The Snail repressor positions Notch signaling in the Drosophila embryo

Dorsal is a maternal regulatory protein that is distributed in a broad dorsoventral gradient in the precellular Drosophila embryo (reviewed by Drier and Steward, 1997). It initiates the differentiation of the mesoderm, neurogenic ectoderm and dorsal ectoderm by regulating a variety of target genes in a concentration-dependent manner (reviewed by Rusch and Levine, 1996). Two lines of cells that straddle the presumptive mesoderm express the regulatory gene sim and ultimately form derivatives of the mesectoderm at the ventral midline of advance-stage embryos (Crews et al., 1988; Nambu et al., 1990; Nambu et al., 1991). Perhaps as little as a twofold difference in the levels of Dorsal determines whether a naïve embryonic cell adopts a mesodermal or mesectodermal fate (Gonzalez-Crespo and Levine, 1993; Ip et al., 1992; Kosman et al., 1991). We have investigated the basis for this precise regulatory switch in cell fate.

Previous studies suggest that sim responds directly to the Dorsal gradient through high-affinity Dorsal-binding sites in the 5′ cis-regulatory region (Kasai et al., 1998). In principle, high and intermediate levels of Dorsal can activate sim in the presumptive mesoderm and mesectoderm, but high levels of Dorsal also lead to the activation of the Snail repressor in the ventral mesoderm (Gonzalez-Crespo and Levine, 1993; Ip et al., 1992). Snail represses sim in the mesoderm, and thereby restricts expression to lateral regions that form the mesectoderm (Kasai et al., 1992; Kasai et al., 1998; Nibu et al., 1998). Twist-binding sites in the sim 5′ regulatory region might work in concert with Dorsal to activate gene expression (Kasai et al., 1998). Dorsal-Twist synergy has been implicated in the formation of the sharp lateral borders of the snail expression pattern that define the boundary between the mesoderm and mesectoderm (Ip et al., 1992). The Dorsal and Twist gradients extend several cell diameters beyond this boundary, yet sim is activated in only a single line of cells (Kosman et al., 1991). Recent studies suggest that Notch signaling helps restrict sim expression to the mesectoderm (Hartenstein et al., 1992; Martin-Bermudo et al., 1995; Menne and Klambt, 1994; Morel and Schweisguth, 2000).

The activation of the Notch receptor triggers the conversion of the Su(H) transcription factor from a repressor into an activator (reviewed by Artavanis-Tsakonas et al., 1999; Bray, 1998; Kadesch, 2000; Mumm and Kopan, 2000). Su(H) is maternally expressed and uniformly distributed throughout the early embryo (Lecourtois and Schweisguth, 1995). It is initially associated with a co-repressor complex consisting of Hairless (H) and possibly dCtBP (Bang and Posakony, 1992; Bray and Furriols, 2001; Morel et al., 2001). Upon signaling, the Notch intracellular domain (Notch^IC) enters the nucleus and interacts with Su(H) (Kidd et al., 1998; Rebay et al., 1993; Struhl and Adachi, 1998; Struhl et al., 1993). The resulting Su(H)-Notch^IC complex functions as a transcriptional activator (Bailey and Posakony, 1995; Fortini and Artavanis-Tsakonas, 1994; Lecourtois and Schweisguth, 1995). It has been suggested that Su(H)-H represses sim in the neurogenic ectoderm, but activation of the Notch receptor in the presumptive mesectoderm permits sim expression, owing to disruption of the Su(H)-H repressor complex (Morel and Schweisguth, 2000). We have investigated the basis for localized Notch signaling in the mesectoderm.

A constitutively activated form of the Notch receptor, Notch^IC (Struhl et al., 1993), was placed under the control of the even-skipped (eve) stripe 2 enhancer. This stripe2-Notch^IC transgene induces ectopic activation of sim and m8. The latter gene is a member of the Enhancer of split [E(spl)] complex that encodes Notch-responsive HES-family transcriptional repressors, which inhibit neurogenesis through the silencing of achaete-scute proneural genes (Bailey and Posakony, 1995; Lecourtois and Schweisguth, 1995; Nakao and Campos-Ortega, 1996; Nellesen et al., 1999). Both the sim and m8 5′ cis-regulatory regions contain optimal, Su(H) binding sites (Morel and Schweisguth, 2000; Nellesen et al., 1999). Nonetheless, the stripe2-Notch^IC transgene differentially regulates the two genes. It induces a stripe of m8 expression, but causes a ‘pyramid’ pattern of ectopic sim staining that corresponds to the spatial intersection between Notch signaling and the Dorsal gradient. Ectopic activation of sim and m8 is inhibited in the ventral mesoderm by the Snail repressor. However, Snail also appears to stimulate Notch signaling. When introduced into mutant embryos that contain low, uniform levels of Dorsal, a stripe2-snail transgene activates sim and m8 expression. These results suggest that Snail functions both to generate a Notch signal and repress Notch target genes, thereby restricting mesectodermal fate to a precise line of cells. We discuss the basis for this dual activity of the Snail repressor and consider other cases where Snail and Snail-related repressors might localize Notch signaling.

