The Snail protein family regulates neuroblast expression of inscuteable and string, genes involved in asymmetry and cell division in Drosophila

The Snail protein family contains a key signature of four to six conserved zinc fingers. These proteins function as transcriptional regulators during embryonic development, cancer formation and apoptosis (Hemavathy et al., 2000; Manzanares et al., 2001). Drosophila Snail was the first member identified and shown to be a repressor essential for mesoderm development (Grau et al., 1984; Nusslein-Volhard et al., 1984; Boulay et al., 1987; Kosman et al., 1991; Leptin, 1991; Ip et al., 1992; Hemavathy et al., 1997). Expression pattern analyses revealed that Snail is also present in embryonic wing disc primodia and neuroblasts (Alberga et al., 1991; Kosman et al., 1991; Leptin, 1991; Ip et al., 1994). Snail acts redundantly with Escargot, another member of this zinc-finger protein family, to control wing disc development (Whiteley et al., 1992; Hayashi et al., 1993; Fuse et al., 1996). However, single and double mutants of snail and escargot do not have a significant phenotype in the developing central nervous system (CNS) (Ashraf et al., 1999).

The absence of a CNS phenotype in the null mutants of snail is due to the redundant function provided by escargot and worniu, the third member of the protein family. These three genes are clustered in 35D1 region of the second chromosome. In deletion mutants that uncover these three genes, the ventral nerve cord is severely underdeveloped, as revealed by analysis of multiple neuronal markers (Ashraf et al., 1999). Some of the early CNS markers affected include fushi tarazu (ftz) and even-skipped (eve). ftz is initially expressed in many ganglion mother cells (GMCs) and later in many neurons (Goodman and Doe, 1993). In the deletion mutants, ftz expression in GMCs is almost abolished, and such defect can be rescued efficiently by transgenic expression of snail, worniu or escargot (Ashraf et al., 1999).

During CNS development, clusters of cells in the neuroectoderm receive instructions from proneural genes to become competent to form neuroblasts (Campos-Ortega, 1993). These proneural genes include the achaete-scute complex, ventral nervous system defective (vnd), intermediate neuroblast defective (ind) and muscle segment homeobox (msh) (Isshiki et al., 1997; Campos-Ortega, 1998; Chu et al., 1998; McDonald et al., 1998; Weiss et al., 1998). Through the process of lateral inhibition, which involves Notch-Delta signaling, one of the cells in each cluster is selected to become neuroblast and delaminates from the ectoderm (Bhat, 1998; Rooke and Xu, 1998). Delaminated neuroblasts have stem cell property, whereby each goes through repeated asymmetric cell divisions to generate multiple GMCs (Lu et al., 2000). Many genes that participate in neuroblast asymmetric division have been identified. For example, bazooka functions in the neuroectoderm to help polarize the cells (Schober et al., 1999; Wodarz et al., 1999). During or soon after delamination, genes such as inscuteable, miranda and staufen are expressed (Ikeshima-Kataoka et al., 1997; Li et al., 1997; Shen et al., 1997). One of the functions of these genes is to control the subcellular localization within the neuroblasts of prospero mRNA and Prospero protein, which are segregated preferentially into GMCs after cell division. Prospero is a key factor in determining GMC fate, regulating the expression of neural genes such as ftz and the single round of cell division that produces postmitotic neurons (Doe et al., 1991; Vaessin et al., 1991; Li and Vaessin, 2000). As snail and worniu have extensive expression in neuroblasts, and GMC and neuronal marker expression is defective in mutants that have the snail family locus deleted, we surmised that Snail family of proteins may function at a regulatory step in neuroblast or GMC development (Ashraf et al., 1999).

