Cell proliferation in the developing renal tubules of Drosophila is strikingly patterned, occurring in two phases to generate a consistent number of tubule cells. The later phase of cell division is promoted by EGF receptor signalling from a specialised subset of tubule cells, the tip cells, which express the protease Rhomboid and are thus able to secrete the EGF ligand, Spitz. We show that the response to EGF signalling, and in consequence cell division, is patterned by the specification of a second cell type in the tubules. These cells are primed to respond to EGF signalling by the transcription of two pathway effectors, PointedP2, which is phosphorylated on pathway activation, and Seven up. While expression of pointedP2 is induced by Wingless signalling, seven up is initiated in a subset of the PointedP2 cells through the activity of the proneural genes. We demonstrate that both signalling and responsive cells are set aside in each tubule primordium from a proneural gene-expressing cluster of cells, in a two-step process. First, a proneural cluster develops within the domain of Wingless-activated, pointedP2-expressing cells to initiate the co-expression of seven up. Second, lateral inhibition, mediated by the neurogenic genes, acts within this cluster of cells to segregate the tip cell precursor, in which proneural gene expression strengthens to initiate rhomboid expression. As a consequence, when the precursor cell divides, both daughters secrete Spitz and become signalling cells. Establishing domains of cells competent to transduce the EGF signal and divide ensures a rapid and reliable response to mitogenic signalling in the tubules and also imposes a limit on the extent of cell division, thus preventing tubule hyperplasia.
While our understanding of cell cycle regulation at the molecular level is detailed and sophisticated, our knowledge of the way in which cell division is patterned in developing tissues is relatively rudimentary. We have used a simple epithelial tissue, Drosophila renal tubules, to understand how cell-specific patterns of gene expression combine with intercellular signalling to regulate the spatial organisation of cell proliferation during embryogenesis.
The majority of embryonic cells in Drosophila go through a tightly regulated programme of cell divisions (Campos-Ortega and Hartenstein, 1997). After cellularisation of the early syncytial embryo at the end of cycle 13, most cells divide a limited number of times, in many cases no more than three. These cells then enter a variable number of endoreplicative cycles, allowing the larva to grow through increases in cell size (Foe et al., 1993). These postblastodermal divisions lack G1 but are strikingly patterned in both space and time through the regulation of G2 by the expression of the Drosophila cdc25 homologue, string (Edgar and O’Farrell, 1990; Edgar et al., 1994). A few tissues in the embryo undergo further divisions, with the introduction of a regulated G1 phase (Richardson et al., 1993; Knoblich et al., 1994). These include cells set aside to construct the adult organism, the nervous system and the Malpighian tubules (Kerber et al., 1998). The four tubules differ in size, the pair that project anteriorly are longer than the posterior pair (see Fig. 3A). Clonal analysis has revealed that tubule cells undergo three to five postblastodermal divisions, which are consistently patterned and result in an average of 107 cells in each posterior and 144 cells in each anterior tubule (Janning et al., 1986; Skaer, 1989). In summary, tubule cell divisions occur in two phases; ‘early’ divisions, lacking G1, that are common to all embryonic cells (known as cycles 14-16) and ‘late’ divisions, complete with G1, that they share only with adult precursors and cells in the developing nervous system (cycle 17 onwards).
In this paper, we analyse how these late cell divisions in the tubules are regulated and patterned. It is already known that cell division in the tubules requires the activity of two signalling pathways, the Wingless (Wg) and the EGF receptor pathways, and is regulated by a specific cell, the tip cell (TC), which is situated at the distal tip of each growing tubule (Skaer, 1989; Skaer and Martinez-Arias, 1992; Baumann and Skaer, 1993). Ablation of the TC, either mechanically or genetically, results in approximately half the normal number of tubule cells (Skaer, 1989; Hoch et al., 1994). Kerber et al. (Kerber et al., 1998) have shown that the TC stimulates division in its neighbours by secreting the EGF ligand, Spitz (Spi), thereby activating the transcription factors Pointed and Seven up, and leading to the expression of string and the G1 regulator, cyclin E.
