Antagonistic activities of Suppressor of Hairless and Hairless control alternative cell fates in the Drosophila adult epidermis

The generation of cell diversity among groups of equipotent cells is a fundamental aspect of the development of multicel-lular organisms. The Drosophila adult peripheral nervous system provides an excellent experimental system in which to address this problem genetically. Each external sense organ of the adult fly is composed of only a few cells that in most cases are the progeny of a single sensory organ precursor (SOP) cell (Hartenstein and Posakony, 1989). SOPs are determined during the late larval and early pupal stages within undiffer-entiated epithelial sheets, the imaginal discs, which will give rise to cuticular structures of the adult fly. In the first step of this determination process, the spatially restricted activities of the achaete (ac) and scute (sc) genes confer upon ‘proneural clusters’ of cells the competence to adopt the SOP fate (Cubas et al., 1991; Skeath and Carroll, 1991). Subsequently, one proneural cluster cell is singled out to become the future SOP, while the remaining cells of the cluster are inhibited from expressing this fate by local cell-cell interactions, referred to as lateral inhibition.

Proper singularization of the SOP requires the activity of multiple loci of the neurogenic group (Dietrich and Campos-ortega, 1984; Hartenstein and Posakony, 1990; Heitzler and Simpson, 1991; Parks and Muskavitch, 1993; Simpson, 1990), as well as the functions of the genes Hairless (H) (Bang et al., 1991) and Suppressor of Hairless [Su(H)] (Schweisguth and Posakony, 1992). H activity is required for the stable determi-nation of the SOP cell fate, and flies carrying null or strong hypomorphic alleles of H display an extensive ‘bristle loss’ phenotype (Bang et al., 1991). In addition, H acts as an antag-onist of neurogenic gene function (Vässin et al., 1985; A. Bang and J. W. P., unpublished results), and an excess of H activity causes phenotypes that closely resemble neurogenic gene loss-of-function phenotypes (Bang and Posakony, 1992). Con-versely, Su(H), a dominant modifier of H (Ashburner, 1982), acts as a neurogenic gene, in that it is required to limit the expression of the SOP fate among proneural cluster cells (Schweisguth and Posakony, 1992).

Following its stable determination, the SOP follows a stereo-typed pattern of division (Hartenstein and Posakony, 1989). The precursor cell for a typical mechanosensory bristle divides to yield two secondary precursor cells; one gives rise to the sensory neuron and thecogen cell, while the other gives rise to the trichogen and tormogen cells. The three non-neuronal cells form concentric sheaths around the dendrite of the neuron and produce the stimulus-receiving cuticular structures. In particu-lar, the trichogen and tormogen cells produce the bristle shaft and the socket that surrounds the base of the shaft, respectively. The adoption of unique cell fates by the four sensory organ cells requires the neurogenic genes Notch (N) (Hartenstein and Posakony, 1990) and Delta (Dl) (Parks and Muskavitch, 1993), as well as H. Partial loss of H function leads to a nearly complete transformation of the trichogen (shaft) cell into a second tormogen (socket) cell, yielding a ‘double socket’ phenotype (Bang et al., 1991; Lees and Waddington, 1942), while a gain of H function results in the converse transforma-tion and produces a ‘double shaft’ phenotype (Bang and Posakony, 1992). A reduction of Su(H) dosage suppresses the H double socket effect (Schweisguth and Posakony, 1992); however, a strict requirement for Su(H) activity in this late cell fate choice has not yet been demonstrated.

The H gene encodes a novel basic protein of unknown function (Bang and Posakony, 1992; Maier et al., 1992). Su(H) is the Drosophila homologue of the mouse J_κ–Recombination signal Binding Protein gene (J_κ–RBP) (Furukawa et al., 1992; Schweisguth and Posakony, 1992). The J_κ–RBP and Su(H) proteins are 82% identical over most of their length, and are likely to share an as yet unidentified biochemical activity involving sequence-specific DNA binding.

