A subset of Notch functions during Drosophila eye development require Su(H) and the E(spl) gene complex

A signalling pathway involved in many cell fate decisions is centred around the transmembrane receptor encoded by the gene Notch (N). Several elements in the Notch signalling pathway have been identified (Artavanis-Tsakonas et al., 1995). In Drosophila these are encoded by, among others, the Delta (Dl), Serrate (Ser), Suppressor of Hairless (Su(H)) and Enhancer of Split (E(spl)) loci. Dl and Ser encode membrane-bound ligands of Notch. Su(H) acts as an intracellular transducer of the signal from the membrane to the nucleus. E(spl) is a nuclear target in the relay of the signal(Artavanis-Tsakonas et al., 1995). E(spl) is a complex locus, including seven genes that encode closely related transcription factors bearing a basic Helix-Loop-Helix (bHLH) motif (m8, m7, m5, m3, mβ, mγ and mδ), and two genes that encode non-bHLH proteins (m4 and groucho; Delidakis and Artavanis-Tsakonas, 1992; Knust et al., 1992; Schrons et al., 1992) (Fig. 1).

In many cases Notch signalling is thought to delay cell differentiation, keeping cells competent to respond to later inductive signals (Coffman et al., 1993; Fortini et al., 1993). Lateral specification, exemplified by embryonic neuroblast commitment, is an example. During lateral specification an individual cell is singled out from a group of equipotent cells and follows a different developmental fate. Notch does not signal in this committed cell, whereas the remaining cells remain uncommitted as a result of Notch signalling (Lieber et al., 1993; Struhl et al., 1993; Jennings et al., 1994).

In addition to delaying differentiation, recent studies reveal examples of inductive Notch signalling. Activation of Notch induces particular cell fates at the developing wing margin (Kim et al., 1995; Couso et al., 1995; Doherty et al., 1996). The different types of N signalling can share signal transduction components. Both lateral specification and inductive signalling at the wing margin require the Su(H) signal transducer, but at the wing margin target genes that act positively to promote cell fates are activated, in addition to the repressors of the E(spl) gene complex (E(spl)-C) (de Celis et al., 1996a,b; Kim et al., 1996). Although the E(spl)-C is expressed at the wing margin, little function has yet been found for this expression (de Celis et al., 1996a,b).

Here we investigate the signal transducers and target genes involved in Notch signalling in the eye imaginal disc. The adult eye of the fly contains 800 ommatidia (unit eyes) arranged in a two-dimensional lattice. Each ommatidium includes eight photoreceptor neurons and twelve accessory cells arranged in a stereotypical pattern, and is founded by a single R8 photoreceptor cell, which induces most of the other cell types (Wolff and Ready, 1993; Tio and Moses, 1997). Expression of the proneural gene atonal (ato) is autonomously required for R8 determination (Jarman et al., 1994).

Notch signalling serves dual roles in the specification of R8 cells (Baker and Yu, 1997). As expected, Notch-dependent lateral specification occurs to restrict ato expression to dispersed single cells that become R8 photoreceptor precursors. In an additional earlier process, Notch enhances ato expression and function as ato transcription becomes autoregulatory and independent of prepattern signals. The initial enhancement of ato expression occurs in many cells, but once ato autoregulation is established, lateral specification starts to limit ato expression to R8 precursor cells. Thus Notch signalling is required at successive steps during R8 specification, initially to promote neural potential and later to suppress it through lateral specification. Consequently the phenotype of loss of Notch gene function varies with time. If Notch (N) function is removed conditionally once ato expression has been enhanced, supernumerary R8 cells differentiate because lateral specification is affected. If N function is absent from the outset, such as in a clone of cells lacking N, little R8 specification can occur. For this reason clones of N null mutant cells in the eye disc almost completely lack neural differentiation, contrasting with the neurogenic phenotype of null mutant embryos (Baker and Yu, 1997) (Fig. 2).

Although Notch signalling is involved in many cell fate decisions during development, the dual role in R8 specification is unusual in that both roles affect the same cell-fate decision in the same cells. The two processes occur consecutively without any clear gap between them. In this study we investigated whether components of the pathway are involved in both processes and whether the E(spl) genes serve as the last step in the relay of the signals. We started this work with two hypotheses in mind. In the first, by analogy with the wing margin, proneural enhancement of ato would require Su(H), but E(spl) genes would only be needed to mediate lateral specification. Our second hypothesis was that both processes could involve E(spl) genes, with different combinations of E(spl) proteins playing distinct roles. For example, some of the bHLH genes, or groucho, could be involved in lateral specification, and others in proneural enhancement. Our data argue against both of these hypotheses. We show that Notch and Delta participate in both proneural enhancement and lateral specification, but that Su(H) and E(spl) are only required for lateral specification, not for proneural enhancement. It follows that the proneural function of Notch is mediated through a distinct signal transduction route.