Clearly, the normal segregation of TCs, which are capable of secreting Spi, underpins this mitogenic signalling. A single TC develops in each tubule primordium from a cluster of cells that expresses proneural genes of the achaete-scute complex (AS-C). All the cells in this proneural cluster (PNC) have the potential to form a TC, but lateral inhibition, which is mediated by the neurogenic genes, represses expression of AS-C genes in all but a single cell in the PNC, the tip mother cell (Hoch et al., 1994). The tip mother cell divides, partitioning Numb asymmetrically between its daughters and the cell inheriting Numb becomes the TC (Wan et al., 2000). Expression of AS-C genes intensifies in the TC, it develops a striking morphology (see Fig. 2D) and later differentiates many characteristics of neuronal cells. The other daughter, the sibling cell (SC), is inhibited from taking the primary fate by further signalling through Notch (N), losing the expression of proneural genes and retaining a normal epithelial morphology (Hoch et al., 1994).
We have previously shown that the number of TCs in the Malpighian tubule can be manipulated genetically; in embryos that lack proneural gene function the tip mother cell is not specified, so tubules develop without the TC or SC. By contrast, neurogenic mutants segregate multiple TCs (Hoch et al., 1994). In the absence of Numb, tubules develop with two SCs but no TC, while the overexpression of numb results in the differentiation of two TCs at the expense of the SC (Wan et al., 2000). As might be predicted, the loss of TCs in embryos without the AS-C results in tubules with approximately half the wild-type number of cells. However other manipulations produce results that are not as easy to explain: alterations of Numb, whether loss or gain of function, have no effect on tubule cell number, suggesting that specification of TC versus SC fate is not important for cell division in the tubules (Wan et al., 2000). In addition we show here that, despite the presence of eight to 12 TCs in N mutant embryos, there is no increase in cell proliferation as would be expected from an increase in the number of signalling cells in these tubules. Instead we find a dramatic reduction in division so that N mutant tubules have as few cells as those that lack tip cells altogether.
We show that both SCs and TCs are sources of the ligand Spitz and that activation of the EGF pathway is required only for ‘late’ divisions (from cycle 17 onwards) in the tubules. We demonstrate that N acts through lateral inhibition to segregate two cell fates within the PNC; the primary TC/SC fate and a second fate in the remaining cells of the PNC; those that are competent to respond to the mitogenic signal. We show that cells are primed to respond to EGF signalling by the expression of two EGF pathway effectors, seven up (svp) and pointedP2 (pntP2), even before signalling through this pathway starts in the tubules. We have analysed the factors involved in regulating the expression of svp and pntP2 and show how they combine to pattern cell division in the tubules by delineating not only the source of the EGF signal but also the population of cells that is competent to respond to it.
MATERIALS AND METHODS
Oregon R flies were used as wild type, unless otherwise specified. The following stocks were used: numb796, N55e11, Df(1)svr and ac3 sc10–1 for loss or severe reduction in proneural gene function; topCO for loss of the EGF receptor; topCO; Df(3L)H99 to protect against cell death (White et al., 1994) in an otherwise EGF receptor mutant background; Df(1)svr; hs-Nintra to express activated Notch in a proneural mutant background; UAS-Nintra to express activated Notch; hs-wg, svp07842 as svp-lacZ; pnt07825 as pntP2-lacZ; wgCX4, argosIΔ7, UAS-argos, UAS-λtop, A37 (a p-lacZ insert in neuromusculin); and UAS-sSpi to overexpress the active form of Spitz. For constitutive activation of the EGF receptor (Clifford and Schüpbach, 1994) in an otherwise proneural or neurogenic mutant background, we drove UAS-λtop in Df(1)svr or N55e11 mutants. Mutant embryos were recognised by selecting against the blue balancer. Df(3L)H99 cell death-deficient embryos were identified by their characteristic head involution defects (White et al., 1994).