Flies were cultured on standard yeast-cornmeal-molasses-agar medium at 25°C. Mutant alleles of Suppressor of Hairless [Su(H); 2-50.5] and the Su(H) deficiency Df(2L)TE35BC-24 are described in Schweisguth and Posakony (1992) and in Lindsley and Zimm (1992). The 8xSu(H) line was derived by recombination and conventional crosses using three independent P[w⁺, Su(H)⁺ ] transposon insertion lines (C1, C3 and C11; Schweisguth and Posakony, 1992). The number of P transposon insertions in the 8xSu(H) line was verified by genomic Southern blot analysis (not shown). Mutant alleles of Hairless (H) are described in Bang et al. (1991), except for H^E31. The E31 allele was isolated in an F₁ screen for H alleles induced by the imprecise excision of the D179 P[w⁺, lacZ] enhancer trap transposon inserted in the 5′ untranslated region of the H coding sequence (Bang and Posakony, 1992). Genomic Southern blot analysis indicates that E31 corresponds to a small deficiency, from the w⁺ minigene through about two-thirds of the H protein coding sequence (unpublished results). For the Su(H)/H epistasis experiment, CyO and TM3 balancer chromosomes carrying a P[w⁺, actin-lacZ] transposon (kindly provided by J. Thomas) were used to balance the Su(H) and H mutant chromosomes. Male flies of the following genotype were analyzed in the FLP recombination experiment (Golic and Lindquist, 1989; Xu and Rubin, 1993): y w hsFLP1/Y; Su(H)^SF8A1-2-29 P[ry⁺; hs-neo; FRT]40A/P[ry+; y+]25F P[ry +; hs-neo; FRT]40A. These were F₁ progeny of a cross between y w hsFLP1; P[ry⁺; y⁺]25F P[ry⁺; hs-neo; FRT]40A females and w¹¹¹⁸; Su(H)^SF8A1-2-29 P[ry⁺; hs-neo; FRT]40A/CyO males [the enhancer trap insertion A-1-2-29 (see below) is fully viable when homozygous, and was used in this exper-iment only as a w⁺ recombination marker]. Flies of the genotype w¹¹¹⁸ were used as wild-type controls. The enhancer trap transposon inser-tions P[ry⁺, lacZ]A37 (kindly provided by Y. Hiromi and C. O’ Kane), P[ry⁺, lacZ]A101 (kindly provided by H. Bellen), and P[w⁺, lacZ]A-1-2-29 (kindly provided by Y. N. Jan) were used as markers for sensory organ precursor cells and their progeny (Bang et al., 1991; Ghysen and O’Kane, 1989; Hartenstein and Jan, 1992; Huang et al., 1991). An achaete-lacZ fusion gene (Van Doren et al., 1992) and the scabrous P[w⁺, lacZ] enhancer trap transposon insertion A2-6 (Mlodzik et al., 1990) were used as proneural cell markers. Mutations and chromosomes not described herein are described in Lindsley and Zimm (1992).

Trans-heterozygous Su(H) and H mutant larvae are identified unam-biguously using the dominant Black cells (Bc) and Tubby (Tb) mutations (Bang et al., 1991; Schweisguth and Posakony, 1992). In some cases, mutant genotypes were also confirmed by observing the characteristic wing pouch phenotypes associated with H and Su(H) mutations (see Fig. 5). In the Su(H)/H epistasis experiment, imaginal discs from double mutant larvae were identified as those lacking the strong β-galactosidase activity staining that results from the expression of the actin-lacZ fusion gene carried by the balancer chro-mosomes.

A 2.8 kb HindIII/XbaI full-length Su(H) cDNA fragment was isolated from plasmid pKS12 (Schweisguth and Posakony, 1992). The HindIII terminus was converted into an XbaI site by filling in using the Klenow polymerase, followed by the addition of a phosphorylated XbaI linker (NEB). The resulting XbaI/XbaI fragment was then cloned into the CaSpeR-Hsp70 transformation vector (Bang and Posakony, 1992) at the unique XbaI site. The resulting P[w⁺, Hs-Su(H)] trans-posable element was introduced into the germline of w¹¹¹⁸ recipient embryos by coinjection with a Δ2-3 helper plasmid (Rubin and Spradling, 1982). Eleven stocks, designated P[Hs-Su(H)]−1 to −11, were established from independent w⁺ G1 adults.

Staged pupae were placed in a humid chamber and subjected to heat shock using a precise temperature-controlled water bath. Heat-shocked animals were then returned to 25°C and allowed to develop. Pharate adults (manually removed from the pupal case) and newly eclosed adults were dissected and prepared for light microscopy as described (Bang and Posakony, 1992).

Histochemical staining for β-galactosidase activity was carried out as described by Romani et al. (1989).

