Sequential fates in a single cell are established by the neurogenic cascade in the Malpighian tubules of Drosophila

The question of how cells become different from one another is best addressed in systems in which such differences can be recognised soon after cell fates are specified. In particular, the segregation of individual cells provides a simple model in which to analyse the mechanisms that underlie both cell specification and differentiation (e.g. the anchor cell in the nematode vulva (Greenwald et al., 1983), individual photore-ceptor cells in the adult eye (Tomlinson, 1988) and individual neuroblasts and sensillum precursor cells in the nervous system of Drosophila (reviewed in Campos- ortega, 1993; Goodman and Doe, 1993; Jan and Jan, 1993). Studies with such systems indicate that local interactions between cells frequently underlie the specification of cells to a particular developmen-tal fate.

We present a novel system in the Malpighian tubules of Drosophila melanogaster in which a single cell, the tip cell, becomes distinctive early in development and subsequently adopts a characteristic fate, so that details of its differentiation can be charted. Tip cells can be followed from their birth to full differentiation in a system that is amenable to genetic analysis, organ culture (Broadie et al., 1992) and cell manipulation (Skaer, 1989), so that the activity of these cells can be analysed at every stage in their life history.

The four Malpighian tubules of Drosophila function as a kidney, excreting nitrogenous waste as uric acid and regulat-ing the ionic balance of the organism. The tubules arise during embryogenesis as four protuberances from the proctodeum (Fig. 1) and grow, first by cell proliferation and then by extensive rearrangement of the cells to produce elongated, blind ending tubes composed of a single-cell-layered epithe-lium. The four tubules are arranged as two pairs, each pair emptying into a common ureter, and are distinct both in their length and arrangement in the organism. The longer, anterior pair project forward from their point of insertion in the hindgut, whereas the posterior pair extend backwards so that their tips become attached to the posterior part of the hindgut (Poulson, 1950; Wessing and Eichelberg, 1978). Shortly after the tubules segregate from the hindgut, the tip cell becomes apparent at the distal end of each growing primordium (Fig. 1E,I), distinctive in that it does not divide at a stage when other cells in the primordium are proliferating. Tip cell ablation results in the failure of division in the remaining cells of the tubule, indi-cating that its presence is required for cell proliferation in the tubule of which it is a part (Skaer, 1989).

The genetic circuitry underlying the segregation of Malpighian tubules from the proctodeum has been charted in some detail (reviewed in Skaer, 1993). Maternally supplied products act through a signalling cascade to specify the embryonic termini, resulting in the expression of the zygotic terminal genes, tailless and huckebein in these regions. Tailless and Huckebein act through the transcription factor, fork head (Weigel et al., 1989), to initiate the expression of Krüppel (Kr) in the posterior domain (Gaul and Weigel, 1990), which forms the proctodeum after gastrulation. The expression of Kr, which encodes a zinc finger DNA-binding protein (Rosenberg et al., 1986), shows a dynamic pattern of expression in the proctodeum, gradually becoming restricted to a ring of cells, which push out to form the tubule primordia (Gaul et al., 1987; Fig. 2). This process of eversion fails in embryos mutant for Kr (Gloor, 1950; Redemann et al., 1988), suggesting that Kr plays a critical role in the specification (Harbecke and Janning, 1989) or segregation (Skaer, 1993) of tubule primordial cells.

Little is known about the subsequent specification and differentiation of cell types in the Malpighian tubules. In this paper, we analyse the processes leading to the specification of the tip cell and show that it is segregated from a cluster of equipotent cells, through a series of inhibitory cell interactions. We identify elements of the genetic hierarchy that underlies tip cell allocation and demonstrate a key role for the proneural genes in conferring tip cell potential and for the neurogenic genes in the segregation of a single tip cell. By following the activity of the tip cell after its selection, we show that this cell adopts two sequential fates, playing an important role in regu-lating cell division during embryogenesis and subsequently differentiating many characteristics typical of neural cells.

