Tumor suppressor genes encoding proteins required for cell interactions and signal transduction in Drosophila

Cell proliferation is controlled by cell-autonomous mechanisms involving cyclins and cell division control genes, but il is also affected by factors originating outside the cell including both diffusible signals and contact-mediated signals from adjacent cells. Disruption of proliferation control in cancer cells is often associated with loss of cell polarity, adhesion and junctions, suggesting that the maintenance of these features is important for the operation of the signalling mechanisms controlling proliferation and differentiation.

An important source of information regarding these aspects of cell biology is provided by the identification and analysis of oncogenes and tumor suppressor genes, and their protein products. At least nine human tumor suppressor genes have now been identified by analyzing the changes in somatic cell DNA that occur during tumor development (see Bryant. 1993 for review). Structural analysis of these genes indicates (hat their protein products are involved in cell adhesion, signal transduction, and cell cycle control, although none of their products had been previously recognized in studies of normal cells. By revealing mechanisms and molecules that have not been identified in studies of normal cells, the study of tumor suppressor genes is contributing not only to our understanding of oncogenesis, but also to basic cell biology.

Recently it has become clear that Drosophila can make an important contribution in this area, because simple traditional genetic procedures can be used to identify mutations that cause overgrowth of specific tissues and thus to identify tumor suppressor genes (Gateff. 1978; Gateff. 1982; Bryant. 1993). Furthermore, molecular genetic techniques available lor Drosophila facilitate the identification and characterization of the gene products. Sequence comparison can then be used to identify human homologs as candidate tumor suppressor genes.

Mutations in most of the known Drosophila tumor suppressor genes are recessive lethals that show either hyperplastic or neoplastic overgrowth of specific tissues, followed by death of the animal during the late larval or early pupal stage. The tumor suppressor genes functioning only in the gonads arc exceptional in that the tumors form in the adult and cause sterility but not lethality. Over 50 tumor suppressor genes have been identified; they function in the embryo, in the developing gonads, in the developing brain, in developing imaginai discs (embryonic tissues that develop inside the larva and differentiate into parts of the adult fly during metamorphosis), and in the developing hematopoietic system. Fight of these genes have been cloned and characterized, and three of them show clear homology to human genes, none of which had been recognized previously as a tumor suppressor gene (Bryant. 1993). In this paper we describe recent findings on these three genes.

The aberrant immune responses (air8) mutation was recovered in a P-elcment mutagenesis screen that led to the identification of over 20 genes in which mutations lead to melanotic tumor formation (Watson et al., 1991). The mutations arc called aberrant immune response (or air) since genetic alterations rather than foreign challenges lead to the activation of cellular defense mechanisms and the formation of melanotic tumors (Fig. 1). The air mutations fall into two classes; the phenotype of the first class (including air8) includes overgrowth of the hematopoietic organs (lymph glands) indicating that the corresponding genes function in cell proliferation control in that tissue. The second class of air mutations has no observable effect on the lymph glands, but the mutant animals develop melanotic tumors associated with specific tissues. Mutations in this class arc thought to result in “autoimmune responses” against the affected tissues (Watson et al., 1991).

The air8 mutation inhibits growth of most larval organs but causes overgrowth of the lymph glands (Fig. 2: Watson et al., 1992). II also causes overproduction of blood cells from the lymph gland (K.L.W., unpublished data) and the precocious transformation of plasmatocytcs to lamellocytes during larval life, an event that normally occurs during the pupal stage (Fig. 3; Shrestha and Gateff. 1982; Rizki and Rizki. 1984). The lamellocytes aggregate to form large melanotic tumors, and death occurs during the late larval stage (Figs 1. 3; Watson et al., 1991; Watson et al., 1992). Transplantation of air8 hemocytes into the abdomens of wild-type females leads to bloating, melanotic tumor formation and significantly more host lethality than wildtype hemocyte injections (K.L.W., unpublished data). Also, explained air8 hemocytes show’ indefinite proliferation in vitro (K.I.,W. and I). Peel, unpublished data). In contrast to its effect on the hematopoietic system, the air8 mutation appears to block cell growth in other tissues. Mitotic recombination clones of air8 mutant cells fail to survive both in the germline and developing imaginai disc cells, indicating a requirement of the normal gene product for cell survival and/or proliferation in these tissues (Watson et al., 1992).

