Suppressor of Hairless-independent events in Notch signaling imply novel pathway elements

The Notch (N) signaling pathway defines an evolutionary conserved cell interaction mechanism that controls the pro-gression of precursor cells into a more differentiated state. Genetic and molecular analyses have identified components of this pathway, which include a transmembrane receptor N, two transmembrane ligands Delta (Dl) and Serrate (Ser), a cyto-plasmic protein Deltex (Dx), and nuclear components including a transcription factor Suppressor of Hairless (Su(H)) and basic helix-loop-helix (bHLH) genes encoded by the Enhancer of split (E(spl)) locus (reviewed in Artavanis-Tsakonas et al., 1995).

Genetic analyses have indicated that Su(H) behaves as a downstream effector of Notch signaling, while biochemical studies have demonstrated that it directly interacts with the intracellular domain of Notch (Fortini and Artavanis-Tsakonas, 1994; Tamura et al., 1995). The expression of at least some of the bHLH proteins encoded by E(spl) has been shown to depend on Notch signaling and the binding of Su(H) on their regulatory sequences (Jennings et al., 1994; Lecourtois and Schweisguth, 1995; Bailey and Posakony, 1995). It is not known whether Notch uses effectors other than Su(H) since, so far, only Su(H)-dependent Notch activities have been identified. A possible exception is implied by the finding that, while the Drosophila mesoderm formation and the expression of the single-minded (sim) gene in the midline depends on Notch, Su(H) loss-of-function mutations do not affect sim expression or mesoderm formation (Lecourtois and Schweisguth, 1995). A biochemical study involving Notch signaling in mammalian tissue culture cells has also raised the possibility of additional effector molecules (Shawber et al., 1996).

Several independent studies have implicated the region of the Notch intracellular domain composed of six tandem cdc10/ankyrin repeats (ANK repeats) in Notch signaling (Rebay et al., 1993; Lieber et al., 1993; Shawber et al., 1996; Roehl et al., 1996). Protein-binding assays involved this Notch domain in interactions with both Su(H) as well as with Dx, which acts as a positive regulator of the pathway (Fortini and Artavanis-Tsakonas, 1994; Diederich et al., 1994; Matsuno et al., 1995). The existence of a tandem ANK repeat motif has been documented in a very large number of proteins, but its general biochemical and biological properties, as well as its role in Notch signaling, are not well understood (Bork, 1993). Here we examine directly the biochemical and genetic activ-ities of the Notch ANK repeats. These sequences are found to mediate homotypic interactions. Moreover, the analysis of a mutation affecting the ankyrin repeats implicates ANK-mediated interactions in modulation of Notch receptor activity. In vivo analysis unravels a hitherto unidentified biological activity associated with the ANK repeats. Molecular markers demonstrate that this activity affects Notch signaling through Su(H)-independent, non-autonomous events suggesting the possible existence of novel Notch pathway elements.

A Notch cDNA encoding amino acids 1890-2108 (Wharton et al.,1985) downstream of a translation initiation codon from the actin 79B gene was cloned into the sev promoter vector. A fragment including the 3′ untranslated portion of the Drosophila alcohol dehydrogenase gene was inserted downstream of the Notch cDNA. Germline trans-formation was performed using standard procedures (Rubin and Spradling, 1982). Two independent transformant lines showing dominant rough eye phenotypes were used for further analyses. Compared to sev-N^act or sev-N^nucl lines, sev-ANKlines showed weaker rough eye phenotypes (Fortini et al., 1993). Antibody staining was carried out as described previously (Fortini et al., 1993). To detect the ANK repeat peptide, mouse monoclonal antibody C479-2B, which recognizes an epitope within the Notch ANK repeats, was used (R. G. Fehon, I. Rebay and S. Artavanis-Tsakonas, unpublished data).Mouse monoclonal antibody, mAb174, a generous gift from Sarah Bray (Cambridge University), was used to detect E(spl) Mδ (Jennings et al., 1994). Scanning electron microscopy, sectioning and staining of eyes were performed as described previously (Fortini et al., 1993).

