The Drosophila EGF receptor controls the formation and specification of neuroblasts along the dorsal-ventral axis of the Drosophila embryo

‘A fertilized egg divides into two cells, these into four and so on, until billions of cells have been generated. Why do these cells not simply form a large heap, each cell like the other and perhaps resembling the original undivided egg cell?

Why, instead, do some cells become nerve cells, or bone cells or liver cells, or any other of the many differentiations of which cells are capable? Why…are they present not in haphazard disorder but in fixed, specific regions organized into a nervous system, bones, a liver, and all the other organs which themselves have fixed and specific locations?’

Curt Stern, American Scientist, 1954

A central theme in developmental biology is to elucidate the mechanisms that specify particular cell types as well as those that govern where and when these cell types form during animal development. The Drosophila embryonic CNS provides an ideal system in which to dissect the developmental programs that create cell-type specific patterns. The segmented portion of the Drosophila CNS arises from a simple uniform sheet of neuroectodermal cells located in the ventrolateral region of the blastoderm embryo (reviewed in Goodman and Doe, 1993). The neuroectoderm is sub-divided along the anteroposterior (AP) and dorsoventral (DV) axes into a precise orthogonal pattern of neuroectodermal or ‘proneural’ clusters. Cell interactions between cells within a cluster direct one cell into the neuroblast fate. These cells enlarge and delaminate into the interior of the embryo to form an invariant pattern of 30 neuroblasts per hemisegment (a hemisegment is a bilateral half of a segment and is the developmental unit of the segmented portion of the CNS). Each neuroblast divides via a specific lineage to produce a characteristic family of neurons and/or glia. In addition, each neuroblast expresses a distinct set of molecular markers. It is clear that within a hemisegment each neuroblast acquires a unique fate based on its position and time of formation. However, the genetic regulatory mechanisms that create the stereotyped neuroblast pattern and specify unique fates to cells at characteristic positions within this pattern remain unclear.

Neuroblast formation occurs in five temporally and spatially distinct waves (S1-S5; Doe, 1992). The first 15 neuroblasts develop during the first two waves, S1 and S2, and form an orthogonal array of four AP rows (1, 3, 5, 7) and three DV columns (medial, intermediate and lateral). 15 additional neuroblasts form during later waves of neuroblast segregation. Neuroblast formation is controlled primarily by the action of two sets of genes. The proneural genes of the achaete-scute complex (AS-C) promote neuroblast formation and are expressed in proneural clusters (Jimenez and Campos-Ortega, 1990; Martin-Bermudo et al., 1991; Skeath and Carroll, 1992). The neurogenic genes suppress proneural gene function and limit the number of neuroblasts that form from each cluster (reviewed in Campos-Ortega, 1993). The activity of the AP-and DV-axis patterning genes are thought to establish the pattern of AS-C-expressing proneural clusters in the neuroectoderm, which in turn dictates where and when neuroblasts form (Martin-Bermudo et al., 1991; Rao et al., 1991; Skeath et al., 1992). The role the segmentation genes play along the AP axis to pattern neuroblast formation is well established; however, less is understood about the mechanisms that pattern neuroblasts along the DV axis.

Within a hemisegment each neuroblast acquires a unique fate based on where and when it forms. Recent genetic and cell transplantation analyses indicate that neuroblasts inherit their identity from the cluster of neuroectodermal cells from which they delaminate (Chu-LaGraff and Doe, 1993; Skeath et al., 1995; Udolph et al., 1995). For example, along the AP axis the wingless and gooseberry segment polarity genes function within the neuroectoderm to promote neuroblast formation and to specify the fate of neuroblasts that develop adjacent to wingless-expressing cells or from gooseberry-expressing neuroectodermal cells (Chu-LaGraff and Doe, 1993; Skeath et al., 1995). In contrast to the AP axis, little is known about the developmental mechanisms that control neuroblast identity along the DV axis.

DER (also referred to as faint little ball and torpedo), which encodes a receptor tyrosine kinase, is a key determinant of cell fate along the DV extent of the neuroectoderm (reviewed in Schweitzer and Shilo, 1997). In DER mutant embryos, lateral cell fates replace ventral cell fates and CNS development is grossly disrupted. Recent genetic, molecular and biochemical analyses indicate that spitz and vein each encode ligands that activate DER independently of one another (Rutledge et al., 1992; Schweitzer et al., 1995; Golembo et al., 1996; Schnepp et al., 1996) and that DER uses the conserved RAS signal transduction pathway to transduce signals to the nucleus (reviewed in Seger and Krebs, 1995). Localized transcription of vein and rhomboid in the ventral domains of the neuroectoderm lead to localized DER activation (Bier et al., 1990; Golembo et al., 1996; Schnepp et al., 1996; Gabay et al., 1997a). rhomboid encodes a membrane-spanning protein (Bier et al., 1990) that appears to mediate DER signaling by helping to process Spitz from its inactive transmembrane form to its active secreted form (Golembo et al., 1996). vein encodes a secreted protein (Schnepp et al., 1996).

