Shar-pei mediates cell proliferation arrest during imaginal disc growth in Drosophila

Cell proliferation must be tightly regulated to ensure the development of properly proportioned organs and tissues. During development, most organ primordia grow by increasing in cell number until the appropriate organ size is attained. Cells then exit from the cell cycle and differentiate into specific cell types. Defects in cell cycle exit result in excess cells and underlie tumor progression. Mechanisms that promote cell cycle exit are thus fundamental to development and understanding them should help us elucidate how tumors progress.

The Drosophila imaginal discs provide an excellent model system to study how cell proliferation is regulated during organ growth(Edgar and Lehner, 1996;Johnston and Gallant, 2002). Imaginal discs are epithelial sacs that differentiate into the external structures of head, thorax and genitalia of the adult fly(Cohen, 1993). Each disc develops from 10-30 precursor cells that proliferate extensively during the larval stages to give rise to approximately 50,000 cells in case of wing and eye discs, before differentiating into the corresponding adult structures during metamorphosis (Bryant,1978). The growth of imaginal discs to specific sizes and shapes is directed by secreted signaling molecules including Decapentaplegic (Dpp), a TGFβ homolog, Wingless (Wg) and Hedgehog (Hh), which act as morphogens to induce patterning and growth (Day and Lawrence, 2000; Lawrence and Struhl, 1996; Serrano and O'Farrell, 1997). Although these factors may still be expressed,imaginal disc cells stop proliferating when discs reach their correct size(Bryant and Levinson, 1985). In addition, transplantation experiments revealed that developing discs transplanted into adult hosts grow until they reach their normal size and shape but do not grow larger than normal size even though they are not forced to differentiate (Bryant and Levinson,1985; Garcia-Bellido,1965). Therefore, mechanisms exist that terminate cell proliferation when discs have reached their correct size.

The generation of discs with stereotypical sizes and shapes is not the result of a predetermination of the number of cell divisions of progenitor cells. This is evident because the size of clones of imaginal cells is variable and thus not determined(Postlethwait, 1978), and cells can compensate for growth defects by extra proliferation. The flexibility of cell lineages is best illustrated by the observation that mitotic clones of cells that are induced to grow faster than their neighbors are significantly larger compared with wild-type clones(Morata and Ripoll, 1975). Notably, after such manipulation of proliferation rates, the final pattern and size of the adult structures are normal. Moreover, discs can regenerate missing parts after surgical manipulation(Bryant, 1978;Bryant and Simpson, 1984) and when ∼75% of the progenitor cells of imaginal discs are killed by X-rays,the remaining cells proliferate and compensate for the loss of cells(Haynie and Bryant, 1977). Hence, cell proliferation is plastic and cells in a developing tissue adjust their proliferation depending on whether more cells are needed to build a normal sized structure (Day and Lawrence,2000; French et al.,1976; Garcia-Bellido and Garcia-Bellido, 1998). However, the molecular mechanisms that direct cells to stop proliferating once the primordium of a structure has reached the correct size are poorly understood.

In principle, defined organ size can be generated either by regulating the extent of cell proliferation or by eliminating superfluous cells through programmed cell death, or both. Only limited amounts of cell death are observed during imaginal disc growth(Milan et al., 1997;Wolff and Ready, 1991),indicating that disc size is primarily, albeit not exclusively, controlled at the level of cell division. Thus, factors must exist that regulate the decision of imaginal disc cells to re-enter or exit the cell cycle to mediate growth control.

The Drosophila eye is particularly well suited to identify factors that regulate cell proliferation. First, the various stages of cell division and differentiation can be accurately followed in eye imaginal discs. Second,defects in growth control and differentiation can be easily scored. In the early growth phase of the eye disc, cell cycles are not synchronized and proliferating cells are evenly distributed throughout the disc(Baker, 2001). Later, during the third larval stage, a wave of differentiation called the morphogenetic furrow sweeps across the eye disc from posterior to anterior(Wolff and Ready, 1993). Cells anterior to the furrow are developmentally uncommitted and divide asynchronously, whereas cells within the furrow arrest in the G1 phase of the cell cycle, synchronize and either start to differentiate into photoreceptor cells as they leave the furrow or undergo one additional round of cell division, referred to as the second mitotic wave before differentiating into the remaining photoreceptor, cone, pigment and bristle cells(Baker, 2001;Dickson and Hafen, 1993;Wolff and Ready, 1993). Thus,different modes of cell proliferation control can be studied with single cell resolution.

