Glial differentiation does not require a neural ... - Semantic Scholar

3189

Development 125, 3189-3200 (1998) Printed in Great Britain © The Company of Biologists Limited 1998 DEV7653

Glial differentiation does not require a neural ground state Roberto Bernardoni, Alita A. Miller and Angela Giangrande* Institut de Génétique et Biologie Moléculaire et Cellulaire, IGBMC/CNRS/ULP-BP 163 67404 Illkirch, c.u. de Strasbourg, France *Author for correspondence (e-mail: [email protected])

Accepted 4 June; published on WWW 21 July 1998

SUMMARY Glial cells differentiate from the neuroepithelium. In flies, gliogenesis depends on the expression of glial cell deficient/ glial cell missing (glide/gcm). The phenotype of glide/gcm loss- and gain-of-function mutations suggested that gliogenesis occurs in cells that, by default, would differentiate into neurons. Here we show that glide/gcm is able to induce cells even from a distinct germ layer, the mesoderm, to activate the glial developmental program, which demonstrates that gliogenesis does not require a ground neural state. These findings challenge the common

INTRODUCTION Nervous system development takes place after positional information has been laid along the dorsoventral axis. In Drosophila, maternal information provided by a gradient of nuclear dorsal protein subdivides the embryo into ventral, lateral and dorsal territories giving rise respectively to mesoderm, neurogenic region and dorsal ectoderm and amnioserosa (Rusch and Levine, 1996). The initial subdivision is then maintained by the expression of zygotic genes. The definition of the neurogenic region relies on decapentaplegic (dpp) and short gastrulation (sog) (Biehs et al., 1996; for a review, see Bier, 1997). dpp in dorsal ectoderm represses genes that induce neural development. sog in lateral stripes blocks dpp and permits neurogenesis. Similarly, dpp homolog BMP4 suppresses neurogenesis in early frog development while chordin, the sog homolog, blocks BMP4 activity (Sasai et al., 1995; Wilson and Hemmati-Brivanlou, 1995). Importantly, once the neurogenic region is defined, precursor cells located at stereotyped positions proliferate and differentiate into neurons and glial cells. We and others have shown that glide/gcm is required for the differentiation of all glial cells in the peripheral and in the central nervous systems except for midline glial cells (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996; for reviews see Pfrieger and Barres, 1995; Anderson, 1995; Giangrande, 1996). glide/gcm codes for a transcription factor (Schreiber et al., 1997) that is thought to activate the glial differentiation program and to repress the neuronal one (Giesen et al., 1997). In glide/gcm embryos, glial cells do not differentiate due to their transformation into neurons; in addition, ectopic expression of glide/gcm within the nervous system results in ectopic expression of repo, an early

view on the establishment of cell diversity in the nervous system. Strikingly, ectopic glide/gcm overrides positional information by repressing the endogenous developmental program. These findings also indicate that glial differentiation tightly depends on glide/gcm transcriptional regulation. It is likely that glide/gcm homologs act similarly during vertebrate gliogenesis. Key words: glial cell deficient/glial cell missing (glide/gcm), Glial differentiation, Cell fate, Drosophila, Neurogenesis

glial marker, and repression of neuronal markers, suggesting that neurons are converted into glial cells (Hosoya et al., 1995; Jones et al., 1995). Thus, despite the apparent fixed array of neurons and glial cells, most cells in the nervous system have the competence to take the glial fate if provided with glide/gcm. On the basis of the gain- and loss-of-function phenotypes, it has been proposed that the neuronal pathway constitutes a default route and that glide/gcm dictates the choice between neuronal and glial fate, which implies that gliogenesis can only occur in the neurogenic region (Anderson, 1995). An alternative model proposes that the competence to adopt the glial fate is common to several cell types and that there is no requirement for a neural ground state, predicting that even cells outside the nervous system may adopt the glial fate if challenged with glide/gcm. To test this model, we used the Gal4 system (Brand and Perrimon, 1993). In this paper we show that ectopic glide/gcm induces expression of glialspecific genes in cells in which dpp normally inhibits neurogenesis and gliogenesis. Importantly, glide/gcm is also able to exert its differentiative function in a different germ layer, the mesoderm, indicating that a single gene is sufficient to activate the glial differentiation program. Mesodermal expression of glide/gcm results in the repression of mesodermspecific genes and in the loss of muscle morphology, concomitant with ectopic expression of several glial markers, suggesting that glial cells are induced at the expense of mesodermal tissue. Thus, glide/gcm can redirect the differentiation programs induced by maternal and zygotic positional information. Altogether, these data strongly suggest that gliogenesis does not require a ground neural state and that glide/gcm transcription must be tightly regulated in order to ensure correct embryonic development.

3190 R. Bernardoni, A. A. Miller and A. Giangrande MATERIALS AND METHODS

RESULTS

Stocks Wild-type Sevelen stock was used as a control. Three independent transgenic UAS-glide/gcm lines, F18A, M21G and M24A (Bernardoni et al., 1997), were crossed with twi-Gal4 (Baylies and Bate, 1996), 24B (Brand and Perrimon, 1993), sca-Gal4 (gift from M. Mlodzik), elav-Gal4 (Luo et al., 1994) or with MD237, which reveals the wild-type pattern of pannier (Heitzler et al., 1996). MD237 expression pattern was followed using a UAS-lacZ line (Brand and Perrimon, 1993). The enhancer trap lines rC56 (Klaes et al., 1994) and AE2 (Auld et al., 1995) were used as markers for lateral glial cells, AA142, as a marker for midline cells (Klaes et al., 1994).