To elicit Gal4-mediated misexpression (Brand and Perrimon, 1993), the following drivers were used: da-Gal4 (ubiquitous); ase-Gal4 (TMC and its progeny, from S. Stein); fkhXB30-Gal4 (Malpighian tubules, from M. Hoch); and CtB-Gal4 (Malpighian tubules (V. S. and S. P.-S., unpublished)).
Antibody staining of embryos was performed by following standard techniques (Hoch et al., 1994). The following antibodies were used: mouse anti-Achaete (1:10 from P. Simpson); mouse anti-BrdU (1:50, DAKO); mouse monoclonal anti-β-Gal (1:1000, Promega); rabbit polyclonal anti-β-Gal (1:10000, Sigma); mouse anti-Cut (1:100, Developmental Studies Hybridoma Bank); rabbit anti-Krüppel (1:500, from C. Rushlow); and rabbit anti-Rhomboid (1:500, from E. Bier). Primary antibody incubations were performed overnight followed by incubation with appropriate biotinylated secondaries and amplification using the Vector Elite ABC Kit (Vector Laboratories, CA). For fluorescent double labelling, we used appropriate secondary antibodies conjugated with Cy2, Cy3 or Cy5. When required, we performed an additional amplification step using streptavidin-conjugated Cy3. All fluorophores were obtained from Jackson Immunochemicals.
In situ hybridisation was performed according to Jowett (Jowett, 1997). cDNA for svp was obtained from M. Mlodzik and the antisense probe was made by digesting with KpnI followed by synthesis with T3 RNA polymerase. A pntP2 specific fragment was obtained from P. zur Lage. The antisense probe was made by digesting with BamHI followed by synthesis with T7 RNA polymerase. Stained preparations were viewed using a Zeiss Axioplan compound microscope and images were captured using a JVC-KY55B digital camera and a Neotech Image Grabber, UK. Fluorescent preparations were viewed using Leica TCS confocal microscope, and figures were prepared using Adobe Photoshop and Canvas.
BrdU incorporation and cell counts
BrdU incorporation was performed either according to Shermoen (Shermoen, 2000) in whole-mount embryos or in dissected preparations according to Skaer (Skaer, 1989). For Malpighian tubule cell counts, embryos were either dissected as above and stained with anti-Cut antibody or tubules were dissected out of stained embryos in 70% glycerol.
The EGF receptor is required for the late cell divisions in the tubules
To assess when EGF pathway signalling is required for tubule cell division, we compared BrdU incorporation in wild-type and EGFR mutant embryos. In wild type, the first cycle after cellularisation, cycle 14, is common to the shared hindgut/tubule anlage. After M14, the tubule cells are specified and behave as a separate population, progressing synchronously through cycle 15 (Ainsworth et al., 2000) (A. G., unpublished). Cycle 16 is restricted to a synchronous group on the posterior side of each tubule (Fig. 1A). After M16 there are 70 (±4) cells in each tubule (Skaer, 1989). With the introduction of G1 in cycle 17, synchrony is lost, the incorporation of BrdU is more sporadic and is limited to distal regions of the tubules (Fig. 1B). In embryos mutant for the EGF receptor (topCO), the early divisions (cycles 14-16) occur normally (Fig. 1C) but there is no incorporation of BrdU from cycle 17 onwards (Fig. 1D). We, therefore, conclude that the EGF receptor is required for cell cycle progression from G1 of cycle 17. Thus, it is the late cell divisions that are driven by TC release of EGF, and these divisions increase the average tubule cell number from 70 (±4) to 124 (±3) cells.
