Notch regulates wingless expression and is not required for reception of the

Transcription

1 Development 121, (1995) Printed in Great Britain The Company of Biologists Limited Notch regulates wingless expression and is not required for reception of the paracrine wingless signal during wing margin neurogenesis in Drosophila Eric J. Rulifson 1,2 and Seth S. Blair 1 1 Department of Zoology and 2 Neuroscience Training Program, University of Wisconsin, Madison, Wisconsin 53706, USA Address for correspondence: Department of Zoology, University of Wisconsin, 250 N. Mills Street, Madison, WI 53706, USA ( rulifson@students.wisc.edu) SUMMARY In the developing wing margin of Drosophila, wingless is normally expressed in a narrow stripe of cells adjacent to the proneural cells that form the sensory bristles of the margin. Previous work has shown that this wingless is required for the expression of the proneural achaete-scute complex genes and the subsequent formation of the sensory bristles along the margin; recently, it has been proposed that the proneural cells require the Notch protein to properly receive the wingless signal. We have used clonal analysis of a null allele of Notch to test this idea directly. We found that Notch was not required by prospective proneural margin cells for the expression of scute or the formation of sensory precursors, indicating Notch is not required for the reception of wingless signal. Loss of Notch from proneural cells produced cell-autonomous neurogenic phenotypes and precocious differentiation of sensory cells, as would be expected if Notch had a role in lateral inhibition within the proneural regions. However, loss of scute expression and of sensory precursors was observed if clones substantially included the normal region of wingless expression. These anti-proneural phenotypes were associated with the loss of wingless expression; this loss may be partially or wholly responsible for the anti-proneural phenotype. Curiously, Notch clones limited to the dorsal or ventral compartments could disrupt wingless expression and proneural development in the adjacent compartment. Analysis using the temperature-sensitive Notch allele indicated that the role of Notch in the regulation of wingless expression precedes the requirement for lateral inhibition in proneural cells. Furthermore, overexpression of wingless with a heat shock-wingless construct rescued the loss of sensory precursors associated with the early loss of Notch. Key words: EGF, neurogenesis, lateral inhibition, pattern formation, compartments, Drosophila, Notch, wingless INTRODUCTION In the imaginal wing disc of Drosophila, sensory mother cells (SMCs), the precursors of the sensory organs, differentiate in a highly stereotyped pattern (Ghysen and O Kane, 1989; Huang et al., 1991; Blair et al., 1992). It is thought that the patterned determination of SMCs occurs via a multistep process (reviewed in Ghysen and Dambly Chaudiere, 1989; Artavanis-Tsakonas and Simpson, 1991). First, positional cues in the developing imaginal disc epithelium specify clusters of proneural cells in a process known as prepatterning. Proneural clusters are defined by the regional expression of proneural genes, largely those of the achaete-scute complex (AS-C) (Romani et al., 1989; Cubas et al., 1991; Skeath and Carroll, 1991), and by the lowered expression of antineural genes, such as extramacrochaetae (emc; Cubas and Modellel, 1992) and hairy (h; Skeath and Carroll, 1991; Blair et al., 1992; Orenic et al., 1993). Next, the clusters undergo refinement, such that only some subset of the proneural cells, often only one, take on the SMC fate, while the other cells of the cluster are maintained as epidermis. As in the insect central nervous system (Doe and Goodman, 1985), refinement and the maintenance of the epidermal fate appears to occur via competitive cell interactions among proneural cells using some form of lateral inhibition (Stern, 1954). The neurogenic group of Drosophila genes, including Notch (N), have been shown to be required for the proper segregation of SMCs from the proneural clusters in some regions of the adult PNS, as well as for proper neuroblast segregation from the embryonic neurectoderm. Loss of N function causes an increased number of neuroblasts to delaminate from the embryonic neurectoderm, with a concomitant loss of epidermal precursors (Lehmann et al., 1983), and analogous overproduction of SMCs within some proneural regions of embryos (Goriely et al., 1991) and imaginal discs (Shellenbarger and Mohler, 1978; Dietrich and Campos-Ortega, 1984; Hartenstein and Posakony, 1990; Heitzler and Simpson, 1991). Such overproduction only occurs within proneural regions and requires AS-C expression; therefore, the neurogenic genes are thought to be epistatic to the prepatterning of proneural clusters and only required for refinement, mediating in some fashion the lateral inhibition between competent, proneural cells. It appears likely that the transmembrane protein encoded by N is a receptor for this lateral inhibition signal. In the notal proneural regions of the wing disc, the N gene is required cell autonomously for the formation of epidermal cells (Heitzler

2 2814 E. J. Rulifson and S. S. Blair and Simpson, 1991), although some controversy still exists whether the same is true within the embryonic CNS (Technau and Campos-Ortega, 1987; Hoppe and Greenspan, 1986; reviewed in Campos-Ortega and Jan, 1991). The expression of a truncated, intracellular form of N leads to antineurogenic phenotypes, as would be expected of a constitutively activated receptor (Fortini et al., 1993, Lieber et al., 1993, Struhl et al., 1993; Rebay et al., 1993). However, N is also involved in a number of other differentiative events in Drosophila (e.g. Hartenstein et al., 1992). In the prospective margin of the wing imaginal disc, it appears that N is required for prepatterning of the margin, as loss-offunction mutations can induce the loss of margin SMCs. It has been suggested that here N is acting as a component of the mechanism by which wingless (wg) signalling is received (Couso and Martinez Arias, 1994; Hing et al., 1994). wg encodes a secreted protein which is a member of the Wnt family of growth factors (reviewed in Nusse and Varmus, 1992). Beginning in mid-third instar, wg is expressed in a narrow stripe of cells along the presumptive wing margin, adjacent to cells that will form the