for probe synthesis was the wingless cDNA wg7121 (Cabrera et al., 1987).

Wandering 3rd instar larvae were selected by the appearance of mature salivary glands, characteristic of the end of the larval period, which are visible as large clear areas through the ventral body wall. They were injected with 10 mM bromodeoxyuridine (BrdU) in Ringers solution, kept for 2 hours and then assayed for β-galactosidase activity and BrdU incorporation. The anterior third of the animal was inverted in phosphate-buffered saline (PBS, pH 7.2) and fixed in PBS containing 4% paraformaldehyde for 5 minutes at 22°C. The sample was then stained in X-gal as described previously (Phillips et al., 1990), washed in PBS and postfixed in Carnoy’s fixative for 15 minutes at 22°C. After rehydration, the sample was incubated for 1 hour at 22°C in mouse anti-BrdU monoclonal antibody with DNase (Amersham, RPN202). Subsequently bound antibody was detected using a biotinylated goat anti-mouse secondary antibody and avidin/biotin/horseradish peroxidase complex (Vector Laboratories). The heads were washed in PBS containing 1% bovine serum albumin (fraction V, Sigma) and 0.1% Triton X-100 (PBT), blocked in PBT plus 2% goat serum for 1 hour and then incubated in secondary antibody in PBT with serum at 4°C overnight. The samples were washed twice in PBT, twice in PBS with 0.1% Tween 20 (PBW) and incubated with avidin/biotin/horseradish peroxidase complex in PBW prepared according to the manufacturer’s instructions. The samples were washed three times in PBW and then incubated 2 to 10 minutes at 22°C in PBW with 0.1% 3, 3′-diaminobenzidine and 0.05% hydrogen peroxide. The reaction was stopped by washing in PBW. Discs were dissected from the inverted heads and mounted in PBS plus 0.1% sodium azide for examination and photography using differential interference contrast microscopy.

A ratio of replication indices between the margin and the pouch of wing imaginal discs was calculated by dividing the proportion of cells that incorporate BrdU in the marginal region by the proportion in the pouch away from the margin. The margin was defined by expression of the wg-lacZ reporter. The sample was taken from the region of the pouch anterior to the M line (see Results) and distal to the distal ring of wg-lacZ expression. Significance was tested by Student’s t-test after arc-sine transformation.

Larvae and prepupae were dissected as for BrdU staining and fixed for 1 hour at 4°C in 0.1 M Pipes (pH 6.9), 1 mM EGTA, 2 mM MgSO₄ with 1% Triton X-100 and 1% formaldehyde (final concentration). Samples were washed and blocked for 2-4 hours at 22°C in PBT, incubated overnight at 4°C in PBT with 10% antiAchaete mouse monoclonal supernatant (Skeath and Carroll, 1991), which was generously provided by M. Haenlin and P. Simpson. The samples were washed and primary antibody binding was detected as described for anti-BrdU immunohistology, except that the secondary antibody incubation was for 3 hours at 22°C and the staining buffer included 0.075% each of CaCl₂ and NiCl₂.