Embryos from wild-type, mutant, and transgenic lines were collected, fixed and then hybridized with dioxygenin-UTP labeled antisense RNA probes as previously described (Jiang et al., 1991). The snail, sim and T3 cDNAs used to produce these probes were previously described (Gonzalez-Crespo and Levine, 1993; Ip et al., 1992; Kosman et al., 1991). The m8 cDNA used to generate antisense RNA probe was a gift from S. Bray. The Delta cDNA was kindly provided by E. Lai.

The construction of the hsp83-Toll^10B-bcd 3′UTR has been previously described (Huang et al., 1997). For the construction of the stripe2-Notch^IC transformation vector, a genomic fragment containing the intracellular domain of Notch (a gift from G. Struhl) (Struhl et al., 1993) was placed under the control of the eve stripe 2 enhancer by cloning it into the AscI site of a modified pCasPeR injection vector. This injection vector contains two tandem copies of an augmented stripe 2 enhancer upstream of a frt-stop-frt cassette (Kosman and Small, 1997). The stripe2-Notch^IC transformation vector was then injected into yw embryos as previously described (Kosman and Small, 1997). The construction of the stripe2-snail, stripe2-snail/hairy and stripe2-snailM1M2 has been described previously (Nibu et al., 1998). Transgenic females carrying the stripe2-snail and stripe2-Notch^IC transgenes were mated with males homozygous for the yeast Flp recombinase under the control of a sperm-specific tubulin promoter. F₁ males containing both the transgene and the Flp recombinase were selected for subsequent matings. The F₂ progeny derived from these males have ectopic snail or Notch^IC expression that is due to the rearrangement of the frt-stop-frt cassette.

The Toll^rm9 and Toll^rm10 mutations cause constitutive, low levels of Dorsal nuclear transport in affected embryos (Anderson et al., 1985). Toll^rm9/Toll^rm10 females were obtained by mating Toll^rm9/TM3, Sb, Ser males with Toll^rm10/TM3, Sb females. Non-Sb, non-Ser F₁ females were collected and mated with yw, flipped stripe2-snail, or flipped stripe2-Notch^IC males. Embryos from this cross were then collected for in situ hybridization. All crosses and collections were conducted at 25°C.

The gd⁷ allele was used to generate gd^–/gd^– females (Konrad et al., 1988), which were mated with yw, flipped stripe2-snail or flipped stripe2-Notch^IC males. Embryos from this cross were collected and fixed for in situ hybridization. All crosses and collections were conducted at 25°C.

Previous studies have indicated a role for Notch signaling in the regulation of sim expression (Hartenstein et al., 1992; Martin-Bermudo et al., 1995; Menne and Klambt, 1994). Removal of maternal Notch⁺ gene activity results in a loss of sim expression, while overexpression of a UAS-Notch^IC transgene with ubiquitous GAL4 driver lines expands the sim pattern (Morel and Schweisguth, 2000). The importance of Notch signaling for mesectodermal specification was confirmed using a stripe2-Notch^IC transgene that produces a localized source of Notch signaling in the early embryo (Figs 1-3).

The eve stripe 2 enhancer directs early expression of Notch^IC at the boundary between the presumptive head and thorax. Expression is initially detected by the onset of nuclear cleavage cycle 14 (Fig. 1A) and persists during gastrulation (data not shown). In situ hybridization assays also detect the endogenous Notch RNA, which is distributed throughout basal regions of the cytoplasm. The stripe2-Notch^IC transgene induces an ectopic stripe of m8 expression (Fig. 1B). Staining might be initially asymmetric, but the stripe becomes uniformly intense in lateral and dorsal regions by the completion of cellularization (Fig. 1B and data not shown). However, the strong expression of Notch^IC in ventral regions (Fig. 1A) is not sufficient to induce m8 expression, probably due to repression by Snail as m8 expression expands into ventral regions of snail^–/snail^– mutant embryos (data not shown).