dorsal and ventral rows of sensory and non-sensory bristles (Baker, 1988; Blair, 1993, 1994; Couso et al., 1993, 1994; Phillips and Whittle, 1993; Williams et al., 1993). This wg expression is required for proneural AS-C gene expression along the margin and for the formation of margin bristles and their precursors (Phillips and Whittle, 1993; Couso et al., 1994). Hypomorphic N alleles also induce the loss of AS-C gene expression and SMCs in the developing anterior wing margin; moreover, such alleles are dominant enhancers of hypomorphic wg alleles, supporting the involvement of N and wg in a common pathway (Couso and Martinez Arias, 1994; Hing et al., 1994). However, the proposed role for N in the reception of the wg signal has not been directly tested; in particular, it is not clear if N is required in the signalling or receiving cells. Therefore, we have performed a clonal analysis of margin development, examining the behavior of cells completely lacking N function. If N is required for wg reception, cells in the proneural region of the margin that lack N should always undergo a cell autonomous loss of or delay in the formation of margin SMCs due to the loss or reduction wg-mediated AS-C gene expression. However, if N was only required for refinement then proneural cells lacking N should overproduce SMCs when adjacent to cells expressing wg. Interestingly, we observed both loss and overproduction of SMCs in N clones along the anterior wing margin. However, as many cells lacking N express normal levels of AS-C genes and overproduce SMCs from the earliest stages of margin development, there is clearly no absolute requirement for N in the reception of the wg signal. Rather, our results show that N is required for the establishment or maintenance of wg expression; the loss of wg expression in cells lacking N may be partially or wholly responsible for the subsequent loss of SMCs. MATERIALS AND METHODS Drosophila stocks Oregon-R was used for wild type. N 55e11 /FM7; N 55e11 is a null allele produced by a 3.5 kb insertion in the 5 coding region, which leads to premature termination of transcripts (Kidd et al., 1986). We have also found that homozygous clones of N 55e11 do not stain with anti-n antibody. In some cases, we balanced N 55e11 over the FM6-12 balancer, which contains a P<lacZ> insertion with ubiquitous expression (kindly provided by M. Hoffmann); this allowed us to identify the balancer histologically when necessary. N ts1 (maintained at 18 C) is a temperature-sensitive allele which is homozygous and hemizygous viable at 18 C yet is lethal and fails to functionally complement N locus mutations at 29 C (Shellenbarger and Mohler, 1978; Heitzler and Simpson, 1991). neur A101 /TM3,Sb; neur A101 is a P<lacZ> enhancer trap insertion in the neuralized locus (Huang et al., 1991; Boulianne et al., 1991). WG1296; neur A101 /TM3,Sb contains the ubiquitously expressed WG1296 P<lacZ> enhancer trap (Blair, 1992a, 1994). 18-2πM (w, P<mini-w + ; hs-πm>5a,10d, P<ry + ; hs-neo; FRT>18A) contains πm insertions at both 5A and 10D on the X chromosome. The πm construct contains a heat-shock inducible c-myc epitope. 18-2πMF is 18-2πM with the heat-shock-inducible FLPrecombinase construct (hsflp3,sb/tm6b,tb) carried on the third chromosome (Xu and Rubin, 1993). 18-2πM; neur A101 /TM3. HS-wg/TM3,Sb contains a heat-shock-inducible wg construct (Noordermeer et al., 1992). wg-lacz/cyo is a lethal P<lacZ> insert in the wg locus (Kassis et al., 1992) that expresses lacz in a nearly identical pattern to endogenous wg (Couso et al., 1993; Blair, 1994). vg enhancer-lacz is a construct (kindly provided by S. Carroll) which is expressed in a stripe straddling the D/V compartmentment boundary in the same cells that express wg during late third instar (Williams et al., 1994). All stocks were maintained at 25 C unless otherwise stated. Mosaic analysis Mitotic recombinant clones were obtained either by γ-rays as described in Blair (1992b) or through use of the FLP/FRT system as described by Xu and Rubin (1993). To make use of the FLP/FRT system, N 55e11 was recombined with an FRT sequence at cytological position 18A (N 55e11,18A). To generate N clones using FLP/FRT, N 55e11,18A/18-2πM; hsflp3,sb/+ female larvae were produced by crossing N 55e11,18A/FM7 virgins to 18-2πMF males. To generate N clones marked with neur A101, wg-lacz, or vg enhancer-lacz using FLP/FRT, N 55e11,18A/FM7 virgins were first crossed to neur A101 /TM3,Sb, wglacz/cyo, or vg enhancer-lacz males and the virgin N 55e11,18A/+; lacz marker/+ progeny were then crossed to 18-2πMF males. FRT recombination was induced by placing the food jars containing larvae in a 38 C water bath for 1.5 hours. N heterozygous female larvae for γ- ray induced clones were generated by crossing N 55e11 /FM7 virgins to 18-2πM or 18-2πM; neur A101 /TM3,Sb or WG1296; neur A101 /TM3,Sb males; clones in irradiated FM7/+ wings are extremely rare. All clones in this study were generated between 48 and 72 hours after egg laying (72-48 hours before pupariation). Temperature-sensitive analysis Appropriate crosses were made to generate the following genotypes: N ts1 /Y; neur A101 /+ or N ts1 /N 55e11 ; neur A101 /+ or N ts1 /N 55e11 ; neur A101 /HS-wg. Larvae were shifted by placing food jars containing second and third instar larvae into a 30 C water bath for specified times. Cell death A slightly modified form of the acridine-orange-staining method of Sprey (1971) was used. Dissected discs were placed for minutes in Drosophila Ringers containing 0.2 µg/ml acridine orange, washed and mounted in Ringers. In order to confirm the identity of N ts1 /N 55e11 discs, one disc from each larva was stained for cell death and the other was stained for either the absence of the FM6-12 marker or the loss of margin anti-wg staining. Immunostaining Larvae were prepared and wing discs and pupal wings were dissected, fixed in a Pipes-formaldehyde fixative, and double stained with primary and FITC or biotin-ritc-streptavidin secondary and tertiary

