Neurogenic and proneural genes control cell fate specification in the Drosophila endoderm

The segregation of the epidermal and neural precursor cells in the Drosophila ectoderm has become a paradigm for the mechanism of cell type specification that involves transcriptional regulators that belong to the basic helix-loop-helix (bHLH) family. The Enhancer of split gene complex (E(SPL)- C) that comprises seven bHLH genes is required for the development of epidermal cells, and the achaete-scute gene complex (AS-C) that comprises four bHLH genes is necessary for the formation of neural cells (for review see Campuzano and Modolell, 1992; Ghysen et al., 1993; Campos-Ortega, 1993). The E(SPL)-C, together with Notch (N), Delta (Dl), neuralized (neu) and several other genes, belongs to a group of genes that has been collectively called neurogenic genes, because in each of the corresponding loss of function mutants an increased number of neural precursors develops (reviewed by Campos-Ortega, 1993). However, in the achaete (ac), scute (sc), lethal of scute (l’sc) and asense (ase) mutants of the AS-C, as well as in mutants for atonal (ato) and daughterless (da) neural precursor cells fail to develop (Garcia-Bellido, 1979; Brand and Campos-Ortega, 1988; Caudy et al, 1988; Jiménez and Campos-Ortega, 1990; Brand et al., 1993; Domínguez and Campuzano, 1993; Jarman et al., 1993a, 1994). These genes therefore have been designated as proneural genes (Ghysen and Dambly-Chaudière, 1989; Romani et al., 1989). More recently it has been shown that the neurogenic mutants also affect the development of many non-epidermal cell types (Cagan and Ready, 1989; Corbin et al., 1991; Ruohola et al., 1991; Hartenstein et al., 1992). That some of the proneural genes also must be involved in more than one developmental process is demonstrated by the requirement of da and sc for sex determination (reviewed by Jan and Jan, 1993a). However, our understanding of the function of the proneural genes in cell type specification has so far been limited to their participation in the development of the nervous system. We show here that both the neurogenic and proneural genes play a key role in cell type specification of the Drosophila endoderm.

The endoderm of Drosophila, as in most other animals, gives rise to part of the epithelial lining of the digestive tract. The endodermally derived gut epithelium is composed of different cell types. A serial subdivision of the gut tube into specialized regions along the anterior-posterior axis can be observed. Each region contains a main (or principle) cell type with distinct structural and physiological properties. In addition there are scattered specialized cells that are intermingled with the principle cells of the epithelium. Examples include the epithelial stem cells of the mammalian intestine and the adult midgut precursors in the Drosophila larval midgut (e.g. Leblond, 1981; Skaer, 1993). Although the structure and physiology of the gut has been studied extensively, little is known about the embryonic origin and specification mechanisms of endodermally derived cell types. We describe here that the Drosophila endoderm segregates into three cell types, the principle midgut epithelial cells (PMECs), the interstitial cell precursors (ICPs) and the adult midgut precursors (AMPs) early in development. The PMECs form the larval midgut epithelium (Tepass and Hartenstein, 1994b), whereas the ICPs and AMPs initially retain their mesenchymal morphology and enter the midgut epithelium later in development (Reuter et al., 1990; Hartenstein and Jan, 1992). The latter two cell types form scattered populations that are mixed with the PMECs (Filshie et al., 1971, Skear, 1993). Here we show that the specification of PMECs, ICPs and AMPs takes place in the absence of mesoderm, and that it requires the activity of the neurogenic and proneural genes. Our observations suggest that proneural genes together with the neurogenic genes form a gene cassette (Jan and Jan, 1933c) that fulfils a similar function in the ectoderm and in the endoderm.

The following mutations, which are described in Lindsley and Zimm (1992) if not otherwise indicated, were used in this study. A chromosome mutant for both twist¹¹¹ and snail^4·26. Mutations of neurogenic genes include neu^IF65, Dl^9P39, and Dl⁵, and the deletions Df(l)N^8lKl, Df(3R)E(spl)^8D06 and Df(3R)E(spl)^Rl. Mutations of proneural genes studied are Df(1)sc^B57 and Df(2L)da^KXl36- As wildtype stock we used Oregon R.