We used Oregon R, N^55e11, da^KX136, Dl^9P, stg^9M53, (provided by the Tübingen stock center), ovo^D1 (a gift from J. Campos- ortega), Df (1) sc^B57 and Df (1) svr (which removes AS-C and the neighbouring genes, elav and vnd) (a gift of Juan Modolell’s laboratory) and the enhancer trap lines A31 (Ghysen and O’Kane, 1989), A37, A434, A101 (Bellen et al., 1989), E7-3-58 and B7-3-32 (Hartenstein and Jan, 1992). The flies were maintained and embryo collections made according to standard procedures.

Germ-line clones were generated by X-irradiation of 36-hour-old larvae from N^55e11/FM7c females crossed to ovo^D1 males, according to the method of Jimenez and Campos- ortega (1982).

Antibody staining of whole-mount embryos was carried out as described previously (Macdonald and Struhl, 1986), using the Vec-tastain ABC Elite-horseradish peroxidase system. NiCl₂ or Ni/CoCl₂ enhancement was used where necessary. The stained embryos were embedded in Araldite in capillaries according to the procedure of Schmidt-Ott and Technau (1992).

We used the following antibodies at the dilutions indicated in parenthesis: mAb22C10 (Zipursky et al., 1984; 1:20 dilution), mAbfasII (Grenningloh et al., 1991; 1:20), mAb9BDelta (gift of Artavanis; 1:1000) mAbachaete (Skeath and Carroll, 1991; 1:50), anti-Krüppel (Gaul et al., 1987; 1:100), anti-β-galactosidase (Cappel; 1:10000), anti-Synaptotagmin (Littleton et al., 1993; 1:50), anti-HRP (Sigma; 1:2500), anti-Synaptophysin (Dianova; 1:100) and anti-Cysteine-string protein (ab 49192; Buchner; 1:10).

Individual tip cells were identified by their distinctive position and morphology or by staining embryos or larvae from the enhancer trap line A37 with X-gal (Ashburner, 1989). Tip cells were filled in the embryo (stages 16 and 17), in newly hatched larvae and in mature third instar larvae. Living preparations or those lightly fixed for 2 minutes in 4% paraformaldehyde were viewed with Nomarski optics and iontophoretically injected with a fluorescent dye under epifluo-rescence; a solution of 2% N-(2-aminoethyl) biotinamide hydrochlo-ride (Neurobiotin (FW 322.47); Vector Laboratories) and 2% Lucifer Yellow. Lucifer Yellow was injected with small (nanoamps) hyper-polarising current pulses under direct observation; to load Neurobi-otin, the polarity of the current was reversed after confirming cell identity with the Lucifer Yellow. After injection, preparations were fixed for 1 hour in 4% paraformaldehyde, washed with PBT, and incubated with a commercial avidin-peroxidase complex (ABC Elite kit; Vectastain) for 30 minutes. The specimens were then reacted with DAB, cleared and mounted in the normal way.

Tubule-specific markers label those cells in the proctodeum that will evert to form the Malpighian tubule primordia (Figs 1, 2). The allocation of these cells is reflected in the expression of Kr during stage 10 (Gaul et al., 1987) (Fig. 2B). The band of cells expressing Kr on the dorsal side of the proctodeum becomes narrow, only 1-2 cells in width, whereas laterally and ventrally the band remains broad (Fig. 2C). This broader band resolves into four groups of 25-30 cells, towards the end of stage 10 (cf. Fig. 2B with C). As the primordia evert (during stage 11), Kr expression is modulated, becoming stronger in a subset of 6-8 cells in each primordium (Fig. 2C). Expression in all but one of these cells gradually fades so that by mid-stage 11, Kr is most strongly expressed in a single cell in each tubule (Fig. 2D-E). Later in stage 11 strong Kr expression is apparent in two cells of unequal size that lie adjacent to one another. This pattern persists until late in stage 12 (Fig. 2F), when expression is once more limited in each tubule to a single cell, the larger of the two daughter cells, which can be identified by its characteristic shape, position and size as the tip cell (Skaer, 1989; 1992) (Fig. 2G,H). These distinctive cells remain at the distal tip of each tubule and continue to express Kr as the tubules elongate and mature. The Malpighian tubules take up characteristic positions in the mature embryo, so that the tip cells of the anterior tubules contact fat body anterior to the ureters and those of the posterior pair contact the visceral mesoderm ensheathing the hindgut close to the anus (Fig. 2I).