The earliest known function for DER in patterning the ventral ectoderm occurs after germ-band extension at approximately stage 10 (reviewed in Schweitzer and Shilo, 1997), when midline cells appear to secrete Spitz (Golembo et al., 1996). Secreted Spitz (s-Spitz) is thought to diffuse bilaterally to pattern the adjacent ventral neuroectoderm. However, recent results indicate that DER signaling is active and regulates gene expression in a discrete DV domain in the neuroectoderm during gastrulation well before the extended germ-band stage (D’Alessio and Frasch, 1996; Gabay et al., 1997a; Yagi and Hayashi, 1997). This suggests that DER may play a role prior to the production of s-Spitz by midline cells to pattern the neuroectoderm along the DV axis. However, loss of DER activity at this time has not been associated with any morphological defects.

In this paper, I show that DER functions to establish the correct pattern, and helps to specify the individual fates, of neuroblasts. Specifically, I show that in DER mutant embryos (1) intermediate column neuroblasts do not form, (2) medial neuroblasts are partially mis-specified and acquire some traits characteristic of lateral neuroblasts, while (3) lateral neuroblasts develop normally. I also show that active DER signaling is restricted to the medial and intermediate neuroectodermal columns. Furthermore, I demonstrate that the concomitant loss of rhomboid and vein yield CNS phenotypes indistinguishable from DER mutant embryos, even though the loss of either rhomboid or vein produces only minor CNS defects. In addition, I show that DER is activated twice in the neuroectoderm of the embryo: once at the onset of (and during) gastrulation by a combination of Spitz group and Vein activities, the second time after germband extension, presumably by Spitz activity produced by the midline. Together these results demonstrate that DER functions along the DV axis to promote the formation, patterning, and individual fate specification of neuroblasts within the Drosophila CNS.

Wild-type patterns of gene expression were examined in Oregon R embryos. Mutant lines used were: DER, alleles flb1K35, flb2C82 and flb2W74, provided by Alan Michelson (Clifford and Schupbach, 1994), and l(2)03033 (Spradling et al., 1995); veindddRy and veinγ3 (Schnepp et al., 1996); spitzIIA, StarIIN (Nüsslein-Volhard et al., 1984); rhomboid7M (Mayer and Nusslein, 1988) and rhomboiddel-1 (Bier et al., 1990); H162 is an enhancer trap line inserted into the seven-up gene (Mlodzik et al., 1990) and is referred to as svp-lacZ. The double mutant rhomboid7M veindddRy was generated by recombination.

Single-and double-label immunohistochemistry and RNA in situ analysis was performed as described in Skeath et al. (1992). For the active MAP kinase antibody, biotinyl tyramide (NEN Life Science Products) was used to amplify the signal following the manufacturer’s protocol. The following antibodies were used at the indicated dilutions: rabbit anti-Eve (1:5000; Frasch et al., 1986); mouse anti-Eve 2B8 (1:50; Patel et al., 1994); mouse anti-Engrailed 4D9 (1:5; Patel et al., 1989); mouse anti-βgal (1:2000; Promega); rabbit anti-βgal (1:15000; Cappel); mouse anti-Pros MR1A (1:4; Spana and Doe, 1995); mouse anti-Ftz (1:1000; Kellerman et al., 1990); rabbit anti-Odd (1:5000; provided by Ellen Ward); rabbit anti-Deadpan (1:300; Bier et al., 1992); and mouse anti-Active MAP kinase (1:2000; Gabay et al., 1997a; Sigma).

To identify mutations that disrupt embryonic CNS development I screened a collection of approximately 1300 second chromosomal lethal P element lines obtained through the Berkeley Drosophila Genome project (Spradling et al., 1995) for defects in the CNS expression pattern of the eve gene (Doe et al., 1988; Patel et al., 1989; Broadus et al., 1995). Among the cells that express Eve in the wild-type CNS are the progeny of two medial neuroblasts, the aCC/pCC neurons (neuroblast 1-1) and the U/CQ neurons (neuroblast 7-1), as well as the progeny of intermediate neuroblast 4-2, the RP2 and RP2 sib neurons. In the screen I uncovered one P element mutation, l(2)03033, that causes a loss of essentially all Eve-positive RP2/RP2 sib neurons (Fig. 1). This P element mapped to cytological position 57F1-2 in the right arm of the second chromosome and was known to be inserted within the DER locus (Spradling et al., 1995). To verify that lesions in DER result in a nearly complete loss of RP2 motoneurons I obtained three additional DER mutants, including the DER null allele, flb^1K35 (Clifford and Schupbach, 1994). Essentially all Eve-positive RP2 motoneurons are absent from embryos homozygous mutant for each DER allele (Fig. 1; Table 1). I used the flb^1K35 null allele for all subsequent experiments; I refer to flb^1K35 embryos as DER mutant embryos.