Germ line clones for a chromosome carrying Df(1)N^8lKl and an ase promoter-lacZ construct (F:2.0; Jarman et al., 1993b) were generated as described (Tepass and Knust, 1993). The Df(l)sc^B57Dl^9P39 double mutants marked with the enhancer-trap insertion B11-2-2 (Bier et al., 1989) were generated by crossing Df(l)sc^B57/+; +/+ ; Dl^9P39/+ X +/Y; B11-2-2/+ ; Dl^9P39/+. Among the offspring of this cross 6.25% of the embryos labeled with B11-2-2 have the genotype Df(l)sc^B57’/Y; B11-2-2/+ ; Dl^9P39/Dl^9P39. These embryos can be recognized by their ameliorated neurogenic phenotype (Brand and Campos-Ortega, 1988). Flies were grown under standard conditions and crosses were performed at room temperature or at 25°C. Egg collections were done on yeasted apple juice agar plates. Embryonic stages are according to Campos-Ortega and Hartenstein (1985).

As cell type specific markers we used the enhancer-trap line B11-2-2 (Bier et al., 1989) and A490.2M3 (Bellen et al., 1989) and promoter-lacZ constructs of the ase gene (F:2.0; Jarman et al., 1993b) and of the lab gene (HZ550 and C₂7.31; Tremml and Bienz, 1992). These enhancer-trap lines and promoter constructs express β-galactosidase which was detected with a polyclonal anti-β-galactosidase antibody (Cappel; dilution 1:2000). A polyclonal anti-ase antibody (Brand et al., 1993) was diluted 1:5000. Antibody stainings and sections of stained embryos were done as described previously (Tepass and Knust, 1993).

Digoxigenin-labeled DNA probes were prepared following manufacture instructions (Genius kit; Boehringer) using full length cDNAs of the AS-C genes ac, sc, l’sc (Cabrera et al., 1987) and the E(SPL)-C genes mβ, mγ, mδ, m3, m5 and m7, and a genomic HindIII-EcoRI fragment that contains the m8 coding region (Klambt et al., 1989; Knust et al., 1992). In situ hybridizations to whole-mount embryos were prepared according to the protocol of Tautz and Pfeifle (1989). Embryos were dehydrated and embedded in a 1:3 mixture of methyl salicylate and Canada balsam.

At the extended germband stage (stages 10 and 11) the endoderm splits up into three cell types, the PMECs, the ICPs, and the AMPs (Fig. 1). This process can be visualized by following the expression of the ase gene (Fig. 2). ICPs are formed exclusively in the distal part of the posterior endoderm; AMPs derive, at a slightly later stage, from the proximal posterior endoderm and the entire anterior endoderm. Both cell populations are clearly distinct from a group of ase-expressing cells that are associated with the foregut and comprise the stomatogastric nervous system (Hartenstein et al., 1994) and cells at the tip of the Malpigian tubules (Hoch et al., 1994; own observations).

During the time at which ICP segregation takes place, the posterior endoderm forms a blind ending sac of epithelial cells. The ICPs assume a bottle shaped morphology with a constricted apical (luminal) surface and a rounded basal cell portion. Curiously, unlike most other delaminating cell populations, the ICPs delaminate towards the apical side of the epithelium, so that they come to lie in the lumen of the posterior endoderm. AMPs and PMECs form at a stage when all endodermal cells lose their epithelial morphology and form solid clusters of apolar, mesenchymal cells. Shortly thereafter, PMECs reorganize into an epithelium which will become the larval midgut epithelium. Both ICPs and AMPs initially remain as mesenchymal cells in the lumen of the forming midgut. The ICPs form a coherent cluster of large cells located towards the center of the midgut and the AMPs are small cells distributed evenly over the entire midgut.

Recent studies have shown that several aspects of endoderm development require interactions between mesoderm and endoderm (Reuter et al., 1993; Tepass and Hartenstein, 1994b; Bienz, 1994). In order to determine whether the specification of PMECs, ICPs and AMPs is also influenced by the mesoderm, we assayed for the development of these cells in twist snail double mutant embryos, which entirely lack the mesoderm (Grau et al, 1984). In this mutant a regular midgut epithelium does not form; instead, the endoderm cells remain as solid mesenchymal clusters (Tepass and Hartenstein, 1994b). However, the segregation into PMECs, ICPs and AMPs still occurs. Similar to what has been shown above for wild-type embryos, in twist snail double mutants a fraction of endodermal cells express ase (Fig. 3). Some of the ase-positive cells form a cluster of large cells (the ICPs), whereas other, small ase-positive cells are scattered over the entire endoderm (the AMPs).