Expression of glide/gcm in mesoderm induces early and late glial-specific genes Three UAS-glide/gcm lines (Bernardoni et al., 1997) were crossed with a twist-Gal4 line. The twist promoter drives Gal4 expression in all presumptive mesoderm and in mesectoderm (Baylies and Bate, 1996), two tissues in which glide/gcm is not expressed. Western analyses performed on embryonic extracts with anti-glide/gcm antibodies revealed that the three lines express different levels of glide/gcm, UAS-glideF18A (F18A) and UASglideM24A (M24A) displaying the lowest and the highest levels, respectively (Fig. 1A). glide/gcm expression under the twist promoter is detrimental to the animal: 100%, 95% and 72% lethality is observed using M24A, UAS-glideM21G (M21G) and F18A respectively, confirming the difference in transgene expression observed in western analyses. The embryonic progeny of these crosses was labelled with a glial marker, anti-repo (Campbell et al., 1994; Halter et al., 1995), which recognizes a nuclear protein that is expressed, like glide/gcm, in all lateral but not in midline glial cells (Scholz et al., 1997) nor in mesoderm. repo expression starts slightly after that of glide/gcm and is abolished in glide/gcm embryos (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). In addition, the repo promoter contains eleven glide/gcm binding sites (Akiyama et al., 1996). The fact that repo is most likely a target of glide/gcm enabled us to use the anti-repo antibody to trace ectopic glide/gcm activity. Ectopic glide/gcm expression in the F18A line (Fig. 2A,B) led to repo expression at the position of the heart, a tissue of mesodermal origin (Fig. 2D). Furthermore, anti-repo labelling

Immunohistochemistry and western blot Embryos were labelled as described by Vincent et al. (1996). Antibodies against the following proteins were used: repo (1:150); MHC (1:1000); D-MEF2 (1:800); β-gal (Promega, Cappel) (1:1000); elav (1:50); HRP (1:500). Secondaries conjugated with Oregon Green, FITC, Cy3 or biotin were used at 1:500. Biotin-conjugated secondaries were revealed using the ABC kit (Vector) as described by Giangrande et al. (1993). Preparations were analyzed using conventional (Axiophot, Zeiss) or confocal (DMRE, Leica) microscopy. Anti-glide/gcm was raised, as described by Bernardoni et al. (1997), against a peptide within the C-terminal region of the protein. Western blot was carried out as described by Bernardoni et al. (1997) using a 1:500 dilution. Signal quantitation was performed on a Biorad GS700 imaging densitometer. RNA analysis In situ hybridizations were performed on embryos after overnight collection as described by Lehmann and Tautz (1994). The probe was as in Bernardoni et al. (1997).

Fig. 1. Levels of glide/gcm expression in the three UAS-glide lines. (A) Western blot analysis in overnight embryonic extracts. WT, wild type. F18A/F18A, M21G/M21G, M24A/+ indicate three independent UAS-glide transgenic lines that have been crossed with the twi-Gal4 stock. The relative amounts of glide/gcm were determined by densitometric analysis to be 16-fold, 18-fold and 30-fold, respectively. Since M24A is a homozygous lethal line, only half of the progeny of the cross expresses glide/gcm ectopically, thus the amount in M24A is 60-fold that of wild type. Molecular mass markers (in kDa) are indicated on the left, glide/gcm protein is indicated by an arrowhead. (B-F) In situ hybridization analysis of glide/gcm transcripts in stage 11 wild-type and mutant embryos. (B) F18A embryo, ventral view. Arrow indicates position of the midline, arrowheads indicate midline glide/gcm expression. (C-F) Lateral views; (C) wild-type, (D-F) transgenic embryos displaying weak, intermediate and strong glide/gcm expression, respectively.

glide/gcm can turn muscles into glia 3191

Fig. 2. twi-glide/gcm and 24B-glide/gcm expression induce anti-repo labelling in mesodermal derivatives. (A-C,E,F) Anti-repo labelling of stage 14 embryos; lateral views. (A) Wild type: repo is not expressed in the dorsal-most part of the embryo (asterisks). (B) twi-Gal4/+; F18A/+ embryo. Weak ectopic glide/gcm induces repo expression (open arrows in all panels) dorsally (D), at the position of the heart (shown in D) and laterally (L), in scattered cells. (C) twi-Gal4/M24A embryo. Strong ectopic glide/gcm results in repo expression at the position of somatic mesoderm throughout the anteroposterior axis and at the position of pharynx muscles (shown in D). (D) Anti-D-MEF2 labelling on wild-type stage 14 embryo. The positions of the heart (H), pharynx (PH) and somatic muscles (SM) are indicated. (E) 24B-Gal4/+; F18A/+ embryo. (F) 24B-Gal4/M24A embryo. Repo-positive cells are fewer in number in E and F compared to those in C and D. Bar 50 µm.

at the position of midline cells, which have a mesectodermal origin, was observed (Fig. 3A,B). The lateral glial identity of repo-positive cells at the position of the midline was confirmed with two enhancer trap lines: AA142, which specifically labels midline cells, and rC56, which labels all central but midline glial cells (Klaes et al., 1994). Indeed, AA142-positive cells are missing at the midline whereas ectopic cells are labelled with rC56 (Fig. 3C-F). Strikingly, ectopic expression in the M24A line resulted in the activation of glial markers in many more cells. For example, repo-positive cells were detected at the position of the pharynx and in metameric stripes along the dorsoventral axis, at the position of somatic muscles (Fig. 2C). In order to quantify the effects of mesodermal glide/gcm activity we counted the number of repo-positive nuclei at the position of somatic muscles in a stage 14 twi-glide embryo. Each stripe contains on average 60 repo-positive nuclei. In a wild-type embryo of a similar stage, each stripe contains on average 130 nuclei expressing D-MEF2, a marker for all muscle precursors (see below). Thus, we estimate that roughly half of the muscle cells express the repo glial marker upon mesodermal glide/gcm activation. Massive repo expression was also observed at the

position of other mesodermal derivatives such as visceral muscles (data not shown). None of the repo-positive cells were positive for the neuronal marker elav (Robinow and White, 1988) (data not shown). The patterns of repo expression observed with both transgenic lines corresponds to a subset of the cells that express β-gal in twi-Gal4; UAS-lacZ embryos (Baylies and Bate, 1996). This raises an important question as to whether glide/gcm is only ectopically expressed in few cells of mesodermal origin or whether only few cells express glide/gcm at levels that are sufficient to activate repo expression. To address this issue, we analyzed the ectopic glide/gcm expression at the cellular level by in situ hybridization. Embryos from all transgenic lines express glide/gcm at ectopic positions, however, the absolute level of expression and the number of cells expressing glide/gcm at high levels depend on the line analyzed. In particular, M24A embryos display strong ectopic expression in many more cells and at higher levels compared to F18A, confirming the results obtained by western blot (Fig. 1). The fact that, in F18A embryos, ectopic glide/gcm is found in stripes while repo is only activated in few scattered cells also confirms that threshold levels necessary to activate