The gene identified by the air8 mutation was the first tumor suppressor functioning in the Drosophila hematopoietic system to be characterized at the molecular level. Sequencing of air8 cDNAs from an adult female library revealed an open reading frame encoding 248 amino acids with 75% identity and 95% similarity to both human and rat ribosomal protein S6 (Fig. 4A: Watson et al., 1992). Two different P-element insertions from different melanotic tumor strains (air8 and WGI288) map to the 5’ transcribed but untranslated region of the Drosophila RpS6 gene. On northern blots, the I kb RpS6 transcript is barely detectable in air8 mutant larvae yet this transcript is abundant in wild-type animals throughout development and in air8 revenant animals that were produced by excision of the P-element (Watson et al., 1992).

Although mammalian S6 has not been recognized as a tumor suppressor, there are several indications that it may play a role in cell proliferation control and tumorigenesis. The human rpS6 gene has been mapped to 9p21, a region in which genetic alterations arc associated with various cancers including acute lymphoblastic leukemia and acute myelogenous leukemia (Antoine and Fried, 1992). In addition, S6 is overexpressed in a variety of tumors such as human colon carcinoma and adenomatous polyps (Pogue-Geile et al., 1991; Barnard et al., 1992). Most intriguingly, the S6 protein contains a cluster of serine residues in the carboxy terminus that shows developmentally regulated and mitogen-induced phosphorylation (Traugh and Pendergast. 1986; Sturgill and Wu. 1991); it is not yet known whether the Drosophila protein is similarly modified. In quiescent mammalian cells, most ribosomes exist as inactive monosomes and S6 is not phosphorylated (Traugh and Pendergast. 1986). However, shortly after cells arc stimulated to proliferate by treatment with scrum growth factors, insulin, tumor promoling agents, or transforming viruses, protein synthesis increases as a result of higher rates of translation initiation (Traugh and Pendergast. 1986: Ballou et al., 1988). S6 becomes multiply phosphorylated in an orderly manner aller mitogen stimulation and during development, and ribosomal subunits containing phosphorylated S6 become preferentially incorporated into polysomes (Traugh and Pendergast. 1986: Heinze et al., 1988). It has been suggested that phosphorylation of specific residues in S6 may control the state of cells by regulating the selective translation of mRNAs (Mutoh et al., 1992). The S6 protein resides in the mRNA binding domain of the 40S subunit where it might be expected that any subtle conformational and/or charge change induced by S6 phosphorylation could have a significant effect on ribosome function (Ballou et al., 1988: Traugh and Pendergast. 1986).

Two distinct families of S6 kinases, pp90^rsk and pp70^S6k, have been shown in vivo and in vitro to phosphorylate S6 directly via two distinct intracellular signalling pathways (Erikson. 1991). In one of these pathways. pp90^rsk and S6 arc downstream targets in a serine/threonine phosphorylation cascade that has been identified in a variety of systems (Lange-Carteret al., 1993). Recently. pp70^S6k function was shown to be necessary for cell cycle progression through GI and its principal target. S6, is phosphorylated during this transition (Lane et al., 1993). These results provide additional support for the hypothesis that S6 phosphorylation may regulate cell proliferation.

The growth inhibition seen in most organs of air8 animals, and the failure of mutant clones to survive in the germ line and imaginai discs, is consistent with a defect in protein synthesis. However, it is surprising to find that tumor formation in the hematopoietic system appears to occur in the absence of the RpS6 transcript and presumably S6 protein. One possible explanation is that a different protein is able to substitute for S6 in the ribosomes of hematopoietic cells, and this hypothesis is supported by the finding that the genomic DNA immediately Hanking the 3′ end of the RpS6 coding region contains two tandem repeals (3RI and 3R2) of a sequence closely related to RpS6 intron 2, exon 3 and 3′ flanking DNA (Fig. 4B). The existence of these repeats can be explained by two unequal crossover events, leading first to a duplication of part of RpS6. followed by duplication of the copy (K.L.W., unpubl.). The intron 2/exon 3 boundaries, the open reading frame and the putative polyadenylation signals are retained in 3RI and 3R2, suggesting that the repeals arc transcribed. Conceptual translation of the repeals predicts peptides showing 62.3% amino acid identity plus an additional 25.4% similarity with S6 (Fig. 4C). However, there are interesting differences between the repeats and S6 including deletions and residue changes that predict S6 isoforms with altered functions. In particular, there are residue alterations in the putative nuclear localization signals and loss of three of the C-terminal serine residues that may be phosphorylated in S6 (Fig. 4C). Preliminary results suggest that the RpS6 repeats arc transcribed (K.L.W.. unpublished data) and if these transcripts are translated, the alternative isoforms might be capable of assembling into ribosomes. Tissue-specific isoforms of ribosomal proteins have not been reported, but if they exist they could explain the tissue-specific phenotype caused by mutations in the RpS6 gene.