The gradual restriction of Kr expression within the tubule primordia suggests a model for tip cell segregation in which a single cell becomes determined from a group of equivalent cells, as in the segregation of neuroblasts and sensory mother cells from proneural clusters in the ectoderm during neuroge-nesis (reviewed in Campos- ortega, 1993; Jan and Jan, 1993). Cells in proneural clusters are marked out by the expression of the proneural genes, which include the four transcription units of the achaete-scute complex (AS-C) : achaete (ac), scute (sc), lethal of scute (l’sc), and asense (ase), and which, with the ubiquitously expressed gene, daughterless (da), are required for the neural potential of cells in these clusters (reviewed in Goodman and Doe, 1993).

We therefore examined embryos stained for products of the proneural genes and found expression in the proctodeum and, later, in cells of the Malpighian tubules (data for ac shown; Fig. 3). During stage 9/10 both transcripts and the protein products of ac, sc and l’sc are expressed at a low level in 4 groups of 6-8 cells in the proctodeum (Fig. 3A,B). An hour later (early stage 11), when the tubule primordia evert, these genes are more strongly expressed but in a single cell in each primordium, always situated in the middle of the posterior-facing side (Fig. 3C,D). From this stage, the other proneural gene, ase, is expressed in the same pattern as ac, sc and l’sc. From mid-stage 11 to early in stage 12, these proneural genes are expressed in two cells, which remain closely paired in each primordium (Fig. 3E). These cells gradually move towards the distal end of the tubules, with expression persisting in only one of them, so that by mid-stage 12, products of the proneural genes are confined to a single cell at the distal tip of each tubule (Fig. 3F). This pattern continues until stage 16, when the expression of ac is lost altogether from the tubules.

We conclude that the cell pair expressing proneural genes arises by division of the first strongly expressing cell rather than by recruitment, since in embryos mutant for string (stg), in which all postblastodermal cell division fails (Edgar and O’Farrell, 1990), only one cell expresses ac from stage 11. Fur-thermore, the expression of β-galactosidase in the enhancer trap lines A37, A101, A434 (Fig. 3 H,I), B7-3-32, E7-3-58, which marks out the tip cell and its close neighbour in wild-type embryos (Fig. 3H), remains in a single cell in a stg mutant background (Fig. 3I).

We have shown that the expression of both ac and Kr is gradually restricted to a single cell in each tubule. Double staining with antibodies to both proteins confirms that these genes are coexpressed in the tip cell lineage (Fig. 3G), since each tubule contains only 2 stained cells early in stage 12 (see Fig. 2F for Kr and Fig. 3E for ac expression). In embryos lacking the function of proneural genes (e.g. hemizygous for Df(1)sc^B57, Df(1)svr or homozyous for da), although Kr is initially expressed in the tubule primordia (Fig. 4C), it decays completely during stage 12 so that no tip-cell-specific expression remains (Fig. 4D). Furthermore, no morphologi-cally distinctive tip cell appears at the distal tip of the tubules in these embryos. In contrast, in embryos mutant for extra-macrochaetae (emc), in which ac is expressed more widely than in the wild type (Martinez et al., 1993), both ac and Kr are expressed in many cells in the tubules (see Cubas et al., 1994 for ac, our unpublished observations for Kr). Thus the late expression of Kr in the tubules, which is normally confined to the tip cells, appears to be under the positive control of the proneural genes.

In embryos hemizygous for Df(1)sc^B57 or Df(1) svr, and therefore lacking the function of proneural genes, the Malpighian tubules remain shorter than in wild-type embryos. Further, the posterior pair are misplaced, lying in the central region of the abdomen, rather than extending distally and attaching to the hindgut. This observation suggests that inter-actions between the tip cell and the hindgut might be required to anchor the mature posterior tubules.