3192 R. Bernardoni, A. A. Miller and A. Giangrande the glial program are only reached in subsets of glide/gcm expressing cells. Interestingly, glide/gcm ectopic expression in M21G is variable, some embryos appeared similar to F18A embryos, some to M24A embryos and others displayed intermediate phenotypes (Fig. 1 and data not shown). This is in agreement with the observations that the marker used to trace the insertion of the transgene, white, is also variably expressed in this line and that twi-Gal4/+; M21G/+ embryos display a variable phenotype with respect to ectopic repo expression. Finally, the number of repo-expressing cells is higher in embryos carrying both M24A and M21G compared to that found in embryos carrying M24A alone (data not shown). In this combination, the repo profile of expression mimicks that of β-gal in twi-Gal4; UAS-lacZ embryos, which also supports the idea that the transgenic lines express threshold levels of glide/gcm (and thus activate repo) only in a subset of mesodermal cells. Two additional glial-specific enhancer trap lines were used to confirm the results obtained with anti-repo: rC56 and AE2, which labels subsets of peripheral and central glial cells (Auld et al., 1995). The expression pattern of AE2 reflects that of gliotactin, a gene coding for a transmembrane protein expressed at late embryonic stages and required for bloodnerve barrier formation. Indeed, ectopic labelling was observed in twi-Gal4/UAS-glide embryos with both markers (Figs 4, 5). The number of AE2-positive cells is lower than that of repopositive cells. This might indicate that only a subset of cells attain late differentiation or that, as gliotactin is normally expressed in a sub population of glial cells, only a fraction of ectopic cells takes that identity. The above results clearly indicate that ectopic glide/gcm is sufficient to induce many aspects of the glial genetic program outside the nervous system, although we cannot formally exclude that ectopic glial cells may not be functional due to the lack of contact with neurons. Indeed, glial and neuronal cells interact with each other during development and ontogenesis, Fig. 3. twi-glide/gcm expression induces lateral glial-specific genes in the midline. Ventral views of stage 13 embryos. (A,B) Anti-repo labelling in (A) a wild-type embryo, (B) twi-Gal4/+; F18A/+ embryo. In the wild type, repo expression is absent in the midline (indicated by the horizontal lines in the insets on the right, which show a higher magnification of the area above the vertical bars in A and B). MM-CBG indicates a class of lateral glia called medial-most cell body glia labelled by anti-repo: one MM-CBG is present per hemisegment in the abdomen (A1), two cells are present in the thorax (T3). In embryos expressing weak ectopic glide/gcm, repo is induced at the position of midline cells, see arrows. (C,D) Anti-β-gal labelling in (C) AA142/+ embryo; (D) twi-Gal4/+; F18A/AA142 embryo. Midline is indicated by an open arrow. Curved arrows show labelling in the salivary glands, present in both wild-type and mutant embryos, while midline-specific labelling is absent in D. (E,F) Antiβ-gal labelling, in (E) rC56/+ embryo and (F) twi-Gal4/+; F18A/rC56 embryo. (G-I) Anti-HRP labelling of stage 14 embryos, ventral view, showing the axonal scaffold organization. (G) Wildtype; (H) twi-Gal4/+; F18A/+, (I) twi-Gal4/M24A embryos, respectively. The white arrowhead indicates a complete commissure fusion, the black arrowhead indicates the almost complete lack of the posterior commissure; the broad arrows indicates the few commissure fibers left in an embryo expressing twi-glide/gcm strongly. Bars (A,B) 50 µm; (E-I, insets) 25 µm; (C,D) 62.5 µm.

maintaining a cross-talk that is necessary for both cell types to function and to survive. Finally, our results show that the degree of glial fate induction depends on the strength of the UAS-glide line. The competence to express glial-specific genes becomes restricted during development The observation that twi-glide/gcm activates glial-specific genes at the position normally occupied by mesoderm prompted us to ask whether the competence to express glialspecific genes becomes more restricted at late stages of development. Indeed, the twi-Gal4 construct drives β-gal expression at early stages, during gastrulation. To express genes late in mesodermal cells we used the 24B-Gal4 line (Brand and Perrimon, 1993; Baylies and Bate, 1996), which starts driving mesodermal expression during germ band

glide/gcm can turn muscles into glia 3193

Fig. 4. Mesodermal gliotactin expression in twi-glide/gcm embryos. Stage 15 embryos: (A,C) AE2/+ (wild type), (B,D) twi-Gal4/AE2; M21G/+, (strong glide/gcm expression). (A,B) Dorsal views, (C,D) ventral views. vc, pg indicate ventral cord and peripheral glia, respectively. Ectopic AE2-positive cells are present at dorsal and lateral positions. Bar 54.8 µm.

elongation, at mid stage 10. By this stage, mesodermal cells have moved dorsally to cover the ectodermal territories. At the end of stage 10, the third mesodermal division has taken place and cells in this layer have started segregating into visceral and somatic mesoderm (see Bate, 1993). In animals carrying the strong UAS-glideM24A transgene and 24B-Gal4, some anti-repo labelling is observed at ectopic positions (Fig. 2E,F). However, the cells expressing repo ectopically are much fewer than those observed using the same transgene with the twi-Gal4 driver. As in the case of the twiGal4 driver, the effect is stronger with high levels of glide/gcm than with low levels. Mesodermal glide/gcm expression inhibits the muscle fate To analyze further the mode of action of glide/gcm, we asked whether ectopic expression of glial-specific genes had occurred at the expense of mesoderm-specific genes. To this end, we determined the expression profile of two markers using mesoderm-specific antibodies in twi-Gal4/UAS-glide embryos. D-MEF2 is expressed in all muscle and cardial progenitors as well as in differentiated cells (Bour et al., 1995; Lilly et al., 1995) whereas MHC is only expressed in differentiated cells (Kiehart and Feghali, 1986). In transgenic animals, both antibodies labelled much fewer cells than in the wild type, indicating that mesodermal glide/gcm expression leads to the repression of muscle-specific genes (Figs 5, 6). To determine whether glial-specific and mesodermal-specific genes are co-expressed, double labelling experiments with a glial