Immunodetection of the ac protein was carried out according to a protocol provided by J. Skeath. Imaginal discs were dissected in phosphate-buffered saline (PBS), fixed overnight in 2% formalde-hyde in ‘Brower fix’ (100 mM Pipes, 2 mM MgSO₄, 1 mM EGTA, 0.1% NP40, pH 6.9), and then washed in PBS containing 0.3% Triton X-100 (PBS-T). The tissue was incubated overnight with the primary antibody (Skeath and Carroll, 1991) diluted at 1:50 in PBS-T, washed in PBS-T over a 1-hour period, incubated for 90 minutes with the biotinylated anti-mouse secondary antibody (Vector) diluted at 1:200 in PBS-T and then washed. All fixation, incubation and wash steps were at 4°C. Peroxidase activity staining was carried out using the Elite kit (Vector) at room temperature over a period of a few hours. Immunostaining using the mouse monoclonal antibodies 22C10 [a specific marker for SOP progeny cells (Hartenstein and Posakony, 1989)] and 202 (anti-Delta) was performed as follows. Dissected tissues were fixed for 20 minutes in 4% paraformaldehyde in PBS at room temperature. The monoclonal primary antibody was diluted at 1:100 and incubated overnight at 4°C or 2 hours at room tempera-ture. Secondary antibody reaction and peroxidase staining were performed using the anti-mouse Elite kit (Vector). For the rabbit polyclonal anti-HRP serum (Cappel), a 1:1000 dilution was used. Anti-rabbit secondary antibody coupled to alkaline phosphatase (Biosys) was used at a 1:1000 dilution. β-galactosidase staining was carried out as described above immediately following the immunos-taining reaction.

Loss-of-function alleles of Su(H) are potent dominant sup-pressors of the H ‘double socket’ phenotype (Schweisguth and Posakony, 1992; unpublished results), suggesting that reduction of Su(H) activity favors the trichogen (shaft) cell fate. Consistent with this interpretation, ‘double shaft’ macrochaetes are occasionally found on flies trans-heterozy-gous for the Su(H) hypomorphic allele HG36 and a null allele (Fig. 1A). We have directly investigated the role of Su(H) in the trichogen/tormogen cell fate choice by generating somatic clones of homozygous Su(H) mutant cells in an otherwise het-erozygous fly, using the FLP/FRT method developed by Golic and Lindquist (1989) and Xu and Rubin (1993). Briefly, the yeast FLP recombinase, expression of which is driven by the Drosophila Hsp70 promoter, catalyzes site-specific recombi-nation between two FRT sites inserted at the base of chromo-some arm 2L, at cytological position 40A (Xu and Rubin, 1993). One of the two FRT-bearing chromosomes is also mutant for Su(H) (at 35B9-10); the null allele SF8 was used in this study. The other homologue carries a P element transpo-son that includes a wild-type copy of the cuticular marker gene yellow (y) (inserted at 25F; Xu and Rubin, 1993), and the experiment is carried out in a y⁻ genetic background. Upon heat induction of FLP-mediated recombination, y⁻Su(H)⁻ mutant clones are generated, and their phenotypes on the adult cuticle can be analyzed.

Two bristle phenotypes are observed in the mosaic flies. When recombination is induced in first and second instar larvae, patches of naked cuticle are found (the cellular defects associated with this phenotype will be presented elsewhere). Recombination events induced later in development (i.e., in third instar larvae) predomi-nantly yield double-shaft bristles. All these double-shaft bristles are y⁻, indicat-ing that they are homozy-gous mutant for Su(H). Double-shaft bristles may be found at every head and notum macrochaete position (Fig. 1B), and at microchaete positions on the notum and abdomen (Fig. 1C). They almost always appear individually, except on the abdomen, where clones containing several double-shaft microchaetes may be observed (Fig. 1C). Thus, it is likely that this ‘double shaft’ phenotype is primarily associated with late-arising somatic clones of Su(H) mutant cells. We conclude that a reduction in Su(H) activity may result in the transformation of the tormogen (socket) cell into a second trichogen (shaft) cell, just as is observed with an excess of H activity (Bang and Posakony, 1992).

In the complete absence of Su(H) function, most or all cells in imaginal disc proneural clusters express the SOP-specific markers A37 and A101, indicating that, like certain genes of the neurogenic group, Su(H) is required early in adult sensory organ development to restrict the expression of the SOP fate (Schweisguth and Posakony, 1992). However, it has not been determined which other characteristics of wild-type SOPs are exhibited by these Su(H) mutant proneural cluster cells.