The pattern of wingless expression is neither spatially uniform nor static during larval development. Our results confirm the principle elements of the transcription pattern of wingless in imaginal discs from wandering third instar larvae as shown by in situ hybridization and autoradiography (Baker, 1988a). A stripe of intensely labelled cells bisects the wing pouch (wp, Fig. 1A), which Baker correctly suggested might lie at the presumptive wing margin. The marginal sensory bristles form on either side of the wing margin, which is also the dorsoventral compartment border (Garcia-Bellido et al., 1973). The stripe of wingless expression (between arrowheads, Fig. 1A,B,D) is two cells wide. Simultaneous detection of neu-lacZ and wingless shows that the stripe lies between the SMCs of the dorsal and ventral rows of the marginal chemosensory bristles (drc, vrc in Fig. 1D). The cells in this stripe form the mechanoreceptor bristles of the anterior margin and the posterior marginal hairs (Hartenstein and Posakony, 1989). In the rest of the wing pouch, cells show an intermediate level of wingless expression. Parts of two rings of winglessexpressing cells can be resolved at the base of the wing pouch in the presumptive hinge region (e.g. at arrows in Fig. 1A-D). The distal ring (Fig. 1D) includes SMCs of the giant sensillum of the radius (gsr) and the dorsal humeral crossvein sensillum (dhcv), and wingless-expressing cells extend from the ring near the anterioposterior border towards the SMCs of the vein 3 sensilla (Fig. 1D, open arrow). The SMCs of the sensilla of the dorsal radius (dr, Fig. 1D) and the ventral radius (arrowhead in Fig. 1C) are between the rings. A group of wingless-expressing cells surrounds the SMCs of the tegula (t, Fig. 1E). A proximal band of expression (3 – 5 cells wide) extends along the A/P border and anteriorly across the presumptive notum (n, Fig. 1A; see also B,C,E). The SMCs of the scutellar (sc, Fig. 1C) and posterior postalar (ppa, Fig. 1E) bristles are within this band while the dorsocentral bristles (dc) are on its proximal edge (Fig. 1B,E).

The RNA expression profile does not differ substantially at the third instar and prepupal stages, or in the embryo, from the pattern of β-galactosidase expression of the wing less enhancer detector (wg-lacZ), stock 1-en-11 (Kassis et al., 1992; also compare Fig. 1A,D and Fig. 2C,D). We have therefore taken the latter to inform us of the wingless transcription pattern. The late third instar pattern (Fig. 2C) begins to emerge in the early third instar when the marginal stripe appears as a row of intensely stained cells in the otherwise uniform expression in the distal half of the wing disc (III₍₁₎; Fig. 2A). During the subsequent six hours of development, staining in four quadrants of the wing pouch decreases to leave a ring and cross, and at 12 hours into the third instar, the notal band of wingless expression has appeared in most wing discs (III₍₂₎; n in Fig. 2B). By 30 hours into the third instar (III₍₃₎) a group of cells on the presumptive tegula begin expressing, completing the final pattern.

The topological transformation that results from eversion of the wing disc can be appreciated by following the position of the principal components of the wg-lacZ pattern. During eversion, the dome-shaped wing pouch (Fig. 2C) is extended and flattened to form a mitten-like shape (Fig. 2D) moving the marginal stripe (arrowheads in Fig. 2C,D) from the face of the disc to the edge and bringing the dorsal (•) and ventral (★) arcs of the rings (Fig. 2C,D) into superposition. The presumptive tegula (t, Fig. 2D) lies on the edge of a deep sulcus which separates the wing from the notum. This topology and staining pattern is maintained until 80 hours after pupariation when the cells of the wing disc have secreted adult cuticle and the bristles of the notum are visible (Bainbridge and Bownes, 1981). This permits direct comparisons between the position of wg-lacZ expression and the sites of sensory organs, which confirm and extend the relationships shown in the wing disc. In addition to the scutellar (sc) and dorsocentral (dc) bristles, the posterior postalar (ppa) bristle forms in the notal band of wg-lacZ expression (Fig. 2E). The epithelium that constitutes the wing hinge is elaborately folded and the cells are densely packed, which contributes to its complexity. The proximal ring extends from the base of the costa across the proximal tip of the radius, distal to the axillary sclerites, ending at the axillary cord and the distal ring extends from the medial costa across the humeral crossvein to the distal radius and then to the base of the alula (Fig. 2F). A branch of expression along vein 3 near the A/P border reaches the anterior crossvein (acv), and the posterior crossvein also stains (pcv, Fig. 2G). The chemosensory bristles of the anterior margin, which lie on the edge of the marginal stripe, also express and can be seen against the background of weaker staining from underlying cells (arrowheads in Fig. 2G).