3 Notch and wingless in the fly wing 2815 antibodies as described previously (Blair, 1992a, 1993, 1994). The biotin-ritc-streptavidin staining gave the better signal-to-noise ratio, and was used for anti-wg. Triple labelling used biotin anti-rabbit IgG and RITC-streptavidin as above, and 3 hours total incubation in 1/200 FITC donkey anti-mouse IgG (minimal cross reaction with rat; Jackson) and 1/200 Cy5 anti-rat IgG (minimal cross reaction with mouse; Jackson). For labelling with anti-β-gal alone, the Vector ABC procedure outlined in Blair (1992a) was used. Wing discs were mounted and examined as described previously and fluorescent markers were visualized using either a standard epifluorescence microscope or a Biorad confocal microscope. Multicolor images were generated using Adobe Photoshop. Primary antibody concentrations were as follows: 1/200 mouse anti-β-gal (Promega); 1/10,000 rabbit anti-β-gal (kindly provided by R. Holmgren); 1/200 rat anti-β-gal (kindly provided by S. Carroll); 1/100 rabbit anti-wg (kindly provided by R. Nusse); 1/1000 rabbit anti-scute (kindly provided by G. Panganiban); 1/100 mouse anti-n (Fehon et al., 1991; kindly provided by R. Fehon); 1/5 mouse anti-cmyc supernatent (supplied by the Univ. Wisc.-Madison Biotechnology Center). RESULTS Chemosensory SMC differentiation on the wing margin The double row of chemosensory SMCs, which give rise to the dorsal and ventral rows of recurved chemosensory bristles of the anterior wing margin, arise as a small number of evenly spaced cells within a diffuse proneural region of scute (sc) and achaete (ac) expression (summarized in Fig. 1). Along the anterior wing margin, ac and sc expression appears at approximately 20 hours before pupariation (BP) (Fig. 1B); the SMCs are first identifiable by their heightened expression of sc and ac relative to the other cells of the cluster (Fig. 1B,C; Romani et al., 1989; Cubas et al., 1991; Skeath and Carroll, 1991). At approximately 12 hours BP, the presumptive SMCs delaminate from the wing disc epithelium and express the gene neur (Fig. 1C,D; Huang et al., 1991). wg expression in the developing wing disc The expression pattern of wg in the wing disc changes during larval development (Couso et al., 1993; Williams et al., 1993; Phillips and Whittle, 1993). From the second to early third instar (approximately hours BP), wg is expressed in a largely ventral portion of the disc. During early to mid third instar (approximately hours BP) wg expression expands to cover most of the presumptive wing blade region. By mid third instar (approximately 24 hours BP), the wg expression becomes limited to a stripe along the full extent of the margin, to two bands of expression that encircle the presumptive wing blade region, and to a broad band of wg expression across the notal region of the disc (Fig. 1; Baker, 1988). The margin expression precedes the appearance of margin AS-C expression, and is stable until at least 5 hours AP. At late third instar this expression is approximately 3-6 cells wide, and straddles the dorsoventral (D/V) lineage compartment boundary. The region of strong ac and sc expression extends a further 3-6 cells to either side of the wg-expressing cells, and the chemosensory SMCs, as well as later-appearing mechanosensory and posterior bristle precursors, arise immediately adjacent to the wg-expressing cells (Blair, 1993, 1994; Hing et al., 1994). The margin stripe of wg expression is required to activate the ac and sc expression and for subsequent SMC and bristle formation along the margin (Phillips and Whittle, 1993; Couso et al., 1994). N is required for margin wg expression The failure of margin SMC development in some loss-offunction N alleles is reminiscent of wg loss-of-function phenotypes (Couso et al., 1994; Couso and Martinez Arias, 1994; Hing et al., 1994). We will demonstrate below similar disruptions of proneural development by a N ts allele and by N clones that overlapped the region of wg expression on the margin. Therefore, we have investigated the possibility that the loss of N function disrupts wg expression. Careful analysis showed that even the reduction in N levels in N 55e11 /+ heterozygous discs (Fig. 2B) was sufficient to sporadically reduce the levels of wg protein along the entire margin from the levels found in phenotypically wild-type (FM7/FM7) late third instar discs (Fig. 2A); the margin stripe appeared thinner and contained occasional breaks. More complete disruption could be observed using a temperaturesensitive genotype, N ts1 /N 55e11. Late third instar discs stained with anti-wg showed further disruption after upshifts to the non-permissive temperature for as little as 3 hours (n=4). After upshifts of 6 (n=12) or 12 hours (n=24), secreted wg protein on the margin was severely reduced (Fig. 2C,C ), as seen by the near absence of the intensely staining wg-containing vesicles normally found at the apical surface (Hing et al., 1994), although some faint diffuse staining was sometimes still detected in the basal epithelium. 24 hour upshifts reduced all staining to undetectable levels (Fig. 2D,D ; n=19). Nor was such disruption limited to the wing margin region; reductions in notal and wing hinge staining were also observed, and antiwg staining in haltere and leg discs was noticably lower than in control discs after a 6 hour upshift (n=7). wg levels were even more completely disrupted in marked N clones observed in late third instar and early pupal wing discs (Fig. 2E,E,E,F,F,F ), as assessed using anti-wg or wglacz. N clones that overlapped the normal region of margin wg expression showed complete, cell-autonomous loss of wg expression within the clone (n=22 using anti-wg; n=18 using wg-lacz). Normally, wg-expressing cells lie on both sides of the D/V lineage compartment boundary (Blair, 1993, 1994), while almost all the N clones were limited to the dorsal or ventral lineage compartments. Interestingly, 31/33 of the clones that defined the D/V boundary for more than 3 cell diameters disrupted the wg stripe on both sides of the D/V boundary, inducing a bilateral gap in the wg stripe both within and adjacent to the clone. This cell nonautonomy was seen in both dorsal and ventral clones with nearly equal frequency. Interestingly, many large clones also caused some distortion of the D/V boundary, appearing to bulge into the opposing compartment (see Fig. 4D,E). In clones that nicked the margin stripe without making substantial contact (more than one or two cell diameters) with the D/V boundary, purely cellautonomous loss of wg-lacz was seen (n=7). Clones that did not intersect the region of wg expression did not disrupt wg, even when the clone lay immediately adjacent to wg-expressing cells. As in the N ts analysis, the disruption was not limited to the wing margin. N clones in the notum region of wg expression (n=8/9), in the leg disc (n=2/2) and in the haltere disc (n=4/4)