The stripe2-Notch^IC transgene also induces expression of sim (Fig. 1C), and, like m8, ectopic expression is excluded from the ventral mesoderm. However, unlike m8, sim is not activated in the dorsal-most regions, but is restricted to a pyramid pattern in ventrolateral regions. This pyramid is detected before the expression of the endogenous pattern (Fig. 2A), and might reflect a requirement for both Notch signaling and Dorsal + Twist activators in the regulation of sim expression. Occasionally, stripe2-Notch^IC induces sim expression in dorsal regions during gastrulation, although staining is stronger in ventral regions containing the Dorsal and Twist activators (Fig. 2B). This ‘Notch-only’ sim activation may depend on high levels of Notch signaling, as it is not seen in transgenic lines that express low levels of Notch^IC.

Ectopic sim expression persists in ventrolateral regions, the presumptive neurogenic ectoderm, during gastrulation and germ band elongation (Fig. 2C). Marker genes that are expressed in the CNS exhibit gaps in the vicinity of this ectopic sim pattern (Fig. 2D), which may reflect a transformation of neurogenic ectoderm into mesectoderm. The persistence of ectopic sim expression in the ventral nerve cord is probably the result of Sim autoregulation (Morel and Schweisguth, 2000; Nambu et al., 1991).

The differential response of m8 and sim to the stripe2-Notch^IC transgene might reflect the difference between a ‘hard-wired’ target gene (m8) that is activated primarily by Notch signaling, and a conditional target gene (sim) that is jointly regulated by Notch and the Dorsal gradient. This issue was examined by comparing the ability of two separate stripe2-Notch^IC transgenic lines to induce ectopic expression of sim and m8 in mutant backgrounds. One of the lines directs strong expression of stripe2-Notch^IC, while the other directs much lower levels of Notch^IC expression based on in situ hybridization assays (data not shown). Each line was introduced into mutant embryos derived from Toll^rm9/Toll^rm10 females. Owing to the mutant Toll receptor, these embryos contain low, uniform levels of Dorsal that are insufficient to activate twist or snail. Neither sim nor m8 expression is detected in central regions of Toll^rm9/Toll^rm10 mutant embryos, though there is staining at the anterior and posterior poles (Fig. 3A; Fig. 5G,J). Introduction of the strong stripe2-Notch^IC transgene into this mutant background induces strong expression of sim (Fig. 3B), whereas the weaker line leads to low levels of expression (Fig. 3C). However, both stripe2-Notch^IC lines are capable of driving strong ectopic expression of m8 (data not shown). The absence of the Snail repressor probably accounts for the uniform induction of sim and m8 expression across the dorsoventral axis. These results also suggest that Notch signaling is sufficient to activate sim and m8 in the absence of Twist.

To determine if Notch signaling was sufficient to activate sim or m8 in the absence of both Twist and Dorsal, the stripe2-Notch^IC lines were crossed into mutant embryos derived from gd^–/gd^– females. These embryos fail to process the Spätzle ligand, and there is a block in Dorsal nuclear transport (Drier and Steward, 1997). As a result, there is no expression of twist, snail or sim in central regions (data not shown, Fig. 3D). However, mutant embryos exhibit weak, broad expression of m8, probably owing to the derepression of the dorsal ectoderm pattern (Fig. 3G); m8 is normally expressed both in the mesectoderm and the dorsal ectoderm of wild-type embryos (Wech et al., 1999) (Fig. 5A). In this background, only the strong stripe2-Notch^IC transgene induces weak expression of sim (Fig. 3E), while the weaker line fails to induce any expression whatsoever (Fig. 3F), suggesting that Dorsal is necessary for sim expression. By contrast, both stripe2-Notch^IC lines induce strong stripes of m8 expression in mutant embryos (Fig. 3H,I).