We show here that the absence of ftz and eve in GMCs and neurons is probably due to misregulation of early steps of neuroblast function. The snail family deletion mutant embryos exhibit normal early neuroblast delamination. The delaminated neuroblasts, however, have significantly lower level of inscuteable RNA expression. The expression of other genes involved in asymmetric division, including miranda, staufen and prospero appears to be normal. Consistent with the defect of inscuteable expression, the asymmetric localization of prospero RNA is disrupted and the strong Prospero protein staining in GMCs, normally a result of asymmetric division, is lost. All of these defects can be rescued by transgenic expression of Snail, Worniu or Escargot. Thus, the establishment of neuroblast asymmetry is partially dependent on the Snail family of proteins. We have also observed that the phenotype of loss of Prospero protein staining in GMCs is more severe in the snail family genes deletion mutants than in the inscuteable mutants. Therefore, Snail family may have functions in addition to the regulation of inscuteable. Accordingly, we have found that expression of neuroblast-specific string RNA and string promoter-lacZ reporters (Lehman et al., 1999) are also dependent on Snail protein family. The Prospero expression in GMCs of snail family mutant embryos can be rescued by transgenic expression of inscuteable and string, suggesting that activation of inscuteable and string are key functions of Snail in neuroblasts. We further demonstrate that the co-repressor interaction motifs (Nibu et al., 1998a; Nibu et al., 1998b) of Snail are essential for rescue of the CNS phenotypes, indicating that Snail probably acts as a repressor in the CNS and activates target genes indirectly. Together, the results support the idea that both neuroblast cell division and asymmetry are regulated by the Snail family of proteins, perhaps by repression of a yet to be identified target gene that normally functions to suppress inscuteable and string transcription.

Fly stocks were maintained at 25°C using standard cornmeal-yeast-agar medium. The y w⁶⁷ stock was used for all P-element-mediated transformation. The transformation constructs were co-injected with the Δ2-3 transposase helper plasmid. Genetic crosses that established the rescue lines were performed as described previously (Ashraf et al., 1999). Briefly, the individual rescue transgenes on the third chromosome were crossed with the osp29 mutant chromosome and stable lines were established. Chromosomes that contain two transgenes were obtained by meiotic recombination. The recombined third chromosome containing two transgenes was then crossed with the osp29 chromosome. The string-lacZ 5.3 transgene is located on the third chromosome and was crossed to the deletion mutant background by genetic crosses. The string-lacZ 6.4 and UAS-string transgenes are located on the second chromosome and were recombined with the osp29 mutant chromosome. Female flies transheterozygous for osp29 and string transgenes were collected and mated with a second chromosome balancer line (y w; BcElp/CyO) males. Male offspring that had red eyes and curly wings were collected and mated with y w; osp29/CyO and sna^HG31/CyO females to test for lethality over the mutant chromosomes. The crosses that produced no straight wing progeny were further tested by examining morphological defects that are similar to snail mutants and by in situ RNA staining for lacZ or string, confirming the presence of transgenes. Df(1)sc^B57, vnd^Δ38, ind^Δ79.3, ind^RR108, msh^Δ68 and insc^P72 mutant alleles were used in gene expression analyses (Kraut et al., 1996; Chu et al., 1998; McDonald et al., 1998; Weiss et al., 1998). UAS-string (Neufeld et al., 1998) was obtained from Bloomington stock center. The deficiency lines have been described in detail elsewhere (Ashraf et al., 1999).

snail cDNA fragments with dCtBP-binding site mutations were isolated from pSK(+)snail M1, M2 or M12 (Nibu et al., 1998a; Nibu et al., 1998b) by digestion with AscI. The fragments were blunted with T4 polymerase and cloned into the KpnI and XbaI (both blunted) sites of pCaSpeR-Snailp vector (Ashraf et al., 1999). This vector contain 2.8 kb of the snail promoter. The generation of transgenic rescue constructs of snail, worniu and escargot were described previously (Ashraf et al., 1999). pSnailpGal4 plasmid was constructed by cloning the Gal4 fragment (KpnI/XbaI) from pGATN into the same sites of pCaSpeR-Snailp vector.