4 2816 E. J. Rulifson and S. S. Blair A B C A P D D V above for N in maintaining wg expression, and the known requirement for wg in margin proneural development (Phillips and Whittle, 1993; Couso et al., 1993), we have grouped the results of the clonal analysis into two classes based on the position of the clone with respect to the normal region of wg expression. In most cases, this was done in double-labeled discs, staining the clone boundary and either the SMCs (neur A101 ) or sc protein; in such discs the clone position was judged in relation to the normal SMCs and the D/V lineage boundary. Clones that did not contact and define the D/V boundary were judged to lie wholly or largely outside the region of normal wg expression, and thus should have caused no or only minor defects in wg expression (see above). Clones that defined the smooth D/V lineage boundary were judged substantially to have intersected the normal region of wg expression; such clones should in most cases have disrupted both dorsal and ventral wg expression (see above). In a smaller number of cases, discs were triple labeled for clone boundaries, sc expression and wg-lacz or vg enhancerlacz; vg enhancer-lacz is expressed along the margin in the same cells as express wg-lacz (Williams et al., 1994), and is similarly disrupted in N clones (Fig. 4C,C ). Our results show that clones that did not substantially disrupt wg expression had neurogenic phenotypes, while clones that disrupted wg expression had a mixture of anti-proneural and neurogenic phenotypes. wg AS-C SMCs Fig. 1. Summary of the anatomy and normal development of the prospective wing margin in late third instar wing discs. (A) The grey shaded regions correspond to the dorsal compartment (D); white and light shaded regions show the ventral compartment (V). The boundary between anterior (A) and posterior (P) compartments traverses the disc and intersects the D/V boundary at the prospective distal tip of the wing blade. In late third instar discs, both dorsal and ventral regions of the wing blade are visible in a single focal plane, separated by the prospective wing margin (region within box). Expression of wg is shown in green; lighter green denotes wg expression at deeper focal planes. (B-D) summarize the events of SMC development on the anterior wing margin in time sequence from earliest (B) to latest (D); see Results for details. The expression of the AS-C on the anterior wing margin is shown in blue; darker shades denote increased levels of expression, and hatched green stripe denotes the overlapping expression of wg. SMCs are identified by neur expression (red). also caused a cell-autonomous reduction of wg-lacz expression, although the degree of loss was not as pronounced as that observed on the margin. SMC differentiation and proneural activation in N null margin clones display dual phenotypes N clones within the wing margin proneural region of third instar and early pupal stages displayed both neurogenic phenotypes, associated with the overproduction of SMCs, and antiproneural phenotypes, associated with the loss of SMCs and of proneural (sc) gene expression. Because of the strong role demonstrated Clones outside the region of wg expression All clones lying within the proneural region but largely outside the region of wg expression exhibited neurogenic phenotypes. Most or all N cells formed SMCs, as assessed using either neur A101 (n=16; Fig. 3), or using anti-sc in discs old enough that the heightened expression of sc observed in normal and mutant SMCs was clearly visible (n=23; Fig. 3A,A,A,C,C ). The density of SMCs was increased relative to that seen in heterozygotic tissue. This effect appeared to be cell autonomous, as ectopic N SMCs could form immediately adjacent to heterozygous proneural cells (see Fig. 3D). In no such clones was sc expression noticably lowered (n=30). As long as the clone lay outside the region of wg expression, its position within the proneural region had no apparent effect upon the neurogenenic phenotype: similar phenotypes were observed in clones that were immediately adjacent to (Fig. 3C, lower arrow) or distant from (Fig. 3C, upper arrow) wg-expressing cells. On occasions, we observed ectopic SMCs located a small distance outside the apparent epithelial boundaries of the N clone. However, these ectopic SMCs were N (lacking the marker) and were located basal to the overlying heterozygotic epithelium; apparently, it is possible in some cases for ectopic SMCs to delaminate and slide under the adjacent epithelium for a few cell diameters, as seen in Fig. 4A. Interestingly, the developmental timing of SMC differentiation in N clones of this type was speeded up relative to that in the heterozygotic epithelium. When N clones were observed at a time when the double row of SMCs on the presumptive wing margin was not yet completely established, the mutant cells expressed high levels of neur A101 before their heterozygotic neighbors (Fig. 3D; n=4). Thus, contrary to its proposed role in the reception of the wg signal, the loss of N does not slow even the initial stages of margin SMC development. Goriely et al. (1991) reported that the differentiation of sensory structures is speeded up in N embryos. This loss-offunction phenotype complements the gain-of-function pheno-

5 Notch and wingless in the fly wing 2817 Fig. 2. Regulation of wg expression by N in the prospective wing margins of late larval wing discs. Anti-wg staining is shown in red. (A) The margin stripe of wg expression in FM7/FM7 (control); (B) wg expression in N 55e11 /FM7. Note that the stripe is generally narrower and of uneven width compared to control in A. (C) wg expression in N ts /N 55e11 ; neur A101 /+ wing disc after a 12 hour temperature shift. Note that wg is significantly reduced on the margin. (C ) Double image of C showing the SMC pattern in green (anti-β-gal staining of neur A101 ). Note the SMC hyperplasia on the margin combined with reduced wg levels. (D) Complete loss of margin wg in N ts /N 55e11 ; neur A101 /+ wing disc after a 24 hour temperature shift (arrow indicates position of the distal tip). (D ) Double image of D showing the SMC pattern in green (anti-β-gal staining of neur A101 ). Note the absence of all margin SMCs, despite the neurogenic defects observed in other proneural regions. (E,E,E and F,F,F ) Single and double images