The late pattern of wingless expression in wing discs suggests that it has a local role in cell behaviour which might be separable from the earlier, more global effect on the distal half of the disc (Phillips and Whittle, unpublished observation). We therefore followed the consequences for imaginal cuticular differentiation of interrupting wingless activity at selected developmental stages using a temperature-sensitive genotype.

A maximum of five of the eleven bristles of the notum is sensitive to non-permissive conditions in the conditional wingless genotype, wg-lacZ/wg^IL called wg-ts. When shifted up before the third instar, wg-ts larvae formed pharate adults lacking specific bristles on the notum including the posterior postalar, the anterior and posterior dorsocentral and scutellar bristles (compare Fig. 3B with the wild-type notum in Fig. 3A). The anterior dorsocentral bristle has a very strong temperature dependence in this genotype, with a sharp transition between stages III₍₃₎ and III_(w) (Fig. 4A). The posterior postalar bristle (ppa) is most sensitive to decrease in wingless activity. The ppa bristles are lost in some wg-ts individuals even at the permissive temperature, and up-shifts as late as the wpp stage still significantly increase the penetrance of this phenotype (Fig. 4B). However, like the dorsocentral and scutellar bristles, the ppa bristles are not rescued by down-shift after the late IIIrd instar, indicating a wingless requirement at this stage. The anterior scutellar bristle (asc) is deleted with high penetrance by up-shifts through the beginning of the IIIrd instar, and a significant level of sensitivity persists through the III₍₃₎ stage (Fig. 4C). The frequency of loss of the posterior dorsocentral (pdc) and scutellar (psc) bristles is much lower and more variable in this wg-ts genotype, but these bristles are completely absent in wg^ts/wg^CX4 as discussed later. After up-shift in the III_(1/2) stage, the number of psc bristles formed is 0.78 of the wild-type number, which increases to 0.94 after shifts in the III₍₂₎ stage, while for the pdc bristles the fraction increases from 0.94 to 1.00 over the same period. Shifts before the IIIrd instar do not increase penetrance of this phenotype. These five sensitive bristles form from SMCs that arise within or adjacent to winglessexpressing cells of the notal band and the decreased sensitivity of asc, psc and pdc bristles coincides with the beginning of wingless-lacZ expression in the notum.

The sensory bristles of the wing margin in the wg-ts genotype are temperature sensitive during the late third instar and early prepupal stages. Up-shifts prior to the wpp stage delete bristles of the triple row (tr) and double row (dr), and down-shifts from the III_(w) stage onwards fail to rescue the normal margin bristle pattern (Fig. 5). When wg-ts larvae are held at the non-permissive temperature through-out the period from the middle third instar (III₍₂₎) to the bpp stage, they develop normal-sized wing blades with wild-type venation but are devoid of marginal structures save an occasional bristle or two of the medial tr or dr located at the distal extremity of the wing blade (arrowhead, Fig. 3D). More bristles in the proximal margin are formed following up-shifts at progressively later stages after III₍₂₎. The posterior row hairs and the long hairs of the alula reappear after down-shift in the III₍₃₎ stage but are still sensitive to up-shift during the bpp stage (data not shown).

wg-lacZ reporter gene expression is similar to the wildtype pattern until after the prepupal stages, when wg-ts larvae are up-shifted at the end of the second instar (compare Figs 2C and 6B). It therefore follows that, although this period includes the temperature-sensitive period for the effects on the wing margin and bristles, concurrent wing less activity is not required for cellular viability or maintenance of wingless transcriptional regulation in the winglessexpressing cells.

The pattern of cell division in the wing disc immediately before pupariation is not uniform (O’Brochta and Bryant, 1985; Schubiger and Palka, 1987). Regions of mitotic quiescence are associated with the site of specification of SMCs (Hartenstein and Posakony, 1989; Usui and Kimura, 1992). As their immediate neighbours stop replication, the SMCs initiate a specialized series of mitoses to produce the sensory organ.