(rC56) and a mesodermal (D-MEF2) marker were performed. The analysis of single sections revealed the presence of three cell types, those that express either the glial or the mesodermal marker, which constitute the majority of the cells, and those that express both markers (Fig. 5G-I), possibly identifying cells with an intermediate fate. These results strongly suggest that mesodermal cells fail to differentiate because they have been transformed into glial cells and that competition between two fates takes place upon ectopic glide/gcm activity. A close inspection of mutant embryos revealed that the muscle layer is severely disrupted (Figs 6, 7). At the end of embryogenesis, somatic and visceral muscles can be easily identified at the inner surface of the epidermis and coating the gut, respectively. In wild-type embryos, body wall muscles display a stereotyped pattern along the anteroposterior axis and are organized in longitudinal, oblique and transverse muscles. In twi-glide/gcm embryos, muscle fibers are reduced in number and poorly organized (Fig. 6). Embryos expressing glide/gcm lack most fibers and display a significant number of round, unfused muscle cells. These results were also confirmed by immunohistochemistry on embryonic sections (Fig. 7). In wild-type embryos, most repo-positive cells are in the central nervous system, with a few cells located along peripheral nerves. In twi-glide/gcm embryos, many repo-positive cells are present at the position of muscles and the body wall muscle layer normally underlying the epithelium is severely disrupted. In addition, gut cells display severe disorganization, most likely as a consequence of defects in visceral mesoderm. Indeed, as the anterior and posterior midgut primordia grow out and migrate toward each other, they move over bands of visceral mesodermal cells and meet before completion of germ band retraction. Subsequently, the endoderm spreads towards the dorsal and the ventral sides to surround the yolk sac as a monolayer epithelium (see Bate, 1993 for a review). The phenotype of tinman embryos, in which the visceral mesoderm is removed and midgut primordia fail to migrate, suggests that visceral mesoderm is a required substrate for this movement (Azpiazu and Frasch, 1993). Likewise, in twi-glide/gcm embryos, gut cells accumulate in clusters instead of surrounding the yolk sac as a monolayer epithelium. Finally, the loss of muscle-specific features (myofilaments, attachment sites and syncitia) in twi-glide/gcm embryos was also confirmed at the ultra-structural level (data not shown).

twi-glide/gcm expression leads to severe defects in the ventral cord The analysis of cross sections showed that the overall ventral cord morphology is altered in twi-glide/gcm embryos. In order to characterize the nature and the origin of the defects, we used several markers. This confirmed the presence of repo-positive cells at the midline and adjacent to it, most likely due to migration (Fig. 7B,C), demonstrating that a high number of cells do not activate the midline fate, even though we cannot exclude that some of them express an intermediate fate. Anti-elav labelling performed to identify neuronal cell bodies showed that elav expression is restricted to the ventral cord and to the peripheral nervous system, demonstrating that the effects of ectopic glide/gcm expression are limited to the glial lineage and that mesodermally derived cells do not simultaneously express glial and neuronal markers (Fig. 7F,G). The elav labelling also showed that the neurons are not compacted into

3194 R. Bernardoni, A. A. Miller and A. Giangrande laterally symmetric hemineuromeres and that the midline cannot be recognized (Fig. 7F,G). This is most likely the consequence of repo expression in the mesectodermal cells, which are normally required to guide commissural growth cones towards and across the midline. To demonstrate this, we used markers labelling the axonal commissures. Indeed, ectopic glide/gcm results in severe defects: lack of the posterior commissure, fusion

of posterior and anterior commissures and lack of both commissures (Fig. 3G-I and data not shown), this last phenotype being observed in embryos expressing high levels of glide/gcm. Interestingly, such defects are similar to those observed in mutants affecting midline cell development. For example, orthodenticle and mutations in the EGF receptor pathway result in absence of specific midline cells and display lack of posterior

Fig. 5. Expression of muscle- and glial-specific genes in the mesoderm of twi-glide/gcm embryos. Lateral views of stage 14 embryos, confocal images. (A,D,G) β-gal labelling of rC56-positive cells is shown in red, (C,F,I-K) D-MEF2 labelling of mesodermally derived cells is shown in green, (B,E,H) double labelling. (A-F,J,K) Total projections obtained by superposition of optical sections. (A-C) wt (rC56/+), same embryo. DMEF2 heart labelling is indicated by an asterisk. (D-I) twi-Gal4/M24A; rC56/+, same embryo. White lines in D indicate the region displayed in (G-I), which show a single optical section to allow a colocalization analysis. Open arrows in G indicate cells expressing only β-gal, arrows in I indicate cells expressing only D-MEF2, arrowheads in (G-I) indicate cells that express both markers. (J,K) twi-Gal4/+; F18A/+, white lines in K indicate the region shown in J. Note that some labelling at the position of the heart is missing (star). The number of D-MEF2-positive cells decreases as the amount of glide/gcm increases (compare K and F). Bars (A-F,K) 50 µm; (G-J) 25 µm.

glide/gcm can turn muscles into glia 3195

Fig. 6. MHC expression in twi-glide/gcm embryos. Stage 17 embryos, ventrolateral views. (A) Wild type: LT, VL, VO, VA indicate lateral transverse, ventral longitudinal, ventral oblique, and ventral acute muscles, respectively. (B) twi-Gal4/+; M21G/+ embryo displaying a weak phenotype. Arrows indicate the absence and the disorganization of some muscle fibers. (C) twi-Gal4/+; M21G/+ embryo displaying an intermediate phenotype consisting of a strong disruption of muscle fibers. Arrowheads in B and C indicate unfused muscle cells. Bar, 50 µm.

commissure and fused commissure phenotypes, respectively (Klambt et al., 1991). Finally, embryos which do not express midline-specific genes (commissureless: Seeger et al., 1993) or embryos in which the midline express genes normally expressed in lateral glial cells (ectopic pointed P1: Klaes et al., 1994), also display a lack of commissure phenotype. Thus, the analysis of the axonal scaffold provides functional evidence for the failure of midline cells to differentiate and supports the hypothesis that these cells are transformed into lateral glial cells. Ectopic expression of glide/gcm in the dorsal ectoderm To confirm that the competence to adopt the glial fate is extended to cells outside the neurogenic region, we examined