To gain insights into the mechanisms that regulate cell proliferation during organogenesis, we conducted a genetic screen in Drosophila to identify mutations that affect adult eye size. We describe the identification and phenotypic characterization of a novel gene, shar-pei(shrp), that is required for cells to terminate proliferation once imaginal discs have reached their correct size. Flies with shrpmutant tissues have enlarged structures that contain more cells of normal size. These overgrowths result from an extended period of cell proliferation,accompanied by a decrease in cell death. Based on these observations, we propose that shrp regulates organ growth by promoting both cell cycle exit and apoptosis. As shrp is not required for terminal cell cycle exit, we conclude that control of proliferation arrest during organ growth is separable from, and probably precedes, cell cycle exit during terminal differentiation. The presence of shrp homologs in other species suggests that the mechanisms that control tissue size are evolutionarily conserved.

shrp mutants were isolated from an eyFLP-mediated EMS mutagenesis screen using a strategy described earlier(Newsome et al., 2000). Adult heads were screened for phenotypes that displayed defects only in size/growth but retained wild-type pattern elements on the head cuticle. A total of sixshrp alleles were recovered, balanced and tested for intra-allelic complementation. To generate mutant clones, FRT82B shrp/TM6B virgin females, carrying either y w eyFLP or y w hsFLP on the X chromosome, were crossed to either y w; FRT82B P[w⁺]ubi-GFP^nls/TM6B(Datar et al., 2000) or y w; FRT82B cell lethal P[w⁺]/TM6B(Newsome et al., 2000). Stocks used for meiotic mapping and male recombination were: y w;P[w⁺]how^E7-3-4(Lo and Frasch, 1997), y w; P[w⁺] crb^j1B5/TM3, Sb(Spradling et al., 1999),y¹; ry⁵⁰⁶ P[SUP] KG05850, y¹;ry⁵⁰⁶ P[SUP] KG02176, y¹; ry⁵⁰⁶P[SUP]CG444^KG03827(http://flypush.imgen.bcm.tmc.edu/pscreen)and ry⁵⁰⁶ P[ry⁺]klg^rN712/TM3,Sb (Spradling et al.,1999). Other stocks used are Df(3R)hh(Mohler and Vani, 1992),Df(3R)M95A (Reuter et al.,1986), Elav-GAL4 UAS-mCD8-GFP(Lee and Luo, 1999),GMR-GAL4 (Freeman,1996) and UAS-p35(Hay et al., 1994).

Flies carrying the shrp¹ allele were crossed to flies carrying one of the five P elements in the 94A-96A region (see fly stocks above). Transheterozygous virgin females (y w/y w; shrp ¹ /P[w⁺]) were collected and crossed to males of the genotype y w eyFLP/Y; st shrp⁴ ca/TM3 P[w⁺] Sb. Recombination betweenshrp¹ and a given P element would produce wild-type chromosomes that lack shrp¹ and the P element. Half of these chromosomes were recovered over the st shrp⁴ cachromosome and the other half over TM3 P[w⁺]. The reciprocal chromosomes carried both shrp¹ and the P element. Recombinant flies of the genotype y w / y w; + / st shrp⁴ ca were the only progeny that did not carry aP[w⁺] and were identified by their white eyes among the rest of the red-eyed progeny. The frequency of white-eyed progeny is thus equal to half the meiotic distance in cM.

Adult flies with heads in which over 90% of cells were mutant, were processed for SEM by using the hexamethyldisilazane (HMDS) method(Braet et al., 1997) with modifications. Flies were fixed for a day in 70% acetone, and washed twice in 100% acetone for 4 hours each. Acetone was then exchanged with HMDS through two washes in 1:1 acetone:HMDS and two washes in 100% HMDS over 2 days. Samples were air dried for 1 day prior to sputter coating with 25 nm platinum alloy. Antibody staining of imaginal discs carried out as described earlier(Halder et al., 1998). The following antibodies were used (dilutions in parenthesis): rat α-Elav(1:30) (O'Neill et al., 1994),rabbit α-Dlg (1:2000) (Cho et al.,2000), rabbit α-Drice (1:2000)(Yoo et al., 2002), guinea-pigα-Dlg (1:2000) (Woods and Bryant,1991), guinea-pig α-Sens (1:1000)(Nolo et al., 2000), mouseα-BrdU (1:50, Becton-Dickinson) and mouse α-CycE (1:50)(Richardson et al., 1995). Secondary antibodies were donkey Fab fragments from Jackson Immuno Research. BrdU incorporation was carried out as described(de Nooij et al., 1996) by incorporating BrdU for 1 hour. For in situ hybridization, DrosophilacDNA clone RE52745 (ResGen Invitrogen Corp.) was used as template to generate DIG-labeled RNA probes (Roche), and in situ hybridization was performed as described (Nolo et al.,2000).

To analyze cell cycle and cell size distribution of shrp mutant and wild-type cell populations, wing imaginal discs containing mutant clones were dissected from transheterozygous larvae of the genotype y w hsFLP;FRT82B shrp/FRT82B P(w⁺) ubi-GFP^nls. Clones were induced 24-48 hours after egg laying (AEL) by administering a heat-shock at 34°C for 30 minutes and discs were dissected 72 hours later. About 80-100 wing discs were dissected in PBS and transferred to 5 ml polystyrene tubes containing Trypsin-EDTA (Sigma, T-4174): PBS 9:1 v/v and 0.5 μg/ml Hoechst 33342 (Neufeld et al., 1998). Cells were dissociated for 4 hours by gentle shaking. The dissociated cells were analyzed on a Becton Dickinson Vantage Fluorescence activated cell sorter(FACS) and more than 50,000 mutant cells were scored for each sample. Data were analyzed with the Cell Quest program. For cell counts, wing discs fromy w hsFLP; FRT82B shrp / FRT82B P(w⁺)ubi-GFP^nls transheterozygous larvae were dissected 48 hours after clone induction, fixed in PEM (100 mM PIPES, 2 mM MgSO₄, 1 mM EGTA, 4% formaldehyde) on ice for 1 hour, washed briefly in PBT (PBS + 0.3%Triton X-100) and incubated with Hoechst (0.02 μg/ml) for 25 minutes. The discs were washed twice in PBT and mounted in Vectashield (Vector Labs). Cell numbers were determined by counting the nuclei (Hoechst-stained) of cells in mutant clones (GFP negative) and associated wild-type twin clone(GFP-positive) on a Zeiss axioplan fluorescence microscope.