Recessive lethal mutations in the dlg gene cause neoplastic overgrowth of the imaginai discs in the larva, followed by death of the animal in the early pupal stage (Woods and Bryant. 1989, 1991). The overgrowing imaginai (.lises, which are normally single-layered epithelia, become disorganized masses and lose the ability to develop into adult parts even after transplantation into normal hosts. Strong alleles disrupt epithelial structure over the entire dise whereas with weaker alleles only part of the tissue loses epithelial structure and overgrows (Figs 5, 6A). When the entire disc is affected, it can grow to at least three times the normal cell number during the extended larval period (Woods and Bryant, 1989). Normal imaginai disc cells arc highly polarized epithelial cells connected by junctional complexes including adherens and septate junctions, but in the imaginai discs of larvae carrying a dlg loss-of-function mutation, apical-basal polarity is almost completely lost and the septate junctions arc missing (D. F. W. and P. J. B., unpublished data). The brain and lymph gland also overgrow in the mutant and the larvae become bloated, but the remainder of the larval organs appear normal.

The dlg gene encodes a 960-amino acid protein (DlgA) that is expressed in epithelial cells, as well as other tissues including the hematopoietic organs and the nervous system, throughout development (Woods and Bryant. 1989; Woods and Bryant. 1991). Immunocytochemical analysis shows that in epithelial cells the DlgA protein is localized in a subapical belt of the lateral cell membrane (Fig. 6B), at the position of the septate junction (Woods and Bryant. 1991). Thus the DlgA protein appears to be localized in septate junctions and is required for their formation. The role of septate junctions is unknown, although they have often been considered the invertebrate equivalent of vertebrate tight junctions (Noirot-Timothee and Noirot. 1980). Tight junctions arc known to provide a transcpithelial barrier to diffusion of small molecules and to prevent diffusion of membrane components between the apical and basolateral membrane domains (Schneeberger and Lynch. 1992). Septate junctions also provide a barrier to transcpithelial diffusion of lanthanum (Green and Bergquist. 1982) and appear to separate membrane domains as indicated by the behavior of lipophilic markers (Wood, 1990).

Sequence comparisons based on the dlg cDNA sequence indicate that the encoded protein includes a series of modules each of which is also found in other proteins (Fig. 7). These domains and their possible functions are as follows:

There are three tandem copies of the DHR motif in the N-terminal half of the predicted DlgA protein (previously reported as a “filamentous” domain based on its sequence similarity to some filamentous proteins: Woods and Bryant. 1991). The motif was named GLGF based on an amino acid sequence present al the N-terminal end of the motif in DlgA and one of its mammalian homologs (Cho et al., 1992). However, since the GLGF sequence itself is absent from most other copies of this motif, we suggest the alternative name DHR (Discs large Homologous Region). A single copy of this domain is present in nitric oxide synthase (NOS; Cho et al., 1992), in a human cytosolic tyrosine phosphatase (Gu et al., 1991), and in the cytoplasmic domain of a human brain transmembrane protein encoded by a gene potentially involved in Friedreich ataxia (Duelos et al., 1993). Most of these proteins are localized al the cell surface, probably in association with the cortical cytoskeleton. Since DlgA and at least two of its mammalian homologs (see below) are also membrane associated, we propose that the DHR repeal is involved in directing this localization.

The OPA trinucleotide repeat is found in many genes that are developmentally regulated or show tissue-specific expression, in a variety of organisms from yeast to mammals (Wharton et al., 1985). It is translated in several reading frames to give tracts that are rich in specific amino acids. In dlii it is predicted to give rise to tracts of serines and alanines.