the effects of ectopic glide/gcm activity in another cell type. Dorsal ectoderm is subdivided into two lineages, the amnioserosa, which maps to the dorsal-most region of the embryo, and, adjacent to it, the dorsal epidermis. Both lineages are specified by the activity of dpp, which suppresses neurogenesis and gliogenesis. We used a pannier-Gal4 (pnrGal4) line (Heitzler et al., 1996), which drives expression in the dorsal epidermis and in the amnioserosa at late stages of embryogenesis (β-gal starts being detected at stage 12). Upon weak expression, ectopic anti-repo labelling was observed in several cell clusters of the dorsal epidermis (Fig. 8A,C,F). Upon strong expression, anti-repo labelling was observed in many more dorsal epidermal cells (Fig. 8E). Dorsal closure did not occur in many pannier-glide/gcm embryos, resulting in swelling of all internal organs (data not shown), which most likely accounts for the lethality observed in such a cross and argues for the presence of defects in dorsal epidermis differentiation. In order to look at the morphology of the ectopic repo-positive cells, the labelling was performed simultaneously with anti-repo and with anti-β-gal antibodies in embryos carrying both UAS-glide and a UAS-lacZ transgene in which the β-gal has a cytoplasmic localization (Brand and Perrimon, 1993). This showed that, indeed, repo-positive cells display a typical elongated glial cell morphology (Fig. 8B,D,G). Note that, due to the uniform β-gal expression in dorsal ectoderm, morphology can only be scored in ectopic repo-positive cells that migrate out of this territory. Finally, the glial identity was also confirmed using the late AE2 marker (Fig. 8H,I). These results show that glide/gcm expression in non-neurogenic regions is sufficient to induce many aspects of glial differentiation. The fact that repo-positive cells were not observed in the amnioserosa of most embryos (Fig. 8) may have several explanations. First, amnioserosa is the region of highest expression of dpp, therefore we may need more glide/gcm to activate glial-specific genes than that provided by the transgene. Second, the pannier-Gal4 construct induces expression rather late in development, which may limit the responsiveness to glide/gcm, as already seen in the mesoderm. Finally, amnioserosa is made up of cells that do not divide. We have observed that ectopic activation of glial-specific genes driven by glide/gcm expression in post-mitotic cells is less efficient than expression in pre-mitotic cells of the neurogenic region. Indeed, embryos expressing glide/gcm early in the neurogenic region or in neurons express ectopic repo at the expense of the elav neuronal marker (Jones et al., 1995; Hosoya et al., 1995). However, early expression in the neurogenic region induces the fate change in many more cells than expression in post-mitotic cells (compare repo and elav labelling between sca-glide/gcm and elav-glide/gcm embryos in Fig. 9). DISCUSSION Glial differentiation does not require a neural ground state Evidence from several sources suggest that a neural ground state is a prerequisite for gliogenesis. First, glial cells, which constitute the majority of cells in the vertebrate nervous system, arise from the neuroepithelium, a territory defined through the combined activity of neural inhibitory and

3196 R. Bernardoni, A. A. Miller and A. Giangrande

Fig. 7. Glial and neuronal organization in twi-glide/gcm embryos. (A-E) Cross sections of stage 16-17 embryos labelled with anti-repo: (A) wild type; (B,C) twi-Gal4/M24A; (D,E) Higher magnification of boxed areas in A and B. sm, vm, gl, vc, y, g indicate somatic muscles, visceral muscles, glia, ventral cord, yolk and gut, respectively. Somatic muscles are absent in B (open arrow). Glial cells revealed with anti-repo are found at the position of somatic muscles and at the position of the midline (which is indicated by a black arrow). Gut cells in B and C are disorganized and do not envelop the yolk. (F,G) Cross sections of stage 16-17 embryos labelled with anti-elav; lc, t, PNS indicate longitudinal connectives, trachea and peripheral nervous system, respectively. (F) Wild type; (G) twi-Gal4/M24A. Note the midline defects: compare black arrow between F and G. Bars (A-C,F,G) 25 µm; (D,E) 10 µm.

permissive factors (for a review, see Bier, 1997). Indeed, only the ventral region of the vertebrate neural tube gives rise to astrocytes and oligodendrocytes, suggesting that the competence to adopt the glial fate is restricted to a subset of cells of the developing embryo (Warf et al., 1991). Similarly, fly glial cells differentiate within the ventral cord at stereotyped positions. Second, gliogenesis and neurogenesis are tightly linked during development: glial and neuronal cells are controlled by the same genes and in many cases have a common precursor (flies: Condron and Zinn, 1994; Nelson and Laughon, 1994; Giangrande, 1994, 1995; Bossing et al., 1996; Schmidt et al., 1997; vertebrates: see for reviews, Jessen and Mirsky, 1992; Stemple and Anderson, 1993; Le Douarin et al., 1994; Miller, 1996). Finally, lack of glide/gcm activity in fly glial precursor cells results in a switch from the glial to the neuronal fate (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996). On the basis of these results, it was thought that glial determinants impose the glial fate on cells displaying a neural default state (Anderson, 1995), implying that glide/gcm would only promote gliogenesis in the neuroepithelium. In this study we show that glide/gcm expression outside the nervous system activates the glial differentiation program. Cells from both dorsal ectoderm and mesoderm, two regions in which neuro- and gliogenesis are normally inhibited, express

early and late glial markers when challenged with glide/gcm. These cells exhibit the glial phenotype without going through a neural state, since they never express neural-specific markers. Importantly, ectopic glide/gcm entails repression of the endogenous fate, suggesting that competition takes place between the local and the glial developmental programs. In addition, we have observed that even temporal cues are bypassed by glide/gcm, since ectopic glial cells differentiate earlier than the endogenous ones (data not shown). Our data clearly show that (i) the competence to adopt the glial fate is not restricted to cells of the neurogenic region, (ii) a neural ground state is not a prerequisite for the activation of the glial differentiation program, (iii) glide/gcm is able to redirect the local differentiation programs induced by positional information, (iv) glide/gcm is necessary and sufficient to activate the glial differentiation program, even though glial cells may require additional cues, such as the presence of nearby neurons, to be functional and/or to survive. Finally, our results establish that glial differentiation tightly depends on the precise spatial and temporal regulation of glide/gcm transcription and explain the glide/gcm phenotype. During embryogenesis glide/gcm is expressed in glial precursors after the neurogenic region has been defined (Hosoya et al., 1995; Jones et al., 1995; Vincent et al., 1996).