To identify novel components of growth control pathways, we performed a genetic screen in adult Drosophila to isolate mutants in which tissue size but not tissue patterning is affected. Because genes involved in growth control may have ubiquitous functions, we anticipated that animals homozygous for mutations in these genes might die during embryogenesis. We therefore screened randomly mutagenized chromosome arms that were made homozygous only in the head using an eyeless enhancer driven Flipase transgene(eyFLP) (Newsome et al.,2000). Mutations in several genes were isolated that resulted in enlarged heads but did not affect patterning. These include mutations in theDrosophila homologs of PTEN and TSC1/2 tumor suppressor genes, which act in cell growth control pathways that affect cell number as well as cell size (Potter and Xu, 2001; Stocker and Hafen,2000; Tapon et al.,2001), mutations in warts/lats, a previously described tumor suppressor gene encoding a Ser/Thr kinase(Justice et al., 1995;Xu et al., 1995) that affect cell number but not cell size and mutations in a previously undescribed gene. We named this gene `shar-pei' because of its folded cuticle phenotype in the head, which resembles the folded skin of Shar-pei dogs.

Mutations in shrp were isolated on chromosome arm 3R using chemical mutagenesis. Complementation tests showed that six mutations(shrp^1-6) that caused a head overgrowth phenotype fail to complement each other. All mutations showed a very similar phenotype and caused early larval lethality. Given the nature of the molecular lesions(see below) it is likely that they are either null alleles or very severe loss-of-function alleles. All experiments involving cell clones were performed with at least three independent alleles.

The heads of flies in which over 90% of cells are homozygous mutant forshrp¹ are proportionally larger than other structures but have a normal overall pattern, including bristles, ocelli and ommatidia (Fig. 1A-F). All mutant fly heads have folded head cuticle and eye tissue(Fig. 1C-F) and over 15% of flies are severely affected (Fig. 1D). Smaller clones generated by heat-shock induced Flipase expression do not exhibit this folding phenotype. Folding may therefore be a secondary consequence of limited space within the pupal case, which does not allow overgrown tissue to fully expand. In addition to producing structures that are too big, shrp mutant cells appear to out-compete wild-type cells. The flies shown in Fig. 1A,B are eyFLP induced genetic mosaics in which clones are genotypically marked white, heterozygous portions are wild type and twin clones were eliminated by a cell lethal mutation to increase the amount of mutant cells in the eye (Newsome et al., 2000). When white cells proliferate at a normal rate, heterozygous red cells contribute about 20% of cells in this type of experiment. In contrast, eyes with white markedshrp¹ mutant clones are predominantly white(Fig. 1B) and contain fewer wild-type ommatidia when compared with a control fly shown inFig. 1A. Hence, the phenotype suggests that shrp mutant cells proliferate more rapidly than wild-type cells (Kirby and Bryant,1982; Moberg et al.,2001; Simpson and Morata,1981).

To test whether shrp affects cell proliferation in tissues other than the head, we induced random clones by heat-shock induced Flipase expression (Xu and Rubin,1993). Such mutant clones resulted in overgrowths on thorax,wings, halteres and legs (Fig. 1G-J, not shown). As observed for the eye and head, these structures differentiated the correct tissue-specific cell types. We conclude that shrp is generally required to restrict the size of imaginal disc-derived adult structures, whereas tissue-specific cell-type specification and differentiation remain unaffected in shrp mutant cells.

To define the developmental basis for the enlarged tissue phenotypes, we focused on patterning and cell proliferation in the developing eye because the eye has a precise pattern of cell types and highly regulated cell proliferation (Baker, 2001;Kumar and Moses, 2000;Wolff and Ready, 1993). We first analyzed the pattern of differentiated photoreceptor cells in adultshrp mutant clones in 1 μm sections(Fig. 2A). We observed eight photoreceptors per ommatidium with a normal trapezoidal arrangement,indicating that this aspect of pattern formation is not affected. However,spacing between individual photoreceptor clusters was significantly increased in shrp⁵ clones when compared with wild-type areas (Fig. 2A arrowhead). To test whether the increased space was due to an excess of interommatidial cells, we stained wild-type and mutant midpupal retinas with an antibody against Discs-large (Dlg), a protein that localizes to apical junctions and hence reveals cell outlines (Fig. 2B,C) (Woods and Bryant,1991). We found that shrp⁴ mutant clones exhibit a dramatic increase in the numbers of interommatidial cells(Fig. 2B) when compared with wild type (Fig. 2C). These extra interommatidial cells differentiated into pigment cells that produced normal pigmentation when clones were induced in a w⁺background (not shown). These data indicate that Shrp regulates cell number but not differentiation in the retina.