Constructs designed to produce fusion proteins with a DNA-binding domain of LexA and with an acidic transcription activation domain are referred to as pLEX and pJG, respectively (Zervos et al., 1993). The interaction trap and β-galactosidase assays were performed as described previously (Matsuno et al., 1995). Interaction was monitored by measurement of β-galactosidase activity in liquid cultures grown in galactose. Basal activity was measured when cells were grown in the equivalent glucose medium. Assays of β-galac-tosidase activity were performed on three or six independent trans-formants. pLEXICN1, pJGICN1, pJGICN2, pLEXANK1-5, pJGANK1-5, pLEXICN1DANK, pJGICN1DANK, pJGICN1827-2109, pJGICN1827-2076, pJGICN1827-1996, pJGICN1827-1963, pJGICN 1827-1921, pJGICN1846-2076, pJGICN1889-2076, pJGANK3-5, pJGANK5, pJGANK1, pJGSTOP and pJGdx are described previously (Matsuno et al., 1995). pLEX-Hairless and pJGH-Hairless contains an entire coding region of Hairless cDNA in EcoRI site of pEG202 and pJGSTOP, respectively. The Hairless clones were a generous gift of Anette Preiss (University of Freiburg). pJGhN-1ANK and pJGhN-2ANK contain cDNAs encoding ANK repeats regions of human Notch-1 (aa 1826-2147) and Notch-2 (aa 1772-2084), respectively, in the EcoRI site of pJGSTOP vector (Ellisen et al., 1991; Stifani et al., 1992). pLEXCactus and pJGCactus contain a cDNA encoding ANK repeats region of Drosophila Cactus protein (aa 173-500) in EcoRI sites of pEG202 and pJGSTOP vectors, respectively (Kidd, 1992). pLEXICN1Su42c contains a mutant ICN1 (aa1827-2259) cDNA which has a base substitution from C to A at 6920 (Diederich et al., 1994). pJGICN11769-2259, pJGSu(H)BS-2 and pJGANK1-6 contain parts of Notch cDNAs encoding aa 1796-1825, 1769-1825 and 1891-2113, respectively, in the EcoRI site of pJGSTOP. Fragments described above were generated by PCR or isolated by restriction digestion.

NTMICN1Myc consists of Notch peptide amino acid 1-107 and 1702-2157 (amino acid numbers according to Wharton et al., 1985). DlTMICN1Myc encompasses Dl amino acids 1-33 and 580-633 as well as Notch amino acids 1791-2157 (amino acid numbers according to Kopczynski et al., 1988). NTMMyc encompasses Notch amino acids 1-107 and 1702-1792. DlTMMyc encompasses Delta amino acids 1-33 and 580-633. Co-localization assays were performed as described previously (Diederich et al., 1994; Fortini and Artavanis-Tsakonas, 1994; Matsuno et al., 1995). Transfected cells were double-labeled with anti-Myc mouse monoclonal antibody and rat anti-Notch extracellular domain polyclonal antibody, Rat-3 (R. G. Fehon, I. Rebay and S. Artavanis-Tsakonas, unpublished data).

The CAT reporter construct and the conditions of the transcription assay (Fig. 4) were described in Eastman et al. (1997).

Truncations of the extracellular domain of Notch result in the constitutive activation of the Notch receptor, triggering the expression of downstream genes in the E(spl) complex (Jennings et al., 1994; Sun and Artavanis-Tsakonas, 1996). The expression of such constructs in the developing eye under the cell-specific promoter sevenless (sev) have proven to be par-ticularly informative in the analysis of Notch signaling (Fortini et al., 1993; Verheyen et al., 1996). We sought to examine the biological activity of the Notch ANK repeats by a similar approach. To that end, sev-driven constructs carrying the ANK repeats were expressed in the developing eyes of transgenic flies and the phenotypes were compared to those induced by the expression of activation forms of Notch (Fig. 1).

The expression of the entire intracellular domain of Notch under the sev promoter (sev-N^act and ^sev-Nnucl in Fig. 1) in the developing eye disc results in specific cell fate changes. In a wild-type eye disc, each ommatidial cluster has four cone cells (Fig. 2D). In sev-N^act discs, clusters containing five cone cells are observed (Fig. 2E). This phenotype is consistent with the idea that this truncated form of Notch inhibits the R7 precursor (in which the sev promoter is active) that acquires R7 fate and, instead, it turns into the additional cone cell (Fortini et al., 1993). In the same disc, the earlier and transient expression of sev-N^act in R3 and R4 precursors results in the transformation of these cells into two R7 cells (Fig. 2K and Fortini et al., 1993). As a result, a rough eye phenotype is observed (Fig. 2 and Fortini et al., 1993).