glide/gcm can turn muscles into glia 3197

Fig. 8. pnr-glide/gcm induces glial-specific genes in dorsal ectoderm. (A,B) UAS-lacZ/+; pnr-Gal4/+ stage 14 embryo (wild type), dorsolateral view, labelled with anti-repo (A) and anti-β-gal (B). Black arrowheads indicate two peripheral glial cells. (C,D,F,G) UAS-lacZ/+; pnrGal4/F18A stage 15 embryo (weak pnr-glide/gcm expression), dorsolateral view, labelled with anti-repo (C,F) and anti-β-gal (D,G). White lines in C and D indicate the region shown in F and G, respectively. (E) M24A/+; pnr-Gal4/+ stage 15 embryo (strong pnr-glide/gcm expression), dorsolateral view, labelled with anti-repo. Note the anti-repo labelling at dorsal positions in C,E,F (arrows), normally devoid of glial cells (compare to A). Note that in pnr-glide/gcm embryos dorsal closure, normally complete by stage 15, is defective, which makes these embryos resemble a stage 14 wild-type embryo (compare with A and B). Note the typical glial elongated shape of the repo-positive cells indicated by the arrowheads in D and G. (H,I) Stage 15 embryos, lateral views, labelled with anti-β-gal. (H) AE2/+ (wild type); (I) AE2/M24A; pnr-Gal4/+ (strong pnr-glide/gcm expression). Broad arrows in I indicate ectopic β-gal-positive cells at dorsal positions. Black arrows in A-D indicate the position of the dorsal rim in wild-type and mutant embryos. Bars (A-E,H,I) 50 µm; (F,G) 25 µm.

In glide/gcm embryos, glial precursors adopt the neuronal fate because this is the only other fate they can take, due to their state of determination. glide/gcm homologs have been identified in vertebrates (Akiyama et al., 1996; Altshuller et al., 1996; Kammerer and A. G., unpublished). Given the conservation of the basic

developmental pathways throughout evolution, it is tempting to speculate that the mechanism of action of these genes is similar to that observed in Drosophila melanogaster. A single gene controls glial cell differentiation The fact that glide/gcm is sufficient to promote the glial

3198 R. Bernardoni, A. A. Miller and A. Giangrande differentiation program ectopically and that the lack of its in the mesoderm (Baylies and Bate, 1996) and that the product removes all glial precursor cells indicates that a single competence of the different mesodermal derivatives to repress gene controls gliogenesis. Indeed, when pointed, a gene the endogenous program upon glide/gcm expression depends expressed in a subset of lateral glial cells, is induced early on the levels of twi present in the cell. For example, heart cells, throughout the nervous system, it induces a glial-specific which contain low amounts of twi, are more apt to repress the marker in some cells, however the overall number of glial cells endogenous fate and to express repo compared to somatic appears unaffected at the end of embryogenesis (Klaes et al., muscle cells, which contain high amounts of twi and require 1994). Furthermore, the mutant phenotype and genetic high amounts of glide/gcm to repress the endogenous fate. analyses indicate that pointed is downstream of glide/gcm The finding that ectopic glide/gcm triggers transcription of (Klaes et al., 1994). This suggests that several genes are downstream glial-specific targets raises an important question activated by glide/gcm and that only the simultaneous about the mechanisms governing cell-specific transcription. activation of all its targets, including repo and pointed, is able How can genes that are normally silent in mesoderm be to induce gliogenesis. It is worth noting that neuronal switched on by the activity of a single transcription factor? An differentiation seems to be regulated in a different way. appealing explanation is that, like other activators (see for Neurogenesis is under the control of a family of related genes, reviews Sauer and Tijan, 1997; Tsukiyama and Wu, 1997; the helix-loop-helix transcription factors of the achaete-scute Wade et al., 1997), glide/gcm may recruit coactivators that complex (ASC). Indeed, although a single member of the remodel nucleosomes in the regions containing glide/gcm family is sufficient to induce ectopic neurons, loss-of-function targets and activate transcription of such genes. Interestingly, mutations of any member do not eliminate all neural precursor glide/gcm is less efficient when induced in mesoderm at late cells, fostering the idea that cooperativity is required between stages, which may be due to the reduced accessibility of the several members of the family (Campuzano and Modolell, chromatin of glide/gcm targets over development. We propose 1992). that, as the mesodermal program is activated, a more stable Despite the differential responsiveness of tissues to repressive state in those chromatin regions is established and glide/gcm, the most responsive one being the nervous system, propagated throughout the cell cycle. The establishment of a it is clear that even cells from mesoderm, a different germ layer, stable structural difference on chromosomal domains has can adopt the glial fate. It will be interesting to determine already been observed during vertebrate development (Wade et whether other genes inducing a cell-specific fate such as al., 1997). A role of the proliferative state in the establishment eyeless (Halder et al., 1995) and MyoD (Weintraub et al., 1991) of a repressed chromatin state is also in agreement with the behave similarly and whether the pan-neural genes (Bier et al., different degree of glial fate induction observed upon 1992; Brand et al., 1993; Roark et al., 1995), which are glide/gcm expression in pre- versus post-mitotic cells of the expressed in all neuronal precursors, convert cell types as nervous system. Thus, understanding the glide/gcm mode of divergent as mesoderm into neurons. The precise role of glide/gcm in the repression of the mesodermal fate is still to be elucidated. The mechanisms that activate twi most likely still operate in the mesoderm, since in twi-glide/gcm embryos twi expression is not modified at detectable levels (R. B. and A. G., unpublished results). Based on the recent finding that D-MEF2, a direct target of twist (Cripps et al., 1998), is repressed in twiglide/gcm embryos, we propose that ectopic glide/gcm interferes with the endogenous developmental program by competing with twist activity. This is also in Fig. 9. glide/gcm expression in the neurogenic region and in neurons. Stage 15, ventral view embryos. agreement with the (A,B) Wild-type, (C,D) sca-Gal4/M24A, (E,F) M24A/+; elav-Gal4/+ embryos. Embryos were doubleobservations that twi labelled with antibodies against repo (A,C,E) and elav (B,D,F) to recognize glial and neuronal cells, respectively. Bar 50 µm. expression is modulated

glide/gcm can turn muscles into glia 3199 action represents an extremely important step towards unravelling the general mechanisms of cell differentiation at a molecular level.