The EGF receptor is also required for tubule cell survival
As the number of tubule cells in the wild type is 70 at the beginning of cycle 17, it is surprising that the final cell number in Egfr or spi mutant embryos is much lower (23-27 cells; Table 1), despite normal progression through cycles 14-16 (Baumann and Skaer, 1993; Kerber et al., 1998) (Table 1). We were able to show that this discrepancy results from cell death: first, because the number of cells in each tubule declines after cycle 16 (from 39±2 to 26±2; Table 1) and, second, because the cell number is 74±1 in topCO embryos that are additionally mutant at the hid locus, in which cell death is blocked (Fig. 1E,F; Table 1) (White et al., 1994). The difference in cell number between topCO embryos at the end of cycle 16 (39±2) and topco/hid double mutants (74±1) indicates, first, that cells start to die in topCO embryos before the tip cell arises (Hoch et al., 1994) and, second, confirms that the EGF receptor is required for cell division only from cycle 17. Because mature topCO embryos have fewer tubule cells than results simply from the loss of the TCs, and this additional loss can be rescued by blocking cell death, we deduce that EGF receptor signalling has an additional function for cell survival, during the earlier phase of cell divisions, before cycle 17.
Notch but not Numb is required for the normal tubule cell number
We have previously shown that embryos lacking Numb have tubules with no tip but two sibling cells. By contrast, the overexpression of numb produces two tip but no sibling cells. Strikingly, neither of these changes in cell fate alters the number of tubule cells (Wan et al., 2000) (Table 1). This suggests that SCs as well as TCs secrete the mitogenic signal Spitz. We therefore analysed the expression of Rho, as a measure of the ability of a cell to process Spi (Golembo et al., 1996; Wasserman and Freeman, 1997; Bang and Kintner, 2000; Urban, 2001). Rho is first detected in the tip mother cell (Fig. 2A,C). After the tip mother cell divides, we detect Rho in the TC and SC (Fig. 2B,D). In situations where two SCs are specified, Rho is detected in both cells (Fig. 2E,F). Thus, although the tip mother cell division is clearly asymmetric in terms of the segregation of Numb and cell fate, Rho is subsequently maintained in both daughters.
As the tip mother cell is selected by refining proneural gene expression to a single cell in the PNC (Hoch et al., 1994), we tested whether Rho expression requires proneural gene function. In embryos that lack the proneural achaete-scute complex (AS-C–/–), Rho expression is lost in the Malpighian tubules (Fig. 2G). Conversely, in neurogenic N mutants, where all cells in the cluster become TCs and express achaete (Hoch et al., 1994), Rho is detected in all the transformed cells (Fig. 2H). We conclude that rho expression is initiated downstream of proneural gene function in the tip mother cell, that rho expression persists when the tip mother cell divides and that this is sufficient to activate secretion of the mitogen, irrespective of the daughter cell fates.
In neurogenic mutants such as N, where multiple Spi-secreting tip cells are specified, we expected to find a greater mitogenic response. Surprisingly, we found that in zygotic N mutants, the tubules fail to grow (compare Fig. 3A with 3B). In fact, there are far less than the wild-type number of cells in each tubule (89±4; Table 1). In N germline clone embryos, the final number of cells is the same as in the tubules of AS-C–/– embryos with no TC/SCs (76±6 compared with 72±3; Table 1). The incorporation of BrdU in N mutant embryos is normal until S16 but from cycle 17 is greatly reduced or entirely absent (compare Fig. 3C with 3D), indicating that the reduction in tubule cell number results from failure of the late, TC/SC-driven cell divisions.
Failure of the late cell divisions in N mutant embryos could be explained in several ways:
(1) The activation of N drives cell proliferation in a cell autonomous fashion.
(2) N is required for the processing of the EGFR ligand.
(3) Cells that are ‘competent’ to respond to EGF receptor activation are not specified in N mutant embryos.
We tested the first possibility by overexpressing the activated form of N in the tubules of proneural gene mutants. Expression of Nintra in AS-C–/– embryos, throughout the period of late cell divisions (from 5 hours to 10.5 hours), does not rescue tubule growth (compare Fig. 3E with 3F) or cell division (Table 1). This result shows that N activation is not sufficient to drive proliferation in the absence of signalling cells, and suggests a more indirect role for N in wild type tubules.
We have tested the second hypothesis by driving the expression of λtop, an activated form of the EGF receptor (Clifford and Schüpbach, 1994), in the tubule cells of N mutant embryos. Forced expression of λtop fails to rescue the late (EGF receptor-driven) cell divisions (Fig. 3G, Table 1), eliminating the possibility that N exerts its function solely by affecting Spi secretion or binding. This observation also confirms that without N function the late, EGFR-driven divisions fail, suggesting that N plays a role in ensuring the response to EGF receptor activation.