There is a single copy of the SH3 (src-hoinology region 3; Musacchio et al., 1992) domain in the predicted DlgA protein. Mutations in the SH3 domain of v-src can restore transforming activity in otherwise defective mutants (Dezélée et al., 1992), indicating that this domain is important for regulation of protein activity. SH3 binds a ten-amino acid proline-rich peptide motif that has been found in several signal transduction proteins (Ren et al., 1993) including one that is similar to the GTPase activating protein (GAP) associated with the ras-relatcd protein rho (Cicchetti et al., 1992). Other binding targets for SH3 include the GTP-hydrolyzing motor protein dynamin (Booker et al., 1993), and guanine nucleotide exchange factors (Olivier et al., 1993; Rozakis-Adcock et al., 1993).

The PEST domain was originally described in proteins that had a high rate of degradation in the cell and was therefore interpreted as a signal for rapid turnover (Rogers et al., 1986). However, this interpretation is open to question since the Notch protein, which contains both PEST and OPA sequences, is apparently stable for several cell cycles (Kidd et al., 1989).

The DlgA protein contains an approximately 179 aminoacid region al the carboxyl terminus with significant sequence similarity lo the soluble enzyme guanylate kinase (GUK) of yeast (Berger et al., 1989; 36% identity) and pig (Zschocke et al., 1993; 35% identity) (Fig. 8). This enzyme transfers a phosphate group from ATP lo GMP, converting it to GDP. If the DlgA protein has GUK activity, it could regulate the production of guanine nucleotides that act as messenger molecules within the cell (Woods and Bryant. 1991). The DlgA GUK domain shows conservation or conservative substitution of all of the amino acids that have been shown to interact with the GMP substrate as well as the B motif that is thought to bind the Mg⁺⁺ cofactor in the yeast enzyme (Fig. 8; Stehle and Schulz. 1990, 1992).

However, in DlgA the A motif GxxGxGK. which in the yeast enzyme binds the ATP phosphate donor, is interrupted by a three amino-acid deletion (Koonin et al., 1992). This deficiency in the ATP-binding site (a domain that is otherwise highly conserved in a wide variety of kinases) suggests that the DlgA protein may have lost the ability to bind ATP and may have acquired a novel function requiring GMP binding. Whatever the exact catalytic function of the DlgA protein, its potential involvement in guanine nucleotide metabolism or regulation suggests a relationship to signal transduction mechanisms known lo control cell proliferation, such as those mediated by pp2l^ras (Woods and Bryant. 1991).

Recent findings indicate that the dlg gene has homologs in mammals and that there is at least one other related gene in

These genes all encode membrane-associated guanylate kinase homologs, which we abbreviate MAGUKs. Each MAGUK consists of three distinct domains (Fig. 7). (1) One or three copies of DHR. (2) an SH3 domain and (3) the guanylate kinase-like domain (GUK). The OPA and PEST domains present in the DlgA protein are not present in the other MAGUKs identified so far.

The first mammalian MAGUK was identified as a partial cDNA sequence (1779) from a collection of cDNAs made from human brain RNA (Adams et al., 1992). The partial sequence shows that 1779 is the human homolog of a gene recently identified in the rat (hat encodes a component of synaptic junctions (PSD-95/SAP90) in the adult brain (Cho et al., 1992; Kisiner et al., 1993). The predicted PSD-95/SAP90 protein shows strong homology (about 67% identity) to DlgA throughout its length and we therefore refer to it as a Dlg-R (discs large-related) protein. Unlike DlgA, which is expressed in a variety of tissues including the nervous system. PSD-95/SAP90 appears to be expressed only in the nervous system in postnatal rats. However, it is intriguing to note that, like the product of dlg PSD-95/SAP90 is localized in a specialized junction, the synaptic junction, that is involved in information transfer between cells. Also like the product of dlg, the PSD-95/SAP90 sequence shows strong conservation of the GMP-binding site but has the same deficiency in the putative ATP-binding site (Fig. 8), suggesting that its catalytic activity, if any, is similar to that of the Drosophila protein (Koonin et al., 1992). We have no information about PSD-95/SAP90 expression in embryonic stages of development, so it is still possible that it is expressed in epithelial cells and is involved in cell interactions controlling proliferation in these early stages.

