Regulation of proneural gene expression and cell fate during neuroblast segregation in the Drosophila embryo

The neuroectoderm of insect embryos consists of a uniform sheet of cells, all of which possess the potential to become neuroblasts (NBs) (Bate, 1976; Doe and Goodman, 1985). Cell ablation studies performed on grasshopper embryos have shown that the fate of individual cells is guided by cell-cell interactions between neighboring cells (Taghert et al., 1984; Doe and Goodman, 1985). In the neurogenic region, one cell from a cluster of 5 –6 cells normally enlarges, delaminates and becomes a neuroblast (NB). If all of the cells of the cluster are ablated, no NB forms. However, if different portions of the cluster are ablated, leaving a number of neuroectodermal cells intact, a NB is always formed. Further, if one waits and ablates only the cell that is enlarging to become the NB, one of the remaining cells forms the NB. Finally, if the NB is ablated just before its first cell division, no new NB is formed. These experiments suggest that (1) the NB arises from these neuroectodermal cells; (2) every cell of the cluster shares a common NB-forming potential (i.e. an equivalence group); (3) local inhibition of the remaining cells by the enlarging NB ensures that only one NB arises from the equivalent group of cells; (4) all cells of the cluster retain their NB-forming potential at least while the NB is enlarging but lose this potential by the time the NB is about to divide. The molecular basis of these cellular events is not well understood. The central questions are: (1) which genes give the neuroectodermal cells their NB-forming potential?; (2) which genes are required to suppress this potential in the remaining cells of the cluster after one cell is chosen to become a NB? and (3) how do these genes interact at the molecular level to specify both the neural fate of the one cell chosen to become a NB and the non-neural/epidermal fate of the other cells?

Genetic analysis of Drosophila neurogenesis and epidermogenesis has identified a modest number of genes that function to allow a single neural precursor to arise from an initially equivalent group of cells (Stern, 1954; Garcia-Bellido and Santamaria, 1978; Garcia-Bellido, 1979; Lehmann et al., 1983). These genes can largely be grouped into two classes: (1) the proneural genes, which promote neurogenesis, and (2) the neurogenic genes, which suppress neurogenesis and facilitate epidermal development (for reviews see Ghysen and Dambly-Chaudiere, 1989 and Jan and Jan, 1990 and references therein). Genetic experiments performed on both Drosophila embryos and adult flies suggest that the proneural genes are required for the initial commitment of cells to the neural fate. In Drosophila embryos and adult flies homozygous for null alleles of the proneural genes of the AS-C, achaete (ac), scute (sc) and lethal of scute (I’sc), fewer than the normal number of sensory structures and neural precursor cells, NBs or sensory mother cells (the cells that give rise to the sensory structures), arise (Garcia-Bellido and Santamaria, 1978; Garcia-Belhdo, 1979; Jimenez and Campos-Ortega, 1979; Dambly-Chaudiere and Ghysen,.1987; Jimenez and Campos-Ortega, 1990). Although the mechanisms by which ac, sc, and I’sc promote neurogenesis appear to be the same, each of these genes directs the specification of largely independent but partially overlapping neural precursor cells (Garcia-Bellido and Santamaria, 1978; Garcia-Bellido, 1979; Jimenez and Campos-Ortega, 1979; Dambly-Chaudiere and Ghysen, 1988; Jimenez and Campos-Ortega, 1990). Conversely, an increased number of sensory structures and neural precursors are found in Drosophila embryos and adult flies carrying either duplications of the AS-C or hypermorphic alleles of ac or sc (Lindsley and Grell, 1968; Garcia-Bellido, 1979; Campuzano et al., 1985; Jimenez and Campos-Ortega, 1990). The transcript patterns of the ac, sc and I’sc genes strongly correlate with NB segregation and sensory mother cell formation (Cabrera et al., 1987; Romani et al., 1989). Molecular studies of these genes have revealed that they each encode proteins that possess a basic helix-loop-helix (bHLH) motif (Villares and Cabrera, 1987), which is found in a number of proteins involved in transcriptional regulation and cell determination. The HLH domain is required for homo- and heterodimer formation between HLH proteins and the basic region just to the amino-terminal side of the HLH domain appears to confer DNA-binding specificity and transcriptional activating properties on dimers of these bHLH proteins (Murre et al., 1989; Davis et al., 1990). In fact, recent results suggest that heterodimers between I’sc and daughterless, a universally expressed bHLH gene also required for neurogenesis, may transcriptionally activate certain genes within NBs (Cabrera and Alonso, 1991).