The absence of the Eve-positive RP2 motoneuron observed in DER mutant embryos could arise from a number of different developmental defects. For example, the intermediate neuroblast that produces RP2 may not form, it may not divide or it may acquire an inappropriate identity and not produce RP2. DER is expressed and is active in the neuroectoderm prior to neuroblast formation (Zak et al., 1990; Gabay et al., 1997a,b) but is not expressed in neuroblasts or their progeny prior to or during RP2 motoneuron formation (Zak et al., 1990). Given this and DER’s demonstrated role in establishing DV patterning, it seemed plausible that DER controls early steps of CNS development by regulating the DV limits of gene expression in the neuroectoderm prior to neuroblast formation.

The initial step of CNS development involves the activation of the AS-C proneural genes in a precise pattern of proneural clusters as gastrulation commences (Cabrera et al., 1987; Martin-Bermudo et al., 1991; Skeath and Carroll, 1992). To investigate whether DER regulates AS-C expression in the neuroectoderm, I followed the expression patterns of the achaete (ac) and lethal of scute (l’sc) genes in DER mutant embryos. Loss of DER causes specific defects to the DV registration of ac and l’sc gene expression in the neuroectoderm; however, I observed no defects to the AP registration of ac or l’sc gene expression. In wild-type embryos during stages 8/9, ac is expressed in the medial and lateral, but not intermediate, clusters of rows 3 and 7 (Skeath and Carroll, 1992); l’sc is expressed in the medial and lateral, but not intermediate, clusters of row 7 and in the medial, intermediate and lateral clusters of rows 1 and 5 (Martin-Bermudo et al., 1991). A single neuroblast subsequently forms from each proneural cluster. In DER mutant embryos ac gene expression expands into the intermediate column in rows 3 and 7 and l’sc expression expands into the intermediate column in row 7; l’sc is expressed normally in rows 1 and 5 (Fig. 2; not shown for l’sc). The lateral limits of ac and l’sc gene expression in the neuroectoderm are unaltered in DER mutant embryos. The changes to the DV registration of ac and l’sc gene expression in DER mutant embryos suggest that neuroectodermal cells in the intermediate column change their fate. Both ac and l’sc are normally expressed in the medial and lateral columns in the affected rows, thus the phenotype is consistent with intermediate cells acquiring either a lateral or a medial fate. D’Alessio and Frasch (1996) recently showed that msh-1, which is expressed exclusively in the lateral column, expands into the intermediate column in DER mutant embryos. In this context, it appears that ac and l’sc expression expand from the lateral column into the intermediate column in the absence of DER.

The defects observed to the pattern of ac-and l’sc-expressing proneural clusters in DER mutant embryos suggested that neuroblast formation would also be altered. I used the anti-Deadpan antibody, which labels all neuroblasts (Bier et al., 1992), to follow neuroblast formation in wild-type and DER mutant embryos. Normally early forming neuroblasts develop in three mediolateral columns by mid-stage 9 (Fig. 3A). However, in DER mutant embryos only two neuroblast columns develop. With respect to the midline these neuroblasts form in the medial and intermediate columns (Fig. 3B). However, within a hemisegment each neuroblast column contains four SI neuroblasts as per wild-type medial and lateral columns; note that only two SI intermediate column neuroblasts form in wild-type embryos (Fig. 3; Table 2). This, together with the apparent transformation of intermediate column cells towards a lateral fate (Fig. 2) and the absence of RP2/RP2 sib, which form from an intermediate neuroblast (Fig. 1; Table 1), suggest that the intermediate neuroblast column does not form in DER mutant embryos.