In AS-C–/– embryos, expression of λtop also fails to rescue the late cell divisions (Fig. 3H, Table 1). This suggests that in addition to specifying the mitogen-secreting cells, proneural gene function is required for the mitogenic response, even when the EGF receptor is activated. These data are consistent with the third hypothesis: that the proneural and neurogenic genes function not only to select the cells that will secrete Spi, but also to specify the cells that will respond to this signal.
The proneural and neurogenic genes specify a second cell fate
To understand how the proneural and neurogenic genes pattern the response to EGFR activation, we analysed the expression of genes involved in transduction of the pathway. It has previously been shown that the orphan nuclear-receptor svp functions downstream of the EGF receptor to promote cell divisions in the tubules (Kerber et al., 1998). In the absence of Svp function, cycE and stg transcription is abolished, with a consequent reduction in EGFR-driven cell divisions (Kerber et al., 1998). We followed these late divisions in the tubules of stage 12 wild-type embryos and found that BrdU incorporation and, hence, cell division is confined within the svp-lacZ domain (Fig. 4A). These results define the svp domain of expression as including those cells which will divide in response to EGFR activation. However, we noticed that the expression of svp-lacZ was initiated in a group of cells surrounding the tip mother cell, before the birth of the TC (Fig. 4B,C) (Kerber et al., 1998). This early onset of svp expression occurs before the late divisions start (cycle 17 onwards), when neither Svp function nor EGFR activation is required for cell proliferation. The pattern of gene expression we observe suggests that the Svp-positive cells surrounding the tip mother cell, derive from the proneural cluster.
To test this hypothesis, we observed the expression of svp-lacZ in embryos lacking proneural gene function. Indeed, in AS-C–/– embryos, the expression of svp is not initiated in the tubules (Fig. 4D). Conversely, in N mutants, where all cells in the cluster adopt the primary (TC) fate, the expression of svp is confined to the transformed cells (Fig. 4E). After the initiation of Spi signalling from the TC/SC, svp expression depends on EGF receptor activation (Kerber et al., 1998). However that the early expression of svp is not dependent on EGFR function is shown in topCO mutants, where svp-lacZ expression is still initiated normally (Fig. 4F), but is not maintained. In AS-C–/– embryos expressing λtop in the tubules, svp expression was not detected (data not shown). Together these data suggest that the initiation of svp depends on the proneural genes but is independent of EGF receptor signalling, which acts only from cycle 17 to maintain svp expression.
These results show that the expression of proneural genes in the tubules not only confers tip cell potential but also initiates the expression of an effector of the EGF pathway, svp. We suggest that this primes cells to divide in response to EGF receptor activation. Proneural genes are therefore required to specify two cell fates in the tubule PNCs; the tip mother cell and cells competent to respond to EGFR activation.
pntP2 expression is initiated in MT primordia by Wg
Svp is not the only effector of the EGF pathway. The ETS domain protein PointedP2 (PntP2) functions downstream of EGFR/Ras signalling (Brunner et al., 1994; O’Neill et al., 1994; Wassarman et al., 1995). This protein contains a single MAPK phosphorylation site (Brunner et al., 1994; O’Neill et al., 1994) and upon phosphorylation, competes with the ETS domain transcriptional repressor, Yan, to activate the expression of target genes (O’Neill et al., 1994; Rebay and Rubin, 1995). In the absence of pnt function, cell proliferation in the tubules is reduced in a manner similar to svp mutants (Table 1) (Kerber et al., 1998).