Once a cell is chosen to become a NB, it inhibits the neighboring cells of the cluster from following suit via a process termed lateral inhibition (for review see Simpson, 1990). Mutations in any one of the neurogenic genes appear to cripple this process and result in neural hyperplasia at the expense of epidermis (Lehmann et al., 1983). Many lines of evidence strongly suggest that the neurogenic genes function in a cell-communication pathway that ultimately suppresses neurogenesis (for review see Campos-Ortega, 1991 and references therein) . However, the molecular basis through which these genes prevent neurogenesis is not well understood. Initially, it was shown that I’sc protein accumulated only within NBs while I’sc RNA was detected in larger cell clusters (Cabrera, 1990). Further, it was shown that in certain neurogenic mutant backgrounds I’sc protein accumulated in all of the cells that expressed I’sc RNA (Cabrera, 1990). This suggested that the neurogenic genes opposed neurogenesis within the other cells of the cluster by preventing proneural protein accumulation within them. More recently, however, using,a different antibody preparation, I’sc protein has beén detected in wild-type embryos in essentially the same pattern as I’sc RNA (Martin-Bermudo et al., 1991). Thus, the dynamics of proneural protein distribution and the effect the neurogenic genes have on proneural protein expression are a matter of some dispute.

Here, we show the ac RNA and protein patterns to be essentially identical and that the dynamics of the ac expression pattern reflect at the molecular level the processes of singling out one cell from an initially equivalent cluster of cells to become a NB and of suppressing the NB-forming potential in the remaining cells of the cluster. We find that the ac protein is first expressed in a segmentally repeating pattern of clusters of 5 –7 ectodermal cells arranged in columns along the ventral neuroectoderm. Even though 5 –7 cells initially express ac protein only one, the NB, retains ac protein expression, while the other cells rapidly lose ac protein expression. Further, we show that once a cell is chosen to become a NB the neurogenic genes are required to suppress ac protein expression in the remaining cells of the cluster, since in embryos lacking any one of five neurogenic genes ac expression is not restricted to a single cell. Instead, all cells of each cluster retain ac expression, enlarge, delaminate and become NBs.

The following fly stocks were used: big brain^ID005, Delta^9Q, Enhancer of split^8D06, neuralised^9LI19, Notch^55e11 and ln(l)y^3PLsc^8R These stocks were obtained from the laboratories of Mark Muskavitch and Spyros Artavanis-Tsakonas, and the Bowling Green and Tiibigen Stock Centers.

A pet3a expression plasmid containing the 0.45 kb Smal to Pstl fragment of the ac coding region was generously provided to us by Tadashi Uemera. The ac protein fragment was prepared for immunizations and ELISA experiments by making a soluble extract from inclusion bodies yielding roughly 15 milligrams of soluble protein per 500 ml initial culture. By SDS-polyacrylamide gel analysis, the ac protein fragment was found to constitute between 80 and 90% of the total protein in these preparations.

Six female BALB/c mice were immunized with the ac protein fragment. For the first boost, 50 μg of protein was emulsified 1:1 with complete Freund’s adjuvant and injected intraperitoneally into the mice. The mice were subsequently boosted with 50 –100 μg of protein in PBS at roughly three to four week intervals. Six days after each boost the mice were tail bled and the sera was tested for reactivity against the ac protein by staining 0 –8 hour old embryos. After five boosts, serum from one mouse stained embryos in a pattern similar to the ac RNA pattern. Spleen lymphocytes from this mouse were fused to NS-1 myeloma cells following established protocols. Roughly 2000 hybridoma supernatants were screened by ELISA for reactivity against the ac protein fragment coated at a concentration of 2 μg/ml. The 158 ELISA-positive supernatants were then screened on 0 –8 hour embryos to determine which hybridoma colonies produced antibodies that recognized the ac protein fragment immuno-histochemically. Of 21 hybridoma colonies that produced antibodies that recognized the ac fragment immunohisto-chemically, 11 were subcloned and kept. Two of these 984A11C1 and 990E5F1, were used in this study.