RNA in situ hybridization was performed as described previously (Hemavathy et al., 1997). Antibody staining was performed essentially as previously described (Ashraf et al., 1999). For localization of Snail in CNS, affinity purified polyclonal antibody (guinea pig) was used at 1:5 dilution. The Prospero monoclonal antibody was used at 1:1 and Hb polyclonal antibody (guinea pig) at 1:400. The secondary antibodies were obtained from the Jackson Laboratory and used at 1:1000 for anti-rabbit, 1:400 for anti-mouse and 1:1000 dilutions for anti-guinea pig. Anti-phosphorylated H3 antibodies (rabbit) were purchased from Upstate Biotechnology and used at 1:200 dilution. Embryo sectioning was performed by embedding the stained embryos in Epon plastic (Hemavathy et al., 1997). The plastic embedded embryos were cut as sections of 3 μm thickness.

Both snail and worniu have extensive expression in neuroblasts, while that of escargot is transient and sparse. Furthermore, based on genetic analysis, snail and worniu have more important role than escargot in the regulation of CNS development (Ashraf et al., 1999). As loss of ftz expression in GMCs was the earliest CNS defect observed, we carefully examined the expression of snail and worniu in GMCs. A better understanding of the patterns should help us to predict their possible functions in either neuroblasts or GMCs. In situ hybridization revealed that worniu RNA, in contrast to the extensive expression in neuroblasts, is present in only a small number of GMCs (Fig. 1A, compare with 1B). Even in later staged embryos, when there should be multiple GMCs surrounding each neuroblast, we could detect the staining in no more than one small cell next to each neuroblast (data not shown). The limited staining in the GMCs is probably due to the segregation of some RNA from the parental neuroblast. Once the GMC is formed, the active transcription of worniu probably ceases. We also examined the protein and RNA expression of snail. The results showed that there is also very limited expression of snail in GMCs. We rarely detected snail RNA-containing GMCs next to neuroblasts (data not shown). Consistent with RNA expression, antibody staining revealed that the protein is predominantly in the neuroblasts (Fig. 1C,D).

We then examined whether the neuroblast expression of snail and worniu is regulated by proneural genes. Such a result would place the snail family in the well established genetic hierarchy that controls early neuroblast differentiation. The scute^B57 deletion mutant uncovers the three pro-neural genes: achaete, scute and lethal of scute. In this mutant, the expression of worniu in neuroblasts was significantly reduced (Fig. 1F). Only a few neuroblasts within each segment exhibited staining, and the expression level was substantially lower than in the wild type. The expression of worniu is also regulated by vnd and ind, such that in these mutant embryos the whole ventral and intermediate columns of staining were missing (Fig. 1G,H). In the msh^Δ68 mutant, no abnormal expression of worniu was detected (data not shown). Previous results have shown that the neuroblast expression of snail is slightly affected in achaete-scute and vnd mutants but is not affected in a daughterless mutant (Ip et al., 1994; Skeath et al., 1994). In ind and msh mutants, we observed Snail protein expression in many neuroblasts but the spatial pattern was rather disorganized (data not shown). In summary, most of the proneural genes tested have profound effects on the expression of worniu, and have detectable but lesser effects on that of snail. The predominant expression of snail and worniu in neuroblasts and their regulation by proneural genes suggest that the snail family genes may have important functions within neuroblasts.

In mutants containing deletions that uncover escargot, worniu and snail, many early neuroblast markers are normal, but ftz expression in GMCs is abnormal (Ashraf et al., 1999). The regulation of ftz depends on Prospero, a homeodomain protein that controls GMC fate (Doe et al., 1991; Vaessin et al., 1991; Li and Vaessin, 2000). Prospero protein and mRNA are preferentially segregated to GMCs from the neuroblast through the process of asymmetric division (Lu et al., 2000). Genes that are involved in asymmetric segregation of Prospero include inscuteable, miranda and staufen. Therefore, we examined the expression of these possible Snail family target genes in neuroblasts.