To assay the fate of the neuroblasts that do form in DER mutant embryos, I used various molecular markers that are expressed in specific subsets of neuroblasts. In wild-type embryos the row 1 medial (1-1) and lateral (2-5) but not intermediate (3-2) S1 neuroblasts express the Odd-skipped (Odd) protein during stage 9 (Fig. 4A) (Broadus et al., 1995). In the absence of DER function two row 1 neuroblasts form and they both express Odd (Figs 3B, 4B) suggesting that these are neuroblasts 1-1 and 2-5. I used Eve expression to determine whether these neuroblasts acquire medial and/or lateral traits. Eve is expressed in the progeny of neuroblast 1-1 (100%; n=438) but not 2-5 (Broadus et al., 1995). In the absence of DER function, neuroblast 1-1 produces Eve-positive progeny in 88.0% of neuroblasts examined (n=457); I never observed Eve-positive progeny in the neuroblast 2-5 position. These results suggest that, for the most part, neuroblast 1-1 acquires its correct fate and are consistent with neuroblast 2-5 acquiring its normal lateral fate. However, 12.0% of the time (n=466) neuroblast 1-1 fails to produce Eve-positive progeny and thus fails to follow its normal developmental path. Neuroblast formation and division in the medial column are essentially normal in DER mutant embryos (Table 2; data not shown). Thus, the simplest interpretation of these data is that loss of DER has no effect on the development of the lateral neuroblast 2-5, causes a low but consistent mis-specification of the medial neuroblast 1-1, and removes the intermediate neuroblast 3-2.

In contrast to Odd, most neuroblasts eventually express Seven-up-lacZ (Svp-lacZ; Doe, 1992). Each neuroblast activates Svp-lacZ at a specific time and expresses Svp-lacZ at a stereotyped quantitative level. In wild-type embryos immediately after S1 neuroblast formation, only neuroblasts 5-2 and 7-4 express Svp-lacZ, and they express Svp-lacZ at high levels (Fig. 5A). In equivalently staged DER mutant embryos most neuroblasts in the 5-2 position (54.3%; n=116) do not express Svp-lacZ (Fig. 5B) while both row 7 neuroblasts express Svp-lacZ strongly (Fig. 5B). Thus, the row 5 medial neuroblast (5-2) fails to acquire one of its specific traits while the row 7 medial neuroblast (7-1) acquires a trait characteristic of the row 7 lateral neuroblast. As development proceeds in DER mutant embryos, most neuroblasts in the 5-2 position eventually activate Svp-lacZ, although relative to wild type they express it at reduced levels. In addition, the neuroblast in the 7-1 position does acquire traits characteristic of its position. For example, in wild-type embryos neuroblast 7-1 divides to produce multiple Eve-positive progeny (99.8%; n=438) (Broadus et al., 1995) and in DER mutant embryos the majority of neuroblasts in the 7-1 position produce Eve-positive progeny (78.4%; n=444). Thus, in the absence of DER function medial neuroblasts can acquire appropriate traits. Nonetheless, a significant fraction of medial neuroblasts still fail to activate their appropriate gene expression profile.

And, in at least one case, a medial neuroblast, neuroblast 7-1, acquires traits characteristic of its lateral neuroblast neighbor. Thus, loss of DER function appears to cause a partial mis-specification of medial column neuroblasts and may cause medial neuroblasts to acquire some traits characteristic of lateral neuroblasts.

At progressively later stages of development, Svp-lacZ can be used to follow the fate of all SI lateral neuroblasts. In wild-type embryos, neuroblast 7-4 activates Svp-lacZ first followed sequentially by the lateral neuroblasts of row 1 (neuroblast 2-5; Fig. 5C), row 3 (3-5) and row 5 (5-6; Fig. 5E). In addition, as one moves posteriorly from neuroblast 7-4 each neuroblast expresses Svp-lacZ at a quantitatively lower level (Fig. 5E). In DER mutant embryos lateral column neuroblasts recapitulate the wild-type expression pattern of Svp-lacZ. For example, the lateral row 7 neuroblast activates Svp-lacZ first, at the time and level diagnostic of neuroblast 7-4 (Fig. 5B). Neuroblasts in rows 1, 3 and 5 then express Svp-lacZ in temporal order and at progressively lower levels as per wild-type lateral neuroblasts (Fig. 5D, F). Thus, lateral neuroblasts appear to form and to develop normally in DER mutant embryos. These results using Svp-lacZ and Odd expression to follow neuroblast fates, taken together with those using Eve, Ac and L’sc expression, indicate that in the absence of DER medial neuroblasts exhibit defects in their specification and intermediate neuroblasts do not form, while lateral neuroblasts develop normally.

