When the control is lacking the role of tumour suppressor genes in cancer development

Transcription

1 J. Biosci., Vol. I9, Number 5, December I994, pp Printed in India. When the control is lacking the role of tumour suppressor genes in cancer development BERNARD M MECHLER Department of Developmental Genetics, Deutsches Krebsforschungszentrum, Im Neuenheimer Feld 280, D-69I20 Heidelberg, Germany MS received 8 March I994 Abstract. The potential to genetically dissect tumorigenesis provides the major reason to study this process in the fruit fly Drosophila. Over the last 30 years genetic analysis has identified some 55 genes in which recessive mutations cause the appearance of specific tumours during development in tissues such as the imaginal discs, the brain hemispheres, the hematopoietic organs or the gonads, Since the normal allele acts dominantly over the mutated allele, these genes are designated as tumour suppressor genes. The estimate of the number of genes that can be mutated to tumour formation may be, however, much higher ranging between I00 to 200. The challenge before this field is how best to identify these genes and elucidate their function. Current molecular procedures, such as mutagenesis mediated by P-element transposon, provide new ways for tagging any gene of interest in Drosophila and thus for cloning it rapidly. Function of the gene product can be inferred by comparing its amino acid sequence with sequences of proteins with known function or can be determined by histochemical and biochemical investigations. Progress in the understanding of tumour suppression in Drosophila is most advanced in the case of genes regulating cell growth in imaginal discs. The imaginal discs are small groups of cells displaying a strong apical-basal polarity and form folded sacs of epithelia which grow throughout the larval life and give rise to the adult tegument during metamorphosis. Tumour suppressor genes regulating cell growth of imaginal discs, such as the lethal(2)giant larvae (l(2)g1), lethal(1)discs large-1 and expanded genes, were found to encode proteins localized in domains of cell to cell contact on the plasma membrane and were thus thought to maintain cell adhesion. However, recent studies of l(2)gl have revealed that the l(2)gl protein is a component of the normal cytoskeleton which can participates to the cytoskeletal matrix underlaying the plasma membrane. These findings indicate that the changes in cell shape and the loss of apical-basal polarity in imaginal disc cells result primarily from alterations in the cytoskeleton structure. Furthermore the neoplastic growth of the mutated cells may be caused by the disorganization of an intracellular communication system that ultimately controls cell proliferation and/or cell differentiation. Keywords. Drosophila; tumour suppressor genes. I. Introduction Cancer is generally considered as a failure in the normal progression of differentiation. As a result, cancer cells escape the mechanisms controlling normal growth and proliferate. When the overgrowth of neoplastic cells reaches a critical threshold, complex syndromes arise and, ultimately, lead to the death of the organism. A genetic basis for cancer is now well established and includes the discovery of proto-oncogenes, whose activation or altered expression is associated with the cancerous state, and tumour suppressor genes which appear to be lost or functionally inactivated. 537

2 538 Bernard M Mechler About 25 years ago, pioneering studies in the field of Drosophila, mouse somatic cell and human genetics revealed that neoplasia may result from loss of function in regulatory genes controlling cell growth and differentiation (Gateff and Schneiderman 1967, 1969; Harris et al 1969; Knudson 1971). Despite these early findings it was only in the mid to late eighties that real insights into the identity of tumour suppressor genes and their mode of action have emerged. In the meantime, oncogenes have occupied the front stage for understanding and explaining the origins of cancer. Oncogenes, however, have shed light on only a part of the scenery. It is the existence of tumour suppressor genes that provides a clear understanding of inherited predisposition in human cancer, a solid basis for investigating genetic alterations in somatic cells producing tissue specific cancers, and new concepts in the regulation of how cells divide and multiply. The main criterium for tumour suppressor genes is loss of function associated with the appearance of cancerous behaviour of cells: the opposite mechanism of oncogene activation. Experimentally, oncogenes were easier to study because activated oncogenes can be transmitted to normal cells which become converted to cancerous cells. The opposite experiment of reverting cancerous cells to normality is a more difficult task, for it requires a sensitive assay selecting cells with either a low proliferating rate or a different morphology. Thus, despite the initial finding that malignancy can be abrogated by fusion with normal cells or by introducing a normal chromosome in the cancer cells, the identification and cloning of tumour suppressor genes were essentially achieved by studying familial cases of inherited cancers. Loss of tumour suppressor genes has been found to be a critical event in the occurrence of a series of human cancer such as retinoblastoma, Wilm's tumour, neurofibromatosis types 1 and 2, familial adenomatous polyposis coli and other inherited forms of colon cancer, Li-Fraumeni syndrome, and multiple endocrine neoplasia. Surprisingly, the losses of the Rb and p53 genes which are key events in inherited retinoblastoma and Li-Fraumeni syndrome, respectively, appear to be also important in other forms of more common cancers, such as colon and breast cancers. In general, the rare inherited cancers seem to be genetically homogenous, whereas the common cancers are heterogeneous involving the loss of more than one tumour suppressor gene. As shown in table 1, eight human tumour suppressor genes have been cloned (Knudson 1993). However, neoplasms are not only limited to humans but occur throughout the animal kingdom, from the most primitive metazoa to invertebrates and vertebrates and can even affect plants. Despite the rarity of spontaneous neoplasms in primitive vertebrates and invertebrates, some of these organisms are particularly suitable for genetic investigations, e.g., the fruit fly Drosophila. Because of its relatively short life span and ease of rearing in the laboratory, the fruit fly is a good model system for studying genetic mechanisms controlling cell proliferation. Furthermore, it is possible to design straight forward screening procedures for isolating mutants of interest. Over the past two decades Drosophila has become the organism of choice for molecular and genetic investigations of eukaryotic biology. Its emergence as an animal model system is closely related to the rapid advances in recombinant DNA technology and ideas established by decades of classical genetics and embryology. One of the principal reasons for the choice of Drosophila resides in the genetic legacy of the earlier studies, bequeathed by Morgan and his colleagues. Over 80