The number of cells in both the anterior and posterior tubules is also affected by the absence of tip cells; with an average of 72±3 (s.e.m.; n=25) and 73±4 (n=18) in embryos hemizygous for Df(1) svr and Df(1)sc^B57 respectively, compared with 124±3 in mature wild-type embryos (Baumann and Skaer, 1993). Our observations indicate that the tip cell is finally selected between 7.5 and 8 hours AEL. At this time, the number of cells in the tubules of wild-type embryos is between 63±3 (7.75 hours) and 77±3 (8.25 hours) (Skaer and Martinez Arias, 1992). Thus the tip cell appears to be active in promoting cell division in its neighbours from the time of its selection. If the tip cells are experimentally ablated from a tubule primordium at this time (7.75 hours), the final number of cells in the tubule is 75±4 (Skaer, 1989). The close correlation between the effects of the loss of proneural gene function (72±3, 73±4) and the experimental ablation of the tip cell immediately after its selection (75±4) on the final number of tubule cells suggests that, in the absence of the proneural genes, the mitogenic function of the tip cells is lost.

The segregation of the tip cell occurs in two steps; first by the selection of a ‘tip mother cell’ from a group of cells expressing ac, l’sc and sc in the proctodeum and subsequently by selection of one of its daughter cells, both of which express proneural genes. During neurogenesis, interactions between cells within each proneural cluster confirm the neural potential of one cell; the remaining cells of the group adopt an epidermal fate. The products of the neurogenic genes (Notch (N), Delta (Dl), neuralised (neu), mastermind, bigbrain, almondex, groucho and Enhancer of split) mediate these interactions, so that in embryos lacking their function extra cells are allocated to the nervous system due to a failure in lateral inhibition (reviewed in Campos- ortega, 1993).

We therefore analysed the Malpighian tubules of embryos lacking the function of neurogenic genes. In embryos mutant for Notch (N) and Delta (Dl), the restriction of Kr and ac expression to a single cell in each tubule primordium fails, so that in mature embryos up to 12 cells in each tubule continue to express Kr and ac (Fig. 4E,F; Table 1). Furthermore, enhancer trap lines that mark out the tip cell and its sibling reveal a similar phenotype when crossed into a N mutant back-ground; supernumerary tip cells segregate from the tubule pri-mordial cells, so that again up to 12 cells in each tubule express β-galactosidase (Fig. 4G; Table 1). In germ-line clones of N, virtually all the tubule cells express markers for the tip cell (Fig. 4H). Interestingly, the tubules of these embryos do not secrete uric acid, indicating that these supernumerary ‘tip cells’ do not differentiate later as normal secretory epithelial cells.

In the developing nervous system, the expression of Dl and neu becomes restricted to the neuroblast, once it segregates from other cells in the proneural cluster, whereas N shows a reduced level of expression in these cells (Heizler and Simpson, 1991; Kooh et al., 1993; Haenlin et al., 1990). We find that both Dl and neu expression is restricted in the tubules, initially to the tip mother cell and subsequently, during stage 13, to the tip cell (Dl, Fig. 4A,B, neu (as reported by the expression of β-galactosidase in A101, Fig. 6A). In contrast, Notch is found on the membrane surface of all the primordial cells and, later, on all the Malpighian tubule cells, with the exception of the tip cell (data not shown). In embryos mutant for N, many cells continue to express Dl at a stage when expression is restricted to a single cell in wild-type embryos (data not shown).

We conclude that the tip mother cell segregates from a group of cells that have the potential to adopt this cell fate and that the selection of the tip mother cell results from cell interactions requiring the expression of the neurogenic genes. Further, our observations suggest that a second interaction, between the daughter cells of the tip mother cell, is mediated by the neurogenic genes and results in the selection of the tip cell. This is indicated by comparing the number of Kr- or ac-expressing cells (i.e. tip cells) in the tubules of embryos mutant for N with the number of cells expressing β-galactosidase, when an enhancer trap line that labels both the tip cells and their siblings is crossed into a N mutant background (Table 1). The similarity between these figures suggests that in the absence of functional Notch both daughters of the tip mother cell adopt the tip cell fate.