At the presumptive wing margin, this pattern of mitotic regulation is dependent upon wingless activity. In wing discs from larvae of the wg-lacZ/en-lacZ genotype (wg+) that are close to pupariation (late III_(w) stage), fewer marginal cells, which express wg-lacZ, show BrdU incorporation than in the rest of the pouch (Fig. 6A between arrowheads). The change in labelling often extends several cells dorsally beyond the region of wingless expression. In discs from sibling wg-ts larvae reared at the non-permissive temperature during the IIIrd instar, suppression of replication is reduced (Fig. 6B). Discs from larvae reared at the restrictive temperature from the end of the second instar were staged by the presence of a region of reduced replication associated with the presumptive third vein (the M line), which forms just after division stops at the presumptive margin (Schubiger and Palka 1987; Hartenstein and Posakony, 1989). The ratio of replication indices between the margin and the pouch in wg-ts discs was five times greater than the ratio in wg+ discs (Students t, P<0.01).

SMCs arise at fixed locations and times during wild-type imaginal disc development as judged by activation of a lacZ reporter in the neuralised gene (Huang et al., 1991). We used the null allele wg^CX4 (Baker, 1987) in the genotype wg^ts/wg^CX4 instead of wg-lacZ so that the expression of the neu-lacZ reporter would not be obscured. This allele shows a more severe phenotype than wg-lacZ in combination with wg^ts or with other viable alleles of wingless (Phillips, unpublished). The SMCs of the wingless-sensitive SOs are not formed in discs from wg^ts/wg^CX4 larvae that have been cultured at the non-permissive temperature during the III instar. The SMCs of most sense organs that are unaffected by the wg-ts genotype, including anp, pnp, psa, tr1 and apa, are present in the wing disc of the wg^ts/wg^CX4; neu-lacZ wandering third instar larvae (arrowheads on presumptive notum, Fig. 7B) or appear after pupariation (asa, data not shown). However, the SMCs of the wingless-sensitive SOs (dc, sc and ppa) are absent. The SMCs of the chemosensory bristles of the margin appear within a triangle formed by the SMCs of the sensilla of the anterior crossvein and ventral sensilla of vein 3 and the campaniform sensillum of the margin (tsm1) in discs from wg+ sibs (arrowheads on wing pouch, Fig. 7A) but are absent in wg^ts/wg^CX4 discs (Fig. 7B). The pdc and psc are always missing in this genotype (n=12) although in wild-type flies they are amongst the first SMCs formed (Huang et al., 1991). These SMCs could not be found at any time during the third instar or prepupal stages. In addition, the SMC of the ps bristle, which is unaffected in the wg-ts genotype, is sometimes absent in wg^ts/wg^CX4 discs although the asa bristle, which normally forms later, is present (data not shown). The SMCs of the dhcv and gsr are also sometimes affected.

Each sensory organ of the wing and notum is formed by the progeny of an SMC arising amongst the cells of a proneural cluster. The dorsocentral bristles require achaete activity while the recurved bristles of the margin and the rest of the wingless-dependent bristles of the notum require scute activity (reviewed in Ghysen and Dambly-Chaudière, 1988). However achaete and scute cross-activate (Martinez and Modolell, 1991; Skeath and Carroll, 1991; Cubas et al., 1991). The distribution of Achaete protein is therefore a measure of proneural activity during the formation of both achaete and scute-dependent sensory organs. We have followed this distribution in wg^ts/wg^CX4, wg+ and wg-ts wing discs after development at the restrictive temperature using a monoclonal antibody against the Achaete protein (Skeath and Carroll, 1991). Wing discs from wg+ wandering third instar larvae showed an age-dependent pattern as previously described (Skeath and Carroll, 1991; Cubas et al., 1991), always including two parallel rows of Achaete-expressing cells corresponding to the precursors of the chemosensory bristles of the anterior margin. In contrast, in wg-ts or wg^ts/wg^CX4 wing discs, we found only occasional isolated clusters of Achaete-expressing cells near the presumptive margin (Fig. 8B) which probably correspond to the ventral sensillum of vein three and the acv (compare with Fig. 7B). In wg^ts/wg^CX4 discs, there is no Achaete staining where the dorsocentral and scutellar clusters would normally form (open arrow in Fig. 8B, compare to 8A), proximal to the proneural clusters of the sense organs which do not depend upon wingless (namely the anp, pnp, psa and apa bristles and the trichoid sensillum 1, Fig. 8B, arrowheads shown in the notum). Achaete protein is also reduced in this area in wg-ts discs (data not shown) but SMCs of the pdc and psc appear as isolated stained cells.