We used mutant embryos collected from deficiency strains that uncovers the 35D1 chromosomal region including the snail family genes [for detail descriptions of the Df(2L)osp29 and other deletions, see Ashraf et al. (Ashraf et al., 1999)]. In wild-type embryos, the expression of inscuteable can be detected in delaminating neuroblasts. After delamination, many neuroblasts show localization of the inscuteable RNA (Fig. 2A,B) (Li et al., 1997). Embryos homozygous for the osp29 deletion, however, had significantly lower level of the RNA and the staining was detected in a much smaller number of neuroblasts (Fig. 2C,D). Transgenic copies of snail, worniu or escargot efficiently rescued the expression of inscuteable RNA (Fig. 2E-H), demonstrating that it is the uncovering of the snail family of genes in the deletion that causes the phenotype. The rescue transgenes were under the control of the 2.8 kb snail promoter, which contains the neuroblast expression element (Ip et al., 1992; Ip et al., 1994; Ashraf et al., 1999). A 1.6 kb snail promoter construct (Ip et al., 1992) that contains the mesoderm element but lacks the CNS element could not rescue the defect (data not shown), demonstrating that expression of the transgenes within neuroblasts is essential for the function.

In contrast to that of inscuteable, the miranda RNA pattern and level were very similar in wild-type and osp29 embryos (Fig. 3A,B), suggesting that the RNA expression of miranda is independent of Snail. The Miranda protein was also present in the mutants, and some localization was detectable but less prominent when compared with that of wild type (Fig. 3C,D). The staufen RNA is expressed ubiquitously, with enhanced expression in neuroblasts of wild-type embryos. The overall RNA level of staufen also appeared normal in the mutant (data not shown). The prospero RNA was similarly detectable in the mutant although expression was delayed; in stage 9 mutant embryos, the level was slightly lower, but in older mutant embryos the staining was stronger and similar to that of wild type (Fig. 3F,H). However, one defect we observed was the loss of localization of the prospero RNA (Fig. 3H). Such localization occurs in wild-type neuroblasts just before mitosis (Fig. 3G), and is dependent on Inscuteable and Staufen (Li et al., 1997). Therefore, the localization defect is consistent with the reduction in inscuteable expression. The subcellular localization defect of prospero was rescued in the presence of transgenic snail family (Fig. 3I), demonstrating that the localization phenotype in the osp29 mutant was caused by loss of snail family genes.

The segregation of Prospero protein into GMCs from neuroblasts is a critical event during asymmetric cell division. As inscuteable plays a role in the segregation of prospero gene products into GMCs, we examined whether there is Prospero protein in GMCs of mutant embryos. Prospero protein staining can be easily detected in many wild type GMC nuclei (Fig. 4A). The staining was largely absent in the deletion that uncovers the snail family locus (Fig. 4B); only a few cells with the size of normal GMCs had clear nuclear staining. A band of cells along the midline also had Prospero staining (Fig. 4B, bracket), but these cells probably represent an expansion of the midline (see also Fig. 3F for clear staining of RNA). It has been well documented that in all snail mutants there is derepression of the mid-line determinant single-minded in the blastoderm stage embryo (Nambu et al., 1990; Kosman et al., 1991; Kasai et al., 1992).

To determine whether there are defects within GMCs in addition to loss of Prospero, we examined the expression of Hunchback (Kosman et al., 1998), which is present transiently in early neuroblasts and later in many GMCs (Fig. 4D). In the deletion mutant, the Hunchback protein in GMCs was also absent (Fig. 4E, bracket), while staining in cells surrounding the amnioserosa appeared normal. Transgenes of snail, worniu and escargot rescued the staining of Prospero and Hunchback (Fig. 4C,F; data not shown), indicating that these GMC determinants are downstream of the Snail family. The results also suggest that the regulation of ftz by the Snail family is indirect, probably through an earlier event such as segregation of Prospero from neuroblast to GMC.