One distinguishing feature of normal SOPs is that they accu-mulate a higher level of proneural proteins than the other cells in the proneural cluster (Skeath and Carroll, 1991). Using an anti-ac monoclonal antibody (a generous gift of J. Skeath and S. Carroll; Skeath and Carroll, 1991), we have found that, in Su(H) mutant imaginal discs, most or all of the cells in the proneural clusters accumulate ac protein at a high level (Fig. 2A). These clusters of strongly ac-positive cells appear identical in size and location to the clusters shown previously to express the SOP-specific marker A101 (Schweisguth and Posakony, 1992).

The elevated accumulation of ac protein in Su(H)⁻ proneural clusters is accompanied by a high level of proneural activity in these cells. An ac-lacZ fusion gene that is expressed under the direct positive control of ac and sc in wild-type discs (Van Doren et al., 1992) is activated at a high level in Su(H) mutant proneural clusters (Fig. 2B). The enhancer-trap marker A2-6, inserted at the scabrous (sca) gene, is expressed specifically in the proneural clusters of wild-type discs, and this expression is largely dependent on ac and sc (Mlodzik et al., 1990; data not shown). Moreover, the single SOPs that arise from wild-type proneural clusters exhibit an elevated level of A2-6 expression relative to the remaining cells of the cluster (Mlodzik et al., 1990). Fig. 2C shows that A2-6 is strongly expressed in most or all cells in Su(H)⁻ proneural clusters, further indicating that proneural activity is high in these cells.

Finally, we examined the expression of the neurogenic gene Dl, using an anti-Dl antibody (mAb 202, generated in the lab- oratory of S. Artavanis-Tsakonis and kindly provided by M. Muskavitch). High levels of accumulation of the Dl protein are detected in most Su(H) mutant proneural cluster cells (Fig. 2D).

We conclude that high levels of ac protein accumulation and proneural gene activity are found together with high-level expression of the neurogenic protein Dl in all, or most, proneural cluster cells adopting the SOP fate in Su(H) mutant wing discs. Recent data from our laboratory indicate that loss of H function has the opposite effect: In H⁻ imaginal discs, the single presumptive SOP cell fails to maintain high levels of ac protein and proneural regulatory activity (A. Bang and J. W. P., unpublished data). These results support the conclusion that H and Su(H) have opposing functions in controlling the expression of the SOP fate (Bang et al., 1991; Bang and Posakony, 1992; Schweisguth and Posakony, 1992).

The dominant modification of H mutant phenotypes by both gain and loss of Su(H) function led Ashburner (1982) to suggest that Su(H) acts as a negative regulator of H activity. The molecular cloning of the Su(H) gene (Schweisguth and Posakony, 1992) allows us to test whether an increase in Su(H) activity in an otherwise wild-type fly may lead to phenotypes similar to those associated with H loss-of-function mutations.

Previously, we described a P element transposon carrying a Su(H) genomic DNA fragment that rescues all aspects of the Su(H) null phenotype in transformed flies, and provides Su(H)⁺ activity at a level quantitatively similar to the endogenous gene (Schweisguth and Posakony, 1992). We have established a transformant line that is homozygous for three copies of this transposon; this line thus carries the equivalent of 8 Su(H)⁺ doses per diploid genome and is called 8xSu(H). As shown in Fig. 3A, many bristles on the head and thorax of 8xSu(H) adult flies exhibit a ‘double socket’ phenotype similar to that observed in H hypomorphic mutants. In addition, specific macrochaetes fail to appear; these correspond to those that are most sensitive to loss of H function (e.g., postvertical and humeral bristles; see Bang et al., 1991). The finding that Su(H) hyperactivity is more effective in interfering with trichogen cell differentiation than with expression of the SOP fate is con-sistent with the greater sensitivity of the trichogen/tormogen decision versus the SOP/epidermal decision to loss-of-function mutations in H (Bang et al., 1991).