The events underlying tip cell allocation are therefore closely similar to the process of neuroblast or sensillum precursor cell selection during neurogenesis.

In wild-type embryos, the tip cell remains morphologically dis-tinctive as the tubule matures and becomes physiologically active (Skaer, 1992). Each tip cell protrudes from the distal end of the tubule but remains firmly anchored in the tubule by a long proximal extension into the lumen (Fig. 5A). The distal extremities of all four tubules are found in characteristic positions in the mature embryo. Those of the anterior tubules become embedded in fat body that is associated, in one case, with the tracheal system and, in the other, with the posterior midgut. The tips of the posterior tubules lie one on either side of the hindgut (see Fig. 2I) and, unlike the anterior tip cells, form a close association with the nervous system. Branches of the paired nerves from abdominal ganglion a9 run up on either side of the hindgut, making contact with the tip cell and inner-vating the visceral muscles (Campos- ortega and Hartenstein, 1985). The posterior tubules first contact the hindgut during stage 15 (Fig. 5B) and then, as the branches of a9 grow along the hindgut during stage 16, the tip cells contact them, extending two short protrusions from their distal surfaces (Fig. 5C,D). These contacts persist, so that the posterior tip cells remain in intimate contact with both the surface of the hindgut and the nerve branches innervating it.

From stage 15, the tip cell starts to express a number of cell surface markers that typify neural cells; antibodies against HRP (Jan and Jan, 1982) as well as the monoclonal antibodies 22C10 (Zipursky et al., 1984) and 44C11 (Bier et al., 1988) highlight only the tip cell in each tubule (data for 22C10, Fig. 5A-D). In addition, the expression of β-galactosidase under the regulation of enhancers in neu (A101) (Huang et al., 1991; Boulianne et al., 1991) (Fig. 6A) and neuromusculin (A37) (Kania et al., 1993) (Fig. 6B), both of which are expressed in neural cells, mark out the tip cell and its sibling.

The later differentiation of the tip cell depends on the normal expression of both the proneural and the neurogenic genes. In embryos lacking the function of the proneural genes, there are no 22C10-positive cells in the Malpighian tubules (Fig. 5E) and in embryos mutant for N, multiple 22C10-positive cells appear at the distal end of the tubules (Fig. 5F).

Together, these observations raise the possibility that the tip cell behaves as a neuron and that the posterior pair interact with the hindgut branch of a9. In order to analyse their arbori-sation, we have filled tip cells with the dye, Lucifer Yellow, in the late embryo and in first and third instar larvae. Exami-nation of filled, living cells, as well as permanent preparations stained through a biotin conjugate (Fig. 6E,F), confirms that the tip cell is a large cell with a long neck-like process extending into the lumen of the tubule (cf. Fig. 5A-D) and that it remains closely associated with the branches of a9 throughout larval life (Fig. 6F). These fills also reveal that the tip cell contacts on the nerve remain short; it does not send a lengthy cytoplasmic process along the nerve. However, antibodies raised against a number of synaptic proteins such as Synaptotagmin (Fig. 6C; Littleton et al., 1993); Synaptophysin (data not shown; Weidemann and Franke, 1985) and the cysteine-string protein (Fig. 6D; Buchner et al., 1988; Zinsmaier et al., 1990)) stain the Malpighian tubules; initially in the whole tubule but, by stage 17, only in the tip cells (Fig. 6C,D). We conclude that the tip cells of the posterior tubules might make synaptic connec-tion(s) with neuron(s) in a9 at the site of contact between the nerve branch and the tip cell on the surface of the hindgut.