Further trimming of the intracellular Notch sequences resulted in a construct, sev-ANK, which permitted the expression of only the ANK repeat region (aa1890-2108, Fig. 1). Expression of this construct in the eye gave a rough eye phenotype distinct from that associated with sev-N^act expression (Fig. 2C). sev-ANKeyes have ommatidial clusters lacking one or two cone cells (Fig. 2F) and an adult eye pho-toreceptor phenotype lacking R7 (Fig. 2L).

The transcription factor Su(H) has been established as an effector of Notch since the expression of bHLH genes encoded by E(spl) depends on both Notch signaling and the presence of Su(H)-binding sites in the promoters (Jennings et al., 1994; Lecourtois and Schweisguth, 1995; Bailey and Posakony, 1995). Truncation of the extracellular domain of Notch results in an activated form of Notch as demonstrated by the fact that N^act and N^nucl expression under the sev promoter induces the activation of Mδ, a bHLH member of the E(spl) complex (Fig. 2H; Sun and Artavanis-Tsakonas, 1996). In contrast, the sev-ANKtransgene fails to induce ectopic Mδ expression (Fig. 2I). This is consistent with the observation that the sev-ANKmutant phenotype is distinct from that induced by N^act.

In sev-ANKeye discs, the overall expression of Mδ is sup-pressed (Fig. 2, compare G and I). Mδ protein is detected only in clusters within the morphogenetic furrow where the sev promoter is not active (Fig. 2I; see below). Four lines of evidence argue against the possibility that this suppression reflects a dominant negative behavior associated with an over-expression of the ANK repeats. First, if overexpression of the ANK repeat peptide was interfering with wild-type Notch receptor function, one might expect that overexpressing the wild-type receptor would suppress the sev-ANK phenotype. Double mutants carrying sev-ANKand sev-Notch (full-length Notch, Fortini et al., 1993) do not show any modification of the rough-eye sev-ANKphenotype (data not shown). Second, the same is true for double mutant combinations between sev-ANK and sev-Deltex (K. Matsuno and S. Artavanis-Tsakonas, unpublished data), a positive regulator of Notch signaling (Table 1). Third, the constitutive activation of Notch via the expression of sev-N^act or sev-N^nucl, rather than suppressing the sev-ANK phenotype, enhances it, suggesting that the ANK repeat peptide does not simply cancel the activity of the activated forms of Notch (Fig. 3). Finally, we find that Su(H)-dependent transcription of reporter genes in S2 cultured cells, while enhanced by N^act, is not affected by the expression of the ANK peptide. Fig. 4 summarizes the results of a reporter assay involving the Suppressor-of-Hairless-dependent transcription of a reporter gene driven by the E(spl) promoter m gamma. Previous work demonstrated that the promoter activity of the E(spl) gene mγ depends on the synergistic action of the intra-cellular domain of Notch and Su(H) (Eastman et al., 1997 and Fig. 4: Su(H)+ICN). Consistent with the idea that the ankyrin repeats exert their action independent of Su(H), we do not observe a modulation of mγ transcription by the ankyrin repeats peptides and Su(H) (Fig. 4, Su(H)+ANK).

The above observations suggest that the ANK repeat peptide triggers novel downstream events, apparently bypassing Su(H)-dependent Notch signaling. Consistent with this inter-pretation is also the finding that sev-ANKand either gain-of-function or loss-of-function Su(H) alleles do not interact genet-ically (Table 1). On the contrary, the sev-N^act phenotype is clearly enhanced by gain-of-function Su(H) alleles and sup-pressed by loss-of-function mutations (Table 1 and Verheyen et al., 1996). Analysis involving Su(H) mutant clones was unin-formative since the Notch pathway is disrupted and they do not survive to adulthood (M. E. Fortini, personal communication). It is finally noted that while a polypeptide encompassing the entire Notch intracellular domain localizes to the nucleus (Struhl et al., 1993; Lieber et al., 1993; Fortini et al., 1993; Jarriault et al., 1995), the ANK repeat peptide was found to be cytoplasmic both in cultured cells and in vivo (data not shown), suggesting that, unlike N^nucl, the ANK repeats is not directly involved in transcriptional regulation with Su(H).