stripe2-Notch^IC transgenes fail to induce m8 and sim expression in ventral regions of wild-type embryos (Fig. 1B,C; Fig. 2A,B), but cause uniform expression in mutant embryos lacking Snail (Fig. 3B,H). Similarly, both sim and m8 are derepressed in ventral regions of snail^–/snail^– mutant embryos (Hemavathy et al., 1997) (data not shown). These results are consistent with earlier models suggesting that the Snail repressor forms the ventral border of the sim expression pattern (Gonzalez-Crespo and Levine, 1993; Kasai et al., 1992; Kosman et al., 1991; Nibu et al., 1998; Rusch and Levine, 1996). To test this idea, Snail was misexpressed in transgenic embryos by placing the snail coding sequence under the control of the eve stripe 2 enhancer (Figs 4, 5).

snail is normally expressed in the ventral mesoderm, but exhibits an ectopic stripe in transgenic embryos carrying a stripe2-snail fusion gene (Fig. 4A,B). This ectopic stripe represses several target genes that are expressed in the neurogenic ectoderm, including rhomboid and Brinker. An example of ectopic repression is shown for sog (Fig. 4C,D). There is a gap in the pattern that corresponds to the location of the ectopic Snail stripe (Fig. 4C,D). A mutant form of Snail that lacks the two dCtBP co-repressor interaction motifs (PxDLSxK and PxDLSxR) fails to repress sog (data not shown) (see Nibu et al., 1998).

The stripe2-snail transgene causes complex alterations in the sim and m8 expression patterns. There is an initial gap in the early m8 pattern (Fig. 5B), followed by ectopic staining in the neurogenic ectoderm (Fig. 5C). The ectopic ventrolateral staining persists in advanced-stage embryos and is associated with a gap in the developing ventral nerve cord (data not shown). The stripe2-snail transgene causes the same type of alteration in the sim expression pattern. There is an initial gap in the pattern (Fig. 5E), but in older embryos ectopic expression is detected in one or two cells in the neurogenic ectoderm (Fig. 5F). These alterations in sim and m8 depend upon the ability of Snail to function as a transcriptional repressor, as neither pattern is altered when the dCtBP interaction motifs are removed from Snail (data not shown).

The preceding results suggest that Snail both represses and activates sim and m8 expression. Additional evidence for this dual activity was obtained by crossing the stripe2-snail transgene into mutant embryos derived from Toll^rm9/Toll^rm10 females. The uniform, low levels of Dorsal that are present in mutant embryos fail to activate snail expression (Fig. 4E), so that the stripe2-snail transgene encodes the sole source of the Snail repressor (Fig. 4F). Though unable to induce snail expression, the low levels of Dorsal present in the mutant embryos are sufficient to induce nearly ubiquitous expression of sog (Fig. 4G). When introduced into this mutant background, stripe2-snail is still capable of repressing sog (Fig. 4H). Mutant embryos that lack the stripe2-snail transgene do not exhibit either m8 (Fig. 5G) or sim (Fig. 5J) expression in middle body regions, although there is residual staining at the anterior and posterior poles. The stripe2-snail transgene leads to ectopic induction of m8 (Fig. 5H) and sim (Fig. 5K) expression. In both cases, staining is detected in the vicinity of the eve stripe 2 pattern, but expression is not uniform. Instead, both genes, especially sim, exhibit patchy ‘salt and pepper’ staining patterns (Fig. 5H,K).

The induction of sim and m8 expression depends on the ability of Snail to function as a transcriptional repressor. Mutant proteins that lack the dCtBP interaction motifs weakly activate m8 and altogether fail to activate sim in Toll^rm9/Toll^rm10 mutants (data not shown). Conversely, a stripe2-snail/hairy transgene that contains the Hairy repression domain continues to induce sim and m8 in mutant embryos (Fig. 5I,L).

It is possible that the stripe2-snail transgene establishes a domain of Notch signaling by repressing regulators of the Notch pathway. One candidate is the Notch ligand Delta, which is broadly expressed in lateral and dorsal regions of cellularizing and gastrulating embryos (Fig. 6A). There is little or no expression in the ventral mesoderm, probably owing to repression by Snail, as the Delta pattern expands into ventral regions of sna^–/sna^– mutant embryos (data not shown). The stripe2-snail transgene causes a subtle attenuation in the Delta expression pattern (Fig. 6B, compare with 6A). There is reduced staining in the vicinity of stripe2-snail, particularly in one or two cells straddling the presumptive mesoderm/mesectoderm boundary (arrowhead, Fig. 6B). It is conceivable that this slight reduction in Delta expression helps trigger Notch signaling (see Discussion).