glide/gcm and the choice of the cell fate Many genes inducing a specific developmental program are expressed and required in more than one cell type (Carmena et al., 1995; Halder et al., 1995; Callaerts et al., 1997). Similarly, glide/gcm is also expressed and required in the presumptive territory of hemocytes, insect scavenger cells that phagocytose dying cells (Bernardoni et al., 1997). A null glide/gcm mutation entails the loss of a subset of these cells. These observations clearly indicate that genes playing an instructive role can affect different cell fates, suggesting a context dependency. Preliminary results suggest that mesodermal expression of glide/gcm does not have the same effects with respect to glial and hemocyte differentiation (R. B. and A. G., unpublished results). In the future, it will be interesting to assess the precise role of glide/gcm in each differentiation program and to determine the molecular cues that dictate the choice between the possible fates. The observation that the glial differentiation program can be activated outside the nervous system and that glide/gcm overrides the developmental programs induced by positional information clearly challenges the common view on the establishment of cell diversity in the nervous system. These results also open new perspectives on the understanding of the molecular mechanisms governing cell differentiation. We thank the Bloomington Stock Center, J. Urban, M. Bates, P. Heitzler, M. Mlodzik, L. Y. Jan and Y. N. Jan for stocks, and D. Kiehart, A. Travers, H. Nguyen and G. Rubin for antibodies. Thanks to S. L. Ang, E. Borrelli, F. Guillemot, P. Lawrence, J. Modolell, P. Sassone-Corsi and P. Simpson for comments on the manuscript. Thanks to G. Duval and V. Vivancos for antibody production, to J. L. Vonesch for confocal microscopy, to R. Walther and C. Hindelang for excellent technical assistance and to B. Boulay and Color System for help with the figures. Confocal microscopy was developed with the aid of a subvention from the French MESR (95.V.0015). This work was supported by the Institut National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, the Centre Hospitalier Universitaire Régionale, the Human Frontier Science Program and the Association pour la Recherche contre le Cancer. R. B. was supported by ARC and EEC fellowships, A. A. M. by FRM and Human Frontier Science Program fellowships.

REFERENCES Akiyama, Y., Hosoya, T., Poole, A. M. and Hotta, Y. (1996). The gcm-motif: a novel DNA-binding motif conserved in Drosophila and mammals. Proc. Natl. Acad. Sci. USA 93, 14912-14916. Altshuller, Y., Copeland, N. G., Gilbert, D. J., Jenkins, N. A. and Frohman, M. A. (1996). Gcm1, a mammalian homolog of Drosophila glial cell missing. FEBS Lett. 393, 201-204. Anderson, D. J. (1995). A molecular switch for the neuron-glia developmental decision. Neuron 15, 1219-1222. Auld, V. J., Fetter, R. D., Broadie, K. and Goodman, C. S. (1995). Gliotactin, a novel transmembrane protein on peripheral glia is required to form the blood-nerve barrier in Drosophila. Cell 81, 757-767. Azpiazu, N. and Frasch, M. (1993). tinman and bagpipe: two homeo box genes that determine cell fates in the dorsal mesoderm in Drosophila. Genes Dev. 7, 1325-1340. Bate, M. (1993). The mesoderm and its derivatives. In The development of

Drosophila melanogaster (eds. Bate and Martinez Arias), vol. II, pp. 10131090, Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press. Baylies, M. K. and Bate, M. (1996). twist: a myogenic switch in Drosophila. Science 272, 1481-1484. Bernardoni, R., Vivancos, V. and Giangrande, A. (1997) glide/gcm is expressed and required in the scavenger cell lineage. Dev. Biol. 191, 118130. Biehs, B., François, V. and Bier, E. (1996). The Drosophila short gastrulation gene prevents Dpp from autoactivating and suppressing neurogenesis in the neuroectoderm. Genes Dev. 10, 2922-2934. Bier, E. (1997). Anti-neural-inhibition: a conserved mechanism for neural induction. Cell 89, 681-684. Bier, E., Vassin, H., Younger-Sheperd, S., Jan, L. Y. and Jan, Y. N. (1992). deadpan, an essential pan-neural gene in Drosophila, encodes a helix-loophelix protein similar to the hairy gene product. Genes Dev. 6, 2137-2151. Bossing, T., Udolph, G., Doe, C. Q. and Technau, G. M. (1996). The embryonic central nervous system lineages of Drosophila melanogaster. Dev. Biol. 179, 41-64. Bour, B. A. O’Brien, M., A., Lockwood, W. L., Goldstein, E. S., Bodmer, R., Taghert, P., Abmayr, S. M. and Nguyen, H. T. (1995). Drosophila MEF2, a transcription factor that is essential for myogenesis. Genes Dev. 9, 730-741. Brand A. H. and Perrimon, N. (1993). Targeted gene expression as a mean of altering cell fates and generating dominant phenotypes. Development 118, 401-415. Brand, M., Jarman, A. P., Jan, L. Y. and Jan, Y. N. (1993). asense is a Drosophila neural precursor gene and is capable of initiating sense organ formation. Development 119, 1-17. Callaerts, P., Halder, G. and Gehring, W. (1997). Pax-6 in development and evolution. Annu. Rev. Neurosci. 20, 483-532. Campbell, G., Göring, H., Lin, T., Spana, E., Anderson, S., Doe, C. Q. and Tomlinson, A. (1994). RK2, a glial-specific homeodomain protein required for embryonic nerve cord condensation and viability in Drosophila. Development 120, 2957-2966. Campuzano, S. and Modolell, J. (1992). Patterning of the Drosophila nervous system: the achaete-scute gene complex. Trends Genet. 8, 202-208. Carmena, A., Bate, M. and Jimenez, F. (1995). lethal of scute, a proneural gene, participates in the specification of muscle progenitors during Drosophila embryogenesis. Genes Dev. 9, 2373-2383. Condron, B. G. and Zinn, K. (1994). The grasshopper median neuroblast is a multipotent progenitor cell that generates glia and neurons in distinct temporal phases. J. Neurosci. 14, 5766-5777. Cripps, R. M., Black, B. L., Zhao, B., Lien, C.-L., Schulz, R. A. and Olson, E. N. (1998). The myogenic regulatory gene Mef2 is a direct target for transcriptional activation by Twist during Drosophila myogenesis. Genes Dev. 12, 422-434. Giangrande, A., Murray, M. A. and Palka, P. (1993). Development and organization of glial cells in the pPeripheral nervous system of Drosophila melanogaster. Development 117, 895-904. Giangrande, A. (1994). Glia in the fly wing are clonally related to epithelial cells and use the nerve as a pathway for migration. Development 120, 523534. Giangrande, A. (1995). Proneural genes influence gliogenesis in Drosophila. Development 121, 429-438. Giangrande, A. (1996). Development and organization of glial cells in Drosophila melanogaster. Int. J. Dev. Biol. 40, 917-927. Giesen, K., Hummel, T., Stollewerk, A., Harrison, S., Travers, A. and Klambt, C. (1997). Glial development in the Drosophila CNS requires concomitant activation of glial and repression of neuronal differentiation genes. Development 124, 2307-2311. Halder, G., Callaerts, P. and Gehring, W. J. (1995). Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science 267, 1788-1792. Halter, D. A., Urban, J., Rickert, C., Ner, S. S., Ito, K., Travers, A. A. and Technau, G. M. (1995). The homeobox gene repo is required for the differentiation and maintenance of glia function in the embryonic nervous system of Drosophila melanogaster. Development 121, 317-332. Heitzler P., Haenlin, M., Ramain, P., Calleja, M. and Simpson, P. (1996). A genetic analysis of pannier, a gene necessary for viability of dorsal tissues and bristle positioning in Drosophila. Genetics 143, 1271-1286. Hosoya, T., Takizawa, K., Nitta, K. and Hotta, Y. (1995). glial cell missing: a binary switch between neuronal and glial determination in Drosophila. Cell 82, 1025-1036. Jessen, K. R. and Mirsky, R. (1992). Schwann cells: early lineage, regulation