In the wild-type eye disc, Mδ expression is detected in a subset of cells with basal nuclei located roughly in columns 5 to 12 (Fig. 5A,B). The Mδ staining that we observe is identical to that described by Baker et al. (1996). In the sev-ANKlines, Mδ expression is suppressed in the same cells (Fig. 2I). In order to accurately determine the cells in which the sev promoter is active relative to those producing Mδ, we used a transgenic line expressing β-galactosidase (β-Gal) under the sev promoter (sev-β-Gal). In sev-β-Gal eye discs, β-Gal protein was detected in the cytoplasm as well as in the nucleus, whereas Mδ was observed only in the nucleus. In order to corroborate the identity and relative position of the sev-expressing cells, we have also examined the nuclear staining of sev Delta intra expressing discs, a nuclear fragment deriving from the Delta protein (data not shown; Sun and Artavanis-Tsakonas, 1996). Fig. 5A-C shows that the β-Gal-expressing cells (green) are distinct from those cells producing Mδ (red), in the basal region of sev-β-Gal eye discs (see also Fig. 5G). An apical focal view (Fig. 5D-F; see also Fig. 5G), which shows cells in which the sev promoter is known to be active (green), lacks Mδ expression. Therefore in the sev-ANKdiscs, Mδ expression is affected by the action of the ANK repeat peptide expressed in neighboring cells under the sev promoter. These observations demonstrate that sev-ANKtriggers downstream events that are independent of Su(H) and influence Notch signaling in a non-autonomous fashion.

Given the distinct biological activity of the ANK repeats, espe-cially the fact that the sev-ANKphenotype is not affected by loss-of-function Su(H) mutations, we were interested in examining the biochemical properties of the ANK repeats, which have been implicated in interactions with both Su(H) and Dx (Fortini and Artavanis-Tsakonas, 1994; Diederich et al., 1994; Matsuno et al., 1995). Earlier data from our labora-tory have indirectly implicated the ANK repeats with the inter-actions between Su(H) and Notch, whereas more recent data have convincingly demonstrated a Notch Su(H)-binding site outside the ANK repeat region (Tamura et al., 1995). We have further examined the binding of Notch and Su(H) using the ‘interaction trap’ system (Zervos et al., 1993).

In agreement with the report of Tamura et al. (1995), Su(H) binds to a subtransmembrane region of Notch (aa 1769-1825) which excludes the ankyrin repeats (Fig. 6, construct 2). However, a Notch peptide (aa 1827-2259) that does not include this subtransmembrane region but instead carries a region C-terminal to it, including the ANK repeats, also binds to Su(H) (Fig. 6, construct 3; Fortini and Artavanis-Tsakonas, 1994). The ANK repeats are necessary for this second binding site, since their deletion eliminates Su(H) binding (compare Fig. 6, construct 3 and 14). On the contrary, a peptide encompassing just the ANK repeats (aa 1891-2113) does not bind to Su(H) (Fig. 6, construct 16, and see Tamura et al., 1995) but is capable of binding to Deltex (Matsuno et al., 1995). The in vivo expression of an ANK repeat polypeptide (aa 1890-2108) elicits the dominant phenotypes described above. These results also support the idea that the activity observed in the eye discs of sev-ANKis independent of Su(H).

We conclude that Drosophila Notch has at least two distinct binding sites for Su(H). The first is independent of the ANK repeats, while for the second the ANK repeat region appears to be necessary but not sufficient. We note that the yeast assay data are consistent with co-localization assays in Drosophila tissue culture cells (Fortini et al., 1994).

The interaction trap study reveals that the ANK repeats are involved in homotypic interactions, a property that may underlie the biological function of the ANK repeats (Table 2, construct 1). Deletion studies define the Notch region encom-passing only ANK repeats 1-5 as necessary and sufficient to mediate these interactions (Table 2, constructs 7, 10 and 14). Also, the Notch ANK repeats from Drosophila interact with the analogous regions of two human Notch homologs, Notch-1 and Notch-2 (Ellisen et al., 1991; Stifani et al., 1992; Table 2, constructs 11 and 12). This interaction seems specific since the ANK repeats of Drosophila Notch do not associate with the ANK repeats of Cactus (Kidd, 1992; Table 2, construct 13), even though the Cactus ANK repeats do show homotypic inter-actions (Table 2, construct 14).