In embryos with reduced or no function of the neurogenic loci E(SPL)-C, N, Dl and neu, midgut development is severely disturbed (Hartenstein et al., 1992; Reuter et al., 1993), and in N mutants the number of AMPs is increased (Hartenstein et al., 1992). Further analysis revealed that in mutants carrying null alleles of Dl and in embryos that lack maternal and zygotic N expression, all endodermal cells express ase (Fig. 4B,D). Expression of lab, which serves as a marker for a subset of PMECS (Reuter et al., 1990; Fig. 4G), is absent in these mutants (data not shown). These findings indicate that PMECs do not develop. Correspondingly, in sections of such mutant embryos, no trace of a midgut epithelium can be found (Fig. 4B,D). Approximately 30-50% of the ase-positive cells in the mutant embryos form a coherent cluster of large cells, while the remaining ase-positive endodermal cells form a solid mass of small cells, suggesting that both ICPs and AMPs develop in excess numbers (not shown).

In embryos mutant for a strong but not amorphic neu allele (not shown) and in embryos that lack only zygotic N function, a small fraction of PMECs develops (Fig. 4C). Embryos carrying deletions removing the entire E(SPL)-C which show a consistent extreme neuralization of the ectoderm display a variable endodermal phenotype (Fig. 4E,F). In some embryos, the large majority of endodermal cells turn into ase-positive AMPs and ICPs, whereas in others a substantial fraction of PMECs appears, which are able to form small epithelial islands (Fig. 4E,I). In conclusion, lack of neurogenic gene function leads to an increase of the number of AMPs and ICPs at the expense of PMECs. Even though they become attached to the visceral mesoderm, AMPs and ICPs are incapable of forming an epithelium.

Midgut development was studied in embryos carrying Df(1)sc^B57, which removes all four genes of the AS-C, and Df(2L)da^KXI36, which deletes the da locus. Because the endoderm expression of ase depends on AS-C and da function (Brand et al., 1993), the enhancer-trap lines B11-2-2 or A490.2M3 were used as general endodermal markers and PMECs, ICPs, and AMPs distinguished on the basis of their size, shape and position (Tepass and Hartenstein, 1994b; Fig. 5). In AS-C mutants, ICPs are absent and the number of AMPs is strongly reduced. Since no cell death could be seen in the developing midgut (not shown) it is likely that the cells that would have developed as ICPs and AMPs in wild type, develop as PMECs instead. The PMECs of AS-C mutants form a structurally normal midgut epithelium. A loss/reduction of ICPs and AMPs also occurs in da mutant embryos. However, in this mutant, cell death is apparent in the developing midgut in both A490.2M3-labeled (Fig. 5F) and toluidine blue stained sections (not shown). This suggests that in da, the development of ICPs and AMPs may be initiated, but is terminated shortly thereafter by apoptosis.

The developmental fate of the ICPs, which had previously been called large basophilic cells because of their staining properties (Campos-Ortega and Hartenstein, 1985), has to date been unknown. The finding that the ICPs are missing in AS-C and da mutants allowed us to address the issue of ICP fate. Using lab as a marker for the copper cells (Hoppler and Bienz, 1994), we can show that this cell type is present in normal amounts in AS-C and da mutants (Fig. 5G-I), but that they are not mixed with unlabled cells as in wild type (not shown) indicating that the ICPs represent the precursors of the interstitial cells and that the copper cells develop as a subpopulation of the PMECs.

In double mutant combinations between neurogenic and proneural genes a prevalance of the proneural and a suppression of the neurogenic phenotype was observed in the neuroectoderm (Brand and Campos-Ortega, 1988; Heitzler and Simpson, 1991). Our analysis of endoderm development in Dl AS-C double mutants yields similar results (Fig. 6). An appar-ently normal midgut epithelium forms in these double mutants suggesting that the PMECs develop and differentiate normally. As in AS-C mutants we find that the ICPs are missing in the double mutants (Fig. 6) and that the number of AMPs is reduced (not shown). Taken together, these observations suggest that the proneural phenotype is epistatic to the neurogenic phenotype in the endoderm, implying that the neurogenic genes are not required for the reorganization of the PMECs into an epithelium once the PMECs are specified.