We further investigated these cell fate changes using an inducible system for overexpression of Su(H). A Hs-Su(H) fusion gene, consisting of a full-length Su(H) cDNA under the control of the Hsp70 heat-shock promoter, was introduced into the genome by P element-mediated germline transformation. The resulting P[Hs-Su(H)] transformant lines do not generally exhibit detectable bristle defects in the absence of heat-shock treatment (Fig. 3B). One interesting exception is line P[Hs-Su(H)]-8, which in heterozygous condition exhibits a fully penetrant ‘double socket’ phenotype at macrochaete positions in the posterior part of the notum (Fig. 3E). We applied a 6-hour heat-shock regimen (3×60 minutes at 37°C, separated by two 90-minute intervals at 25°C) to P[Hs-Su(H)] pupae at 16, 20 or 24 hours APF; i.e., during and following the division of microchaete precursor cells but well after the macrochaete cells have begun their differentiation (Hartenstein and Posakony, 1989). The resulting pharate adults exhibit a bristle phenotype very similar to the H double socket phenotype at most notum microchaete and macrochaete positions (Fig. 3C). This effect is still detectable when the heat shock is commenced at 26 hours APF, 10 hours after the division of the microchaete secondary precursors. Such late heat shocks cause many microchaete shafts to be shortened and thickened, with socket-like structures at their base (data not shown). No effect on either macrochaete or microchaete differentiation is detectable with heat shocks applied after 32 hours APF. Thus, the critical developmental period during which Su(H) overexpression is able to produce the double socket effect in microchaetes (16-32 hours APF) is in good agreement with the period of microchaete accessory cell differen-tiation (Hartenstein and Posakony, 1989).

The cellular basis of the double socket defect was examined in pupal nota of the P[Hs-Su(H)]-8 transfor-mant line (Fig. 3E). At wild-type macrochaete positions, using mAb 22C10 as a marker in 24 hours APF pupae, the trichogen cell is detected as a strongly stained, subepidermal, polyploid cell, while the tormogen cell lying above it appears more faintly stained (Hartenstein and Posakony, 1989; Fig. 3F). At double socket macrochaete positions in the same pupae, by contrast, mAb 22C10 detects two and only two lightly stained, polyploid cells in the same epidermal plane (Fig. 3F′). This indicates that the trichogen cell is transformed into a second tormogen cell. We also used the enhancer trap insertion P[w⁺, lacZ]A-1-2-29 as a specific marker for the trichogen and tormogen cells (Hartenstein and Jan, 1992). In wild-type pupae at 40 hours APF, two β-galactosidase-positive polyploid nuclei are detected at the pSC position; the tormogen cell nucleus appears in the epidermal plane (Fig. 3G), while the trichogen cell nucleus is located subepidermally (Fig. 3G′). In P[Hs-Su(H)]-8 pupae, again only two β-galactosidase-positive polyploid nuclei are observed at the ‘double socket’ pSC position; however, these two nuclei are in the same (epidermal) plane (Fig. 3H). Finally, a single neuron is revealed by an anti-HRP polyclonal antiserum at the pSC position in both wild-type and P[Hs-Su(H)]-8 pupae (Fig. 3G′,H′). These results indicate that the P[Hs-Su(H)] ‘double socket’ phenotype results from the specific transformation of the shaft-producing subepider-mal trichogen cell into a second socket-producing epidermal cell (tormogen), while the sensory neuron/thecogen half of the lineage does not seem to be affected.

An earlier, shorter heat shock (90 minutes at 36.5°C) applied at 6 hours APF, prior to the first division of the microchaete SOPs (Hartenstein and Posakony, 1989), results in the loss of nearly all notum microchaetes (Fig. 3D). Macrochaetes appear to be unaffected by this treatment, consistent with the earlier specifi-cation of macrochaete SOPs in late third instar larvae and early pupae (Hartenstein and Posakony, 1989; Huang et al., 1991). We determined the cellular basis of the microchaete loss phenotype by following the development of microchaete SOPs in P[Hs-Su(H)] flies using the A101 enhancer-trap marker (Fig. 4). A101/P[Hs-Su(H)]-1 and control A101/TM6 pupae were heat shocked at 6 hours APF (90 minutes at 36.5°C), and their dissected nota stained for β-galactosi-dase activity at 14 or 24 hours APF. At 14 hours APF, nearly all microchaete SOPs of A101/TM6 pupae are A101-positive, while only a few weakly stained cells are detected in A101/P[Hs-Su(H)]-1 pupae (Fig. 4A,C). At 24 hours APF, again only a reduced number of A101-expressing cells are observed at microchaete positions in P[Hs-Su(H)]-1/A101 pupae compared to the full pattern of bristle cells in A101/TM6 individuals (Fig. 4B,D). We conclude that, by the criterion of A101 expression, the microchaete loss phenotype observed in P[Hs-Su(H)] adults most likely results from a failure of SOP determination. Thus, both loss of H function (Bang et al., 1991) and hyperactivity of Su(H) lead to adult bristle loss by interfering with the expression of the SOP cell fate.