If the misregulation of inscuteable in the deletion mutant is the cause of the loss of Prospero and ftz expression in GMCs, the expression of inscuteable should correct the defects even in the absence of Snail family of proteins. We crossed a line carrying an inscuteable transgenic construct driven by the 2.8 kb snail promoter into the osp29 deletion genetic background. However, the rescue of Prospero expression in GMCs was variable and not nearly as strong as those embryos expressing the snail family transgenes (data not shown, but see Fig. 7). This suggests that inscuteable may not be the only important target gene of Snail. Another line of evidence supporting the idea of an additional target gene comes from the comparison of the phenotypes in osp29 and inscuteable mutant embryos. In inscuteable mutants, the Prospero crescent is formed but the mitotic spindle rotation is randomized (Kraut et al., 1996; Li et al., 1997). As a result, the Prospero protein frequently is present both in neuroblasts and GMCs (Fig. 3J). This phenotype is less severe than the almost total loss of Prospero GMC staining in osp29 deletion mutant (Fig. 4B). Therefore, we surmised that in addition to the misregulation of inscuteable, there may be other defects that lead to the more severe phenotype in the deletion mutants.

One possibility that may explain the more severe phenotype in snail family deletion mutants is additional defects in cell division. Neuroblasts are arrested at the G2/M transition at the embryonic cell cycle 14 (Edgar and O’Farrell, 1990; Edgar, 1995). After delamination, a pulse of string (which encodes a Cdc25 phosphatase homolog) expression in neuroblasts drives the cells to enter mitosis. We examined the expression of string RNA in whole-mount mutant embryos, but the result was ambiguous, owing to the dynamic, high level expression in ectoderm and other tissues, which obscures the signal in the neuroblast cell layer (data not shown) (Edgar et al., 1994). We therefore used tissue sectioning in order to better view the expression of string in neuroblasts. The sections clearly showed expression of string RNA in wild-type neuroblasts at stage 9 embryos (Fig. 5A). There are consistently three to four neuroblasts on each side of the midline that exhibit staining. This neuroblast expression appeared very faint in the osp29 mutant embryos (Fig. 5B), and most sections did not show staining in neuroblasts while expression in ectoderm appeared normal. The presence of wor and esg transgenes in the deletion mutant background led to accumulation of string RNA in some neuroblasts (Fig. 5C), suggesting a positive role for Snail family in regulating string expression.

We also used string promoter-lacZ reporters that have more specific expression in neuroblasts to confirm the above observation. Promoter analysis demonstrated that many modular regulatory elements are present in a 50 kb region of the string locus (Lehman et al., 1999). The string 6.4 line contains a 6.4 kb genomic fragment which is originally located approximately 10 kb from the transcription start site of string. The string 5.3 line contains 5.3 kb of DNA, which is originally located approximately 20 kb from the transcription start site. Both lines have strong expression in early and late neuroblasts (Fig. 6A,B,G,H) (Lehman et al., 1999). In osp29 deletion mutant, the lacZ expression of string 5.3 was almost undetectable (Fig. 6C,D). The string 6.4 lacZ staining was also largely abolished (Fig. 6I,J). The expression of string 5.3 was partially rescued and the string 6.4 was efficiently rescued by snail family transgenes, even when the transgenes were heterozygous (Fig. 6E,F,K,L). The results obtained together demonstrate that transcription activation of string is at least partly under the control of Snail family of proteins.

If regulation of string is an important downstream event of Snail family of proteins, then cell division of neuroblasts should be affected in the absence of these proteins. We therefore examined the mitotic process by staining for phosphorylated histone H3, which reveals condensed chromosomes. In wild-type embryos, although the neuroblasts do not exhibit highly synchronized mitosis, anti-phosphoH3 staining can be detected in multiple cells (Fig. 7A). In the osp29 mutant embryos, such staining was consistently reduced (Fig. 7B). The use of prospero RNA to mark the neuroblast layer and the use of tissue sectioning (Fig. 7C-F) provided further support for the idea that the mutant embryos had reduced mitosis in neuroblasts.

