Within the Drosophila embryo, the formation of many neuroblasts depends on the functions of the proneural genes of the achaete-scute complex (AS-C): achaete (ac), scute (sc) and lethal of scute (l’sc), and the gene ventral nervous system defective (vnd). Here, we show that vnd controls neuroblast formation, in part, through its regulation of the proneural genes of the AS-C. vnd is absolutely required to activate ac, sc and l’sc gene expression in proneural clusters in specific domains along the medial column of the earliest arising neuroblasts. Using ac-lacZ reporter constructs, we determined that vnd controls proneural gene expression at two distinct steps during neuroblast formation through separable regulatory regions. First, vnd is required to activate proneural cluster formation within the medial column of every other neuroblast row through regulatory elements located 3′ to ac; second, through a 5′ regulatory region, vnd functions to increase or maintain proneural gene expression in the cell within the proneural cluster that normally becomes the neuroblast. By following neuroblast segregation in vnd mutant embryos, we show that the neuroectoderm forms normally and that the defects in neuroblast formation are specific to particular proneural clusters.
The cellular complexity of the central nervous system (CNS) of the Drosophila embryo arises as a result of the generation, specification and asymmetric division of neuroblasts (NBs; reviewed by Campos-Ortega, 1993). Neuroblasts (NBs) are singled out from equivalence groups termed proneural cell clusters within the ectoderm and then delaminate into the interior of the embryo. Concomitant with the process of NB delamination, spatial cues apparently bestow unique identities upon individual NBs (Chu-LaGraff and Doe, 1993). NBs segregate in an invariant spatiotemporal sequence, consisting of five waves (SI-SV; Campos-Ortega and Hartenstein, 1985; Doe, 1992), to form a rigidly stereotyped sub-epidermal pattern of neural stem cells. After delaminating, NBs divide repeatedly to create the neurons and glia that make up the larval CNS.
The process of NB formation depends on the functions of the proneural genes of the AS-C and the gene vnd (Jimenez and Campos-Ortega, 1979, 1987; White, 1980; Ghysen and Dambly-Chaudiere, 1988). Removal of the genetic function of either the AS-C or vnd results in the loss of roughly 25% of all segmental NBs (Jimenez and Campos-Ortega, 1990). Since embryos doubly mutant for the AS-C and vnd lack roughly 50% of all segmental NBs, it has been suggested that the AS-C and vnd control the formation of non-overlapping sets of NBs (Jimenez and Campos-Ortega, 1990).
A large body of work has illuminated the developmental regulatory mechanisms that control both the expression of the proneural genes of the AS-C and NB formation (reviewed by Ghysen and Dambly-Chaudiere, 1988; Campos-Ortega and Jan, 1991,Simpson, 1990; Artavanis-Tsakonas and Simpson, 1991; Jimenez and Modolell, 1993). ac, sc and l’sc each encode basic helix-loop-helix type transcription factors (Villares and Cabrera, 1987; Alonso and Cabrera, 1988; Martin-Bermudo et al., 1993) and are initially expressed within the primordium of the embryonic CNS, the neuroectoderm, in a reproducible array of cell clusters, termed ‘proneural clusters’, from which single NBs later arise (Cabrera et al., 1987; Romani et al., 1987; Martin-Bermudo et al., 1991; Skeath et al., 1992). Expression of one or more of these genes within the cells of a cluster appears to bestow on these cells a general neural competency. A ‘proneural cluster’ constitutes an equivalence group where all cells can, although only one cell will, become a NB. The generation of the initial clustered pattern of proneural gene expression is governed by the combined action of the gene products of the pair-rule and dorsal-ventral patterning genes. These proteins presumably act through a vast array of region specific enhancers found within the AS-C to carve out clusters of ac, sc or l’sc gene expression at precise and reproducible coordinates within the Drosophila embryo (Martin-Bermudo et al., 1991 and 1993; Skeath et al., 1992). Within each cluster, the cell that comes to express the proneural gene(s) to the highest level is favored to become the NB or neural stem cell (Cubas et al., 1991; Ruiz-Gomez and Ghysen, 1993). Once chosen, the NB, which retains proneural gene expression, initiates a cell-communication pathway (lateral inhibition) mediated by the gene products of the neurogenic genes. The process of lateral inhibition removes proneural gene expression and, hence, neural competency from the remaining cells of the cluster (Lehmann et al., 1983; Cabrera, 1990; Skeath and Carroll, 1992; reviewed by Campos-Ortega, 1993 and Simpson, 1990).