The extra interommatidial cells could be due to excess cell proliferation,increased spacing of photoreceptor clusters during patterning, lack of apoptosis or a combination thereof. In wild-type flies, interommatidial cells are initially produced in excess but the extra cells are later eliminated by apoptosis during pupal development in a process that requires cell-cell signaling (Rusconi et al.,2000). This system generates a very precise retinal lattice. To determine whether shrp mutant cells initiate the apoptotic program,we stained shrp mosaic pupal retinas with an antibody that detects the activated form of Drice, a caspase that triggers the apoptotic program and specifically marks cells undergoing apoptosis(Yoo et al., 2002). We detected many Drice-positive cells in wild-type retinal tissue, but none were found in shrp³ mutant territories(Fig. 2D,E). Importantly, all Drice-positive cells were wild type. This suggests that the apoptotic pathway is blocked in shrp mutant cells and that this block occurs upstream of Drice activation. We conclude that shrp mutant cells do not receive or are resistant to signals that induce apoptosis.

To test directly whether lack of apoptosis is sufficient to produce theshrp mutant phenotype, we compared the phenotype of shrpmutant retinas with that of wild-type retinas in which apoptosis was blocked by expressing the apoptosis inhibitor p35(Hay et al., 1994). Ectopic expression of p35 eliminates most, if not all, normally occurring cell death in the retina (Hay et al.,1994) and results in extra interommatidial cells(Fig. 2F). However, the number of additional cells is significantly less than that observed inshrp⁴ mutant clones.(Fig. 2B). Therefore, while lack of apoptosis allows additional cells to survive, it is not sufficient to explain the amount of extra interommatidial cells generated in shrpmutants.

To investigate whether the extra interommatidial cells are due to abnormal ommatidial spacing during patterning, we stained developing mosaic eye imaginal discs for the neuronal marker Elav(Robinow and White, 1988) and the R8 marker Senseless (Sens) (Frankfort et al., 2001; Nolo et al.,2000). Elav is expressed in all differentiating photoreceptor cells and outlines differentiating photoreceptor clusters, while Sens is a marker for early pattern formation and ommatidial spacing, as well as R8 photoreceptors. Mutant ommatidial clusters have normal numbers of differentiating photoreceptor cells per ommatidium and are initially spaced correctly (Fig. 2G-I). However,at later stages in more posterior clones, spacing between photoreceptor clusters is increased (Fig. 2J-L). Thus, early retinal pattern formation is normal inshrp mutants.

To test directly whether shrp affects cell proliferation, we monitored the distribution of cell cycle progression in mutant third larval eye discs by bromodeoxyuridine (BrdU) incorporation(Fig. 3). In wild-type discs,BrdU-incorporating cells are randomly distributed in front of the morphogenetic furrow (Fig. 3A). In the furrow, cells synchronously arrest in G1 and do not incorporate BrdU(Fig. 3A, arrow). Posterior to the furrow, cells go through a synchronous S phase in the second mitotic wave,revealed as a band of cells incorporating BrdU(Fig. 3A, arrowhead). Few BrdU-positive cells are found posterior to the second mitotic wave(Fig. 3A).shrp¹ mutant cells also synchronize their cell cycles in the furrow and progress normally through the second mitotic wave(Fig. 3B). However, in contrast to wild-type cells, shrp¹ mutant cells still display BrdU incorporation after the second mitotic wave(Fig. 3B, asterisk). The extra DNA synthesis is followed by cell division, as revealed by ectopic expression of phosphorylated histone H3 (PH3), which marks chromosomes during mitosis(not shown). This phenotype of shrp is cell autonomous, because only mutant cells undergo extra rounds of cell proliferation(Fig. 3C,D), and all territories of mutant cells show the excess interommatidial cell phenotype in pupal retinas, whereas non-mutant tissue appears wild type. Extra cell proliferation continues into the pupal stage but ceases by 24 hours after pupariation (not shown). Double labeling with BrdU and antibodies against Elav to detect differentiating photoreceptor cells revealed that only Elav-negative cells incorporated BrdU (Fig. 3E,F). Therefore, shrp is required to arrest cell proliferation in developmentally uncommitted cells after the second mitotic wave, but is not required for cell cycle arrest of differentiating photoreceptor cells. The ectopic proliferation produces extra interommatidial cells, which together with the lack of apoptosis, are sufficient to explain the overgrowth phenotypes observed in pupal and adult retinas.