show the bilateral loss of wg (red) from both dorsal and ventral compartments in marked N clones (green) that abut the D/V boundary. (E,E,E ) Two N clones (arrows in E ) marked by the absence of πm expression (anti-c-myc) that caused bilateral gaps in margin wg expression (arrows in E). (F,F,F ) A N clone (arrow in F ) marked by the absence of anti-n staining that caused a bilateral gap in margin wg expression (arrow in F). Note that the domineering nonautonomous effect of N clones that abut the D/V boundary arises from both dorsal and ventral clones. Scale bar, 50 µm. types obtained using truncated and constituitively active forms of N, which produce a delay in cell fate choice (Fortini et al., 1993; Coffman et al., 1993), suggesting that N extends in some sense the period of competence during cell-cell interactions. Thus, in this class of clones, there was no absolute requirement for N in the reception of the paracrine wg signal, as wgdependent sc expression and SMC formation was not lost or delayed in cells lacking N. Clones overlapping the region of wg expression Clones that overlapped the region of wg expression for substantial distance showed a mixture of neurogenic and antiproneural phenotypes. N cells distant from wg-expressing cells showed loss of SMCs and sc expression, while N cells near wg-expressing cells retained sc expression and showed neurogeneic phenotypes. Thus, large clones that lay along the D/V boundary for a distance of 6 or more cells differentiated fewer than normal SMCs in the center of the clone while, at the edges of the clone, within 2-3 cells of regions presumed to be expressing wg, extra SMCs were formed (n=9/9 for doublelabeled clones marked using neur A101 ; Fig. 3B,C). Similar results were observed using anti-sc (n=12 for double-labeled discs; Fig. 4B,B,B ; n=5 for triple-labeled discs; Fig. 4D,E). As would be expected from the the bilateral loss of wg associated with these clones (see above), SMCs and sc expression was lost in the heterozygotic proneural region across the D/V boundary from such clones (Fig. 3B, 4B,D). As noted above, some of these clones distorted the D/V compartment boundary, appearing to bulge into the opposing compartment. Our results suggest that much or all of the anti-proneural phenotype is due to the loss of wg expression, as N cells within 2-3 cell diameters of wg-expressing cells retained the

6 2818 E. J. Rulifson and S. S. Blair Fig. 3. SMC differentiation (as seen by neur A101 expression, labeled with anti-βgal) in marked N clones within the anterior wing blade region of late third instar wing discs. (A,A,A ) SMC differentiation in a N clone (arrow) in a N /πm; neur A101 wing disc, resulting from mitotic recombination at approximately 72 hours AEL. (A) N clones are identified by lack of πm expression (red; anti-c-myc). (A ) Expression of neur A101 (green); note the SMC hyperplasia within the N clone which intersects the wing margin proneural region (arrow). (A ) A double image of A and A ; the hyperplasia is limited to only N cells. (B) SMC hypoplasia within a N clone in a N /WG1296; neur A101 /+ disc, resulting from mitotic recombination at approximately 60 hours AEL. The disc is labelled with anti-β-gal (dark); the brightness associated with the N clone (dashed outline) is a result of counterstaining with fluorescent nuclear stain to emphasize the absence of WG1296. The dorsal clone clearly abuts the D/V compartment boundary; SMCs, recognized by more intense lacz expression, are for the most part lost from the clone and from the adjacent ventral margin proneural region. The SMCs seen at the tip of the clone (arrow) are likely within the clone. (C) A double image, marked as in A, showing a N clone resulting from mitotic recombination at approximately 72 hours AEL (arrow). This N clone has an intermediate phenotype that displays SMC hyperplasia in the anterior of the clone (top of clone) and SMC hypoplasia in the posterior of the clone. There is no apparent defect in the ventral margin SMCs adjacent to the clone. (D) A double image, marked as in A, showing a N clone resulting from mitotic recombination at approximately 72 hours AEL that cell autonomously causes SMC hyperplasia and the early differentiation SMCs in the margin proneural region, prior to the expression of neur A101 in adjacent proneural cells. Scale bar, 50 µm. neurogenic phenotype. We would expect, therefore, that clones that lay along the D/V boundary for shorter distances and thus caused smaller regions of disrupted wg expression should also show more neurogenic phenotypes, as now all the N cells will be close to wg-expressing cells. Examination of the triplelabeled discs suggest that this is indeed the case; clones that induced short bilateral disruptions in wg expression of less than 6 cell diameters often showed purely neurogenic phenotypes (Fig. 4E). Similarly, unilateral defects in SMC formation were rare. In only 2 clones, double labeled for neur A101 and the clone boundary, were anti-proneural defects observed in a fraction of the N cells without there being an associated loss of SMCs in the opposite compartment (Fig. 3C). These clones only contacted the D/V boundary for 2-3 cell diameters; while the degree to which wg expression was disrupted in these clones is unknown, it may be that enough wg expression was retained to induce normal SMC on that side. These observations (summarized in Fig. 7) support the hypothesis that N is required on the margin for two distinct processes. First, N is required within the zone of wg expression along the D/V boundary to establish or maintain wg expression. Since wg is in turn required for proneural gene expression on the margin (Phillips and Whittle, 1993; Couso et al., 1994), this loss of wg may be partially or wholly responsible for the antiproneural N phenotype. Second, N is required the properly patterned segregation of SMCs from the proneural region. Temperature shift analysis of N ts In order to clarify the temporal requirement for N, we used the temperature-sensitive allele, N ts1, and found comparable phenotypes to the mosaic analysis. Loss of N function during the later stages of the third instar resulted in excess SMC differentiation, a neurogenic phenotype. However, loss of N function at early to mid-third instar, before activation of the proneural genes, resulted in loss of SMCs on the presumptive wing margin. A previous study using N ts was unable