We therefore asked whether early expression of pnt as well as svp is required to prime the mitogenic response in tubule cells. pntP2 is initiated in the posterior side of each tubule during stage 10 (Fig. 5B). This domain is characterised by high levels of wg expression, which are required for the normal development of AS-C expression in the PNC, as it develops within this domain (Wan et al., 2000). The domain of wg and pntP2 expression is slightly wider than the PNC (Fig. 5A) and pntP2 expression is initiated well before EGFR activity is required for tubule cell divisions. The expression of pntP2 persists in this posterior domain when the tip mother cell is specified (Fig. 5C). In wgCX4 mutant embryos, tubule expression of pntP2 is completely abolished (Fig. 5D), showing that Wg signalling is required to initiate its expression. Conversely, the overexpression of wg, using a hs-wg construct, results in expansion of pntP2 expression to the anterior side of the tubule primordium and elevation of expression to high levels (Fig. 5E). Thus, Wg is necessary and sufficient to activate the expression of pntP2 in the tubules.
EGF receptor signalling regulates tubule cell number by two routes
The importance of EGFR signalling from the TC for tubule cell division is already known (Baumann and Skaer, 1993; Kerber et al., 1998). We show that the pathway is activated to promote cell division by the production of the ligand Spi from both the tip and sibling cell, from their birth early in stage 12. EGFR activation stimulates passage through the G1-S-G2-M cycles in the tubules. Accordingly, EGF signalling is required for the late cell divisions in the tubules, from G1 of cycle 17.
We show that in addition to promoting cell division, the EGF receptor is required for tubule cell survival. In AS-C–/– tubules that lack the tip and sibling cells, the number of cells never increases above 70-75 (Hoch et al., 1994). However in embryos mutant for the EGF receptor, for spi or doubly mutant for rho and Star (S) (and therefore lacking active ligand), tubules contain fewer than 70 cells [Egfr, 26; spi, 23; rho and S, 24 (Kerber et al., 1998); data for anterior tubules only]. We were able to restore EGF receptor mutant tubules to 74 cells by inactivating cell death, confirming a dual requirement for EGF signalling in both the division and survival of tubule cells. This experiment also reveals that the loss of EGFR-driven cell divisions results in the same final tubule cell number as the loss of the TC and SC in AS-C–/– embryos (72 cells), confirming that the source of the mitogenic signal is solely the TC/SCs. However, loss of active Spitz, whether by removal of the gene or failure to process inactive ligand, results in tubule cell death that is just as extensive as that caused by receptor loss. These results show that the ligand for cell survival is Spi. However the source of Spi for cell survival is not the TC/SCs, as in AS-C–/– tubules, division arrests after cycle 16 (70 cells) (Skaer, 1989) and the final cell number in the mutants is 72, indicating that tubule cells do not die in the absence of TC/SCs.
Our results correlate with the observation of Clifford and Schüpbach (Clifford and Schüpbach, 1992) that cell death plays a major role in the manifestation of the top (Egfr) final embryonic phenotype. However, Kerber et al. (Kerber et al., 1998) suggest that the EGF receptor is not required for cell survival in the MTs. Their analysis is based on TUNEL assays and Acridine Orange staining, which may not have revealed the rather gradual cell death we recorded.
Patterning cell divisions in the Malpighian tubules
Previous studies have shown that cell division in the tubules is regulated by Wg and EGF signalling, and that the segregation of tip and sibling cells by lateral inhibition within a proneural cluster of cells ensures restricted secretion of Spi (Skaer and Martinez-Arias, 1992; Hoch et al., 1994; Kerber et al., 1998; Wan et al., 2000). Deregulation of EGF signalling, whether by the overproduction of signalling cells [loss of function in neurogenic genes or extra macrochaetae (emc) (Ellis, 1994), overexpression of AS-C (Hoch and Jäckle, 1998)], or by hyperactivation of the pathway [removal of argos, overexpression of sspi (Kerber et al., 1998)], does not produce a great increase in tubule cell number but instead reveals remarkable stability in the regulation of cell division (Table 1). These data suggest the presence of powerful feedback restraints. The establishment of a restricted domain of cells, competent to respond to the mitogenic EGF signal and close to its source, identifies such a restraint.