The FRT82 Dl Ser chromosome was derived by standard genetic crosses from the Dl^rev10 and Ser^RX106 strains described previously (Baker and Yu, 1997). FRT82 Df(3R)gro^b32.2 and FRT82 Df(3R)E(spl)^BX22 chromosomes were described previously (Heitzler et al., 1996; Treisman et al., 1997), as was Su(H)^SF8FRT40 (Schweisguth, 1995). For chromosome arm 3R the LacZ marker was p[construct D]96A (Tio and Moses, 1997). For chromosome 2L the LacZ marker was armLacZ.2L (Vincent et al., 1994); a hs-π-myc marker was also used (Xu and Rubin, 1993). The hsFLP1 transgene was used (Xu and Rubin, 1993). The UAS-E(spl)bHLH transgenes were prepared as described previously (de Celis et al., 1996a), and the h^H10 Gal4 driver has been described before (Ellis et al., 1994). We have observed weaker phenotypes towards the posterior of the eye disc using h^H10 to express these and many other genes, so this may be a property of the h^H10 driver. Whereash^H10/+; UAS-mδ/+ flies had mild rough eyes, similar combinations with m3, m5, m8, mβ or mγ were phenotypically wild type. Only the mδ, m5 and mβ genotypes have been examined as double homozygotes.

Clones were induced by heat-shocking larvae (1 hour, 37-38°C) of the following genotypes: (1) hsFLP/+; FRT82 p[construct D]96A/FRT82 Dl^rev10e Ser^RX106; (2) hsFLP/+; FRT82 p[construct D]96A/FRT82 kar²ry⁵⁰⁶P[gro⁺ry⁺] Df(3R)gro^b32.2; (3) hsFLP/+; FRT82 p[construct D]96A/FRT82 Df(3R)E(spl)^BX22; (4) hsFLP1/+; armLacZ.2L FRT40/Su(H)^SF8FRT40.

Antibody stainings were performed as described (Baker and Yu, 1997). Rabbit anti-Boss antibody was kindly provided by H. Kramer (Kramer et al., 1991) and rabbit anti-Atonal antibody by A. Jarman and Y.-N. Jan (Jarman et al., 1994). Monoclonal antibodies specific for β-galactosidase (mAb40-1a) and Elav protein (mAb7E8A10) were obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242, USA, under contract N01-HD-7-3263 from the NICHD. These monoclonal antibodies were developed by J. R. Sanes and G. M. Rubin, respectively. Secondary antibodies were obtained from Jackson Immunoresearch. HRP-conjugated goat anti-mouse and goat anti-rabbit antibodies were used as well as Cy3-conjugated goat anti-mouse and FITC-or Cy2-conjugated goat anti-rabbit and goat anti-rat.

Scanning electron microscopy was performed as described (Kimmel et al., 1990).

Previous work established an autonomous, proneural function for Notch, which is required to enhance the expression and function of ato as eye differentiation begins (Baker and Yu, 1997). In contrast to the neurogenic phenotype seen in N null mutant embryos, N mutant clones in the eye disc lacked most neural differentiation, demonstrating an N requirement preceding lateral specification (Fig. 2). The Notch ligand Dl was nonautonomously required, so that the proneural enhancement was rescued in some Dl mutant cells by nearby cells wild type for Dl. Another ligand encoded by the Serrate gene appeared not to be required for eye differentiation (Sun and Artavanis-Tsakonas, 1996; Baker and Yu, 1997).

The ligand encoded by Ser is also expressed in the eye disc (Baker and Yu, 1997). Because Dl and Ser function redundantly in some tissues and can regulate each other’s expression, it was necessary to examine cells mutant for both genes to determine the effects of losing Notch ligands completely (Micchelli et al., 1997; Panin et al., 1997). FLP-mediated recombination was used to induce clones of cells homozygous null for both ligands. Phenotypically, clones of Dl⁻Ser⁻ mutant cells resembled those previously described for Dl alone. Neural differentiation was prevented as in N clones, except that near the clone margins nonautonomy was observed (Fig. 3). Another marker that labels all photoreceptor neurons, the Elav protein, was also examined, with similar results (not shown). Where neural differentiation was rescued near the clone margins, excessive numbers of R8 cells were observed (Fig. 3). This reflects rescue of the proneural function of Notch but insufficiency of signal during lateral specification, and was observed before for Dl mutant clones (Baker and Yu, 1997).