Despite the detailed knowledge of the regulation of and the roles played by the ac, sc and l’sc genes during NB formation, little is known about the influence vnd has on this process, except that it is required for 25% of all NBs to segregate (Jimenez and Campos-Ortega, 1990). For example, is vnd directly involved in neurogenesis or does it influence the nervous system by affecting some other aspect of development? If direct, does vnd act in parallel with or in tandem with ac, sc or l’sc to create proneural clusters? Does it act at a later step to specify which cell within each proneural cluster becomes the NB, or like ac, sc or l’sc does it function more than once during the process of NB formation? In order to begin to elucidate the role vnd plays during the initial steps of embryonic neurogenesis, we determined its effect on the initial pattern of ac, sc, and l’sc proneural clusters and of the ten SI NBs that segregate from these proneural clusters. To our surprise, we found that vnd is required for the formation of ac, sc, and l’sc proneural clusters and is specific to the medial column of every other NB row. We used two different ac reporter constructs to determine that vnd controls proneural gene expression at two distinct steps during NB formation through separate regulatory regions. First, vnd acts, directly or indirectly, through regulatory regions 3′ to ac to activate gene expression in proneural clusters within the medial column; then, through regulatory regions 5′ to ac, vnd functions either to heighten proneural gene expression in the cell that becomes the NB or to maintain proneural gene expression in this cell. By following NB segregation in embryos mutant for vnd we show that the effect of vnd on SI NB segregation parallels its effect on proneural cluster formation.
MATERIALS AND METHODS
The following fly strains were used: B40 (Martinez and Modolell, 1991) and 101H10 (Skeath et al., 1992) ac-lacZ reporter transformant lines, vnd6, Df(1)svr; Df(1)y3PLsc8R, Df(1)scB57, FM7-ftzlacC balancer stock; Oregon R was the wild-type strain used (See Lindsley and Zimm (1992) for descriptions of mutant stocks). These strains were obtained from the labs of Juan Modolell, Peter Gergen, Kalpana White, Michael Young, Y.N and L.Y. Jan and from the Bloomington Stock Center.
In situ hybridization and immunochemistry
ac, sc, l’sc and lacZ transcripts were detected in appropriately staged embryos by hybridization with digoxigenin-labeled antisense RNA probes specific for each transcript made with the Genius 4 kit (Boehringer-Mannheim) following the protocol of Jiang et al. (1991). Double-antibody labeling experiments were carried out as described by Skeath et al. (1992). For detection of sna protein two monoclonal antibodies 1SN G5 and 2SN 5H (kindly provided by Audrey Alberga; Mauhin et al., 1993) were used together, each at a 1:10 dilution; for detection of en expression the INV4D9 monoclonal antibody (kindly provided by Nipam Patel, Carnegie Institute of Embryology, Baltimore, MD) was used at a 1:10 dilution. For detection of hb protein expression a polyclonal antiserum (kindly provided by James Langeland, University of Wisconsin-Madison) was used at a dilution of 1:400. For detection of β-galactosidase, a polyclonal antiserum was used at a dilution of 1:1000. In the experiments where ac, sc and l’sc transcript and sna protein expression were examined in vnd mutant embryos, an FM7-ftzlacC blue balancer was used to identify unambiguously mutant embryos. In order to analyze 101H10 reporter gene expression in the absence of vnd function, we crossed a stock carrying 4 copies (this stock contains homozygous viable inserts on the second and third chromosomes) of the 101H10 construct to Df(1)svr females. Virgin females from this cross were then mated to males carrying four copies of the 101H10 construct and the 101H10 expression pattern was examined in their progeny. Mutant embryos were identified by co-labeling with l’sc and lacZ antisense probes; those embryos that did not express l’sc were of the Df(1)svr genotype. B40 expression in vnd mutant embryos was determined by crossing virgin females heterozygous for vnd to B40/CyO males and analyzing lacZ expression in the resulting embryos. Approximately one quarter of these embryos exhibited the loss of lacZ gene expression within the medial column.