In neurogenic mutants, cells in the PNC all adopt the mitogen-secreting tip cell fate at the expense of cells competent to divide. Therefore, in spite of excess signal, cell divisions fail beyond cycle 17. Strikingly, neither AS-C–/– nor neurogenic mutant tubules are rescued by the expression of an activated EGF receptor protein, confirming that the patterning of tubule cell fate within the PNC is a prerequisite for mitogenic activation in the tubules. Thus, while the primary function of the competence domain is to facilitate cell cycles in an embryo in which the majority of cells have exited the division cycle, its existence also sets a spatial limit on the ‘late’ tubule cell divisions.
These observations raise the question of whether the competence domain or a gradient of Spi activity normally limits cell division in the tubules. Tight regulation of the EGF pathway has been demonstrated in other tissues, both through the graded distribution of ligand and the activity of negative feedback loops mediated by Argos and Sprouty (Schweitzer et al., 1995; Hacohen et al., 1998; Casci et al., 1999; Kramer et al., 1999; Reich et al., 1999). We do not know whether Sprouty acts in the Malpighian tubules, but the graded concentration of sSpi and the spatial restriction of ligand activity by Argos could act together to impose some limit on cell division in the distal region of the tubule (Kerber et al., 1998).
However, several arguments suggest that this is not the case, but that it is the size of the competence domain that limits the extent of cell division. Firstly, clonal analyis by Janning et al. (Janning et al., 1986) indicates that tubule cells divide up to five times (cycles 14-18). Division of cells in the competence domain (12-14 cells) generates 45 in the posterior tubules and 70 in the anterior. Thus, every cell in the domain would have to divide during cycles 17 and 18 in the posterior tubules and at least one further cycle, involving the division of a subset of cells, would be required in the anterior tubules (Skaer and Martinez-Arias, 1992). This argument suggests that all competent cells do divide. Second, over activation of the EGFR pathway (by expressing sSpi, λ-top, activated Ras or in argos mutants) produces no increase in cell proliferation in the tubules (Table 1). Again, this suggests that it is the competence domain, rather than the extent of EGF signalling, that patterns late division in the tubules. Third, the expression of svp throughout the tubules from stage 10 does result in an increase in tubule cell number (Kerber et al., 1998) (Table 1). Ectopic Svp at stage10/11 would result in an expansion of the number of cells expressing both svp and pntP2 and therefore of the competence domain. Extra cell division would then result from cells stimulated by EGFR activation that, in a normal embryo would fail to respond.
Thus, in wild-type embryos the competence domain imposes a limit on tubule cell division and, should mitogenic signalling become deregulated, the extent of the competence domain imposes an effective brake on tubule hyperplasia.
Cell cycles in the Malpighian tubules are pre-patterned by developmental cues
We show here that several layers of pre-pattern are imposed on cell cycles in the tubules by early developmental circuitry (Fig. 6). Wg activates the expression of pntP2 in a subset of tubule cells (encompassing the PNC), thus imposing a first restriction on cells competent to respond to the EGF pathway. Second, the expression of the proneural genes induces expression of svp in the PNC. Third, cells are selected from the PNC, start to express rho and are then able to secrete the mitogen Spi. The remaining PNC cells, a subset of pntP2-expressing cells that also express svp, are the only cells primed to respond to the mitogen.
The initiation of svp and pnt2 expression is independent of EGF receptor activation; in a topCO mutant, both pntP2 and svp are expressed in a subset of tubule cells from stage 11 (V. S., unpublished and Fig. 4F, respectively). pntP2 expression is dependent on Wg. svp expression is promoted by moderate levels of proneural gene expression, while rho expression depends on high levels of proneural gene activity. We propose that lateral inhibition represses the expression of proneural genes, thus ensuring that svp but not rho is activated in the competence domain. We also show that Wg, by activating pntP2 expression, acts together with N and the proneural genes to prime cells close to the source of secreted Spi to be responsive to it.