We further demonstrate by genetic rescue experiments that the severe CNS defects are likely due to a combination of loss of inscuteable and string expression. Similar to the results obtained for inscuteable, transgenic expression of string alone had weak and variable effect in the rescue of Prospero expression in GMCs (data not shown). When both inscuteable and string were simultaneously expressed in neuroblasts of osp29 mutants using the UAS-Gal4 system, clear staining of Prospero in many cells resembling GMCs was observed (Fig. 7H). The staining was particularly apparent alongside the expanded midline, characteristic of mutant embryos with no Snail function in early mesoderm. The results support the idea that both inscuteable and string are relevant targets of the Snail family.

A clearly demonstrated in vivo function of Snail is transcriptional repression. The repression function is mediated through the recruitment of dCtBP (Drosophila C-terminal binding protein), which acts as a co-repressor for Snail to regulate target genes such as rhomboid, lethal of scute and single-minded (Nibu et al., 1998a; Nibu et al., 1998b). There are two conserved P-DLS-R/K motifs in Snail (Fig. 8), as well as in Worniu and Escargot, and they have been shown to be critical for recruiting dCtBP (Nibu et al., 1998a; Nibu et al., 1998b; Ashraf et al., 1999; Hemavathy et al., 2000). Mutations of these motifs abolish the repressor function of Snail in the blastoderm. To gain insight into the molecular mechanism of how Snail regulates CNS development, we introduced into the osp29 deletion background transgenic copies of snail which had the dCtBP interaction motifs mutated. M1 contains the N-terminal motif mutation and M2 contains the C-terminal motif mutation (Fig. 8). The expression of inscuteable and ftz was examined. The assay showed that the double mutant (M12) lost most of the ability to rescue (Fig. 8G,H), and M1 had lost some ability to rescue (Fig. 8C,D). However, M2 functioned quite efficiently, closer to that of the wild-type protein, to rescue inscuteable and ftz expression (Fig. 8E,F, compare with Fig. 2F for inscuteable and with Ashraf et al. (Ashraf et al., 1999) for ftz). These results demonstrate that the dCtBP interaction motifs are essential for the Snail function in the CNS, consistent with the idea that Snail acts as a repressor in neuroblasts to regulate gene expression. Thus, the activation of inscuteable and string by the Snail family may be indirect.

We have demonstrated that the Snail family of proteins function within neuroblasts to control CNS development through regulation of two determinants in the asymmetry and cell division pathways (Fig. 9). One of the downstream events is the regulation of inscuteable expression. In the absence of Snail family of proteins, inscuteable RNA level is substantially reduced. The mutant embryos also lose the asymmetric localization of prospero RNA within the neuroblasts. Inscuteable is required to anchor Staufen, which in turn binds and localizes prospero RNA (Li et al., 1997). Thus, it is conceivable that Snail family regulates the expression of inscuteable, which is essential for the asymmetric localization and segregation of prospero gene products. The loss of prospero gene products in GMCs is probably responsible for the loss of ftz expression we observed previously (Ashraf et al., 1999). A recent publication also reported that in deletions that uncover the snail family genes, the apical crescent of Prospero protein is formed but the basal localization is absent (Cai et al., 2001). This loss of basal accumulation of Prospero, as well as the randomization of spindle rotation, has been shown to be caused by the misregulation of inscuteable, owing to the loss of the Snail family (Cai et al., 2001). Our observations are consistent with and complementary to those conclusions.

While the loss of inscuteable may help to explain the localization defects of prospero gene products, the regulation of inscuteable is not the only function of Snail family of proteins. Together with Miranda and Staufen, Inscuteable helps to localize prospero RNA and Prospero protein (Li et al., 1997; Shen et al., 1997; Lu et al., 2000). Inscuteable is also required for spindle orientation (Kraut et al., 1996; Kaltschmidt et al., 2000). Previous reports have shown that in inscuteable mutants, the apical crescent of Prospero is still formed, but the transport to basal side is much delayed. In the meantime, the spindle rotation during metaphase is randomized. As a result, Prospero protein sometimes segregated correctly into GMCs. In the osp29 deletion mutant embryos, however, Prospero protein staining in GMCs is largely absent. Thus, the phenotype of deleting snail family is much more severe than that of inscuteable. As the expression of miranda appears normal, additional target genes are probably regulated by the Snail family.