Cyclin E is limiting for S-phase initiation and progression during imaginal disc development and several tumor suppressor genes negatively regulate its activity or levels (de Nooij et al.,1996; Duman-Scheel et al.,2002; Lane et al.,1996; Moberg et al.,2001; Neufeld et al.,1998; Richardson et al.,1995). Cyclin E levels are upregulated inshrp¹ mutant cells in the second mitotic wave and posterior to it (Fig. 4A-Carrows). Elevated levels were also observed just anterior to the second mitotic wave, although this effect was not as pronounced. The effect on Cyclin E is cell autonomous and observed in most or all mutant cells, even though only a fraction of them are actively progressing through S phase(Fig. 3C,D). Thus, the effect of Shrp on cell proliferation arrest may be mediated by regulating the levels of Cyclin E.

Our data show that although shrp mutant cells are able to exit the cell cycle during cell differentiation, they are delayed in arresting cell proliferation at the end of eye imaginal disc growth. To determine whethershrp has a function in uncommitted cells anterior to the morphogenetic furrow, we wanted to measure whether mutant eye discs were already larger than wild-type before ommatidial clusters are specified. Because initial spacing of photoreceptor clusters is normal in shrpmutant eye discs (Fig. 2G-I),the final number of ommatidia provides a measure of the number of cells present in mutant eye discs before R8 cells are specified in the morphogenetic furrow. We thus determined and compared the numbers of ommatidia in wild-type and mutant retinas (n=18 each) by counting clusters of photoreceptor cells revealed by Elav-Gal4 driven GFP expression(Lee and Luo, 1999). Mutant retinas contained an average of 913 ommatidial clusters (s.d.=40), whereas wild-type retinas had an average of 776 (s.d.=45) photoreceptor clusters(Fig. 5). The two groups are significantly different by t-test (P<0.001). We conclude that shrp mutant eye discs are already larger than normal at the time when the positions of ommatidia are specified in the morphogenetic furrow. Shrp thus functions in uncommitted cells anterior to the morphogenetic furrow.

To test whether shrp affects the rate of cell proliferation during the growth phase of imaginal discs, we compared cell numbers in mutant clones and their associated twin clones in third instar wing discs(Fig. 6A). To reduce variability in the proliferation rate of wild-type twin clones, we used isogenized FRT 82B ubi-GFP chromosomes to generate mitotic clones. Cell numbers in shrp³ mutant clones were almost always larger than their twin clones, and the difference in cell numbers was significant when assessed using a t-test (P<0.001). The same experiment with a second allele, shrp⁴, showed similar differences (not shown). By contrast, cell numbers in clones of the isogenized wild-type chromosome on which the shrp mutations were induced during the mutagenesis screen were similar and not significantly different from the corresponding ubi-GFP/ubi-GFP twin clones(Fig. 6B). Based on these cell counts and assuming exponential proliferation, the cell division rate ofshrp mutant cells is 1.10 times faster than that of wild-type cells. Our data thus indicate that shrp mutant cells proliferated more. This phenotype is also manifest in mosaic adult eyes, whereshrp¹ mutant cells out-compete wild-type cells(Fig. 1A,B). Determination of the distribution of cell cycle phases in third instar wing discs using FACS analysis (Neufeld et al.,1998) showed that the population of shrp⁴mutant cells has the same distribution of cell cycle phases as the wild-type cells (Fig. 6C). Thus,shrp mutant cells do not accelerate a particular step in the cell cycle. Rather, mutant cells show an even acceleration of cell cycle progression.

Manipulating the activity of cell growth regulators such as components of the insulin receptor signaling pathway results in larger organs because of more and larger cells (Johnston and Gallant, 2002; Potter and Xu,2001; Prober and Edgar,2001; Stocker and Hafen,2000; Tapon et al.,2001; Weinkove and Leevers,2000). To determine whether shrp also affects cell size,we stained mosaic wing discs with antibodies against Dlg to detect apical cell outlines. Cells in shrp³ mutant clones showed normal cell sizes, as judged by cell outlines (Fig. 6E-G) and had normal height as judged by the thickness of the wing disc epithelium in the mutant clones (not shown). In addition, rhabdomeres of mutant photoreceptor cells were of normal diameter(Fig. 2A), and shrpmutant cone and pigment cells are of normal size at the pupal stage(Fig. 2B,C). Furthermore,forward light scatter (FSC) data, a measurement of cell size collected by FACS analysis confirmed that mutant cells have normal size(Fig. 6D). Therefore, Shrp does not regulate cell size. Rather, extra proliferation of shrp mutant cells is induced by stimulation of cell growth and cell cycle progression,resulting in balanced growth and extra cells that are of normal size.

To identify the shrp gene, we first mapped the lethality associated with the mutations through deficiency mapping. We found that theshrp mutations failed to complement the deficienciesDf(3R)hh (93F;094D) and Df(3R)M95A (94D;095A), placingshrp within the 94D interval. Male recombination mapping further mapped shrp to a 150 kb interval between klingon(Butler et al., 1997) andhedgehog (Lee et al.,1992). Meiotic recombination mapping with several P elements as dominant markers in the region(http://flypush.imgen.bcm.tmc.edu/pscreen)(Fig. 7A), further placedshrp within a 90 kb interval (Fig. 7B). Based on the annotation of the Drosophila genome(Adams et al., 2000), we amplified and sequenced six predicted and conserved open reading frames in that region (Fig. 7B) and found mutations in CG13831 in all shrp alleles(Fig. 7C,D). The full-lengthshrp cDNA encodes a protein of 607 amino acids with two WW domains(Fig. 7C,D). WW domains are protein-protein interaction domains that bind to short proline-rich motifs,functionally resembling SH3 domains(Macias et al., 2002).