Overview of proneural gene expression and SI and SII NB segregation
Our work has focused on elucidating the genetic regulatory mechanisms that govern proneural gene expression during early neurogenesis in the Drosophila embryo. In this report, we focus on the role of the vnd gene in regulating the expression patterns of the proneural genes ac, sc and l’sc during SI NB segregation in the Drosophila embryo. We will first summarize briefly the patterns of ac, sc and l’sc gene expression during, and the pattern of NBs generated by, the first two waves of NB segregation. Previous studies have mapped precisely the clustered expression of ac, sc and l’sc prior to and concomitant with SI NB segregation (Martin-Bermudo et al., 1991; Skeath and Carroll, 1992; Skeath et al., 1992). The registration of proneural gene expression in the neuroectoderm prior to SI NB segregation is best described using an orthogonal grid of 12 squares (see Fig. 1A; left). Before SI NB segregation, ac and sc are co-expressed in four cell clusters (roughly 5-7 cells are found in each cluster) per hemisegment - the medial and lateral clusters of rows B and D (Fig. 1A; left and Fig 2A); whereas l’sc is expressed in a circular pattern of eight adjoining proneural clusters - the medial, intermediate and lateral clusters of rows A and C and the medial and lateral clusters of row D (Fig. 1A; left and Fig. 2B). l’sc expression comes on slightly later in the intermediate cluster of row A than in all other SI clusters; in the embryo shown in Fig. 2B,l’sc has yet to be activated in this cluster. Within every cluster proneural gene expression is quickly restricted to a single cell, the presumptive NB. The registration of proneural gene expression is retained by the NBs that segregate from each cluster (Fig. 1A; right) with one exception: the lateral NB of row B (NB3-5) expresses l’sc even though the cluster from which it arises does not (Martin-Bermudo et al., 1991).
Shortly after the SI NBs (Fig. 1B; left) have formed, a second wave of NB segregation occurs (SII; Campos-Ortega and Hartenstein, 1985; Doe, 1992). In contrast to SI NBs, which form predominantly in the medial and lateral columns (Fig. 1B; left), the majority of SII NBs form within the intermediate column (Fig. 1B; right). SII NBs arise from proneural clusters expressing only the l’sc gene (Martin-Bermudo et al., 1991). The prior segregation of the SI NBs allows one to predict unambiguously where SII NBs will form within the NB array. For example, the SI NBs MP2, 3-5, 3-2 and 5-3 encircle the location where the SII NB 4-2 will arise (Fig. 1B; right; Fig. 4).
vnd controls proneural cluster formation within highly specific AP and DV domains
Although the AS-C and vnd are thought to control the formation of largely complementary sets of NBs, the SI NB MP2 does not form in the absence of either the AS-C or vnd (Jimenez and Campos-Ortega, 1990). This suggested to us that the genes of the AS-C and vnd may interact to control the formation of this and other NBs. To investigate this possibility, we determined the expression patterns of the ac, sc and l’sc transcripts during SI NB segregation in vnd mutant embryos. vnd was found to be required for each transcript to be expressed within the proneural clusters of the medial column of rows B and D (Fig. 2; data not shown for sc). The loss of vnd gene function had very little effect on proneural gene expression medially in rows A or C and no effect on gene expression in the lateral column (Fig. 2C,D). Since both ac and sc are normally expressed only in the medial and lateral columns of rows B and D (Fig. 2A), there is a complete absence of ac/sc expression in the medial column in embryos mutant for vnd (Fig. 2C); no change to ac/sc expression is observed within the lateral column. Similarly, in vnd mutant embryos, l’sc expression is lost from the medial, but not the lateral, column of row D (inset to Fig. 2D). l’sc is, however, still expressed in a normal pattern in rows A and C (compare Fig. 2C and 2D), although its appearance within these rows is delayed slightly. The apparent absence of l’sc expression in the lateral column of row D in Fig. 2D is due to the slightly older age of this embryo in relation to the one shown in the inset. l’sc expression normally disappears from row D before it does from rows A or C. The normal pattern of l’sc expression in rows A and C, the normal appearance of the ventral midline and the formation of most of the medial and intermediate SI NBs of rows A and C (see below) in vnd mutant embryos suggest that vnd is not required within the medial column for cell viability or fate specification. Rather, the specific absence of proneural clusters from the medial column of rows B and D indicates that vnd is required along precise domains of both the anterior-posterior (AP) and dorsal-ventral (DV) axes to activate proneural cluster formation.