The expression of the E(SPL)-C and AS-C genes in the neuroectoderm is tightly regulated temporally and spatially and correlates with the segregation of neural and epidermal precursors (Cabrera et al., 1987; Romani et al., 1987, 1989; Cubas et al., 1991; Martín-Bermudo et al., 1991; Skeath and Carroll, 1991, 1992; Knust et al., 1987, 1992). In the endoderm the bHLH genes of the E(SPL)-C (m0, my, m§, m3, m5, m7, and m8, which corresponds to the E(spl) gene) are expressed in very similar patterns, with the exception of m5 for which no expression was detected. Expression is first seen in late stage 9/early stage 10 embryos in the distal posterior endoderm, rapidly followed by a uniform expression in the entire anterior and posterior endoderm. During late stage 10/early stage 11, E(SPL)-C expression disappears, first from the distal posterior endoderm, then from the remaining endoderm (Fig. 7A-C). Weak expression of some of the E(SPL)-C transcripts was seen in the ICPs at stage 11.

The AS-C genes are expressed in distinctly different patterns in the endoderm. We were unable to detect ac transcript in the endoderm, although the same embryos showed intense labeling of proneural clusters in the neuroectoderm. l’sc expression (Fig. 7D-G) appears in the distal posterior endoderm at the end of stage 8. During stage 9 uniform expression is seen in the entire endoderm. l’sc mRNA disappears from the endoderm during stage 10, initially from the distal posterior endoderm, shortly thereafter from the remaining parts of the endoderm. Weak l’sc signal remains in the ICPs at stage 11. The pattern of sc expression follows that of l’sc with a temporal offset (Fig. 7H-N). sc expression, which is initially (early stage 10) homogeneous, increases in the presumptive ICPs before they segregate from the distal posterior endoderm. Slightly later, the level of sc in the presumptive AMPs is also increased. The pattern of sc expression reveals that the presumptive ICPs and AMPs are evenly spaced before their segregation. They are separated by a single cell diameter (Fig. 7I). Expression remains strong in both ICP and AMP cells during and after their segregation until mid stage 11, while the PMECs lose sc transcript. ase, as described above, is expressed exclusively in the ICPs (late stage 10 to mid stage 11) and AMPs (early stage 11 to late stage 12) during and after their segregation.

To study the interactions between neurogenic and proneural genes in midgut development we examined sc expression in neurogenic mutants (Fig. 8). In Dl mutant embryos the level of sc expression is uniform throughout the endoderm and substantially increased compared to presumptive ICPs and AMPs in wild-type embryos (Fig. 8A,B). These findings suggest that the normal activity of neurogenic genes negatively regulates proneural gene transcription in all endodermal cells, i.e. initially also in those cells that accumulate sc in wild type and develop as AMPs or ICPs.

Embryos carrying deletions of the E(SPL)-C show a variable increase in sc expression (Fig. 8C,D). Thus, in some E(SPL)-C mutant embryos high levels of sc expression occurred in the entire endoderm, while in other mutant embryos only scattered patches of endoderm cells showed elevated sc expression. This finding is consistent with our phenotypic observations, which show that in Dl mutants PMECs are missing altogether, whereas in E(SPL)-C mutants the number of PMECs is variably reduced. It seems likely that other, yet unidentified factors, possibly additional bHLH genes located outside of the E(SPL)-C, cooperate with the E(SPL)-C bHLH genes in order to suppress proneural gene activity in the endoderm.

The diverse cell types found in the epithelial lining of the Drosophila larval midgut are specified by at least two different mechanims. During an early endoderm autonomous mechanism involving the activity of the neurogenic and proneural genes, three cell types, the AMPs, ICPs and PMECs, are generated. In neurogenic mutants the number of AMPs and ICPs is strongly increased at the expense of PMECs. In proneural mutants, however, ICPs are absent and the number of AMPs is strongly reduced. The AMPs and the ICPs alone, in contrast to the PMECs, are not capable of forming an epithelium as can be seen in neurogenic mutants. After the PMECs have formed an epithelium, they are regionally specified to form several structurally distinct subpopulations. Previous studies have shown that the specification of at least some of these different PMEC fates depends on interactions between the visceral mesoderm and the endoderm (reviewed by Bienz, 1994).