The yeast-based interaction trap analysis was corroborated using Drosophila cultured cells. Full-length Notch was co-expressed with Myc-tagged proteins with and without the Notch ANK repeats (shown schematically in Fig. 7A). The relative subcellular localization of the two proteins was then monitored. When full-length Notch was co-expressed with a Myc-tagged fragment containing the ANK repeats (NTMICN1Myc and DlTMICN1Myc in Fig. 7A), the two polypeptides co-localized in the majority of cells (Fig. 7D and E). Co-localization was independent of the Myc-tag and depended on the presence of the ANK repeats since Myc-tagged polypeptides consisting of the Notch or Delta trans-membrane domain (NTMMyc and DlTMMyc) did not co-localize with Notch (Fig. 7,F,G).

Insight into the possible functional consequences of the homotypic interaction involving the Notch ANK repeats was provided by N^Su42c, the only point mutation known to map in the Drosophila Notch ANK repeats (Diederich et al., 1994). Table 2 summarizes the results of an interaction trap study showing that the N^Su42c mutation does not significantly affect the interaction of Notch with Deltex (compare constructs 19 and 20), while it strengthens Notch homotypic interactions (compare constructs 17 and 18). The N^Su42c mutation was isolated as a suppressor of the lethality associated with certain heteroallelic combinations of the Abruptex (Ax) mutations (Diederich et al., 1994; Xu et al., 1990). Since negatively com-plementing combinations of Ax alleles appear to reflect an overactive state of the Notch receptor, it is reasonable to assume that strengthening the ANK homotypic interactions leads to the reduction of Notch signaling.

The development of the compound eye in Drosophila rests upon sequential cell fate choices that gradually recruit precursor cells into ommatidial assemblies (Wolff and Ready, 1993). Genetic studies have demonstrated that Notch signaling controls cell fates throughout the development of the eye (Cagan and Ready, 1989). Given the precise cellular architec-ture of the eye disc, this preparation has proven useful in analyzing the activity of mutant Notch forms. Expression of activated forms of the Notch receptor in the eye have helped evaluate the action of Notch signaling at a single cell level (Fortini et al., 1993). Moreover, the phenotypes resulting from the expression of such mutant Notch forms contributed to the dissection of the Notch pathway (Verheyen et al., 1996). Sig-nificantly, known elements of the pathway such as Dl, master-mind, deltex and Su(H) were also identified as modifiers of activated N expressed under the sev promoter, validating the use of these gain-of-function phenotypes as a Notch pathway dissection tool (Verheyen et al., 1996). We were thus encour-aged to use the same approach in combination with biochem-ical protein interaction assays to further dissect the function of the intracellular domain of Notch, in particular the activity of the ANK repeats region.

The protein interaction studies reported here clearly reveal the existence of two distinct Su(H)-binding sites on the Notch receptor, resolving what on the surface appeared to be contra-dictory results reported in independent studies (Fortini et al., 1993; Tamura et al., 1995). Our analysis indicates that one of the Su(H)-binding sites maps outside the ANK repeats, whereas for the second site the ANK repeats appear necessary but not sufficient. In vitro assays involving cultured cells have shown that Su(H) can be sequestered in the cytoplasm by virtue of its binding to Notch (Fortini and Artavanis-Tsakonas, 1994). Upon the interaction of Notch with its ligand Delta, Su(H) translocates into the nucleus (Fortini and Artavanis-Tsakonas, 1994). Nuclear translocation of Su(H) can also be triggered by the overexpression of Dx in the same cell, for which the ANK repeats are both necessary and sufficient in order to associate with Dx (Matsuno et al., 1995). Therefore, despite the distinct binding sites of Su(H) and Dx on Notch, there is an interplay between these molecules (Matsuno et al., 1995).

The protein interaction assays that we describe suggest that the ANK repeats are involved in homotypic interactions. Extrapolating from what is commonly believed to be true in other transmembrane receptors, e.g., the EGF receptor, as well as the genetic behavior of certain Notch mutant alleles, e.g., the Abruptex (Ax)mutations, it is thought that Notch acts as a homomultimer (Foster, 1975; Portin, 1975). Specific Axalleles, which involve point mutations in the extracellular EGF homol-ogous region, can be viable in a homozygous form, yet they are lethal in trans heterozygous combinations. This negative complementation could be explained by homotypic interac-tions between the two mutant Notch molecules. However, direct evidence demonstrating the existence of homotypic interactions between Notch molecules had been elusive. The mutation N^Su42c, which maps within the ANK repeats, was in fact isolated as an intragenic suppressor of the lethality asso-ciated with two negatively complementing Ax alleles (Diederich et al., 1994). The mutant phenotype is paralleled by stronger homotypic interactions in the yeast assay, suggesting an involvement of the ANK repeats in the oligomerization of the Notch receptor. Given these observations, it is reasonable to postulate that homotypic interactions of the ANK repeats can modulate Notch signaling activity.