Our observations reveal striking parallels between neurogenesis and tip cell segregation (summarised in Fig. 7). The domain from which neural or tip cells can arise is marked out by early acting genes. Within this domain, the expression of the proneural genes reflects the competence of cells to adopt a neural or tip cell fate; a com-petence that is revealed by mutations in neurogenic genes. The neurogenic gene products mediate competitive interactions between competent cells so that one adopts the neural or tip cell fate, expressing enhanced levels of Delta, Neuralised and the proneural gene products and reduced levels of Notch. This cell segregates from its neighbours and divides from a few to many times in the nervous system but only once in the Malpighian tubules.

In the tubules, the sequence of events underlying lateral inhibition appears to be repeated in that both daughter cells of the tip mother cell express proneural genes and a second competitive interaction, involving N, determines which daughter finally adopts the tip cell fate. In a similar way, the progeny of sensillum precursor cells in the PNS compete for the primary, neural fate, so that without N function multiple neurones differentiate, at the expense of bristle cells (Harten-stein and Campos- ortega, 1986). In the grasshopper CNS, the daughters of ganglion mother cells are born as an equivalence pair, in which the final cell fate is decided by cell interactions (Kuwada and Goodman, 1985). Whether the final identity of neurons results from such interactions in Drosophila and whether the neurogenic genes play a role in them is not yet known (see Goodman and Doe, 1993).

The activity of the neurogenic genes is required for the normal development of many other tissues in the embryo, including those of mesodermal (Corbin et al., 1991; Ruohola et al., 1991; Hartenstein et al., 1992; Bate et al., 1993) and endodermal origin (Hartenstein et al., 1992). In some of these instances, the neurogenic genes are involved in the selection of cells to a particular fate from a competent cluster (Corbin et al., 1991; Hartenstein et al., 1992; Bate et al., 1993).

Tip cells are selected in the tubule primordia early in stage 12 and eventually differentiate during stage 17 to express numerous markers characteristic of neural cells. Although many of these markers are not diagnostic of nerve cell differ-entiation (Jan and Jan, 1982; Zipursky et al., 1984), both the monoclonal antibody 44C11 and the antibody to Synaptotag-min have previously been reported to label only neural cells (Bier et al., 1988; Littleton et al., 1993). In addition many of the enhancer trap lines that express β-galactosidase in elements of the PNS (Hartenstein and Jan, 1992) also label the progeny of the tip mother cell (our unpublished data). The involvement of proneural and neurogenic genes in tip cell specification and the co-expression of so many neural markers, in particular several proteins associated with synapses, suggests that the tip cell may develop a neural function later in development. We have established that each posterior tip cell makes intimate contact with a branch of the nerve a9 and has a lengthy process into the lumen of the tubule (Fig. 5A-D). This topology would be consistent with its function either as a sensor of tubule activity or as a regulator of tubule function.

During the time between their selection and the onset of differentiation as neural cells, the tip cells are not quiescent. Unlike the remaining cells of the tubules, tip cells do not divide but are active in promoting the division of their neigh-bours. Evidence for this activity has been based on the con-sequences of ablating tip cells in growing tubules (Skaer, 1989). We have now confirmed this role for the tip cells by removing them genetically. In the absence of members of the AS-C, tip cells fail to segregate, as judged by their failure to appear at the distal tips of the tubules, the failure of Kr expression in a single cell in each tubule and by the absence of a cell in the tubule expressing neural markers at the end of embryogenesis. The phenotype of the Malpighian tubules in embryos lacking the function of the AS-C is a reduction in the number of cells in the mature tubules to the level in wild-type tubules at the time of tip cell selection (Skaer and Martinez Arias, 1992). This number is indistinguishable from the final number of cells in tubules of cultured embryos, in which the tip cells have been experimentally ablated at the time of their selection (Skaer, 1989). These results suggest that, from the time that they arise, the tip cells have a decisive influence over cell proliferation in the developing tubules. Whether other ‘preneural’ cells, such as neuroblasts, ganglion mother cells or sensillum precursor cells, have a similar mitogenic influence is not known. However, one class of mature nerve cells does have mitogenic activity; ingrowing photoreceptor axons stimulate proliferation of neuroblasts in the lamina anlage of the optic lobe during the development of the adult visual system in Drosophila (reviewed in Meinertzhagen, 1993).