Adjacent R8 cells bunch very tightly together at their apices. Larger R8 cell clusters seen in other genotypes are often associated with deep folds in the epithelium. In summary, no role could be identified for Ser, even in the simultaneous absence of Dl.

E(spl) bHLH genes have been shown to be transcriptionally activated as a direct consequence of Notch signalling and, along with the corepressor protein Groucho, to mediate inhibition of proneural genes in the nucleus (Jennings et al., 1994; Paroush et al., 1994; Oellers et al., 1994; Dawson et al., 1995; Jarriault et al., 1995; Nakao and Campos-Ortega, 1996; Fisher et al., 1996; Giebel and Campos-Ortega, 1997). At the wing margin E(spl) genes are activated by N, but this expression appears to be dispensable for wing margin formation (de Celis et al., 1996a,b). In the eye, Notch-dependent expression of mδ and mγ accompanies repression of ato expression, suggesting that at least these two of the E(spl) bHLH genes contribute to R8 patterning during lateral specification (Baker et al., 1996; Dokucu et al., 1996). In addition, mδ and perhaps mγ are also transiently expressed prior to lateral specification (Baker et al., 1996), and the m7, m8 and mβ genes are transcribed in distinct patterns that remain uncharacterized in detail for lack of specific antibodies (de Celis et al., 1996a). Thus particular E(spl) bHLH proteins might mediate proneural Notch signalling as well as or instead of lateral specification.

Clones of cells deleted for portions of the E(spl) complex were used to define its role more precisely. The E(spl)^b32.2 deficiency deletes all seven bHLH genes and m4 (Fig. 1; Schrons et al., 1992). Partial gro function was supplemented in our experiment by a linked gro⁺ transgene (see Materials and methods). E(spl)^b32.2gro⁺ homozygous cells displayed a cell autonomous neurogenic phenotype quite unlike that of N or Dl mutant clones. Antibodies against Boss or Elav proteins each labelled a much greater number of cells within the clone than in the surrounding wild-type tissue (Fig. 4, and data not shown). Some clones were difficult to photograph because neurogenic regions often seem to fold in on themselves and crease the eye disc. Because neurogenesis can still occur, it appears that the proneural function of Notch can proceed without any E(spl)-C bHLH genes, whereas N function in lateral specification is severely impaired.

As recently reported (Treisman et al., 1997), we find that clones of cells homozygous for E(spl)^BX22 are also neurogenic in phenotype (data not shown). E(spl)^BX22 affects gro and the bHLH genes m5, m7 and m8 (Fig. 1). It follows that gro is also dispensable for the proneural function of Notch, although it is probably required in lateral specification.

Forced expression experiments were performed to define further the role of particular bHLH proteins. The hairy^H10 enhancer trap was used to drive GAL4-dependent transgene expression anterior to and within the morphogenetic furrow. Fig. 5B shows the double homozygote for both h^H10 and UASmδ, expressing mδ protein during the requirement for ato. The eyes contained few facets and were greatly reduced in size. Eye imaginal discs contained few ommatidia (Fig. 5D). The defect was associated with reduction or absence of ato expression in the morphogenetic furrow (Fig. 5H). These findings indicate that mδ protein is capable of repressing ato expression, as occurs during lateral specification.

Not all E(spl)bHLH proteins repressed ato. Fig. 5 shows that eye disc patterning occurred almost normally in h^H10/h^H10;UASm5/UASm5 homozygotes and in h^H10/h^H10; UASmβ/UASmβ homozygotes. Both h^H10/h^H10; UASm5/UASm5 homozygotes and h^H10/h^H10;UASmβ/UASmβ homozygotes died as pupae without differentiating adult structures, whereas h^H10/h^H10;UASmδ /UASmδ homozygotes survived to pharate adulthood and produced occasional adult escapers. Because of the earlier lethality due to m5 or mβ expression it is difficult to attribute the lack of effect on eye development to lower expression levels than for mδ, and we instead conclude that the mδ protein is qualitatively distinct from m5 and mβ proteins in its ability to inhibit ato expression.

Recent studies have identified Su(H) as a common component in Notch signal transduction pathways. Ligand binding (Delta or Serrate) to Notch activates Su(H), which can shuttle between the cytoplasm and the nucleus and act as a transcription factor. Activated Su(H) turns on a number of downstream target genes mediating Notch signalling in lateral specification or inductive processes (Fortini and Artavanis-Tsakonas, 1994; Jarriault et al., 1995; Lecourtois and Schweisguth, 1995; Bailey and Posakony, 1995; Kim et al., 1996; Eastman et al., 1997). Since the E(spl)-C was not required for proneural N signalling in the eye it was possible that other effector genes were transcribed in response to Su(H) activation, as happens at the wing margin.