The apparently normal segregation of AMPs, ICPs and PMECs in embryos without mesoderm is a novel feature of endoderm differentiation in Drosophila. We have not formally proved that the segregation of AMPs and ICPs from the remaining endodermal cells, the PMECs, is a process that requires only endoderm autonomous functions. Two arguments, however, strongly support this notion. First, we found that this segregation process takes place in embryos that do not have mesoderm. Secondly, the neurogenic and proneural genes whose normal function is required for the segregation process are expressed in the endoderm at the time when the segregation takes place (Romani et al., 1987; Knust et al., 1987; Hartley et al., 1987; Kidd et al., 1989; Godt, 1990; Kooh et al., 1993; Brand et al., 1993; this work). The dynamics of the expression of the proneural genes and the E(SPL)-C in relationship to ICP/AMP segregation is similar to the dynamics observed for the expression of these genes during neuroblast segregation from the neuroectoderm, where an autonomous requirement for several of the neurogenic and proneural genes has been demonstrated by mosaic analysis (reviewed by Campos-Ortega, 1993; Simpson et al., 1993).

Several bHLH genes of the E(SPL)-C, l’sc and initially also sc, are expressed uniformly throughout the endoderm. This represents a fundamental difference to the expression pattern of these genes in the neuroectoderm. Here the proneural genes are activated in small groups of cells, called proneural clusters, by the anterior-posterior and dorsal-ventral patterning genes (Martin-Bermudo et al., 1991; Skeath et al., 1992). Each proneural cluster forms an equivalence group from which a single neural precursor is selected while the remaining cells form epidermal precursors (Heitzler and Simpson, 1991; Ghysen et al., 1993). Since the proneural cluster from which a single neural precursor emerges is small, the position of the neural precursor cells is mainly determined by the prepattern provided by the anterior-posterior and dorsal-ventral patterning genes (Fig. 9A).

The uniform expression of proneural and neurogenic genes in the endoderm suggests that a prepattern does not exist in this structure prior to the specification of the AMPs, ICPs and PMECS. This implies that the pattern of AMPs and ICPs is solely determined by the proneural and neurogenic genes. The result is a simple pattern of rather regularly spaced segregating cells. The fact that ICPs and AMPs are separated by a single cell diameter is consistent with the inhibitory signal provided by the Dl protein acting as a membrane bound and not as a diffusible factor (Vâssin et al., 1987; Kopczynski et al., 1988; Fehon et al., 1990). We propose that the control of the segregation of the AMPs and ICPs from the PMECs represents a simple, two step process in which a uniform activation of the proneural genes is followed by a uniform activation of the neurogenic genes (Fig. 9A).

Our findings show a striking similarity between the endoderm and neuroectoderm with respect to the expression pattern and mutant phenotype of the proneural and neurogenic genes. In both tissues the genes of the neurogenic/proneural group are required for the segregation of different cell fates where the cells that depend on proneural gene function make up a rather small number compared to the cells that require neurogenic gene function. In both tissues l’sc, sc and the E(SPL)-C are activated uniformly in the epuivalence groups that give rise to neural precursors or AMPs/ICPs, respectively. Later, the expression of l’sc and sc increases in the segregating cells and decreases in the remaining cells of the equivalence groups (PMECs or epidermal precursors; Cabrera et al., 1987; Martin-Bermudo et al., 1991; Vâssin et al., 1994; this work).

Furthermore, the genetic and molecular interactions between neurogenic and proneural genes appear to be similar in endoderm and neuroectoderm. In neurogenic proneural double mutants (Brand and Campos-Ortega, 1988; Heitzler and Simpson, 1991; this work) the proneural phenotype prevails and the development of both epidermis and PMECs is restored. This suggests that neurogenic gene function is mediated through the proneural genes in both the neuroectoderm and the endoderm. Similar to the proneural clusters in the neuroectoderm (Brand and Campos-Ortega, 1988; Skeath and Carroll, 1992), the neurogenic genes repress proneural gene expression in the endoderm as indicated by the increased transcript level of sc in Dl and E(SPL)-C mutants. The uniform high level of sc expression in the endoderm of neurogenic mutants indicates a lack of both mutual and lateral inhibition similar to that in the neuroectoderm (Simpson, 1990; Ghysen et al., 1993).