vnd activates proneural gene expression through separate regulatory regions at two distinct levels: proneural cluster formation and NB segregation
In order to define more precisely the influence of vnd on proneural gene expression, we determined the expression pattern of two ac-lacZ reporter constructs in embryos mutant for vnd. These two reporter constructs contain either 10 kb of DNA immediately 3′ to ac (101H10; Skeath et al., 1992) or 3.8 kb of DNA immediately 5′ to ac (B40; Martinez and Modolell, 1991). Within the embryonic CNS, 101H10 drives gene expression in row B, and to a lesser extent, in row D proneural clusters (Fig. 3A; Skeath et al., 1992); while B40 initially directs gene expression within the medial and lateral column of row B, generally in single cells - the presumptive row B SI NBs - before or during NB segregation (Fig 3C; Panganiban, G., Skeath, J. B. and Carroll, S. B., unpublished data; data not shown). The expression of both constructs in the embryonic CNS is independent of AS-C function (Panganiban, G., Skeath, J. B. and Carroll, S. B., unpublished data). Thus, any alterations to the patterns driven by these constructs in embryos mutant for vnd cannot be attributed to the loss of proneural gene expression that occurs in this background.
In vnd mutant embryos, the expression patterns driven by each construct are altered. In the case of the 3′ 101H10 reporter gene fusion, lac-Z expression in the medial proneural clusters is almost completely abolished; a very low level of striped lacZ expression can still be observed that connects the lateral clusters of rows B and D in these embryos (Fig. 3B). Similarly, 5′ B40 reporter gene expression is absent from the medial, but not the lateral presumptive NBs of row B (Fig. 3D). This is not due simply to the failure of these NBs to form because the expression of B40 is not obligately coupled to NB formation (Panganiban, G., Skeath, J. B. and Carroll, S. B., unpublished data). In the absence of the entire AS-C (Df(1)scB57), the medial SI NB of row B (MP2) hardly ever forms (2.5%; n=38 hemisegments; see also Jimenez and Campos-Ortega, 1990); yet, in Df(1)scB57 embryos B40 drives gene expression in the cells that would normally become MP2 (Panganiban, G., Skeath, J. B. and Carroll, S. B., unpublished data). These results show that vnd controls ac (and most likely proneural gene expression in general) at two distinct levels. First, through regulatory regions 3′ to ac, vnd directly, indirectly or cooperatively activates ac gene expression within the medial proneural clusters of rows B and D. Then, through regulatory regions 5′ to ac, vnd appears to increase or maintain proneural gene expression in the cell that becomes the NB. As the level of proneural gene activity appears to be a key determinant of which cell in a cluster becomes the NB, vnd may be involved in singling out which cell within the cluster becomes the NB.
vnd and the genes of the AS-C control the formation of overlapping sets of SI NBs
The dramatic alterations to proneural gene expression in vnd mutant embryos led us to determine the effects of removing vnd or AS-C gene function on SI NB formation. Such experiments have been performed previously (Jimenez and Campos-Ortega, 1990), however, the presence or absence of specific SI NBs was not assayed. To follow SI NB segregation in vnd and AS-C mutant embryos, we used monoclonal antibodies generated against the snail protein product (generously provided by Audrey Alberga; Mauhin et al., 1993). The snail gene is expressed in all NBs as they arise within the developing CNS (Fig. 4B; Alberga et al., 1991; Kosman et al., 1991). Identical results were obtained using an antibody directed against the hunchback (hb) protein product (J. B. S. data not shown) which also marks all NBs (Jimenez and Campos-Ortega, 1990). In vnd mutant embryos, the medial SI NBs of rows B and D (MP2 and NB 7-1) do not form (Fig. 4G, arrows); whereas all lateral NBs form normally (Fig. 4G). At a low frequency (<25%; n>70 hemisegments) NBs from the medial and intermediate columns of rows A and C (NB 2-2, 3-2, 5-2, and 5-3) do not arise. This variable loss of NBs from rows A and C may result from the observed delay of l’sc gene expression in these rows or from an effect of removing vnd function that is independent of AS-C function. In embryos carrying the deficiency Df(1)y3PLsc8R, there is no proneural gene expression within the proneural clusters of row B (Martin-Bermudo et al., 1991; Skeath et al., 1992) and the medial NB (MP2) of row B forms roughly 14% of the time (79 hemisegments scored), while the lateral NB (NB 3-5) forms less than 50% of the time (79 hemisegments scored; arrow-heads, Fig. 4F); all other SI NBs form normally. In the complete absence of the AS-C, the medial NB of row B (MP2) rarely forms and that of row D (NB 7-1) forms less than 10% of the time (n=38 hemisegments).