The correct spatial pattern of differentiation in an epithelium like that in the wing imaginal disc will include signalling events between adjacent cells which constrain the possible alternative fates that they can assume (Greenwald and Rubin, 1992). We have correlated the precision of the pattern of wingless transcription with the assignment of particular cell fates in the presumptive wing margin and for peripheral sense organs on the notum, and we propose that the wingless protein mediates changes in the competence of cells to undergo fate restrictions including commitment to form particular sense organs.

Cells of the peripheral sensory organs and of the central nervous system follow a similar developmental programme including progressive specification of precursor cells followed by differentiative proliferation. Sensory organs of the imaginal thorax and head, excepting the eye, derive from precursor cells called sensory mother cells (SMCs) in the imaginal disc epithelia. Most thoracic sensory organs, excepting the mechanosensory bristles of the wing margin, depend either on scute or achaete (reviewed by Ghysen and Dambly-Chaudière, 1988). Transcription of one of these genes is intitiated in each proneural cluster, and then, in the presence of active product, transcription of both genes is enhanced by autoactivation and cross-regulation (Romani et al., 1989; Cubas et al., 1991; Skeath and Carroll, 1991; Martinez and Modolell, 1991). One cell in the cluster becomes the SMC concomitant with a further increase in achaete (ac)/scute (sc) expression, then the neurogenic gene neuralised is activated in this cell whilst ac/sc expression in adjacent cells of the cluster is repressed.

A transcriptional regulatory mechanism is implicated in the specification of the SMCs (reviewed in Campuzano and Modolell, 1992). The distribution of multiple stimulatory and inhibitory factors including the products of ac, sc and extramacrochaetae (emc) influence the pattern of SMCs (Romani et al., 1989; Cubas et al., 1991; Skeath and Carroll, 1991; Cubas and Modolell, 1992). achaete and scute encode DNA-binding helix-loop-helix (bHLH) proteins (referred to here as AC and SC respectively) which associate with other bHLH proteins in heterodimers to promote DNA binding and transcriptional activation (Villares and Cabrera, 1987; Murre et al., 1989; Cabrera and Alonzo, 1991). The Extramacrochaetae product is an HLH protein lacking a DNA-binding domain which inhibits AC/SC DNA binding and bristle formation (Van Doren et al., 1991; Garrell and Modolell 1990; Ellis et al., 1990).

We find that wingless is necessary to specify a subset of the wing and notum sensory organs which arise in or adjacent to wingless-expressing cells. Inactivation of wingless suppresses the earliest manifestations of sensory organ development but this block is reversible until the stage when the SMCs would normally be established. The timing and developmental phenotype are compatible with a role in regulation of AC/SC activity. Without wingless, development of wingless-sensitive sensory organs is arrested before SMC formation, as judged by neuralised-lacZ expression, and AC is not detectable either in single intensely labelled cells characteristic of SMCs or in proneural clusters in the region of the affected sense organs. Either wingless is required upstream of the transcriptional activation of the ac and sc genes or in the process of auto/cross-regulation which amplifies ac/sc expression in the proneural cluster.

At present, we cannot exclude either of these hypotheses, but we discuss later several observations that suggest that wingless might regulate AC and SC post-transcriptionally. We have summarised our own and other published data on the events during the progress of SMC formation in Fig. 9.