While the initiation of rho expression in the TC and SC depends on high levels of proneural gene expression, its maintenance must be independent of the activity of these genes, at least in the SC, which loses proneural gene expression soon after it is born (Hoch et al., 1994). In other systems, activation of the EGF receptor from a primary source has been shown to stimulate the secretion of the EGF ligand from neighbouring cells (Golembo et al., 1996; Golembo et al., 1999; Wasserman and Freeman, 1998). A similar positive-feedback loop could maintain signal amplification in the tubules with Spi from the TC activating rho expression in the neighbouring SC. However, as both the TC and SC are born expressing rho and tubule cell division is unaffected by the specification of two TCs (UAS-numb) or SCs (numb mutants), it seems more likely that signalling between the TC and SC is reciprocal and acts to maintain rho expression in both cells, thereby ensuring continued high levels of secreted signal.. Indeed, results obtained from TC ablation (Skaer, 1989) support this model. The TC was removed mechanically and resulted in tubules with only 75 cells (Table 1), suggesting complete loss of secreted Spi so that all the late cell divisions failed.
We do not suggest that the expression of ac and the establishment of the PNC are alone sufficient for cell division. Indeed in other situations both the SOP and PNC are mitotically quiescent while they express ac (Usui and Kimura, 1992; Cubas et al., 1991), and in the developing wing, Ac acts to repress the transcription of string, thus ensuring that cells do not divide (Johnston and Edgar, 1998). Rather, we suggest that initial, moderate expression and subsequent repression of proneural gene expression is a precondition for cell division in the tubules, but only in the presence of mitogenic signal.
A common strategy for patterning cell fate?
The selection of founder cells, which organise tissue development by signalling to recruit their neighbours has been demonstrated in Drosophila by mosaic analysis of developing leg bristles (Tobler, 1966) and has since been shown to occur in the developing eye (Cagan and Ready, 1989; White and Jarman, 2000), PNS (Hartenstein and Posakony, 1989; Reddy et al., 1997; zur Lage et al., 1997) and muscles (Bour et al., 2000; Ruiz-Gomez et al., 2002). In each case, the signalling properties are established by the expression of specific proneural genes. In the eye (Tio et al., 1994; Freeman, 1996) and chordotonal organs (zur Lage and Jarman, 1999), though not necessarily in other systems, signalling is through the EGFR pathway, resulting from the initiation of rho expression in the founder cells. We have shown, in the Malpighian tubules, that responsive cells, as well as the progenitors of the signalling cells, segregate from the PNC. This pattern has also been demonstrated in adult chordotonal organs (zur Lage and Jarman, 1999) and in the recruitment of multiple cell fates in the developing eye (Cagan and Ready, 1989; Baker and Zitron, 1995; Flores et al., 2000; White and Jarman, 2000; Xu et al., 2000). Evidence from the pattern of recruitment, as well as from changes in cell fate in mutants, indicates the existence of a similar mechanism in the development of the embryonic chordotonal organs (zur Lage et al., 1997), somatic muscles (Bour et al., 2000; Ruiz-Gomez et al., 2002) and adult olfactory sensilla (Reddy et al., 1997).
These results suggest that, while the segregation of single cells from an equivalence domain is a unifying theme in the generation of tissues from a wide range of organisms (Sternberg, 1993; Simpson, 1997), PNCs in specific tissues have developed an additional function: to establish a second cell fate that cooperates with the first to implement the subsequent programme of tissue differentiation.
We thank the following for kindly providing us with flies and reagents: Alfonso Martinez-Arias, Ethan Bier, Michael Hoch, Andrew Jarman, Petra zur Lage, Chris Rushlow, Trudi Schüpbach, Pat Simpson, Stefan Stein, Ursula Weber and the Developmental Studies Hybridoma Bank. We are grateful to Michael Bate, Barry Denholm, Peter Lawrence and Pat Simpson for helpful discussions and comments on the manuscript, and to Phil Ingham for advice. V. S. thanks Sasikanth Acharya for initiating an interest in Drosophila development. This work was funded by The Wellcome Trust (S. P.-S., A. G. and H. S.), the MRC (S. W.) and an INLAKS Foundation scholarship to V. S.