Database searches revealed that parts of Shrp are conserved in humans, mice(WW45) (Valverde, 2000), andC. elegans (T10H10.3) (Fig. 7C,D). In addition to the highly conserved WW domains, these proteins share a C-terminal domain that is specific for Shrp and not found in other proteins (Fig. 7C,D). Phylogenic analysis by neighbor joining of 202 WW domains(http://www.Bork.EMBL-Heidelberg.DE/Modules/ww/)revealed that the two WW domains of Drosophila Shrp and vertebrate WW45 are more closely related to each other than to other WW domains. Furthermore, only single copies of this gene were detected in all four species. We conclude that Shrp, WW45 and T10H10.3 are orthologous.

The alleles shrp^1-5 have point mutations that result in STOP codons, which truncate the proteins N-terminal to the WW domains(Fig. 7C,D). The sixth allele(shrp⁶) has a 20 bp deletion that results in a frameshift between the WW domains and the Shrp-specific domain, resulting in the addition of 76 unique residues thereby effectively removing the Shrp specific domain. All six alleles are purely recessive and homozygous lethal at the first/second instar stage, and show the same lethal phase when heterozygous overDf(3R)hh. In addition, all alleles show similar phenotypes in mitotic clones in discs and adults. Thus, they all appear to be null alleles forshrp function. Because the frameshift alleleshrp⁶ is recessive and behaves as a null allele, the position of its lesion suggests that the conserved C-terminal domain is essential for Shrp function. In situ hybridization of imaginal discs revealed that shrp is ubiquitously expressed in the eye, wing and leg discs(Fig. 7E,F). This is consistent with our findings that shrp is required in all imaginal disc-derived tissues for proper cell proliferation arrest.

Adult flies with shrp mutant cell clones have enlarged structures,which contain more cells of normal size. This phenotype results becauseshrp mutant cells proliferate faster than wild-type cells and because they do not terminate proliferation when imaginal tissues have reached their normal size. For example, shrp mutant cells in the eye disc undergo extra rounds of cell proliferation after the second mitotic wave, when wild-type cells do not normally proliferate anymore, resulting in extra interommatidial cells. In addition, Shrp effects cell survival. In wild-type,extra interommatidial cells are removed by apoptosis, while extrashrp mutant cells are not. Thus, Shrp has a dual function in promoting cell proliferation arrest and apoptosis. The effects of Shrp on apoptosis and cell proliferation may be separate functions, because inhibition of apoptosis and stimulation of cell cycle progression are not necessarily linked (Asano et al., 1996;Du et al., 1996b;Moberg et al., 2001;Richardson et al., 1995).

Pattern formation appears to progress normally in shrp mutant clones. In the eye, clones show the normal complement and morphology of photoreceptor and cone cells. Clones on head, thorax, halteres and wings show normal patterning of bristles and other tissue-specific structures such as ocelli, wing margin bristles, haltere specific sense organs and tissue-specific cell type differentiation. In addition, mutant cell clones in wing discs respect both the AP and DV compartment boundaries and clone borders within compartments are jagged, indicating that cell affinities are not affected by the loss of shrp function. In summary, Shrp specifically regulates cell number by promoting cell proliferation arrest and apoptosis,but is not required for pattern formation or cell type differentiation.

Shrp function is specifically required for the timely proliferation arrest of developmentally uncommitted cells, but not for terminal cell cycle exit during cell differentiation. In shrp mutant eye discs, cell cycle arrest in differentiating photoreceptor cells still occurs normally, and ectopic BrdU incorporation is confined to developmentally uncommitted cells. Consistent with this observation, we did not observe duplicated photoreceptors or cone cells in the eye or duplicated bristles on the head thorax or along the wing margin. Thus, shrp is not required for cells to exit the cell cycle during terminal differentiation. Rather, shrp is required for proliferation arrest before cells are induced to differentiate into specific cell types. Therefore, more precursor cells are generated inshrp mutants, which then differentiate normally but produce adult organs that are too big. For example, shrp mutant eyes have more ommatidia, indicating that mutant eye discs contain more cells than normal when ommatidial cell clusters are specified. The observation thatshrp mutant clones produce overgrowths on disc-derived structures,including eyes, heads, wings, halteres, legs and thoraxes suggests thatshrp function is ubiquitously required. The ubiquitous expression pattern of shrp in imaginal discs supports this conclusion. In the eye, we also observe an increase in the number of pigment cells per ommatidium, while the other cell types are present in normal numbers per ommatidium. This is because cell types are specified sequentially during eye development. First, some cells are specified as photoreceptors, while the other cells remain uncommitted. These uncommitted cells ectopically proliferate in shrp mutants and produce more interommatidial cells. Successively, the remaining cell types are recruited in normal numbers, but at the end, too many cells are leftover which differentiate into excessive numbers of pigment cells, the last cell type to differentiate.