Immunohistochemical detection of ac protein was carried out as described in Carroll et al. (1988). mAb 984A11C1 was used at a 1:3 dilution. For double-labelling studies, we essentially followed the protocol of Kania et al. (1990). Embryos were first incubated overnight with mAb 990E5F1 diluted 1:1 in PBT (l ×PBS; 0.1% Triton X-100; 1% BSA). After extensive washing for one hour, the embryos were incubated with biotinylated horse antimouse (Vector) for two hours at 4°C and then, after another hour of washes, the embryos were incubated with streptavidin-horseradish peroxidase conjugate (BRL) for one hour at 4°C. After 30 minutes of washes in PBT and another 30 minutes of washes in 100 mM Tris-HCl pH 6.8 the stain was developed in 100 mM Tris-HCl pH. 6.8 with 0.5 mg/ml of diaminobenzidene (DAB) and 0.002% H₂O2. After the reaction was stopped by the addition of 5 μl of sodium azide, the embryos were washed five times in PT (1 ×PBS; 0.1% Tween-20). The embryos were then stripped of the first set of antibodies by incubating them in 200 mM glycine-HCl pH 2.2 for five minutes. Glycine was removed from the embryos by five washes with PT. The embryos were then reblocked in PBT for three hours. For double labelling embryos with ac and en, embryos were then incubated overnight in mAb INV4D9 (kindly provided by Nipam Patel) diluted 1:1 in PBT. After extensive washing for one hour, the embryos were incubated for two hours at 4°C with alkaline-phosphatase conjugated to goat anti-mouse (Fisher Biotech). After another hour of washes, the stain was developed as described in Kania et al. (1990). For double labelling embryos with ac and hb, stripped and reblocked embryos were incubated overnight with a rabbit anti-hb antiserum (kindly provided by James Langeland) used at 2 μg/ml in PBT. Detection of hb expression was performed as described above for ac with the following exception: staining was developed with 0.5 mg/ml of diaminobenzidene (DAB) in the presence of 0.03% (wt/vol) Co²⁺ and Ni²⁺ ions. After staining was completed all embryos were washed five times in PT, transferred to PBS and 10% glycerol and then mounted. In situ hybridization was carried out as described in Tautz and Pfeifle (1989). Embryos were hybridized with 50 μl (1 ng//tl) of a DNA probe from the 2 kb EcoRI fragment of an ac cDNA, generously provided by Carlos Cabrera.

In order to localize the proneural ac protein with high resolution in developing embryos, we generated a number of monoclonal antibodies (mAbs) specific for the ac protein. Using these mAbs we observe that the ac protein first accumulates in a segmentally repeated pattern of clusters of 5-7 ectodermal cells (late stage 8 as defined by Campos-Ortega and Hartenstein, 1985; Fig. 1A,C) and that ac protein is localized primarily within the nucleus (Fig. 2B, note the similarity between the localization of the nuclear engrailed protein and the ac protein). Two medial and two lateral clusters are found per hemisegment (brackets; Fig. 1A). Shortly after the establishment of the cluster pattern, one cell in each cluster, the future NB, comes to express ac most intensely and delaminates towards the interior of the embryo while the other cells of the cluster remain in the ectodermal cell layer and continue to express ac (data not shown). Next, the cells within the ectodermal cell layer lose ac expression (as no cell death occurs in the neuroectoderm during these stages, these cells probably become epidermal cells; Campos-Ortega and Hartenstein, 1985), while the delaminated NB retains it (mid-late stage 9; Fig. 1B,D). Thus, every cluster rapidly resolves to a single enlarged cell, the NB, which still expresses ac (mid-late stage 9; Fig. 1B,D). The fate of each cell within a cluster is strongly correlated then, not with the initial presence of the ac protein, but with the fate of ac expression in that cell: cells that retain ac expression become NBs; those that lose ac expression lose their NB-forming potential and become epidermal.

Once a NB has delaminated away from the ectoderm and before it undergoes any divisions, ac expression in the NB is extinguished. The NBs in the posterior region of each segment are the first to lose ac expression (early stage 10; Fig. IE); shortly thereafter the two anterior NBs also extinguish ac expression (data not shown). Interestingly, the dynamics of ac protein distribution within the neuroectoderm of the Drosophila embryo precisely parallel the dynamics of ac/sc protein expression previously observed within the notai regions of the wing imaginai disc (Cubas et al., 1991; Skeath and Carroll, 1991). Thus, as is found in the genesis of the peripheral nervous syStern of the adult fly, ac appears to be involved within the embryo in the initial commitment of an ectodermal cell to become neural but not in the maintenance or final differentiation of that cell type. After stage 10, ac is expressed in the peripheral and then again in the central nervous syStern in rapidly changing and spatially intricate patterns (Skeath, J. and Carroll, S. unpublished observations).