To examine more closely the effect loss of DER has on the fate specification of medial neuroblasts I analyzed carefully the fate of MP2 in DER mutant embryos. I focused on MP2 because multiple molecular markers uniquely identify MP2 (Doe, 1992; Broadus et al., 1995; Parras et al., 1996; Skeath and Doe, 1996). MP2 normally localizes the Prospero protein to its nucleus and expresses the Fushi-tarazu (Ftz) and Odd proteins (Fig. 6; Table 1). At equivalent developmental stages no other neuroblasts express Ftz or Odd and all other neuroblasts localize Prospero to their cell cortex. In addition, MP2 divides nearly 1.5 hours after delamination while all other neuroblasts divide nearly immediately after delamination. In DER mutant embryos MP2 is often incorrectly specified. Approximately 50% of the neuroblasts in the MP2 position localize Prospero to the cell cortex (Fig. 6B1) or divide prematurely (Fig. 6B2; Table 1) and fail to activate Ftz or Odd (Fig. 6; Table 1). The remaining fraction of MP2s express Ftz and Odd and localize Prospero protein to the nucleus as per wild-type embryos (Fig. 6; Table 1). Thus, removal of DER causes half of all MP2s to acquire traits characteristic of other neuroblasts. Based on results detailed in the previous section, MP2 may acquire traits characteristic of neuroblast 3-5, its lateral neighbor. However, the present lack of markers specific for neuroblast 3-5 precludes a test of this hypothesis. Nonetheless, these results combined with those using Svp-lacZ demonstrate that DER helps to specify medial neuroblast fate and indicate that DER-independent mechanisms also function to specify medial neuroblast fate.

spitz and vein are expressed in the early embryo and appear to function in independent pathways to activate DER during embyrogenesis (Rutledge et al., 1992; Schnepp et al., 1996; Schweitzer and Shilo, 1997). Thus, it is possible that either Spitz or Vein or both activate DER to promote neuroblast formation and specification. rhomboid and Star encode for transmembrane factors (Bier et al., 1990; Kolodkin et al., 1994) that appear to promote the production of s-Spitz and to act in the same linear pathway as spitz (Golembo et al., 1996). To investigate the extent to which spitz, rhomboid and Star, as well as vein participate in DER-mediated control of early CNS development, I assayed neuroblast formation and specification in embryos singly mutant for each gene (Figs 1-3; Tables 1, 2). Early CNS development is essentially normal in embryos singly mutant for spitz, Star or vein. ac expression is restricted correctly to the medial and lateral columns (data not shown); an Eve-positive RP2 motoneuron forms at least 94% of the time (Fig. 1; Table 1) and medial neuroblast specification appears normal (Table 1). Embryos that lack rhomboid function exhibit more severe, yet still relatively mild, CNS defects. For example, ac expression expands incompletely into the intermediate column (Fig. 2C); three neuroblast columns form but there is a weak decrease in the formation of intermediate neuroblasts and their progeny (Fig. 3), and there is a weak mis-specification of medial neuroblasts (Table 1).

Although embryos singly mutant for any one spitz group gene or vein do not display an appreciable early CNS phenotype, it is possible that the spitz group genes and vein act together to mediate full activation of DER. Removal of one group or the other may cause only mild CNS defects because the other remains to activate DER. Alternatively, additional factors may activate DER to promote early CNS development. To determine the CNS phenotype of removing both spitz group activity and vein, I constructed a fly stock doubly mutant for rhomboid and vein. I chose rhomboid instead of spitz or Star because loss of rhomboid causes more severe CNS phenotypes than loss of either spitz or Star. Removal of rhomboid and vein produces CNS defects virtually indistinguishable from those observed in DER mutant embryos (Table 1): ac expression expands completely into the intermediate column in rows 3 and 7 (Fig. 2); only two neuroblast columns form (Fig. 3); the RP2 motoneuron almost never forms (Fig. 1); and, roughly half of MP2s are mis-specified (Table 1). The severity of these phenotypes correlates with the null DER allele, flb1K35, and are stronger than the moderate DER alleles, flb2W74 and flb2C82 (Table 1). The only minor difference I observe is a slightly greater decrease in the formation of the medial neuroblasts, MP2 and 5-2, in rhomboid vein mutant embryos relative to DER mutant embryos. These data suggest that the activity of the spitz group and vein are sufficient to account for all signals that activate DER during early CNS development.

DER signals through the conserved RAS pathway (reviewed in Seger and Krebs, 1995). One of the final effectors of the RAS pathway is MAP kinase. MAP kinase is activated by dual phosphorylation of threonine and tyrosine residues by MEK (Seger et al., 1992; Cobb and Goldsmith, 1995). The recent production of a monoclonal antibody that specifically recognizes the active, dual phosphorylated form of MAP kinase allows one to follow in situ the activation pattern of receptor tyrosine kinase pathways such as DER (Gabay et al., 1997a,b). To identify when and in which cells DER signaling is required to promote early CNS development, I assayed for the presence of active MAP kinase in wild-type, DER, vein, spitz and rhomboid vein mutant embryos (Fig. 7).