The activation of Notch leads to the induction of E(spl) genes such as m8, which encode transcriptional repressors that block the expression of proneural genes in the Achaete-Scute complex (Bailey and Posakony, 1995; Lecourtois and Schweisguth, 1995; Nakao and Campos-Ortega, 1996; Nellesen et al., 1999). T3, or lethal of scute, is normally expressed in a series of lateral stripes in the neurogenic ectoderm of wild-type embryos (Kosman et al., 1991). T3 stripes are expressed throughout mutant embryos derived from Toll^rm9/Toll^rm10 females (Fig. 6C). The stripe2-snail transgene creates a gap in this staining pattern (Fig. 6D), which might help define a zone of Notch signaling, as Achaete-Scute activators can inhibit Notch target genes (Heitzler et al., 1996).

snail is initially expressed in a relatively broad pattern that extends into ventral regions of the presumptive neurogenic ectoderm. This pattern is refined during cellularization, and the final borders coincide with the boundary between the presumptive mesoderm and mesectoderm (data not shown). The refinement process is also observed in transgenic embryos that contain an ectopic anterior-posterior Dorsal nuclear gradient (Fig. 7A-C). Before nuclear cleavage 14, the snail expression pattern exhibits a ‘fuzzy’ border (Fig. 7A). This border is refined by the completion of cellularization (Fig. 7B), and sim expression is detected shortly thereafter (Fig. 7C). Perhaps the early snail refinement process serves to control the temporal onset of sim expression. When broad, the Snail repressor keeps sim off, but after refinement sim can be activated in the domain where snail was transiently expressed.

This study provides further evidence that Notch signaling is essential for the formation of the mesectoderm at the boundary between the mesoderm and neurogenic ectoderm. Two different Notch target genes were examined: m8 expression appears to depend almost exclusively on Notch signaling, whereas sim is a conditional Notch target gene that is activated only in cells containing Dorsal. Evidence is presented that Snail functions as both a repressor and an indirect activator of Notch signaling. In particular, a transient stripe of the Snail repressor creates a domain of Notch signaling in apolar embryos that contain low, uniform levels of Dorsal. We discuss how Snail induces Notch signaling and also represses Notch target genes, and thereby specifies localized lines of sim and m8 expression in the mesectoderm.

A crucial finding of this study is that a stripe2-snail transgene induces ectopic expression of m8 and sim in both wild-type and Toll^rm9/Toll^rm10 mutant embryos, suggesting that the Snail repressor is actually playing a positive role in Notch signaling. Importantly, this stimulatory activity depends on the ability of Snail to function as a transcriptional repressor. Mutant forms of the stripe2-snail transgene that contain single amino acid substitutions in the two repression domains (PxDLSxK and PxDLSxR) fail to induce sim and m8 expression in either wild-type or Toll^rm9/Toll^rm10 mutant embryos (data not shown). By contrast, a stripe2-snail/hairy transgene that contains the Hairy repression domain continues to activate both sim and m8 in mutant embryos (see Fig. 5I, L).

The localized Snail repressor restricts Notch signaling to the mesectoderm of early embryos, presumably by directly repressing Notch target genes. Indeed, the sim 5′ regulatory region contains a series of high-affinity Snail repressor sites (Kasai et al., 1992). It is conceivable that Snail restricts Notch signaling in other developmental processes. For example, after its transient expression in the ventral mesoderm of early embryos, snail is reactivated in delaminating neuroblasts at the completion of germ band elongation (see Fig. 2D). At this stage, Notch signaling subdivides the neurogenic ectoderm into neurons and ventral epidermis. Notch is selectively activated in epidermal cells, where it induces the expression of E(spl) repressors that silence Achaete-Scute proneural genes (Bailey and Posakony, 1995). The localized expression of the Snail repressor in delaminating neuroblasts might help ensure neuronal differentiation by inhibiting Notch-specific target genes. Removal of snail along with two related linked zinc-finger repressors (Worniu and Escargot) leads to a reduction in the number of CNS neuroblasts (Ashraf et al., 1999; Cai et al., 2001).