In order to investigate the role of Su(H), clones of cells homozygous for an apparent null allele of Su(H) were generated by FLP-mediated recombination. In the eye imaginal disc Su(H)⁻ mutant cells were associated cell autonomously with neural hypertrophy (Fig. 6). Many of the ectopic neural cells were R8 photoreceptors, based on expression of the R8-specific protein BOSS (Fig. 6C,D). It appeared that, like the E(spl)-C, Su(H) was required for lateral specification but not for R8 differentiation. To confirm this conclusion ato expression was examined. In wild type, initial broad expression of ato protein is replaced by R8-specific expression that persists for 6-8 hours (3-4 columns of ommatidia) and then fades (Jarman et al., 1994; Fig. 5G). Fig. 6E-J shows that whereas ato expression begins normally in Su(H) mutant cells, ato expression is maintained in many more R8 cells than in wild type, indicating failure of lateral specification. Expression of ato then fades from Su(H) mutant R8 cells at the same time as from wild-type cells. Thus, like the E(spl)-C, Su(H) is required for lateral specification but not for the proneural function of Notch in the retina.

Interestingly, although many extra R8 precursors form in Su(H) mutant clones, not all Su(H) mutant cells maintain ato expression or subsequently express the R8-specific Boss protein. Instead clusters of R8-like cells often seem interspersed with non-R8 neurons (Fig. 6D,J). As noted by Jarman et al. (1995), ato expression in wild type first becomes patterned into regular ‘intermediate groups’ of about ten ato-expressing cells before resolving to individual R8 precursors. Our results support previous conclusions that initial spacing of intermediate groups is not part of the N-dependent lateral specification process (Baker and Zitron, 1995; Lee et al., 1996), and so does not depend on E(spl) or Su(H).

For many of the developmental decisions mediated by Notch, the signal transduction pathway is known in outline. Activated Su(H) protein transcribes target genes, which include bHLH proteins from the E(spl)-C during lateral specification, or other positive factors during inductive processes (Artavanis-Tsakonas et al., 1995; Lewis, 1996). Two successive roles of N signaling have been described during R8 cell specification in the developing eye. First, N is required for the full level of ato expression and function, so that neural differentiation fails in N null mutant clones. Later N signalling represses ato expression during lateral specification. Extra R8 cells are produced if N is inhibited during lateral specification with a temperature-sensitive allele (Baker and Yu, 1997). We have now determined the respective roles of the Su(H) locus and E(spl)-C. Our major conclusion is that while lateral specification of the R8 cells requires Su(H) and E(spl), earlier proneural signalling requires neither and must differ from both lateral specification and induction by using other signal transduction components whose identity is not yet known.

The phenotype of E(spl) mutant cells is dramatically different from that of N mutant clones (Fig. 4). E(spl) mutant cells show an autonomous neurogenic phenotype in which nearly all cells differentiate as neurons, many of which are R8 photoreceptor cells. This showed that E(spl)-C was not required for neural differentiation and cannot be essential for the proneural enhancement for which N and Dl are required. This conclusion applies to all the E(spl) bHLH genes, gro and m4, even although at least some of these genes seem to be expressed in response to the proneural N signalling (Baker and Yu, 1997; Baker et al., 1996). However, lateral specification failed in E(spl) clones, showing that some of the bHLH genes and probably gro are required for lateral specification of R8 cells. This is consistent with the characterization of E(spl) bHLH proteins and gro as repressors that inhibit proneural gene expression and function in other parts of the nervous system (Jennings et al., 1994; Paroush et al., 1994; Oellers et al., 1994; Dawson et al., 1995; Jarriault et al., 1995; Nakao and Campos-Ortega, 1996; Fisher et al., 1996; Giebel and Campos-Ortega, 1997).

At the developing wing margin, inductive Notch signalling can occur in the absence of E(spl), but is still dependent on Su(H) to induce expression of other target genes. Our data showed that proneural enhancement in the eye differed and occurred independently of Su(H). Later, Su(H) was required for lateral specification mediated by the E(spl)-C. In Su(H) mutant cells, ato expression was initiated normally but subsequently persisted in too many cells, leading to ectopic R8 cell differentiation and neural hypertrophy (Fig. 5). We showed previously that the proneural enhancement is mediated by the intracellular domain of Notch (Baker and Yu, 1997). Taken together these findings imply a distinct signaling mechanism downstream of Notch that permits proneural enhancement to occur independently of Su(H) function, and which also does not require the E(spl)-C bHLH genes or gro.