During early embryogenesis, active MAP kinase is present in two temporally and spatially distinct patterns within the neuroectoderm in wild-type embryos (Fig. 7A,D). Prior to and during gastrulation active MAP kinase is first found in two broad bilaterally paired longitudinal bands of cells that run down the length of the neuroectoderm (Fig. 7A; Gabay et al., 1997a,b). At stage 10, active MAP kinase is again expressed in the neuroectoderm in the most medial neuroectodermal cells that flank the midline (Fig. 7D). In DER and in rhomboid vein mutant embryos active MAP kinase is not present in either pattern (not shown). In embryos singly mutant for either vein or spitz, the first wave of active MAP kinase appears in its normal pattern although at reduced levels (Fig. 7B,C). However, the second wave of active MAP kinase is absent in spitz, but normal in vein, mutant embryos (Fig. 7E,F). Thus, early CNS defects correlate with the absence of the first but not the second wave of active MAP kinase as clear CNS defects occur in DER and rhomboid vein but not in spitz mutant embryos.

To determine the precise limits of the initial wave of active MAP kinase in the neuroectoderm I performed double-label immunofluorescence and cell count analyses. Immnofluorescence studies using Twist protein expression as a mesodermal marker (Thisse et al., 1988) indicate that active MAP kinase is excluded from the mesoderm, although in rare cases exceptional cells can be found that express both proteins (Fig. 7G-I). The medial to lateral width of the active MAP kinase stripes at stage 7 is 9.0 cells (s.d. 0.91, n=29; Fig. 7J). To estimate the width of the medial, intermediate and lateral columns at the same developmental stage I counted the width of the ac-positive medial and lateral clusters and the ac-negative intermediate clusters in similarly staged embryos. Note at this stage ac-positive cells in the medial column also abut mesodermal cells (Skeath and Carroll, 1994). From this analysis, I estimated that the width of the medial column is 5.7 cells (s.d. 0.66, n=20), the width of the medial and intermediate column is 8.8 cells (s.d. 0.76, n=25) and the width of the neuroectoderm (medial to lateral column) is 12.6 cells (s.d. 0.84, n=10). These data indicate that active MAP kinase is present in the medial and intermediate but excluded from the lateral neuroectodermal column (Fig. 7J).

DER signaling first becomes active prior to gastrulation in the medial and intermediate neuroectodermal columns. The results in this paper indicate that within the medial column DER functions to help specify the fate of medial neuroblasts. Within the intermediate column DER functions to promote neuroblast formation. Conversely, DER function is dispensable for the development of lateral column neuroblasts. Note that Udolph et al. (1998) have reached similar conclusions through the use of cell transplantation analyses.

To date research on DER-dependent regulation of DV patterning in the neuroectoderm has focused predominantly on the midline as the source of the signal, s-Spitz, that activates DER (reviewed in Schweitzer and Shilo, 1997 and references therein). The results in this paper show that DER acts before midline cell production of s-Spitz to pattern the neuroectoderm. (1) In DER mutant embryos, defects in neuroectodermal gene expression and CNS development occur before midline production of s-Spitz (this paper and D’Alessio and Frasch, 1996; Yagi and Hayashi, 1997). (2) The presence of DER-dependent active MAP kinase staining in the medial and intermediate columns prior to and during gastrulation correlates with the defects observed in the formation and specification of medial and intermediate neuroblasts in DER mutant embryos. (3) Early CNS defects correlate with the absence of the first but not second wave of active MAP kinase in the neuroectoderm: in DER mutant embryos both waves of active MAP kinase are absent and severe CNS defects arise, while in spitz mutant embryos only the second wave is absent and CNS development is grossly normal. (4) CNS development is essentially normal in single-minded mutant embryos in which midline cells do not form (unpublished observations). Thus, midline cell production of s-Spitz appears not to play an early role to form, pattern and specify neuroblasts along the DV axis. However, s-Spitz clearly plays a crucial role in later patterning events in the neuroectoderm (Schweitzer and Shilo, 1997 and references therein). Furthermore, as shown by Schnepp et al. (1996) and confirmed and extended in this paper, Vein acts with the spitz group to activate DER in the neuroectoderm prior to and during gastrulation. The simplest interpretation of these data is that two ligands, Vein and s-Spitz, activate DER prior to gastrulation within the medial and intermediate columns of the neuroectoderm. DER activity would then promote the formation of intermediate column neuroblasts and the fate specification of medial column neuroblasts. Thus, with respect to early DV patterning of the neuroectoderm one must now consider that DER acts at two separate times in two overlapping cell populations to pattern the neuroectoderm, and that two activating ligands act independently of each other to activate DER during the first wave of DER signaling.