The requirement for wingless activity in the development of wingless-dependent SOs can be satisfied by expression during or shortly before the period of SMC specification as defined by neuralised-lacZ expression but not by earlier activity (Fig. 9). For the asc, adc and ppa bristles, this period is substantially later than the beginning of proneural expression in the surrounding cluster of cells (indicated by ^* in Fig. 9; Cubas et al., 1991; Cubas and Modolell, 1992).

Therefore wingless is required after the initiation of proneural gene transcription. The limit to its ability to rescue the formation of particular bristles by down-shift appears to anticipate the beginning of differentiation in that the last effective down-shift for each sensory organ ends 12–15 hours before the first SMC division (Fig. 9). The fact that wingless activity can rescue sensory organ development until the end of the normal period of psc and pdc sensory organ specification makes clear the persisting role for wing less in proneural expression. The wingless product is required continuously to enhance the activity of factors that promote proneural expression.

When we compare the wingless bristle phenotype to that of other mutations that affect thoracic sensory organs, there is only one similar pattern. The ‘pecking order’ of loss of thoracic bristles after interrupting wingless is not the same as that in mutations of achaete, scute, Enhancer of split, extra macrochaetae, Notch, Delta or Hairless (de Celis et al., 1991). The wingless-sensitive bristles do not depend upon a common set of proneural cluster-specific regulatory regions in the scute and achaete genes (Ruiz-Gómez and Modolell, 1987; Leyns et al., 1989). However, the pattern closely resembles the spatial requirement for zw3-sgg. Clones homozygous for embryonic lethal alleles of zw3-sgg (sgg^—) induced during larval development produce supernumerary sensory organs throughout the wing blade and around some bristles of the notum. In these mosaics, most of the extra notal bristles arose close to the dorsocentral and scutellar bristles, others were around the posterior postalar and the presutural bristles but far fewer were close to the other notal bristles (Simpson and Carteret, 1989). sgg^— clones in the wing blade form triple row bristles, double row bristles or posterior row hairs depending upon their position in the wing (Simpson et al., 1988). zw3-sgg activity suppresses ectopic development of bristles which are normally dependent upon wingless activity in regions where wingless is expressed. The spatial correspondence between wingless expression and the pattern of ectopic bristles formed by sgg^— mosaics becomes more striking in the absence of the repressive effect of extramacrochaetae (Simpson and Carteret, 1989). Under these conditions, ectopic bristles in sgg^— clones cover the entire area of the notal wingless stripe including the area anterior to the dorsocentrals and extend into the prescutum (Fig. 2E, arrow), without increasing the number of bristles outside this region. Simpson and Carteret suggest that zw3-sgg is acting during lateral inhibition, but Blair (1992) has shown that sgg^— clones do induce ectopic regions of achaete expression in the anterior wing blade, a region where neither achaete nor scute are normally detectable. We propose that zw3sgg, the Drosphila homologue of the mammalian serine kinase GSK3 (Bourouis et al., 1990; Siegfried et al., 1990, 1992; Woodgett, 1991), antagonises a wingless-dependent activation of the proneural genes. This might occur by modification of the protein products of these genes or by modification of a common activator of these proteins, for example, daughterless (Dambly-Chaudière et al., 1988). We are now testing the genetic and molecular interactions between wingless and zw3-sgg. Independent evidence for a kinasemediated regulation of AS-C gene activity has been reported by Cabrera (1990, 1992). Antibodies were raised against synthetic peptides from lethal-of-scute and achaete which contain consensus phosphorylation sites for casein kinase II. These antibodies only recognize protein in the sensory mother cells although other antibodies show that the protein is present throughout the proneural cluster (Martin-Belmudo et al., 1991) suggesting that the proteins might be phosphorylated when less active.