The relationship of these wild-type and mutant NB patterns to proneural cluster formation is illustrated by inspection of the dynamics of sna protein expression during NB segregation. In wild-type embryos, the antibody initially marks all neuroectodermal cells (Fig. 4A), but only NBs remain labeled (Fig. 4B). This is best seen during SII NB segregation. The SII NB 4-2 arises from a cell cluster located between the SI NBs MP2, 3-5, 3-2 and 5-3 (Doe, 1992; Chu-Lagraff and Doe, 1993; Fig. 1B). Before NB 4-2 segregates, a cluster of ‘sna’ positive neuroectodermal cells is clearly visible above the position where NB 4-2 will form. These cells are in the ectodermal (Fig. 4C), not the NB (Fig. 4D), cell layer. Once NB 4-2 has formed, ‘sna’ expression is restricted to this cell (Fig. 4E). sna protein is therefore expressed in proneural clusters and follows identical expression dynamics as the ac, sc and l’sc genes. Importantly, these sna/hb ‘proneural clusters’ arise normally in the absence of AS-C and/or vnd function. For example, in Df(1)y3PLsc8R and vnd (Fig. 4F, arrowheads; Fig. 4G arrows) embryos, sna positive ‘proneural clusters’ are found in the neuroectodermal cells of the medial column of rows B and D and the medial and lateral columns of row B, respectively, even though in these backgrounds NBs rarely delaminate from these positions. This demonstrates that the neuroectodermal expression of sna/hb arises normally in these mutants but the subsequent restriction of their expression to just the NB requires AS-C and vnd function. The generalized neuroectodermal expression of the sna/hb marker proteins in the absence of AS-C or vnd function may reflect a general neural competency of these cells that normally requires the function of the AS-C and vnd genes to be realized.
We have focused on the role of the gene vnd during NB formation in the Drosophila embryonic CNS. We have shown that vnd is required to activate the expression of the ac, sc and l’sc genes in proneural clusters in precise AP (rows B and D) and DV (medial column) domains. vnd controls proneural gene expression at two distinct steps during neuroblast formation through separable regulatory regions (Fig. 5): (i) vnd activates proneural cluster formation through regulatory regions 3′ to ac; and (ii) vnd apparently increases proneural gene expression within the cell that becomes the NB through regulatory regions 5′ to ac. Finally, we demonstrated that the effect of removing vnd function on proneural cluster formation is paralleled by its effect on NB formation - SI NBs from the medial column of rows B and D do not form.
vnd controls proneural cluster formation and the level of proneural gene expression in the presumptive NB
We have shown that vnd is required for proneural cluster formation during SI NB segregation in the medial column of rows B and D, but not in rows A or C, or in the intermediate or lateral columns (Fig. 5). That vnd is required for proneural cluster formation at all was surprising on two accounts. First, it had been suggested that the genetic activities of vnd and of the AS-C were required for the formation of non-overlapping sets of NBs (Jimenez and Campos-Ortega, 1990); a regulatory dependence of AS-C expression on vnd gene function was not expected. Secondly, the generation of the initial pattern of ac and sc proneural clusters can be explained on the basis of the combinatorial action of the gene products of the pair-rule and DV genes on the ac and sc genes (Skeath et al., 1992). That is, it did not seem necessary to invoke the existence of other genes to explain how the initial pattern of proneural clusters arises within the Drosophila embryonic CNS. The spatial specificity (medial column rows B and D) of vnd action on proneural cluster formation clearly shows that vnd does regulate, directly or indirectly, the initial activation of proneural clusters.
In addition to its role in proneural cluster formation, vnd may also help single out the cell within the cluster that becomes the NB (Fig. 5). This hypothesis arises from the properties that the B40 reporter construct displays in wild type, AS-C, and vnd mutant embryos. Initially, the B40 construct is activated during late stage 8 at low levels generally in single or, occasionally, two cells, within the ectodermal cell layer of the medial and lateral columns of row B (Panganiban et al., 1994). In each location, this cell, or one of the two cells, delaminates and enlarges to become a row B SI NB. B40 appears to be active in the cell that will become the NB before it segregates. Furthermore, expression of B40 within the cells normally destined to become the row B NBs does not require that these cells become NBs. For example, in embryos that harbor a deletion for the entire AS-C, the medial SI NBs of row B rarely ever form. However, in this background B40 drives gene expression within the cells that would normally give rise to these NBs (Panganiban et al., in preparation). Thus, in the absence of NB formation, the cells that would normally become the medial NBs of row B are still singled out, even though they appear to require AS-C activity to become NBs.