Several prior studies have given indications that Su(H)-independent signalling might occur. These have included studies of the C. elegans gene lag-2, a Su(H) homolog, which did not employ null mutations (Lambie and Kimble, 1991), and effects in several systems of misexpressing ankyrin-repeat portions of the Notch intracellular domain (Roehl and Kimble, 1993; Shawber et al., 1996; Matsuno et al., 1997). As this portion of Notch binds Su(H) poorly (Tamura et al., 1995), these misexpression effects might be Su(H)-independent. However, an additional interaction between Su(H) and the ankyrin-repeat region of the receptor has now been implicated in at least some of these examples (Fortini and Artavanis-Tsakonas, 1994; Roehl et al., 1996; Kato et al., 1997). Our data, and also a study of the role of Notch in regulating single-minded gene expression in the embryonic mesectoderm (Lecourtois and Schweisguth, 1995), show definitively through studies of loss-of-function mutations that there are functions of the wild-type Notch protein that occur in apparently Su(H) null mutant cells but not in N null mutant cells, and so must be independent of Su(H).

Although we did not find distinct proneural functions of any E(spl)-C genes, there was evidence for other differences between them, based on forced expression. We find that mδ represses ato expression and function much more efficiently than either m5 or mβ do (Fig. 5). Because m5 or mβ expression causes earlier lethality than mδ expression, it is difficult to account for this except by functional differences between distinct bHLH genes. Lack of point mutations affecting individual E(spl) bHLH proteins has led to the conclusion that these genes are redundant, but the gene complex has nevertheless been conserved through evolution (Schrons et al., 1992; Maier et al., 1993). It may be that the seven bHLH proteins show overlapping functions, not identical ones. The distinct expression patterns of individual genes suggest they may differ in function and so contribute to the specificity of N responses in particular tissues (de Celis et al., 1996a).

Taken together with expression studies that show mδ protein spatially replacing ato protein during lateral specification of R8 cell fate (Baker et al., 1996; Dokucu et al., 1996), our results point to a major role for mδ in this particular instance of lateral specification. However, both mδ and mγ, which is expressed similarly in the eye, can be deleted from the genome without affecting eye development (The et al., 1997). At least one other E(spl) bHLH protein must be able to substitute. A candidate is m8, whose overexpression and mutation in the E(spl)^D mutation can suppress eye neurogenesis in the presence of the N mutant allele spl (Campos-Ortega and Knust, 1990).

Both proneural N signalling and lateral specification require activation of N by Dl but no role can be demonstrated for Ser, even in the absence of Dl. Cells mutant for Dl, or for both Dl and Ser, fail to undergo neural differentiation although the phenotype is rescued nonautonomously near the boundary with wild-type cells (Fig. 2). Proneural signalling seems to be rescued over a greater range than lateral specification, so that neurogenic clusters of R8 cells can differentiate from cells where proneural enhancement occurred but lateral specification did not. Because of this greater range for proneural signalling, described previously for Dl mutant cells (Baker and Yu, 1997), we had considered the possibility of a relay in which signalling by Dl from wild-type cells activated Ser to transmit a proneural signal further into the clone, but the similar phenotype of Dl,Ser double mutant cells to that of Dl mutants alone rules out this model. The basis of the greater range for proneural signalling than lateral inhibition remains unknown.

The independence of proneural enhancement from Su(H) and E(spl)-C may suggest explanations for how these two functions of N can occur sequentially in the same cells. It is possible that the unidentified proneural signalling pathway may act on ato more directly than Su(H), which has first to activate E(spl) expression, so ensuring that proneural enhancement precedes lateral specification. Alternatively, the two pathways might be activated by different levels of N activation. Proneural signaling would be replaced by lateral specification when rising levels of N activation become sufficient to activate Su(H). Finally, elevated ato expression might contribute directly to lateral inhibition, if E(spl) genes require both Su(H) and a proneural activator (ato) for transcription.

We thank P. Alifragis, S. Bray, Y. Chen, P. Heitzler, M. Mlodzik, K. Moses, P. Simpson, G. Struhl, J. Treisman and T. Xu for Drosophila strains, and the Analytical Imaging Facility at AECOM for use of the Biorad MRC600 Confocal Microscope. Supported by grants from the NIH (No. GM47892 to N. E. B.), from the Human Frontiers Science Program (No. RG490/94 to C. D.), and by the award of a UNESCO Travel Fellowship (to P. L.). We thank S. Bray and M. Caudy for comments on the manuscript. P. L. thanks Rafael Fernandez for hospitality in New York.

A Su(H)-independent function for N was also reported by Wang et al. (1997) Development124, 4435-4446.