A second human MAGUK is the heavily palmitoylated protein p55, which was purified from red blood cell membrane ghosts and shows striking similarity to (he DlgA protein in the GUK region, the SH3 region, and one DHR repeal (Ruff et al., 1991; Bryant and Woods. 1992). p55 is expressed in most cell types in addition to blood cells and thus could be involved in cell growth control. We have recently cloned a Drosophila p55 homolog (D. F. W and P. J. B.. unpublished data) and the partial predicted amino acid sequence shows approximately 60% identity to that of human p55. Thus, there are two types of MAGUK in both mammals and Drosophila, suggesting that these proteins may have distinct and conserved functions. The ATP-binding site in p55 does not contain a deficiency like that seen in DlgA. but it does show a K lo R substitution (Fig. 8) that might affect enzyme activity. The p55 gene is located near the lip of the long arm of the X chromosome, between the genes encoding the blood-clotting protein factor VIII and the enzyme glucose-6-phosphate dehydrogenase (Metzen-berg and Gitschier. 1992). No tumor suppressor genes have been discovered in this chromosome region, but at least 19 genetic diseases are associated with the interval and one of them, dyskeratosis congenita, is associated with precancerous lesions (leukoplakia) and a high risk of cancer (McKay et al., 1991; Marsh et al., 1992).

The p55 and PSD-95/SAP90 proteins were purified by sub-cellular fractionation, and in both cases several co-purifying proteins were identified (Table 1). This association may result from specific interactions between the MAGUK protein and the co-purifying molecules, or it may simply be a result of similar physico-chemical properties. The former explanation seems more likely for at least some of these proteins since they remain tightly associated with the MAGUK even after treatment with sarcosyl, a relatively harsh detergent (Cho et al., 1992). Furthermore, some of these proteins have Drosophila homologs that are found, along with DlgA, at the septate junctions (Table 1). These results suggest that the MAGUKs arc integral components of large protein networks that are restricted to specific junctional complexes or other membrane-associated structures in the cell.

Most of the MAGUK-interacting proteins thus far identified appear to be part of the cytoskeletal network, perhaps reflecting the importance of these proteins in maintaining cyloarchilecture. Other associated proteins include cell adhesion molecules (FasIII; Patel et al., 1987) and signal transduction proteins (serine/threoninc kinases; Ruff et al., 1991; Cho et al., 1992). In the case of the synaptic density, where PSD-95/SAP90 is localized, the kinase activity (type II Ca²⁺/calmodulin-dependcnt protein kinase) has been shown to be necessary for induction of long-term potentiation of synaptic transmission (Malinow et al., 1989). There are probably additional proteins associated with the MAGUKs, including some expected to interact with the SH3 domain, whose identities and functions remain to be elucidated. Furthermore, if MAGUK proteins arc involved in transduction of extracellular signals, it seems likely that transmembrane proteins will be involved upstream of them in the signalling pathway. The identity of these elements has yet to be discovered.

In lethal mutations of the fat locus all of the imaginai discs show hyperplastic overgrowth, in that the overgrowing imaginai discs retain their single-layered epithelial structure and their ability to differentiate (Bryant et al., 1988; Mahoney et al., 1991). The gene potentially encodes a 5147-amino acid protein, and it is transcribed in the embryonic ectoderm and larval imaginai discs (Mahoney et al., 1991). The predicted fat gene product is a large transmembrane protein with strong sequence homology to calcium-dependent cell adhesion molecules (cadherins: Fig. 9). However, the product is much larger than typical cadherins, with 34 extracellular domains corresponding to the four domains of typical cadherins. as well as four EGF-like repeats and other cysteine-rich regions in the extracellular domain. The cytoplasmic domain shows no clear homology lo that of typical cadherins. The structure of the predicted protein suggests that cell adhesion, dependent on a cadhcrin-like molecule, is required for the cell communication that controls tissue growth. Parts of the imaginai disc epithelium are shed as vesicles in fat mutants (Bryant et al., 1988). supporting the idea that these mutations partially disrupt cell adhesion.

Cadherins have not typically been considered as tumor suppressor gene products, but recent work is suggesting strongly that the E-cadherin (= uvomorulin, homologous to L-CAM) gene located at I6q22.1 (Nall et al., 1989) may correspond to a tumor suppressor gene identified by loss of heterozygosity in this genetic region in hepatocellular, breast, and prostate carcinomas (Sato et al., 1990; Zhang et al., 1990; (artcr ci al.. 1990; Tsuda ci al., 1990; Field, 1992). E-cadherin expression is lost in head and neck and hepatocellular carcinomas (Schipper et al., 1991; Shimoyama and Hirohashi. 1991). and E-cadherin loss is correlated with invasiveness of carcinoma and epithelial cell lines (Behrens et al., 1989; Frixen et al., 1991). Non-invasive carcinoma and epithelial cell lines become invasive when treated with anli-E-cadherin antibodies (Behrens et al., 1989; Frixen et al., 1991); and invasiveness of carcinoma cell lines can be prevented by transfection with E-cadherin cDNA (Frixen et al., 1991). These results arc highly suggestive of a role for this cell adhesion molecule in tumor or metastasis suppression. The finding that mutations in the fat gene cause overgrowth supports the idea that the cadherin losses reported in cancer cells do, indeed, contribute to the malignant phenotype.