to demonstrate a neurogenic N phenotype in the wing margin and other proneural regions of the wing blade (Hartenstein and Posakony, 1990). Under our slightly different conditions, however, strong neurogenic effects could be observed, both along the margin and elsewhere in the disc. Late third instar discs and early pupal N ts1 /N 55e11 ; neur A101 /+ wings were observed after a 12 hour upshift to the non-permissive temperature. In all cases we observed a massive overproduction of the margin SMCs, along with similar overproductions in other proneural regions of the wing disc (Figs 2C,5A; n=7). However, when late third instar and early pupal N ts1 /N 55e11 ; neur A101 /+ wing discs were examined after up-shifts for the previous 24 (n=25; Figs 2D,5B), 36 (n=4) and 48 (n=5) hours, we observed extreme SMC hypoplasia of the wing margin, despite the strong hyperplasia observed in other proneural regions of the disc. As shown above, up-shifts for the last 12 or 24 hours of

7 Notch and wingless in the fly wing 2819 Fig. 4. sc expression in marked N clones within the anterior wing blade region of late third instar wing discs. All clones shown resulted from mitotic recomination induced at approximately 60 hours AEL. N clones are marked by the absence of πm staining (red; A,A,C,C,D,E) or of anti-n staining (red; B,B ). Anti-sc staining is shown in green (A,A,B,B,C,D,E). (A,A,A ) N clones within the margin proneural region, not in contact with the D/V boundary, cell autonomously showed heightened levels of sc and delaminated as clumps of extra SMCs (arrows); compare to the double row of normal SMCs, identified by increased sc levels. (A ) The ectopic SMCs that appear yellow are N cells that have apparently moved basally, lying under the wild-type epithelium (they are in a different focal plane than the πmexpressing cells but appear not due to an artifact of the image overlay of A and A ). (B,B,B ) A dorsal N clone which abutted the D/V boundary; except for some sc expression in the bottom corner of the clone, sc was lost in most N cells (left arrow) as well as in ventral cells (right arrow in B ) adjacent to the N clone. (C) The wg-expressing cells (purple) were marked using vg enhancer-lacz (see Materials and Methods). (C ) Whether these N clones were in contact with the wg stripe (lower arrow), or were a few cell widths away from the stripe (upper arrow), they both displayed neurogenic phenotypes as in A. (D,E) N clones that intersected the wg stripe (marked with wg-lacz; shown in purple) and produced a loss of wg both in the clones and across the D/V boundary from the clones. (D) This clone shows a mixed phenotype of sc expression. A neurogenic phenotype was seen where N cells were within 2-3 cell widths of wgexpressing cells, yet a loss of sc and SMCs was seen within the interior of the clone (arrow). (E) A neurogenic phenotype in a N clone that intersected the wg stripe for about 6 cell widths. In this clone the N cells differentiating as SMCs were all within about 3 cell widths of wg-expressing cells. Scale bar, 50 µm. larval development were both sufficient to severely reduce wg expression. Thus, the different effects upon neurogenesis are likely due to the timing of proneural development. The normal margin pattern of sc is established at approximately 20 hours BP (Cubas et al., 1991), prior to the 12 hour upshift, and thus it is likely that the late loss of wg induced in such wings is not sufficient to disrupt sc expression or the subsequent formation of margin SMCs. However, loss of N and subsequent loss of wg before activation of sc leads to a failure of proneural development and differentiation of SMCs.

8 2820 E. J. Rulifson and S. S. Blair Fig. 5. Margin SMC differentiation in N ts wing discs after nonpermissive temperature shifts. (A,B) neur A101 was stained with anti-β-gal and visualized with Vector ABC staining (dark); (C) neur A101 was stained with anti-β-gal and fluorescently labeled (light). (A) Late third instar N ts /N 55e11 ; neur A101 /+ wing disc after a 12 hour shift. The wing margin showed extreme SMC hyperplasia along the margin (arrow) and in other proneural regions. (B) Late third instar N ts /N 55e11 ; neur A101 /+ wing disc after a 24 hour shift. Note the total absence of margin SMCs (arrow), even though other proneural regions show characteristic neurogenic defects. The cluster of SMCs marked V is a ventral group and not part of the anterior wing margin proneural region. (C) Late third instar N ts /N 55e11 ; neur A101 /HS-wg wing disc after a 1.5 hour heat shock at 38 C followed by 22.5 hour shift (24 hours total). Note that the SMCs of the margin were present and display the neurogenic phenotype of a later and shorter temperature shift (compare to A). Scale bar, 50 µm. Expression of wg rescues the N ts antiproneural phenotype To test whether the antiproneural effects with the N ts were due to the loss of wg function, we expressed wg independently of N function using the HS-wg construct. N ts1 /Y; HS-wg/+ males were crossed to virgin N 55e11 /+; neur A101 /+ females; the resultant female late third instar and early pupal wing discs were observed after having been temperature shifted for the previous 24 hours; a 1.5 hour 38 C heat shock at the beginning of this period was used to activate the HS-wg. N ts1 /N 55e11 ; neur A101 /? discs were recognized due to the strong neurogenic phenotypes observed in non-margin proneural regions. It is expected that approximately half of these should carry the HSwg construct. 12/23 of these discs showed antiproneural phenotypes on the margin identical to those observed above after a 24 hour temperature shift. However, in 10/23 cases the wing margin showed restoration of margin SMC development (Fig. 5C). Discs from earlier stages had scattered SMCs along the margin, whereas very early pupal discs showed nearly complete restoration of the neurogenic phenotype, producing supernumerary SMCs in the same general pattern as the margin proneural cluster (compare Fig. 5C to A). The variation in the degree of rescue may reflect variation in the timing of the HS-wg induction. Identically treated N ts1 /N 55e11 ; neur A101 discs which lacked HS-wg never displayed neurogenic defects on the margin. Therefore, we conclude that the exogenous expression of wg in these experiments was able to restore lost wg function, inducing SMC formation along the margin in the absence of N function. Cell death and the N phenotype It is clear that cell death plays a role in the adult loss-of-margin phenotypes observed in some N alleles (e.g. Jack and DeLotto, 1992). Some of the distortions in the D/V lineage boundary induced by N clones could also be caused by cell death. We were therefore interested in knowing to what extent cell death could account for the loss of wg expression observed in late third instar and early pupal stages. Obviously increased cell death was not observed in N ts1 /N 55e11 discs after 12 hour temperature upshifts (8 discs; Fig. 6A), despite the reduction of margin wg expression observed in all such discs (Fig. 2C). 