As previously mentioned the relationship between the domains of proneural gene transcription and protein accumulation has been a matter of dispute. To clarify the relationship between ac RNA and protein expression, we compared these patterns by in situ hybridization and immunohistochemical inspection of embryos at the same stage of development. At the ectodermal cluster stage, the ac RNA pattern is no broader than the protein pattern (compare Fig. 1C to Fig. 1G), and both patterns quickly resolve to label single NBs from each cluster (compare Fig. ID to Fig. 1H). Thus, the ac RNA and protein patterns appear to be essentially identical as is most likely the case with I’sc (Martin-Bermudo et al., 1991). This suggests that the spatial control of these proteins is largely transcriptional, not post-transcriptional (Cabrera, 1990). The original I’sc antibody was raised against an epitope containing a putative tyrosine phosphorylation site (Cabrera and Alonso, 1988; Cabrera, 1990), and thus may not recognize a phosphorylated form of I’sc. It is possible that dephosphorylation of this site, which is also present in the ac and sc proteins (Villares and Cabrera, 1987), may occur preferentially in NBs and be involved in the activation of proneural proteins.

The ac-positive NBs represent a subset of the first population of NBs to segregate from the neuroectoderm. The anteroposterior (AP) and dorsoventral (DV) registration of these neuroblasts and the ectodermal cell clusters from which they arise was determined by double labelling embryos with antibodies specific for the ac protein and the protein encoded by the segment polarity gene engrailed (en) (DiNardo et al., 1985; Patel et al., 1989; Fig. 2); and for ac and the protein encoded by the segmentation gene hunchback (hb; data not shown), which is expressed in most if not all NBs (Jimenez and Campos-Ortega, 1990). Two of the four ac clusters per hemisegment, one medial (ventral) and one lateral (more dorsal) cluster, are completely contained within each stripe of en-expressing cells (arrows Fig. 2A,B); the other two clusters are found midway between adjacent en stripes (arrowheads; Fig. 2A,B). It may be important developmentally that the width of the ac clusters equals the width of the en stripe. This correspondence holds from the onset of ac expression until the clusters begin to resolve.

By the time the four ac-positive clusters per hemisegment have resolved to four NBs (mid-late stage 9) the NBs are arranged in three longitudinal columns: medial (m), intermediate (i), and lateral (1) (Hartenstein and Campos-Ortega, 1984). The m and 1 columns consist of four NBs and the i column consists of two NBs each per hemisegment (Fig. 2D, data not shown). In the m and 1 columns, ac is expressed in every other NB (arrows, Fig. 2C,D). Taken together with the observations on embryos double labelled for ac and en, this demonstrates that in each hemisegment ac is expressed in the second most anterior pair of NBs and in the most posterior set of NBs, but not in the intervening rows (Fig. 2D). It has been suggested that the AS-C genes could function to specify NB identity (Cabrera et al.,1987) . Although no formal evidence exists to support this idea (see Martin-Bermudo et al., 1991), the fact that only four out of the first ten NBs express ac protein (Fig. 2C,D) suggests these genes, alone or in combination with other genes, could perform such a role. With the recent increase in the number of specific markers for NBs and their progeny, it may soon be possible to determine if the genes of the AS-C do, in fact, specify NB identity (Martin-Bermudo et al., 1991). For example, if the loss of ac, sc or I’sc protein expression from a particular NB or the directed misexpression of one of these genes in a NB which normally does not express this gene alters the pattern of gene expression within that NB or its progeny, this would suggest a function for the AS-C genes in specifying the fate of that NB.

The exclusive retention of ac expression within one cell, the NB, of an ectodermal cell cluster raised the question of how proneural protein expression is eliminated from the other cells of the cluster, thereby removing their potential to become NBs. The available evidence suggests the neurogenic genes epidermalize these cells via a cell communication pathway that ultimately opposes neurogenesis (Lehmann et al., 1983; de la Concha et al., 1988; Brand and Campos-Ortega, 1989; for reviews see Artavanis-Tsakonas, 1988; Ghysen and Dambly-Chaudiere, 1989; Jan and Jan, 1990; Simpson, 1990; and Campos-Ortega, 1991). In order to determine the relationship between neurogenic gene function and ac protein expression, we assayed the distribution of ac protein in embryos mutant for five neurogenic genes. Homozygous Notch (N^55c11-, data not shown), Delta (DI⁹®), Enhancer of split (E(spl)^8D06), big brain (bib¹¹⁹⁰⁵), and neuralised (neu^9Lll9) mutant embryos exhibit similar effects on ac expression: the restriction of ac expression from a cluster to a single cell does not occur; instead, most to all cells of the cluster retain ac expression at high levels, enlarge, delaminate and apparently become NBs (compare Fig. 3B,C,D to Figs 3A, ID). It appears, then, that one key function of the neurogenic genes is to silence proneural gene expression within the non-segregating cells of the initial ectodermal clusters, thereby allowing epidermal development.