3200 R. Bernardoni, A. A. Miller and A. Giangrande of proliferation and control of myelin formation. Curr. Opin. Neurobiol. 2, 575-581. Jones, B. W., Fetter, R. D., Tear, G. and Goodman, C. S. (1995). glial cell missing: a genetic switch that controls glial versus neuronal fate. Cell 82, 1013-1023. Kiehart, D. P. and Feghali, R. (1986). Cytoplasmic myosin from Drosophila melanogaster. J. Cell Biol. 103, 1517-1525. Klaes, A., Menne, T., Stollewerk, A., Scholz, H. and Klambt, C. (1994). The ets transcription factors encoded by the Drosophila gene pointed direct glial cell differentiation in the embryonic CNS. Cell 78, 149-160. Klambt, C., Jacobs, J. R. and Goodman, C. S. (1991). The midline of the Drosophila nervous system: a model for the genetic analysis of cell fate, cell migration, and growth cone guidance. Cell 64, 801-815. Le Douarin, N. M., Dupin, E. and Ziller, C. (1994). Genetic and epigenetic control in neural crest development. Curr. Op. Gen. Dev. 4, 685-695. Lehmann, R., and Tuatz, D. (1994). In situ hybridization to RNA. In Methods in Cell Biology, Vol. 44 (eds L.S.B. Goldstein and E. A. Fyrberg), pp. 575598. San Diego: Academic Press. Lilly, B., Zhao, B., Ranganayakulu G., Paterson, B. M., Schulz, R. A. and Olson, E. N. (1995). Requirement of MADS domain transcription factor DMEF2 for muscle formation in Drosophila. Science 267, 688-693. Luo, L., Liao, Y. L., Jan, L. Y. and Jan Y. N. (1994). Distinct morphogenetic functions of similar small GTPases: Drosophila Drac1 is involved in axonal outgrowth and myoblast fusion. Genes Dev. 8, 1787-1802. Miller, R. H. (1996). Oligodendrocyte origins. Trends Neurosci. 19, 92-96. Nelson, H. B. and Laughon, A. (1994). Drosophila glial development is regulated by genes involved in the control of neuronal cell fate. Roux’s Arch. Dev. 204, 118-125. Pfrieger, F. W. and Barres, B. A. (1995). What the fly’s glia tell the fly’s brain. Cell 83, 671-674. Roark, M., Sturtevant, M. A., Emery, J., Vassin, H., Grell, E. and Bier, E. (1995). scratch, a pan-neural gene encoding a zinc finger protein related to snail, promotes neuronal development. Genes Dev. 9, 2384-2398. Robinow, S. and White, K. (1988). The locus elav of Drosophila melanogaster is expressed in neurons at all developmental stages. Dev. Biol. 126, 294-303. Rusch, J. and Levine, M. (1996). Threshold responses to the dorsal regulatory

gradient and the subdivision of primary tissue territories in the Drosophila embryo. Curr. Op. Gen. Dev. 6, 416-423. Sasai, Y., Lu, B., Steinbeisser, H. and De Robertis, E. M. (1995). Regulation of neural induction by the Chd and Bmp4 antagonistic patterning signals in Xenopus. Nature 376, 333-336. Sauer, F. and Tijan, R. (1997). Mechanisms of transcriptional activation: differences and similarities between yeast, Drosophila, and man. Curr. Opin. Gen. Dev. 7, 176-181. Schmidt, H., Rickert, C., Bossing, T., Vef, O., Urban, J. and Technau, G. M. (1997). The embryonic central nervous system lineages of Drosophila melanogaster. Dev. Biol. 189, 186-204. Scholz, H., Sadlowski, E., Klaes, A. and Klämbt, C. (1997). Control of midline glia development in the embryonic Drosophila CNS. Mech. Dev. 62, 79-91. Schreiber, J., Sock E. and Wegner, M. (1997). The regulator of early gliogenesis glial cells missing is a transcription factor with a novel type of DNA-binding domain. Proc. Natl. Acad. Sci. 94, 4739-4744. Seeger, M., Tear, G., Ferres-Marco, D. and Goodman, C. S. (1993). Mutations affecting growth cone guidance in Drosophila: genes necessary for guidance toward or away from the midline. Neuron 10, 409-426. Stemple, D. L. and Anderson, D. J. (1993). Lineage diversification of the neural crest: in vitro investigation. Dev. Biol. 159, 12-23. Tsukiyama, T. and Wu, C. (1997). Chromatin remodeling and transcription. Curr. Op. Gen. Dev. 7, 182-191. Vincent, S., Vonesch, J. L. and Giangrande, A. (1996). glide directs glial fate commitment and cell fate switch between neurones and glia. Development 122, 131-139. Wade, P. A., Pruss, D. and Wolffe, A. P. (1997). Histone acetylation: chromatin in action. Trends in Biochem. Sci. 22, 128-132. Warf, B. C., Fok-Seang, J. and Miller, R. H. (1991). Evidence for the ventral origin of oligodendrocyte precursors in the rat spinal cord. J. Neurosci. 11, 2477-2488. Weintraub, H., et al. (1991). The myoD gene family: nodal point during specification of the muscle cell lineage. Science 251, 761-766. Wilson, P. A. and Hemmati-Brivanlou, A. (1995). Induction of epidermis and inhibition of neural fate by Bmp-4. Nature, 376, 331-333.