Without the activity of the proneural genes, both the mitogenic capacity and the neuronal characteristics of the tip cell are lost. As well as failure of tubule cell proliferation in embryos lacking these genes, neural markers that in wild-type embryos highlight the tip cell, fail to stain the tubules. Fur-thermore, the posterior pair of tubules do not become attached to the hindgut or contact a9. The process of tip cell allocation therefore initiates both pathways of differentiation but, whereas the mitogenic capacity develops immediately after the tip cell is specified, differentiation of neural characteristics is delayed. Neural differentiation is preceded by the decline of ac expression in the tip cells. The persistence of ac expression in neural precursors in the nervous system is variable (Cubas at al, 1991; Ruiz Gomez and Ghysen, 1993) but the relationship between these patterns of expression and the onset of neural differentiation is not yet understood.

Cubas et al. (1991) suggest that the disappearance of the ac-sc protein from the sensillum precursor cell in the PNS is a prerequisite for its division. This contrasts with the tip mother cell, which divides while still expressing ac; final tip cell differentiation occurs after the allocation of the tip cell from the sibling pair and further refinement of ac expression to a single cell.

Our results suggest that a subset of cells in the posterior region of the gut has the capacity to differentiate a neural fate. This mirrors the neurogenic potential of cells in the stomodeum from which the stomatogastric nervous system segregates (Campos- ortega and Hartenstein, 1985; Hartenstein, 1993). Whether the activity of the proneural and neurogenic genes are involved in this process is not yet known.

Stüttem and Campos- ortega (1991) analysed the neural potential of ectodermal cells by transplanting them heterotopi-cally. They concluded that cells in the ectoderm differ in their neural potential, with cells from the stomodeal anlage showing an intermediate capacity to become neural and the cells of the proctodeal anlage showing a very low potential. This finding is consistent with our observation that only four groups of 6-8 cells express ac and sc in the whole proctodeum.

The specification and differentiation of tip cells show clear parallels with neurogenesis. They occur in a strikingly simple system, which therefore provides a model in which to analyse the cell activities involved, as well as the role of genes that underlie them. However, the tip cell also demonstrates an unusual feature: the acquisition of an important role during embryogenesis, which is followed by the differentiation of a fate distinct from any other cell in the tissue of which it is a part. The challenge that this cell offers is to understand how the programme initiated by a single process of cell allocation unfolds so that the patterns of gene expression that generate each fate in turn are regulated in a precise temporal sequence.

We thank the following for providing us with fly stocks and anti-bodies: S. Artavanis, H. Bellen, M. Brand, E. Buchner, J. Campos- ortega, Dianova, C. Goodman, A. Jarmann, C. O’Kane, J. Modollel. H. S. acknowledges with gratitude the assistance and advice of J. Campos- ortega, A. Carpenter, F. Grawe, S. Hopkins and A. Martinez Arias in generating germ-line clones, the involvement of E. Andrew and N. Mullett in the analysis of N mutants and of J. Rodford in preparing Fig. 7. M. H. would like to thank S. Elend for technical assistance. We also thank the members of our respective laboratories for their advice and encouragement and would like in particular to thank J. Campos- ortega, A. Martinez Arias, M. Ruiz-Gomez, S. Maddrell, P. Simpson and A. Wessing for helpful discussions and M. Bate, M. Baylies, A. Gampel, M. Gonzalez Gaitan, U. Tepass and N. Tublitz who also made invaluable comments on the manuscript. We also thank C. Doe and V. Hartenstein for sharing unpublished infor-mation with us. This work is supported by the Wellcome Trust (H. S., K. B.) and the Deutsche Forschungsgemeinschaft (SFB 271 to M. H.; Ja 312/6-1 to H. J.).

H. S. would like to record that C. Cabrera predicted the combined role for proneural and neurogenic genes in tip cell allocation at Boldern in 1991.