Even though the resolution of ac-positive cell clusters does not occur in neurogenic mutants, the temporal regulation of ac is normal. Just after mid-late stage 9, ac expression in embryos mutant for any one of the neurogenic genes is lost from the posterior region of each segment leaving two clusters in place of the two anterior NBs found in WT embryos (compare Fig. 3E,F to Fig. IE). Shortly thereafter, as is observed in WT embryos, ac expression is removed from the anterior region of each segment (data not shown). Except in neu^9L119 embryos (Fig. 3E), the number of ac-positive cells per cluster remains relatively constant until these cells cease ac expression (Fig. 3F). The number of cells per cluster in neu^9L119 embryos increases from 5 –7 cells (mid-late stage 9; Fig. 3C) to 5-12 cells (early stage 10; Fig. 3E). Division of cells within the cluster, late derepression of ac expression, or recruitment of adjacent cells into proneural clusters could account for the increase in ac-positive cells in neu^9L119 mutant embryos.

Our results argue strongly that the neurogenic genes function to silence proneural gene expression in the non-segregating cells of the ectodermal cell cluster. However, a number of questions remain as to the exact mechanism/pathway through which these genes accomplish this end. In what order do the neurogenic genes act within the lateral inhibition pathway and what are the physical interactions that occur between their gene products to perpetuate the inhibitory signal? A series of experiments have shown that DI appears to act as the signal that passes on the lateral inhibitory signal from one cell to another via its physical interaction with the receptor trans-membrane protein Notch (Fehon et al., 1990; Heitzler and Simpson, 1991). It will be important to determine biochemically where the rest of the neurogenic genes fit into the pathway. Further, it will be critical to understand how in molecular terms the neurogenic genes silence proneural gene expression. The E(spl) complex appears to act in the last step of lateral inhibition (de la Concha et al., 1988) and encodes several bHLH proteins (Klambt et al., 1989). The gene products of the E(spl) complex may then remove proneural gene expression from cells initially competent but not chosen to become NBs. Given their structure, the E(spl) complex gene products could perform this function by sequestering transcriptional activators of the proneural genes in ‘poisoned’ heterodimers (Benezra et al., 1990; Ellis et al., 1990; Garrell and Modolell, 1990) incapable of transcriptional activation or they could block proneural gene transcription directly by binding to the control regions of these genes.

The precise and reproducible AP and DV pattern of ac-expressing clusters within each segment suggest that the segmentation genes that establish segment number and polarity and the dorsal-ventral genes which specify the DV pattern of the embryo could directly regulate the initial AP and DV limits of ac expression. In fact, preliminary results implicate the pair-rule (but not the segment polarity genes) as being the primary determinants of I’sc (Martin-Bermudo et al., 1991) and ac expression along the AP axis, while the dorsal-ventral genes appear to repress ac expression in the dorsolateral ectoderm (Skeath, J., Panganiban, G, and Carroll, S. unpublished data). In order to determine directly how the early pattern-forming genes regulate the expression of the ac gene it will be crucial to define cisacting elements of ac which respond to these genes; and then to determine via in vitro DNA-protein binding and in vivo reporter mutagenesis experiments which of the early pattern-regulating genes act directly on ac to establish its initial expression pattern.

In conclusion, the dynamics of ac protein expression vividly illustrate the early cellular and molecular events of neurogenesis and the roles of the neurogenic genes in facilitating epidermal development. Several key issues remain with regard to the regulation of the spatial pattern of NBs and the specification of NB identity. It will be important to determine which genes establish the repeating pattern of ectodermal cell clusters, what the specific roles are of each neurogenic gene in lateral inhibition of proneural gene expression, and, which genes specify the identity of each NB and its progeny. The elucidation of the mechanisms involved in regulating ac should aid us in integrating our knowledge of the roles of the genes involved in the specification of cell types with those involved in guiding the overall organization of the embryo.

The monoclonal antibodies were generated in the University of Wisconsin’s Hybridoma Facility and we gratefully acknowledge Carol Sinaiko of the University of Wisconsin Hybridoma Facility for help in this endeavor. We owe a special debt to Tadashi Uemera for kindly providing the ac expression plasmid. We thank Zhen Davis, James Williams and Jane Selegue for their help with the RNA in situ protocol; Chris Doe, Allen Laughon, Sarah Crittenden, Teresa Orenic and Grace Panganiban for their comments on the paper and Fernando Jimenez for communication of results prior to publication; Kathy Vorwerk for technical assistance; Leanne Olds for help with the figures; and Jamie Wilson for help with the preparation of the manuscript. This work was supported by a National Institutes of Health predoctoral traineeship (GM-07215) to J.B.S., a National Science Foundation Presidential Young Investigators Award, the Shaw Scholar’s Program of the Milwaukee Foundation, and the Howard Hughes Medical Institute.