36 hour upshifts did result in increases in the amount of cell death observed along and near the margin (2 discs; Fig. 6B); nonetheless, the region of cell death did not extend along the entire margin, despite the fact that identical upshifts removed the all margin wg expression (compare to similar phenotype of 24 hour upshift in Fig. 2D). Thus, the loss of wg does not appear to be caused by cell death alone. It is possible that some or all of the cell death observed along the margin after 36 hour upshifts was caused by the loss of wg expression. However, the increases in cell death were not limited to the regions of the disc normally expressing wg, suggesting some additional, wg-independent role for N in cell survival. DISCUSSION In this paper, we have provided evidence using null clones and a temperature-sensitive allele that N has a dual function in the

9 Notch and wingless in the fly wing 2821 A B B' wg AS-C SMCs N - Fig. 6. Acridine orange staining of cell death associated with the loss of N function and margin wg expression. Fluorescent labeling indicates cell death. (A) Late third instar N ts /N 55e11 ; neur A101 /+ wing disc following a 12 hour temperature shift. Note that there is very limited cell death on the wing margin (arrow), even though wg expression is dramatically reduced in such discs (Fig. 5C). (B) Late third instar N ts /N 55e11 ; neur A101 /+ wing disc after a 36 hour shift. Note that although there is substantial cell death throughout the disc, cell death on the wing margin (arrow) it is limited to the distal tip and does not extend along the entire margin, even though wg expression is completely lost in such discs (compare to 24 hour shift of N ts /N 55e11 ; neur A101 /+ wing disc in Fig. 5D). Fig. 7. Summary of clonal analysis of N. Color coding is same as Fig. 1 and as summarized below the illustration. N clones are shown as outlined white areas in a grey, heterozygous background. (A) Neurogenic defects outside the wg stripe. N clones that intersect the region of AS-C expression, without substantially intersecting the region of wg expression, cell autonomously induce the formation of supernumerary SMCs. Although not shown here, such clones also elevate AS-C levels in proneural cells (Fig. 4A,C) and induce the premature differentiation of SMCs (Fig. 2D). (B) Antiproneural defects. Both dorsal and ventral (not shown) N clones that lie on the D/V compartment boundary induce the cell-autonomous loss of wg. Such clones also lose AS-C expression and SMCs, except where N cells are within 2-3 cell widths of wg-expressing cells, in which case AS-C expression the overproduction of SMCs are often observed. Such clones also induce the domineering nonautonomous loss of wg and AS-C expression and of SMCs across the D/V boundary from the clone. (B ) Neurogenic defects within gaps in the wg stripe. N clones that contact the D/V boundary, but for short distances of about 6 cell widths, produce gaps in the wg expression just as in B. However these N clones display only neurogenic phenotypes as many N cells appear to receive the wg signal from cells expressing wg, 1-3 cell widths away. Some disruption of the proneural region and SMC pattern across the D/V boundary is associated with these clones and may likely be due to the reduction of wg signal in the region of the clone. process of wing margin neurogenesis. On the one hand, we were able to demonstrate a strong requirement for N in establishing or maintaining wg expression along the margin, as well as a less severe requirement in maintaining wg expression in other imaginal regions. N clones that overlapped the normal margin region of wg expression disrupted that expression, and this loss was associated with the loss of the wg-dependent margin proneural regions. On the other hand, N functions cell autonomously, after activation of the proneural genes, to maintain the non-smc state of proneural cells, as the loss of N from these cells led to a neurogenic overproduction of margin SMCs. We were unable, however, to demonstrate any requirement for N in the reception of the wg signal, since N cells adjacent to wg-expressing cells were not blocked or delayed in their proneural development. The loss of wg expression may be wholly responsible for the antiproneural N phenotype, since wg expression along the margin plays a critical role in the formation of margin cell types and patterns of gene expression. The wg signal appears to be sufficient to specify a margin-like fate in the wing blade: loss of the shaggy-zeste white 3 kinase mimics reception of the wg signal in the embryo (Siegfried et al., 1992), and results in the formation of margin-like cells (Simpson et al., 1988; Ripoll et al., 1988; Perrimon and Smouse, 1989; Blair, 1992b) and patterns of gene expression (Blair, 1992b, 1994) in the wing. Loss of wg expression along the margin results in the loss of proneural gene expression and SMCs (Phillips and Whittle, 1993; Couso et al., 1994). The loss of wg expression that we observed after the loss of N function (and its reduction in N hypomorphs) also provides an explanation for the recently observed synergistic interactions between hypomorphic N and wg alleles during wing margin development (Couso and Martinez Arias, 1994; Hing et al., 1994; Fig. 2). The loss of wg expression appears to be the critical difference