We propose that Snail functions as a gradient repressor to restrict Notch signaling (summarized in Fig. 7D). In precellular embryos, the initial snail expression pattern is broad and extends into the future mesectoderm. During cellularization, the pattern is refined and snail expression is lost in the mesectoderm and restricted to the mesoderm. The early, broad snail pattern might create a broad domain of potential Notch signaling by repressing components of the Notch pathway, such as Delta and T3. After cellularization, Notch signaling is blocked in the presumptive mesoderm by sustained, high levels of the Snail repressor. However, Notch can be activated in the mesectoderm because of the loss of Notch inhibitors repressed by transient expression of the Snail repressor. According to this model, the dynamic snail expression pattern determines both the timing and limits of Notch signaling.

The results obtained in Toll^rm9/Toll^rm10 mutant embryos can be interpreted in the context of this Snail gradient model. The stripe2-snail transgene produces transient expression of the Snail repressor when compared with the endogenous gene. Consequently, the snail stripe creates an early zone of potential Notch signaling in Toll^rm9/Toll^rm10 by repressing Delta, T3, and other components of the pathway (Fig. 6). Perhaps the initially intense expression of the stripe2-snail transgene inhibits the activation of m8 and sim, but these genes are activated as expression from the transgene diminishes. Previous studies lend support to the idea that low levels of Snail can repress some target genes such as T3, while failing to repress others (Hemavathy et al., 1997).

We do not wish to imply that repression by a Snail gradient is the sole basis for positioning Notch signaling. Previous studies suggest that expression of neurogenic genes such as neuralized are also important for the restricted expression of sim and m8 within the mesectoderm (Hartenstein et al., 1992; Martin-Bermudo et al., 1995). Perhaps Neuralized and Snail act separately to establish precise lines of Notch signaling.

Notch, like other signaling pathways, is not dedicated to a particular developmental process (Artavanis-Tsakonas et al., 1999). While first identified as an agent of neurogenesis, it has been shown to play a role in the dorsoventral patterning of the wing imaginal disk, and the specification of the R7 photoreceptor cell in the adult eye (Cooper and Bray, 2000). We have provided additional evidence that Notch signaling specifies the mesectoderm at the ventral border of the neurogenic ectoderm in the early embryo. The regulation of sim may provide insights into how the Notch signaling cassette can perform so many disparate functions.

The analysis of Toll^rm9/Toll^rm10 embryos suggests that Dorsal functions synergistically with Notch signaling to activate sim expression. A stripe2-Notch^IC transgene induces strong sim expression in these embryos, even though they contain low levels of Dorsal and lack Twist. However, the same transgene barely activates sim when crossed into embryos that lack both Dorsal and Twist. By contrast, m8 is strongly expressed in these mutants, indicating m8 is primarily activated by Su(H)-Notch^IC and does not require Dorsal (Bailey and Posakony, 1995; Kramatschek and Campos-Ortega, 1994; Lecourtois and Schweisguth, 1995; Nellesen et al., 1999).

Perhaps the low levels of Dorsal present in the presumptive mesectoderm are not sufficient to activate sim. Instead, activation might rely on protein-protein interactions between Dorsal and the Su(H)-Notch^IC complex within the sim 5′ cis-regulatory region. sim contains a number of optimal Su(H) recognition sequences (Morel and Schweisguth, 2000; Nellesen et al., 1999); these might help recruit Dorsal to adjacent sites. By contrast, the stripe2-Notch^IC transgene appears to be sufficient to activate m8, even though it contains fewer optimal Su(H) binding sites than the sim 5′ cis-regulatory region (Morel and Schweisguth, 2000; Nellesen et al., 1999). Perhaps m8 is ‘poised’ for activation by ubiquitous bHLH activators that are maternally expressed and present throughout early embryos (e.g. Daughterless and Scute). Notch signaling might trigger expression upon binding of the Su(H)-Notch^IC complex. By relying on ubiquitous bHLH ‘co-factors’, Notch signaling may be sufficient to activate m8 in diverse cellular contexts. Accordingly, the differential regulation of sim and m8 by Notch signaling is combinatorial and depends on the distribution of distinct co-factors.

We thank Gary Struhl and Sarah Bray for sending Notch^IC and m8 cDNAs, respectively. We also thank Eric Lai in the Rubin laboratory for Delta EST clones and mutant flies, Steve Beckendorf for helpful suggestions, and Yutaka Nibu for help with the figures. We are grateful to Angela Stathopoulos for critically reviewing the manuscript, and for providing the embryos used in Fig. 7. This work was funded by a grant from the NIH (GM 46638).