How do individual neuroblasts acquire unique fates? Neuroblasts form from equivalence groups called proneural clusters that arise in an invariant orthogonal pattern in the neuroectoderm (Martin-Bermudo et al., 1991; Skeath et al., 1992). Each neuroblast acquires a unique fate and neuroblasts appear to acquire their identity from the neuroectodermal cluster from which they form (reviewed in Goodman and Doe, 1993). Thus, each neuroectodermal cluster should be programmed to produce a specific neuroblast. Recent results suggest a model whereby the superimposition of the activities of the segment polarity genes along the AP axis and the activities of DER, vnd and msh-1 along the DV axis could bestow a unique identity, in the form of differential gene expression or activity, upon each neuroectodermal cluster and thus on each neuroblast (Fig. 8).

Along the AP axis the segment polarity genes are expressed in transverse rows that divide each hemisegment into discrete AP domains. Within these domains segment polarity genes promote the formation and fate specification of neuroblasts (Patel et al., 1989; Chu-LaGraff and Doe, 1993; Zhang et al., 1994; Skeath et al., 1995; Bhat, 1996; Dunman-Scheel et al., 1997). For example, gooseberry enables row 5 neuroblasts to acquire fates different from neuroblasts in the adjacent anterior row (Skeath et al., 1995). Along the DV axis, the activity/expression of DER, vnd and msh-1 divides the neuroectoderm into the three mediolateral domains from which the three SI neuroblast columns form (Fig. 8). vnd and msh-1 expression is restricted, respectively, to the medial and lateral columns (Jimenez et al., 1995; Mellerick and Niremberg, 1995; D’Alessio and Frasch, 1996; Isshiki et al., 1997) and DER signaling is active in the medial and intermediate columns. vnd may enable medial neuroblasts to acquire fates different than intermediate neuroblasts, as the vnd phenotype is consistent with a duplication of intermediate fates at the expense of medial fates (Skeath et al., 1994; J. McDonald, C. Doe, D. Mellerick Dressler, personal communication; F. Jimenez and K. White, personal communication). msh-1 appears to help specify lateral neuroblast fates, as loss of msh-1 duplicates intermediate neuroblasts while lateral neuroblasts fail to develop correctly (Isshiki et al., 1997; Buescher and Chia, 1997). As shown in this paper, DER promotes the formation of intermediate neuroblasts and helps to specify the individual fate of medial neuroblasts. DER may function at or near the top of the hierarchy that controls neuroblast formation and fate along the DV axis, as DER activity restricts msh-1 expression to the lateral column (D’Alessio and Frasch, 1996) and may modulate Vnd activity medially, since Vnd contains five consensus MAP kinase sites (D. Mellerick-Dressler, personal communication). Clearly additional genes likely act with and/or downstream of DER, vnd and msh-1 to maintain this subdivision and potentially to subdivide the neuroectoderm further.

The expression patterns and phenotypes of the segment polarity genes and DER, vnd and msh-1 suggest that these genes play similar roles to pattern, form and specify neuroblasts along the AP and DV axes, respectively. Note that the superimposition of the expression/activity of these genes subdivides the neuroectoderm into an orthogonal pattern of cell clusters (Fig. 8). This pattern of cell clusters matches very closely, and likely represents, the actual pattern of neural equivalence groups, ‘proneural clusters’, from which individual neuroblasts arise. Each cluster expresses a unique combination of these patterning genes and the unique gene expression profile of a cluster likely determines or, at least, helps determine, the fate of each neuroblast that will later form from that cluster. The subsequent process of Notch-mediated lateral inhibition would then act within each cluster to produce a uniquely fated neuroblast (reviewed in Campos-Ortega, 1993).

The model can explain the early CNS defects observed in DER mutant embryos (Fig. 8). Loss of DER appears to remove the boundary between the intermediate and lateral columns and to cause intermediate column cells to acquire lateral cell fates. For example, msh-1 and ac gene expression appear to expand from the lateral into the intermediate column in DER mutant embryos (D’Alessio and Frasch, 1996; Yagi and Hayashi, 1997; this paper). AP subdivision of the neuroectoderm is normal at this stage. Thus, in a given AP row, cells within the intermediate and lateral columns appear to express the same combination of genes (Fig. 8). In each row, one early forming neuroblast forms from this domain and it acquires a lateral fate (Fig. 8). Thus, it is likely that this region represents a lateral neural equivalence group of twice its normal width. Notch-mediated lateral inhibition would then act within cells of this group to choose one as the neuroblast. vnd, which is initially expressed normally in DER mutant embryos (Gabay et al., 1996), likely specifies the fate of medial column cells and thus promotes the formation and, to some extent, the proper fate specification of medial neuroblasts in the absence of DER. Thus, a failure in the process of DER-mediated subdivision of the neuroectoderm appears to result in the formation of two rather than three neuroblast columns. An alternative, but not mutually exclusive, model to DER dependent subdivision of the neuroectoderm, is that DER normally functions to oppose Notch pathway activity during lateral inhibition in the neuroectoderm. DER signaling has been shown to oppose Notch function in other tissues (e.g. see Miller and Cagan, 1998). Thus, loss of DER function might increase effective Notch signaling and result in an overall decrease in neuroblast formation. This model alone, however, does not readily explain the specific loss of intermediate column neuroblasts in DER mutant embryos.