10 2822 E. J. Rulifson and S. S. Blair between the neurogenic and antiproneural N phenotypes, since the expression of wg using a HS-wg construct was sufficient to rescue the antiproneural phenotypes observed in N ts flies (Fig 5C). It is interesting that such ubiquitous expression did not result in the formation of margin SMCs throughout the wing, despite the unpatterned expression expected of the HS-wg construct. In these wing discs, the ectopic wg was expressed at much lower levels than those normally found along the margin (not shown); if there was some residual, temporary wg expression along the margin, it may be that the wg concentration only reached threshold levels in that region. Or it may be that cells near the margin are especially sensitive to the wg signal. N is not required for reception of the paracrine wg signal The similarities in N and wg phenotypes, as well as the synergistic interactions between N and wg hypomorphs, have suggested the possibility that N acted downstream of wg as part of the mechanism by which the wg signal was received by neighboring cells (Couso and Martinez Arias, 1994; Hing et al., 1994). However, if N was required for this process then all cells lacking N would be expected to lose or, at the least, be slowed in their acquisition of the proneural and SMC fates. In contrast, we have shown that cells within margin clones completely lacking N function may still express AS-C genes, as long as the clone does not significantly disrupt wg expression (Fig. 4A,C ). Moreover, such N clones induced a typical neurogenic phenotype from the earliest stages that could be assayed, producing an excess number of SMCs in advance of their phenotypically wild-type neighbors (Fig. 3D). This is in marked contrast to cells lacking downstream elements of the wg signalling pathway (for pathway, see Siegfried et al., 1993). Cells in the margin proneural region lacking armadillo, for instance, do not express AS-C genes (Blair, unpublished observations) or form bristles (Couso et al., 1993). Even in N clones showing an antiproneural phenotype, apparently caused by the loss of wg expression inside the clone, N cells at the edges of the clone could still express AS-C genes and make SMCs. It is likely that such cells are rescued because they are receiving the wg signal from cells outside the clone. The rescue of the N ts antiproneural phenotype with heat-shock-driven wg expression (Fig. 5C) further indicates that N is not required for reception of the paracrine wg signal. It is possible that, while not absolutely required, N may modulate or be modulated by reception of the wg signal. For example, it has recently been suggested that the dishevelled product, an element of the wg reception pathway, binds to and modulates the activity of N directly (Axelrod, Matsuno, Manoukian, Artavanis-Tsakonas and Perrimon, personal communication. One unexpected finding of our clonal analysis was the domineering non-autonomous effect of N clones on neighboring, N /+ cells. In our experiments most N clones were limited to either the dorsal or ventral lineage compartment of the wing. Since the lineage boundary lies in the center of the region of margin wg expression (Blair, 1993, 1994), these clones should not include wg-expressing cells or SMCs in the other compartment. Nonetheless, N clones that abutted the D/V boundary induced bilateral loss of wg and AS-C expression and of SMCs. There are obviously many ways in which N could be involved in setting up or maintaining wg expression along the margin, both autonomously and non-autonomously. In the following sections, we will present for discussion two different models. Model 1: N functions during dorsal-ventral transcompartmental induction The expression of wg along the margin straddles the D/V lineage compartment boundary (Blair, 1993, 1994). The D/V boundary, as defined by the dorsal-specific expression of the apterous (ap) transcription factor (Cohen et al., 1992; Bourgouin et al., 1992), appears at mid-second instar (Williams et al., 1993), well before the margin-specific expression of wg. It is thought that ap acts in a selector-like fashion, its expression (or lack thereof) controlling both the dorsoventral identity of wing cells and the separation of dorsal and ventral lineage compartments (Diaz-Benjumea and Cohen, 1993; Blair et al., 1994). Moreover, the boundary between cells that express and do not express ap plays a special role during wing development, as cells at normal and ectopic ap boundaries form margin-specific cell types (Diaz-Benjumea and Cohen, 1993) and express margin-specific genes (Williams et al., 1994), including wg (Blair, unpublished observations). One way of modeling the behavior of the ap boundary is to posit the existence of dorsal and ventral signals and receptors, which act across the compartment boundary to induce localized margin-specific specification of neighboring cell types. For instance, only ventral cells might be capable of responding to a dorsally expressed morphogen, and vice versa; or cells might require some critical levels of a combination of dorsal and ventral morphogens for margin-like differentiation. This type of transcompartmental induction is similar in principle to the hedgehog-mediated induction that acts across the anteriorposterior lineage boundary in the wing and leg (Basler and Struhl, 1994; Tabata and Kornberg, 1994), except that the induction must operate in two directions. In recent studies, it has been proposed that fringe (Irvine and Wieschaus, 1994) and Serrate (Couso, Carroll, Knust, and Martinez Arias, personal communication) products may act as dorsally expressed morphogens. If we imagine that N is somehow required for induction between the dorsal and ventral compartments, then cells at the boundary which lack N might no longer maintain their marginspecific identity, and thus would lose margin-specific wg expression, as we have observed. Such loss might also account for the distortions to the D/V boundary induced by some N clones. The putative role of dorsally expressed Serrate in D/V signalling (Couso, Carroll, Knust, and Martinez Arias, personal communication) is particularly intriguing in this regard, as it has been shown that N can bind the Serrate product in vitro (Rebay et al. 1991). However, it is unlikely that N could be acting as a ventral-specific receptor in transcompartmental induction, as we saw loss of wg expression in both dorsal and ventral clones (Fig. 2E ). Rather, N might be involved in induction in both directions, being required at the boundary for communication between cells on either side. If this was so, it could account for the loss of wg expression that we observe both within the N clone and in cells immediately across the D/V boundary. Model 2: N functions in the wg autoregulatory pathway We have shown that N is not required for cells adjacent to the