3 When the control is lacking in cancer development 539 Table 1. Cloned suppressor genes in human tumors. years of investigations have provided powerful and extremely accurate analytical techniques designed for this organism which are unsurpassed by any other animal system. A further advantage of Drosophila is the size and complexity of its genome which is intermediate between those of prokaryotes and mammals. Such a combination makes Drosophila particularly good for studying the molecular biology of development, provides tools for dissecting the components of the regulatory pathways which underlie cell proliferation and differentiation, and ultimately may elucidate the biochemical functions of the macromolecules that participate in these pathways. Knowledge about these mechanisms is accumulating rapidly and a recent survey of the genome of Drosophila (Lindsley and Zimm 1992) reports genetic and molecular data on more than 4000 genes. As increasing numbers genes will be molecularly cloned, more knowledge about the signals governing developmentally expressed genes is likely to be acquired. Although the existence of tumour suppressor genes was first postulated more than two decades ago, it has only been in the recent years that insights into their identity and function have emerged. Eight human tumour suppressor genes have been so far identified. However, the number of tumour suppressor genes may be much larger. There are at least 50 distinguishable human hereditary cancers, each possibly attributable to a different gene (Knudson 1993). In Drosophila, more than 40 genes have been so far identified by mutations causing tissue overgrowth (Gateff 1978; Gateff and Mechler 1989; Mechler and Strand 1990; Mechler 1991; Watson et al 1991) and recent genetic analyses have revealed that mutations in many more genes may induce specific tissue overgrowth during Drosophila development (Torok et al 1993b).

4 540 Bernard M Mechler 2. Cell proliferation during Drosophila development Before being committed to neoplastic growth, a cell must be able to duplicate all its essential parts. This ability is normally limited to undifferentiated cells that are either in a proliferative phase or quiescent but can readily be recruited into mitosis. Drosophila displays a very precise programme of cell proliferation during its development (Campos-Ortega and Hartenstein 1985). The embryo is initially a syncytium in which thirteen rapid cycles of synchronous nuclear division occur at approximately 10 minute intervals. Then the nuclei migrate to the cortex and become cellularized at the interphase of cycle fourteen. From this developmental phase the cell cycle lengthens and the following divisions occur in complex mitotic domains according to a specific temporal sequence which is coordinated by a complex pattern of gene expression. This leads to the morphogenesis of specific tissues within the embryo. After 24 h, the embryo hatches and contains about 50,000 cells which can be subdivided in to two classes those that form the larval tissues and grow by expansion of the cell volume with endoreplication of DNA in the absence of mitosis, and those that will constitute the imaginal cells which are destined to form the adult organism and are not themselves necessary for the survival of the larva. The later cells, designated as imaginal cells, resume their proliferation in the middle of the first larval instar and continue to divide throughout the three larval instars up to the larval-pupal transition, arresting only as terminal differentiation occurs. In the adult, only the germline cells and their associated gonadal tissues are actively dividing, as well as some hematopoietic cells. Thus, manifestation of cell overproliferation can become noticeable at the end of both major periods of cell proliferation: late embryogenesis and the larval to pupal transition phase. Mutations giving rise to tumorous growth have been identified at the end of either one or the other period. 3. Identifying overgrowth mutations in Drosophila In Drosophila, scorable overgrowth mutations are usually recessive because dominant mutations will be eliminated almost immediately after they have arisen if they interfere with the life of the individual. Another advantage of Drosophila is the availability of a system of balancer chromosomes. Balancer chromosomes contain multiple inversions preventing crossing over and carry other mutations which are homozygous lethals during embryogenesis. Such a system allows geneticists to maintain recessive mutations in permanent cultures without control and selection. Thus, when a chromosome carrying an overgrowth mutation is placed in conjunction with a balancer chromosome, such genetical configurations produce only heterozygous viable and fertile animals and, in addition, provide a constant supply of homozygous animals with overgrown tissues. Three different experimental approaches can be used for identifying mutations causing tissue overgrowth: (i) direct observation of the overgrown organs by dissection of the mutant animals, (ii) production of tumours following transplantation of mutant tissues into the abdomen of adult flies, (iii) induction of growth abnormalities in mosaic tissues following somatic recombination.