The requirements of Shrp are distinct from those of genes required for cell cycle exit during terminal differentiation. Dacapo (dap), aDrosophila homolog of the Kip family of cyclin-dependent kinase inhibitors, is induced when cells exit the cell cycle prior to terminal differentiation (de Nooij et al.,1996; Lane et al.,1996). In dap mutant embryos, cells go through one extra round of cell division just prior to terminal differentiation. In the developing eye, dap expression is induced in differentiating photoreceptor cells, but not in developmentally uncommitted cells(de Nooij et al., 1996;Lane et al., 1996). Adultdap mutant eye clones do not show gross abnormalities or extra cells,but rare escapers show duplications of bristles on notum and wing margin(Lane et al., 1996). Downregulation of positive cell cycle regulators such as Cyclin E(Crack et al., 2002;de Nooij et al., 1996;Du and Dyson, 1999;Knoblich et al., 1994;Richardson et al., 1995;Richardson et al., 1993) and other negative regulators of cell proliferation such as Rbf(de Nooij et al., 1996;Du and Dyson, 1999;Du et al., 1996a) may act redundantly with Dap to promote cell cycle arrest. Nonetheless, while Dap is upregulated specifically in cells that withdraw from the cell cycle prior to terminal cell differentiation, Shrp is required in developmentally uncommitted cells during the growth phase of organs before terminal differentiation. The requirements for Dap and Shrp functions are thus spatially and temporally distinct.

We propose that the arrest of cell proliferation during imaginal disc development is controlled by several genetically separable mechanisms. First,cells stop proliferating when imaginal discs have reached their correct sizes(Bryant and Levinson, 1985;Garcia-Bellido, 1965). This process requires Shrp function. Second, cells permanently exit the cell cycle during terminal differentiation. Because terminal cell cycle exit is part of cell differentiation, it is directly regulated by patterning mechanisms that determine the identity and position of each individual cell. This regulation is governed by tissue-and cell-type specific enhancers of cell cycle regulators such as dacapo, cyclin E and string, which all have complex cis- regulatory regions(Duman-Scheel et al., 2002;Jones et al., 2000;Lehman et al., 1999;Liu et al., 2002;Meyer et al., 2002). Similarly, patterned regulation of cell cycle progression may occur before terminal differentiation, as is observed in the second mitotic wave in the eye and in the zone of non-proliferation along the presumptive wing margin in the wing disc (O'Brochta and Bryant,1985). None of these processes are affected in shrpmutants. Thus, the direct control of cell cycle progression by patterning mechanisms acts epistatically to the control of cell proliferation observed during organ growth and can impose cell cycle arrest on cells that otherwise may continue to proliferate. Therefore, shrp mutations do not deregulate cell proliferation of terminally arrested cells, and cells differentiate normally. In summary, Shrp identifies a molecular mechanism that is required to arrest cell proliferation during organ growth and that appears to be distinct from the ones used to arrest cells during terminal cell differentiation.

What are the downstream effectors of Shrp that are deregulated inshrp mutants and induce cell proliferation? shrp mutant clones behind the second mitotic wave in eye discs show elevated levels of Cyclin E. Notably, Cyclin E was elevated in all developmentally uncommitted cells of the clones, apparently irrespective of the phase of the cell cycle. The effect on Cyclin E levels may thus be direct and not just a reflection of the ectopic cell proliferation observed in mutant clones. Precise regulation of Cyclin E expression and activity is crucial as ectopic expression of Cyclin E induces entry into S phase and limited cell proliferation in imaginal discs and embryos (Knoblich et al.,1994; Neufeld et al.,1998; Richardson et al.,1995). Several other negative regulators of cell proliferation directly regulate the levels of Cyclin E activity. Dap directly inhibits Cyclin E/Cdk2 complexes (de Nooij et al.,1996; Lane et al.,1996), and Archipelago is required for degradation of Cyclin E(Moberg et al., 2001). The regulation of Cyclin E is thus likely to be an important downstream effect of Shrp function.

Ectopic expression of Cyclin E alone, however, is not sufficient to generate the number of extra cells observed in shrp mutant tissues(Neufeld et al., 1998;Richardson et al., 1995). Artificial acceleration of the cell cycle by ectopic expression of specific cell cycle regulators such as E2F accelerates cell division, but does not stimulate cell growth rates, and cells divide when they are smaller(Neufeld et al., 1998). This results in an increase in cell number and a concomitant decrease in cell size,yet does not affect the overall tissue size. Thus, cell cycle progression is not sufficient to drive cell and organ growth. Conversely, stimulating cell growth alone produces larger organs, but also affects cell size(Johnston and Gallant, 2002;Potter and Xu, 2001;Prober and Edgar, 2001;Stocker and Hafen, 2000;Tapon et al., 2001;Weinkove and Leevers, 2000). For example, artificially stimulating the activities of Ras, Myc or insulin receptor signaling produces more and bigger cells and thus bigger but otherwise normal flies. Thus, cell proliferation during organ growth requires coordinate stimulation of cell cycle progression and cell growth to produce normal sized cells. Because shrp mutant cells maintain normal size,Shrp appears to be required to regulate cell growth and cell cycle progression coordinately. Thus, Shrp probably regulates other targets driving cell cycle and cell growth in addition to Cyclin E.