The finding that both loss- and gain-of-function mutations of H and Su(H) have opposite effects on the SOP/epidermal and trichogen/tormogen cell fate decisions raises the question of their epistatic relationship. To address this issue, we examined SOP deter-mination in the imaginal discs of late third instar larvae carrying null mutations in both genes, using the A37 enhancer-trap as an SOP marker. In imaginal discs, A37 is specifically expressed in SOP cells and their progeny (Bang et al., 1991; Ghysen and O’Kane, 1989; Huang et al., 1991; see Fig. 5A). Single A37-expressing cells are occasionally detected in H mutant wing discs at the dorsal radius (dR) and posterior scutellar (pSC) positions, and at the prospective anterior wing margin (Fig. 5B; see also Bang et al., 1991). By contrast, five large clusters of A37-positive cells are consistently observed in Su(H) mutant wing discs at positions where single SOPs normally arise (Fig. 5C). These clusters appear to correspond to the ventral radius sensillum (vR), the second campaniform sensillum of vein L3 (L3-2), the giant sensillum of the radius (GSR), the dorsal radius sensillum (dR), and the posterior postalar bristle (pPA). Faint β-galac-tosidase staining is sometimes detected at the pSC cluster position. In the Su(H); H double mutant, four to six clusters of A37-expressing cells are reproducibly observed (Fig. 5D). These include positions where lacZ activity can be detected in the H mutant disc (dR and pSC), as well as positions where lacZ expression is observed in Su(H) but not H mutant discs (vR, L3-2, GSR, pPA). In these latter positions, however, A37-positive cells appear significantly reduced in number in the double mutant. It appears, first, that cells that normally strictly require H function to express the A37 marker may adopt the SOP fate in the absence of both H and Su(H) activity; and second, that some of the cluster cells that would express the SOP fate in a Su(H) mutant disc apparently fail to do so when H function is absent as well. This interpretation of our results indicates that no strict epistatic relationship can be established between H and Su(H) for the specification of the SOP fate.

Two novel effects of Su(H) mutant conditions on the differen-tiation of adult mechanosensory bristles are described in this study: A loss-of-function ‘double shaft’ phenotype, in which the bristle develops with two shafts and no socket, and a gain-of-function ‘double socket’ phenotype, in which two sockets appear at the expense of the shaft. These differentiative defects appear to result, respectively, from the transformation of the tormogen (socket) cell into a second trichogen (shaft) cell, and the converse.

The opposite phenotypic effects have been described for H: A partial loss of H⁺ activity results in the development of many double-socket bristles, while H overexpression produces double-shaft bristles (Bang et al., 1991; Bang and Posakony, 1992). Taken together, these data indicate that a high level of Su(H) activity or a low level of H promote the socket fate, while, conversely, a low level of Su(H) or a high level of H favor the shaft fate. The finding from these and earlier studies (Bang and Posakony, 1992) that the trichogen/tormogen cell fate decision may be altered well after the birth of these cells indicates a continuing requirement for antagonistic Su(H) and H activities during their differentiation.

As bristle differentiation proceeds, the cell body of the trichogen comes to lie beneath that of the tormogen (Harten-stein and Posakony, 1989; Lees and Waddington, 1942). This rearrangement is similar in part to the delamination of embryonic neuroblasts; shaft-producing cells and neuroblasts both loosen their contact with the apical surface to move beneath the socket-producing and epidermal cells, respec-tively. We suggest that the choice between the trichogen and tormogen cell fates shares significant similarities with other neurogenic decisions, both in its genetic requirements (Posakony, 1994) and in its possible association with changes in cell-cell adhesion (Hartenstein et al., 1992). In this respect, it is interesting to note that ‘double shaft’ and ‘double socket’ phenotypes have both been observed in specific N loss-of-function (de Celis et al., 1991) and gain-of-function (see Fig. 5D of Rebay et al., 1993) conditions, respectively.