DER helps to specify the fate of medial neuroblasts, as loss of DER causes a failure of medial neuroblasts to display characteristic gene expression profiles and cell behaviors. Furthermore, in at least one case, loss of DER function causes a medial neuroblast to acquire traits characteristic of a lateral neuroblast; I never observed medial neuroblasts that acquired traits characteristic of intermediate neuroblasts. DER is clearly not the sole factor that promotes medial neuroblast fate as medial neuroblasts can develop normally in the absence of DER function. In fact, vnd may be the critical factor that determines medial neuroblast fates as its loss of function phenotype is consistent with a duplication of intermediate fates at the expense of medial fates (Skeath et al., 1994; J. McDonald, C. Doe, D. Mellerick Dressler, personal communication; F. Jimenez and K. White, personal communication). DER does not regulate the initial activation of vnd expression (Gabay et al., 1996), but as noted DER may affect medial neuroblast fate through modulation of Vnd activity.

With respect to MP2 specification, loss of DER produces an essentially identical phenotype to that caused by replacing the endogenous expression of ac and scute in MP2 with that of l’sc (Parras et al., 1996; Skeath and Doe, 1996). However, loss of DER function does not change the AP registration of AS-C gene expression. Thus, the activity of DER and the specific combination of proneural genes expressed in a neuroblast may act in parallel to specify medial neuroblast fate. If these pathways are partially redundant this might explain the observation that only a fraction of medial neuroblasts are mis-specified in the absence of DER function or upon mis-expression of the AS-C genes. In the future, it will be instructive to remove DER function in embryos in which l’sc gene expression replaces that of ac and sc in MP2 to determine the effect of removing both inputs on medial neuroblast fate.

Many additional factors likely contribute to neuroblast specification. For example, the initial phases of DER signaling and msh-1 expression within the neuroectoderm disappear prior to the segregation of late forming neuroblasts. Thus, additional factors must act along the DV axis to pattern and to specify the fate of these neuroblasts. Also, genes other than AS-C proneural genes are expressed in neuroblasts and the proneural clusters from which they form (Duffy and Gergen, 1990; Chu-LaGraff et al., 1995; McDonald and Doe, 1997). Some of these genes, like huckebein and runt, are known to control specific aspects of the fate of individual neuroblasts, while the effect of other such genes on neuroblast specification awaits investigation. Future experiments that address these and other questions will lead to an increasingly lucid picture of the genetic regulatory mechanisms that pattern and specify the fate of neuroblasts in the Drosophila embryonic CNS.

The initial phase of this work was initiated in the laboratory of Chris Doe and I am grateful for his support. I also thank Todd Laverty and Gerald Rubin and the BDGP for the use of the second chromosomal lethal P element collection, and Julie Gates and Carl Thummel for their generosity in sending me the collection. I thank Manfred Frasch, Barry Condron, Siegfried Roth, Harald Vaessin, Ellen Ward, Nipam Patel, Eric Spana and Ian and Diane Duncan for antibodies. In addition, I thank Kathy Matthews and the Bloomington Stock Center for providing many stocks and Amanda Simcox, Alan Michelson and Shigeo Hayashi for sending stocks and for helpful discussions. I also thank Julie Broadus, Chris Doe, Jocelyn MacDonald, Dervla Mellerick-Dressler, Fernando Jimenez and Kalpana White for communication of unpublished results. Finally, I thank Raphi Kopan, Ross Cagan, Susan Spencer, Chris Doe, Julie Broadus, Grace Panganiban, Teresa Orenic, Hemlata Mistry and Ellen Ward for critical comments on the manuscript. This work was supported by a grant from NINDS (RO1-NS36570). J. B. S is supported by the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation Award, DRS-9 and by a HHMI Research Resources Program for Medical Schools Junior Faculty Award #76296-538202.

Similar results have been obtained by Y. Yagi, T. Suzuki and S. Hayashi as those presented here.