Several other mutations have been described that fail to arrest imaginal disc growth and were thus classified as tumor suppressor genes(Gateff, 1994;Turenchalk et al., 1999;Watson et al., 1994). These include discs-large (dlg)(Woods and Bryant, 1991),lethal giant larva (lgl)(Gateff, 1978) andscribble (scrib) (Bilder and Perrimon, 2000), encoding proteins which form an architectural complex localized to septate junctions. Mutations in these genes disrupt septate junctions and apical-basal polarity of epithelial cells and result in neoplastic overgrowth of imaginal discs(Bilder et al., 2000;De Lorenzo et al., 1999;Johnston and Gallant, 2002). Mutations in a second group of Drosophila tumor suppressor genes cause hyperplastic overgrowth of imaginal discs that retain their single layered epithelial structure. These include warts/lats, which encode a kinase that regulates the activity of Cdc2(Justice et al., 1995;Tao et al., 1999;Xu et al., 1995);fat, a large Cadherin (Mahoney et al., 1991); hyperplastic discs, a E3 ubiquitin ligase(Mansfield et al., 1994); anddiscs overgrown, a Drosophila homolog of Casein kinase Iδ/ϵ (Zilian et al.,1999). The imaginal disc overgrowth in mutants of both groups occurs during an extended larval period, and embryonic requirements for these genes appear to be provided by maternal contributions(Bilder et al., 2000;Bryant et al., 1988;Mansfield et al., 1994). By contrast, homozygous shrp mutant animals die as first/second instar larvae, which do not show gross morphological defects. Thus, zygotic expression of shrp is required for early larval viability, whereas that of other tumor suppressor genes is not. Cells homozygous mutant for neoplastic or hyperplastic tumor suppressor genes generally differentiate abnormally and show defects in cell morphology and/or pattern formation(Agrawal et al., 1995;Bilder and Perrimon, 2000;Bryant et al., 1988;Justice et al., 1995;Martin et al., 1977;Woods and Bryant, 1991;Xu et al., 1995;Zilian et al., 1999). These phenotypes are thus different from those of shrp mutant cell clones,which overproliferate but differentiate with normal cell morphology and patterning. In addition to these differences, clones of cells homozygous mutant for shrp proliferate more rapidly and have reduced apoptosis,while cells mutant for most other tumor suppressor genes have reduced viability and a decreased rate of cell proliferation(Bryant, 1987;Mansfield et al., 1994;Woods and Bryant, 1989). Onlyfat and warts/lats mutant cell clones proliferate faster(Garoia et al., 2000) (M. K. S., unpublished), similar to shrp mutant cells. However, the phenotypes of shrp, fat and warts/lats differ, asfat and warts/lats affect the morphology and pattern of adult tissues in addition to cell number. Therefore, shrp affects cell number more specifically than these other mutants, and future work will have to establish whether and how Shrp interacts with other tumor suppressor gene products to control tissue size.

In summary, our studies provide evidence that Shrp functions in the decision of imaginal disc cells to terminate proliferation and to exit the cell cycle once the correct disc size is attained. The determination of the effectors of Shrp action should reveal mechanisms by which cell growth and cell cycle progression are coordinately regulated during organ growth and how cells arrest proliferation once organs have reached their correct size. The presence of Shrp homologs in mouse and human suggest the existence of a conserved organ size control mechanism in mammals. The characterization of Shrp function should therefore provide valuable insights into the mechanisms that underlie tissue size regulation and cause disproportionate growth and tumorigenesis when defective.

We thank Konrad Basler, Andreas Bergmann, Michael Brodsky, Barry Dickson,Bruce Edgar, Margaret Fuller, Yasushi Hiromi, Graeme Mardon, Jym Mohler, Hanh Nguyen and the Bloomington Stock Center for Drosophila stocks; and Bruce Hay, Peter Bryant, Kwang-Wook Choi, Helena Richardson, Terry Orr-Weaver and the Developmental Studies Hybridoma Bank for antibodies. We also thank Richard Behringer, Wei Du, Benjamin Frankfort, Randy Johnson, Jeannette Kunz and Tom Neufeld for helpful discussions and comments on the manuscript. We thank Kenn Dunner and Wendy Schober for technical help with the SEM and FACS analysis, which, with DNA sequencing, were carried out at M. D. Anderson core facilities supported by NCI cancer center support grant CA16672. We also thank Leisa McCord for help with artwork, and Ji Yuan Ni, Zhi Hong Chen and Deepa Murali for technical assistance. P. R. H. is an EMBO fellow. This work was supported by HHMI funding to H. J. B. and a March of Dimes Basil O'Connor Starter Research Award to G. H.