From the above described behavior of the B40 reporter construct, we can conclude that there is an AS-C independent process that acts prior to, or concomitant with, the process of NB commitment through regulatory regions 5′ of ac to increase selectively ac gene expression within the cells that become the medial NBs of row B. Since the cell within the proneural cluster that has the highest level of proneural gene activity is favored to become the NB (Cubas et al., 1991), then the factor(s) that activates B40/ac gene expression within the presumptive NBs of row B may be involved in singling out which cell becomes the NB. We have shown that the loss of vnd gene function abolishes B40 gene expression in the presumptive medial NBs of row B. vnd may act through regulatory regions 5′ to ac to increase ac gene expression within one cell of the proneural cluster and may therefore be involved in singling out this cell as the NB.
The function and regulation of the vnd gene
It appears that vnd acts at two levels during the process of NB formation and is required in a distinct subset of cells. With regard to proneural cluster formation, we can envision two ways in which vnd could influence ac gene expression through the 3′ regulatory region in the medial row B and D clusters (Fig. 5). First, vnd might be spatially regulated in a manner similar to ac and sc and selectively expressed in these clusters. Alternatively, vnd could be expressed in a broader domain but act as an obligate cofactor with the axis-patterning genes that activate proneural gene expression at specific anteroposterior positions within the medial column.
The expression of the B40 reporter gene in vnd mutant embryos suggests that vnd also functions later during NB segregation to help single out the cell that becomes the NB (Fig. 5), again selectively in the medial column of rows B and D. It is possible that like ac, sc and l’sc, vnd is expressed initially in cell clusters and then restricted to single cells. Retention of vnd expression in one cell would then increase proneural gene expression and favor this cell to become the NB. In this scenario, the dynamics of proneural gene expression would largely reflect those of the vnd gene. Clearly, the elucidation of the function, expression, and regulation of the vnd gene and its role in proneural cluster formation, NB segregation and NB identity (see below) awaits the cloning and characterization of the vnd gene product.
Regardless of the mechanism by which vnd activates proneural cluster formation and NB segregation, the fact that it selectively affects cells within the medial column of rows B and D raises the possibility that vnd is involved in specifying the medial identity of specific NBs. For example, in row B, the MP-2 and 3-5 SI NBs each arise from ac/sc-expressing clusters, MP-2 being the medial NB (Fig. 1). It is possible that vnd has a direct role in distinguishing the identity of MP-2 from 3-5 and other SI NBs. The selective role of vnd raises the question of whether other, as yet unidentified genes act within rows A and C and/or within the lateral and intermediate columns to activate proneural cluster formation and NB formation and/or to specify NB identity. Since the loss of the AS-C and vnd removes roughly 50% of all NBs (Jimenez and Campos-Ortega, 1990), it is possible that other genes act to promote neurogenesis and to confer NB identity. Some of these could be zygotic genes, such as vnd, or there could be some maternal contributions to early neurogenesis. For example, the maternal contribution of the neurogenic gene Notch is sufficient to mask the zygotic role for Notch during the first, but not the subsequent waves of NB formation within the embryonic CNS (Struhl et al., 1993; see also Perrimon et al., 1986; Noll et al., 1993). The identification of other specific ‘proneural genes’, if they exist, should clarify issues concerning both the generation of the full complement of NBs and perhaps the determination of NB identity.
We thank Juan Modolell and Sonsoles Campuzano for providing AS-C cDNA clones and the B40 ac-lacZ reporter construct; and the late Carlos Cabrera for providing AS-C cDNA clones. We are grateful to Audrey Alberga for providing monoclonal antibodies generated against the Snail protein, before publication. We thank Chris Doe for his comments on the manuscript and permission to use his NB map as part of Fig. 1. We also thank Dervla Mellerick-Dressler for her critical review of the manuscript. We thank Kathy Vorwerk for expert technical assistance; Leanne Olds and Stephen Paddock for help with constructing the figures; and, Jamie Wilson for help with the preparation of the manuscript. This work was supported by a National Institutes of Health (NIH) predoctoral traineeship (GM-07215) and a Damon Runyon-Walter Winchell Cancer Research Fellowship (DRG 1279) to J. B. S., a NIH postdoctoral fellowship (GM-14441) to G. P., a National Science Foundation Presidential Young Investigators Award, the Shaw Scholar’s Program of the Milwaukee Foundation, and the Howard Hughes Medical Institute.