We have shown here that in a Su(H) mutant wing disc, high levels of ac protein accumulation and of proneural regulatory activity are detected in the same cluster pattern that we described earlier for the expression of the SOP-specific marker A101 (Schweisguth and Posakony, 1992). We have further observed a high level of Dl protein accumulation in the identical cluster pattern in Su(H) mutant discs. This result is consistent with recent data indicating that wild-type SOPs exhibit elevated Dl expression (A. Bang, J. Kavaler, and J. W. P., unpublished observations; A. Parks and M. Muskavitch, unpublished observations). Thus, several important character-istics of normal SOPs are shared by clusters of cells in Su(H) null imaginal discs, strongly supporting our earlier conclusion that a complete loss of Su(H) function results in the commit-ment of all, or most, imaginal disc proneural cluster cells to the SOP fate (Schweisguth and Posakony, 1992).

Somatic mosaic studies in adult flies indicate that Dl acts non-cell-autonomously and is an essential component of the inhibitory signal that prevents the determination of more than one SOP cell within a proneural cluster (Heitzler and Simpson, 1991). Moreover, higher Dl activity in a cell appears to confer a higher capacity to inhibit neighboring cells (Heitzler and Simpson, 1991). Accordingly, the simultaneous presence of high levels of ac and Dl in multiple neighboring cells in Su(H) proneural clusters might be viewed as somewhat paradoxical. One simple interpretation is that Su(H) is normally required for proneural cluster cells to respond to the Dl-mediated inhibitory signal, so that, in the absence of Su(H) function, this signal becomes ineffective at antagonizing proneural gene expression and function.

Several lines of evidence firmly support our conclusion that H and Su(H) encode antagonistic activities that have opposite effects on the SOP versus epidermal cell fate decision in imaginal discs. First, H and Su(H) have opposite null pheno-types with respect to SOP development. Second, both loss of H function and overexpression of Su(H) cause adult bristle loss and, in both cases, the developmental basis for this phenotype is the failure to specify and/or execute the SOP cell fate. Third, bristle loss resulting from Su(H) overexpression is strongly enhanced by reduction of H function (unpublished results).

Our phenotypic analysis of Su(H); H double mutant larvae indicates that proneural cluster cells may adopt the SOP or epidermal fates in the absence of both functions. Thus, in Su(H) mutant wing discs, some cells require H⁺ activity to express the SOP fate, while other cells do not. Similarly, in the absence of H⁺ activity, Su(H)⁺ function is required for some, but not all, cells to express the epidermal fate. These findings suggest that Su(H) and H act primarily to stably establish alter-native cell fates, rather than to specify cell identities initially. The lack of a clear epistatic relationship between H and Su(H) for SOP determination further implies that these two genes do not act sequentially in a strictly linear genetic pathway con-trolling this process.

Considering that H and Su(H) transcripts appear ubiqui-tously distributed in the third instar wing imaginal disc (Bang and Posakony, 1992; Schweisguth and Posakony, 1992), both proteins may be present in all proneural cluster cells. It is formally possible that H and Su(H) are constitutively active in all the cells of the cluster, and that alterations in their relative levels only secondarily affect an asymmetry established inde-pendently. Alternatively, the activities of the H and/or Su(H) proteins may be spatially regulated within the proneural cluster in response to a lateral inhibitory signal from the presumptive SOP cell. We have previously suggested that H acts within the presumptive SOP to make it resistant to inhibitory signaling by its neighbors in the proneural cluster (Bang and Posakony, 1992). It is possible that Su(H) is activated in the non-SOP cells of the cluster; this could lead to the specific inhibition of H activity in these cells, which would thus be prevented from stably adopting the SOP fate.

We are very grateful to M. Ashburner, H. Bellen, E. Bier, Y. Hiromi, Y. N. Jan, J. Roote, G. Rubin, J. Thomas, and T. Xu, and the Bloomington Stock Center, for fly stocks. We thank S. Artavanis-Tsakonas, S. Benzer, S. Carroll, M. Muskavitch and J. Skeath for gen-erously providing us with antibodies. We thank A. Bang and all the members of the lab for their help in the course of this work, and A. Bang, A. Bailey, M. Leviten and F. Maschat for critically reading the manuscript. This work was supported by the CNRS, a Lavoisier fel-lowship from the Ministère des Affaires Etrangères, and a postdoc-toral fellowship from the Human Frontier Science Program (F. S.), and by NIH grant GM46993 to J. W. P.