In some cases, loss of heterozygosity for a tumor suppressor gene in individual cells of a heterozygote, caused by experimentally induced mitotic recombination, leads to overgrowing clones of cells in specific tissues. This has been demonstrated for both dlg (Woods and Bryant. 1991) and fat (Mahoney et al., 1991), and the phenomenon raises the possibility of identifying new tumor suppressor genes by directly examining the effects of mutations in mitotic recombination clones. Recent advances in Drosophila genetics make screening mitotic recombination clones very efficient. The screen takes advantage of transgenic flies developed by Golic and Lindquist (1989) in which a yeast target-specific FLP recombinase gene has been placed under the control of a Drosophila heat-shock promoter and integrated into the Drosophila genome. Yeast FLP recombinase target sequences (FRTs) also have been integrated into several proximal sites on Drosophila chromosomes. Heal shocking flies carrying the FLP recombinase construct and homozygous for a FRT site leads to a high frequency of mitotic recombination al the FRT site (Golic. 1991). To use this system for screening, chromosomes arc mutagenized in the parental generation and then made homozygous in mitotic recombination clones of the F₁ generation. The phenotype of the mutant clones is then examined, and chromosomes that give overgrowing clones arc recovered by breeding from the affected F₁ flies.

We have recovered a deficiency for the warts gene (wts) using this screen. Mitotic recombination clones homozygous for such deficiencies produce spectacular outgrowths from the bod) surface (Fig. 10). The clones are larger than normal and tend to split into several rounded fragments, whereas control clones are intact, but irregular and elongated on both legs and wings. The large size of wts clones indicates that the constituent cells undergo more divisions than normal, and the rounded shape of mutant clones suggests that the division of mutant cells is not preferentially oriented as it appears to be in wild-type imaginai discs (Bryant and Schneiderman. 1969). The texture of the cuticle in wts clones is very unusual -each polygonal cell outline is marked by a thick ridge of cuticle (Fig. 10). The apparent intrusion of cuticle between cells suggests that there may be a failure of adhesion or other abnormality in the apical region of the epithelial cells. The wts gene has been localized lo 100A3-7 on chromosome 3 based on the cytology of a series of overlapping deficiencies that give the II/.V phenotype in mitotic recombination clones (R. W. J and P. J. B., unpublished data).

The mitotic recombination screen is efficient because it allows immediate detection of candidate chromosomes in Ft flies without the necessity of breeding each Ft individual as is required in other screens for lethal mutations. The screen is ideally suited to the identification of tumor suppressor genes that function in imaginai discs.

The preceding examples show’ that the Drosophila model system can be used effectively to identify and characterize tumor suppressor genes that have human homologs. Perhaps not surprisingly, these genes encode proteins at various positions in the cell which might act at different points in the pathways of signalling and signal transduction that control cell proliferation. In the case of epithelial cells, a requirement for cell adhesion in proliferation control is suggested by the overgrowth caused by mutations affecting the fat transmembrane protein and possibly the fat membrane-associated protein. This requirement is consistent with experimental work indicating that contactdependent cell interactions control proliferation in epithelial tissue (Bryant. 1987). The involvement of guanine-nucleotide mediated signal transduction is suggested by the nature of the predicted protein produced by the dlg locus. In the hematopoietic system, little experimental work has been done on the mechanisms controlling cell proliferation and differentiation, but the evidence from the RpS6 gene suggests that phosphorylation cascades and translational control mechanisms will turn out to be important in these cells. The cloning and characterization of more tumor suppressor genes in Drosophila seems likely to provide a wealth of new information on how cell proliferation is controlled by cell signalling in a variety of tissues and organs.

The authors’ research is supported by grants from the National Institutes of Health and the National science Foundation.