The similarity between the neuroectoderm and the endoderm extends to the function of da. In da mutants all neural precursors segregate from the neuroectoderm but do not continue to develop and degenerate (Brand and Campos-Ortega, 1988; Vâssin et al., 1994). In the endoderm of da mutants we find evidence that AMPs and ICPs die soon after they have segregated. Expression of ase, which is restricted to the AMPs and ICPs, is not seen in AS-C or da mutant embryos (Brand et al., 1993; own observations) suggesting that the bHLH protein encoded by da, which cooperates with the bHLH factors of the AS-C in order to control target gene expression (e.g. Murre et al., 1989), might also do so in the endoderm.

It has been speculated that, in contrast to the neurogenic genes (see Introduction), the proneural genes are specifically required to promote neural developmental pathways (for review see Ghysen and Dambly-Chaudière, 1988; Jan and Jan, 1993a). This notion, regarding the genes of the AS-C, is inconsistent with the results described here which show that specification of endodermally derived ICPs and AMPs, two cell types unrelated to the nervous system, require proneural gene activity. Since the function of the neurogenic and proneural genes is apparently not related to the terminal fate of the involved cells the question arises of how the specific cell fate decisions are made. The proneural genes have to act in concert with other factors to commit cells to a particular developmental pathway (Fig. 9B). Candidates for these additional factors are genes that are expressed in a tissue-specific or positional-specific manner as, for example, the anterior-posterior and dorsal-ventral patterning genes in the neuroectoderm and the gene serpent (srp) in the endoderm. In srp mutants, the endoderm is homeotically transformed into ectodermal portions of the gut (Reuter, 1994). The expression of proneural genes plus srp may specify the ICPs and AMPs, and endoderm cells in which proneural genes are repressed by lateral inhibition and which only express srp would develop as PMECs.

In both neuroectoderm and endoderm, the neurogenic genes promote the development of epithelial cells (epidermis and PMECs, respectively), whereas the proneural genes are required for the development of mesenchymal cell types (neural progenitors and AMPs/ICPs, respectively). The hypothesis that neurogenic and proneural genes promote the development of epithelial versus mesenchymal cell morphology, respectively (Fig. 9B), is corroborated by the functional characterization of homologs of the neurogenic and proneural genes in vertebrates. The expression of an activated form of the Xenopus N homolog, Xotch, causes an increase in epithelial neural tube tissue and abolishes the development of the mesenchymal neural crest (Coffman et al., 1993). Uniform activation of N in Drosophila leads to the loss of mesenchymal neural precursors and presumably to an increase in epithelial neuroectodermal cells (Struhl et al., 1993; Rebay et al., 1993; Lieber et al., 1993). Furthermore, the myoD family of transcriptional regulators is structurally related to and operates apparently in a similar manner to the Drosophila proneural genes (Jan and Jan, 1993a). Rudnicki et al. (1993) found that myoD together with myf-5, a second bHLH gene of the myoD family, promotes the segregation of mesenchymal myoblasts from the epithelial somite.

Epithelial and mesenchymal cells are the two basic components of animal embryos (e.g. Bard, 1990; Tepass and Harten-stein, 1994a). The segregation of a homogeneous cell population into these two cell types occurs repeatedly in each germlayer throughout embryonic development. The data reported here and the results of others suggest that the activity of the neurogenic and proneural (and possibly the myoD family) genes controls this segregation process. The genetic program that leads to the differentiation of an epithelial cell structure is suppressed in an embryonic cell in which proneural genes are active and a mesenchymal program is realized. In adjacent cells that derive from the same equivalence group and in which neurogenic genes suppress proneural gene activity, an epithelial program is realized by default.

We are grateful to Mariann Bienz, Michael Brand, Sonsoles Campuzano, Andrew Jarman, Fernando Jimenez, and Elisabeth Knust for providing DNA probes, antibodies, and fly stocks which facilitated this study. We thank the Bowling Green and Bloomington Drosophila Stock Centers for sending fly stocks. We are grateful to Dorothea Godt for stimulating discussion. We appreciate critical reading of the manuscript by Dorothea Godt and members of the Hartenstein laboratory. This work was supported by NIH Grant NS29367 to V. H. and by a HFSPO fellowship to U. T.