The transgenic analysis of the ANK repeats revealed an unexpected activity which acts in a non-cell-autonomous fashion. It should be noted that studies involving Glp-1, a C. elegans counterpart of Notch, indicated that a peptide con-sisting of the ANK repeats as well as short flanking sequences (52 amino acids N-terminal to the ANK repeats and 33 amino acids to C-terminal portion) mimicked gain-of-function mutations (Roehl and Kimble, 1993). More recent results demonstrated that this activity of GLP-1 requires the presence of the ANK flanking sequences (Roehl et al., 1996). Deletion analyses in Drosophila have shown that constructs lacking the ANK repeats behave as dominant negative mutations, mimicking loss-of-function phenotypes of N (Rebay et al., 1993; Lieber et al., 1993). These constructs, however, involve sequences immediately flanking the ANK repeats whose function is unknown. The ANK repeat peptide that we study here does not include flanking sequences, does not bind to Su(H) and fails to induce Mδ expression. Consistent with our findings, the expression of an analogous ANK peptide deriving from rat Notch-1 shows a CBF1 (a mammalian homologue of Su(H))-independent activity (Shawber et al., 1996).

Activation of Notch, as evidenced by the expression of the entire intracellular domain, results in the expression of Mδ which, in turn, depends on a downstream effector Su(H) (Jennings et al., 1994; Lecourtois and Schweisguth, 1995; Bailey and Posakony, 1995). The Ankyrin repeats fail to induce Mδ activity, suggesting that their biological action is indepen-dent of this effector, yet the ANK peptide activity affects the Notch pathway, since ANK expression in one cell was found to suppress Mδ expression in a neighbor. The simplest way to account for these observations is to postulate the existence of a yet to be identified effector whose expression is modulated by sev-ANK. As a consequence, the Notch pathway in a neigh-boring cell is inactivated (‘X’ in the model depicted in Fig. 8).

In this model, Notch is involved in at least two signaling events: one is dependent on Su(H), the other is not. The first requires the entire intracellular domain and induces the expression of E(spl) bHLH genes in a cell-autonomous, Su(H)-dependent fashion. The second depends on the expression of the ANK repeats only and does not involve Su(H). Cells expressing the ANK repeats are capable of antagonizing Notch signaling in the neighboring cells. The simplest way to explain this non-autonomous, antagonistic action of the ANK sequences is to suggest that the ANK-expressing cells down-regulate the endogenous Delta activity. Consequently, the Notch receptor in the neighboring cell is inactivated and Mδ expression is down-regulated. It is therefore not necessary to postulate novel antagonistic activities on the surface of the ANK-expressing cells. Irrespective of the molecular nature of the surface changes induced by the ANK sequences, the action of the ANK repeats must trigger the modulation of an effector, i.e., ‘X’. Since the phenotype associated with ANK expression cannot be modified by either the overexpression of Dx, which binds to the ANK repeats, or by wild-type Notch, we postulate that the effector X is unknown. The possibility that this unknown component down-regulates endogenous Dl is partic-ularly interesting since genetic evidence in C. elegans and Drosophila has suggested the possible existence of a regula-tory loop in the Notch pathway (Wilkinson et al., 1994; Heitzler and Simpson, 1996).

It is worth pointing out that yeast two-hybrid-based screens for proteins capable of interacting with the ANK repeats of Lin-12 resulted in the identification of a protein that shows homology to EMB5, a chromatin-associated protein in mammals (Hubbard et al., 1996). At this point, we do not know if this protein mediates the activity of the ANK repeats. However, the rough eye phenotype associated with the ANK repeat expression make genetic modifier screening possible, providing the means to further dissect the novel Notch down-stream events suggested by the action of the ANK repeats.

The yeast strain, EGY40, and vectors pEG202, pJG4-5 and pSH18-34 were generous gifts from Roger Brent. We thank G.M. Rubin for the sev-β–Gal transformant line. We thank Grace E. Gray, Esther Verheyen, Karen Purcell and Tian Xu for comments on the manuscript and Laurent Caron for his help with the confocal microscope. S. A.-T. is supported by the Howard Hughes Medical Institute and by NIH. grant NS26084.