The antagonism between wingless and zw3-sgg activity that we have identified in imaginal disc cells also exists during embryonic segmentation and may be a general characteristic of wingless signal transduction. The embryonic cuticular phenotypes of zw3-sgg and wingless are opposites; the former has no ventral denticle bands, retaining only the ‘naked’ regions (Perrimon and Smouse, 1989), whilst wing less mutants have a lawn of denticles (Nüsslein-Volhard et al., 1984). Heat-shock-induced ectopic expression of a wingless cDNA causes stable ectopic expression of engrailed and wingless, and a ‘naked’ phenotype (Noordermeer et al., 1992). Ectopic wingless activity is also responsible for this phenotype in the mutation naked (Dougan and DiNardo, 1992). Siegfried et al. (1992) have shown that zw3-sgg acts as a repressor of engrailed autoactivation and that wingless inactivates this function.

The proposed relationship between sensory organ development and changes in cell cycle regulation (Garcia-Bellido and Merriam, 1971; Hartenstein and Posakony, 1989; Huang et al., 1991) has been confirmed by simultaneous detection of BrdU incorporation and of the neuralised-lacZ reporter activity (Usui and Kimura, 1992). We find wing less is necessary for the normal decrease in cell proliferation at the wing margin. The murine Wnt1 was originally identified on the basis of its oncogenic phenotype in mouse mammary tissue (Nusse and Varmus, 1982). In addition wingless is required for cell proliferation in the primordia of the fly Malpighian tubules (Skaer and Martinez Arias, 1992). In these cases, activity results in increased proliferation. However, over-expression of Wnt1 does not increase proliferation in most other cell types (unpublished data of Shackleford and Varmus, reported in Nusse and Varmus, 1992). Like other growth factors (Sporn and Roberts, 1988), the response to Wnt signals appears to be context-dependent. Regulation of proliferation in the MQC cells is not dependent upon scute or achaete activity (Usui and Kimura, 1992). Therefore continued proliferation at the margin in the absence of wingless is independent of decreased ac/sc activity. The converse causal relationship, that proneural expression is dependent upon cessation of cell division, has not been tested.

We have presented evidence for a spatially localised role of wingless in the late stages of imaginal disc patterning. Mosaics of the wg¹ allele showed that the earlier global role of wingless in distalization of imaginal discs was not cellautonomous (Morata and Lawrence, 1977) but we know that wg¹ complements wingless alleles that interrupt wing margin integrity (Phillips, unpublished). Mosaics homozygous for the lethal allele wg^L2 induced during embryogenesis were associated with losses of wing margin (Baker, 1988b) but clones initiated postembryonically showed no such effect. This apparent contradiction may be resolved either if the perdurance of wingless after embryonic expression is long or if there is local non-autonomy at the presumptive DV border, because mosaics initiated after the DV compartmentation event (Garcia Bellido et al., 1973) would not interrupt the wingless stripe of expression in both compartments simultaneously.

We interpret our experiments as evidence that wingless participates in the transduction of positional cues that arise in the main axes of the disc. wingless activity is necessary to establish the locations at which particular SMCs form. However, it is not a global signal, because it is not necessary for all SMCs or for the normal veination pattern. The wingless-sensitive SOs arise in very precise positions relative to wingless expression suggesting that their siting might be wingless-concentration dependent. For instance, the dorsocentral bristles develop just outside the region of intense wingless transcription whilst most of that proneural cluster is distal to the SMCs (Cubas et al., 1991) and therefore within the wingless stripe of expression. Either wingless or another gene coordinately regulated with wingless may inhibit SMC formation within the stripe. wingless expression in the mature wing disc can be characterized as a series of concentric rings or arcs centred at the distalmost point of the disc. We propose that wingless acts as a link between proximodistal patterning processes, which determine its own expression pattern, and specification of part of the SO pattern. Because wingless can provide either a paracrine or an autocrine signal (Nusse and Varmus, 1992), this linkage would extend the influence of the cellautonomous transcription factors which regulate wingless to nearby cells providing a gradient of information for positioning sensory organs.

We acknowledge project grant support from the Science and Engineering Research Council of Great Britain, and are grateful to M. Haenlin and P. Simpson for the gift of the achaete antibody, and to J. Kassis for the wg-lacZ reporter stock.