Transcriptional regulation of string is a key event in controlling embryonic cell cycle 14-16 (Edgar, 1995). At this stage, the maternal and zygotic supplies of other cell cycle regulators drive all postmitotic cells through S phase. Maternal String, however, has been depleted or degraded during early divisions. Thus, the embryonic cells including neuroblasts are arrested at G2/M transition. Once the neuroblasts have delaminated, a pulse of zygotic string transcription leads to the activation of M phase and the cells go through mitosis. We have shown that the expression of string is defective specifically in the neuroblasts, and such defect can be partially rescued by the Snail family of proteins. We also find that two neuroblast-specific regulatory elements of string are at least partially dependent on Snail family for activity. The results strongly indicate a positive involvement of Snail family of proteins. The regulation of string predicts a neuroblast cell division defect, and staining for phosphorylated H3 in the osp29 deletion supports such an interpretation. Most importantly, while transgenic expression of either inscuteable or string did not rescue Prospero expression in GMCs substantially, the combination of both transgenes consistently rescued Prospero staining in GMCs in the snail family deletion background. Our results support the idea that both inscuteable and string are important downstream targets, and that Snail family has an important role in modulating the asymmetry and cell division of neuroblasts.

Although inscuteable and string are at present the two most proximal downstream targets, there is no evidence that they are direct targets of the Snail family transcription factors in neuroblasts. Published results support the theory that Snail is a transcriptional repressor. Snail, Worniu and Escargot all contain two dCtBP-binding motifs (Ashraf et al., 1999), and mutations of the dCtBP interaction motifs in Snail abolish the repressor activity at blastoderm stage (Nibu et al., 1998a). Thus, our results of dCtBP binding motif mutants of Snail suggest that Snail family of proteins function as repressors in neuroblasts. We attempted to examine the expression of some neural markers in dCtBP mutant embryos, but the severe morphological defects in post-gastrulation stages precluded a conclusive interpretation [dCtBP also functions as co-repressor for segmentation determinants such as Krüppel and Knirp (Nibu et al., 1998b; Keller et al., 2000)]. Although we favor the role of Snail family of proteins as repressors, it is formally possible that they can activate target gene expression. First, repression of known target genes by Snail in the early embryos is not sufficient to explain the gastrulation phenotype associated with the snail mutants (Hemavathy et al., 1997). Second, dCtBP can act as an anti-repressor by antagonizing the Groucho co-repressor function when binding the Hairy repressor (Phippen et al., 2000). Third, snail family genes are expressed at approximately the same time as that of inscuteable and string, leaving very little time for the transcription and translation of an intermediate regulator. Therefore, it is also possible that the Snail-dCtBP interaction can lead to direct activation of inscuteable and string. An analysis of the promoters of inscuteable and string and the associated proteins will shed some light on the regulatory mechanism.

We thank Bill Chia, Xiaohang Yang, Fumio Matsuzaki and Chris Doe for sharing reagents and unpublished observations regarding the asymmetric division pathway; Joseph Weiss and Matt Scott for ind mutant flies; Kalpana White for the vnd mutant flies; Yoshimasa Yagi for msh mutant flies; Ethan Bier for scratch and deadpan plasmids; John Reinitz and David Kosman for Hb antibody; Bruce Edgar for string transgenic lacZ lines; Rachel Kraut for inscuteable mutant and UAS-inscuteable flies; and Mike Levine for snail mutant plasmids. We also thank Kirugaval Hemavathy for affinity purified Snail antibody, Xiaodi Hu for generating dCtBP mutant transgenic lines, and Xiaodi Hu and Gregory Hendricks for help with embryo sectioning. Comments on the manuscript by David Arnosti are sincerely acknowledged. This publication was made possible by grant HD36240 from NICHD; its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NICHD. Y. T. I. was a Scholar of the Leukemia and Lymphoma Society of America.