5 When the control is lacking in cancer development 541 The first approach has been most productive for identifying mutations causing tissue overgrowth (Gateff and Mechler 1989; Mechler and Strand 1990; Mechler 1991). The large majority of overgrowth mutations so far identified affects the development of the imaginal tissues, in other words the tissues which grow during larval life. Mutant animals are easily recognized because the growth of the tumorous tissues is usually accompanied by developmental arrest at the end of the larval to pupal transition phase. As a consequence the larval life of the mutant animals is extended over several days and the tumorous tissues can reach a considerable mass which is readily observed upon dissection. Mutant larvae with brain or imaginal disc tumours become either bloated or "giant" and transparent, and those with blood neoplasia become opaque white or completely melanized. In the most extreme cases of blood neoplasia hemocytes are massively released from the hematopoietic organs, invade the entire larval body and destroy most other tissues. By contrast, mutations affecting control of cell proliferation in embryos do not give rise to visible tumours and, in this case, neoplastic growth can only be assayed by transplantation (Gateff and Schneiderman 1974). For this purpose, fragments of embryonic tissues are implanted into the abdomen of adult flies. Under these conditions, the neoplastic cells proliferate rapidly and kill the host in 3 14 days. By contrast, the normal tissues grow only moderately and exert no deleterious effect on the host. This procedure provides an efficient system for recognizing autonomously growing tissues, and particularly neoplastic tissues, but may be insufficient for detecting potentially tumorous tissues with limited capacity for autonomous cell proliferation. The third approach allows the identification of mutations which are usually zygotic lethals but have the capacity to produce abnormal cell proliferation when homozygosity is induced by mitotic recombination. In this assay mitotic recombination is induced in imaginal tissues either by ionizing radiations (for review see Becker 1976), or by the activation of a heat inducible site-specific yeast recombinase gene inserted in the Drosophila genome (Golic 1991). As a result, a single cell may become homozygously mutated in a background of heterozygous cells and, depending upon the developmental phase, this founder cell will generate a clone of cells which colonize one or several organs. If recombination occurs during the early larval stage, the ensuing clone will be confined only to one organ such as a wing, a leg, or an eye where growth abnormalities can be easily scored. Analysis by somatic recombination has allowed the identification of a series of mutations which can cause a global increase in the number of imaginal cells and can alter their pattern of differentiation (Schubiger and Palka 1987; Diaz-Benjumea and Garcia Baldo 1990; Garcia-Bellido and de Celis 1992). However, none of these mutations can produce massive growth of tissues displaying neoplastic properties. 4. The tumour phenotype Drosophila tumours can be classified into two broad categories; neoplasia and hyperplasia (Gateff 1978; Bryant 1987; Gateff and Mechler 1989; Mechler 1991). The difference between both types of overgrowth is particularly visible in the case of the imaginal discs which form distinct groups of undifferentiated cells

6 542 Bernard M Mechler present in the larvae and give rise to various parts of the adult body during metamorphosis. These groups of cells can be identified in the newly hatched larvae as discs of approximately 40 cells in each one. They grow throughout the larval life and form folded sacs of epithelia which are essentially constituted of a monolayer of columnar cells. Other internal organs with proliferating cells, such as the brain hemispheres and the haematopoietic organs, may also be affected by neoplasia or hyperplasia. However, due to the structure of these organs, it is difficult, on morphological criteria, to assign the tumours into one or the other category. 4.1 Neoplastic growth Neoplasia of the imaginal discs, are characterized by a massive proliferation of cells which disrupt the mono-layered epithelial structure of the imaginal discs, converting them into amorphous masses of tissues resembling tumours. The imaginal discs present in the cephalic region of the larvae frequently fuse with one another and the mass of tumorous discs may reach several times their normal size. In addition, neoplastic cells lose their apical-basal polarity and are cuboidal rather than columnal in shape. Other tissues, such as the brain hemispheres and the hematopoietic organs can also give rise to neoplasia. Like vertebrate tumours Drosophila neoplasms exhibit a range of characteristics: (1) cell overproliferation, (ii) altered cell morphology, (iii) loss of differentiation capacity, (iv) invasiveness and (v) transplantability. These features can help to define neoplasia in Drosophila, but only some tumours display all of them (for review, see Gateff and Mechler 1989). The main features of Drosophila neoplasms are the overgrowth of the neoplastic tissues which form amorphic structures and have lost their capacity to differentiate. Further grades in the tumour phenotype can be established. For example, malignant neoplasms grow aggressively after transplantation and kill the host, whereas benign neoplasms are unable to proliferate or grow only moderately. Furthermore, benign neoplasms form usually compact tumours which remain limited within the organ of origin and do not invade the surrounding tissues. However, some tumours which should be classified as benign because they are unable to grow autonomously after transplantation exhibit a very aggressive pattern of growth in situ, invading the surrounding tissues. Brain neoplasms are characterized by a gross enlargement of the optic lobes resulting from the uncontrolled proliferation of optic neuroblasts and ganglion mother cells of the anlagen of the adult optic centers of the larval brain. In tumorous brains these cells do not differentiate optic neurons and display a disorganized pattern of growth. Neoplastic growth of the hematopoietic tissues is characterized by a massive outgrowth of the five to seven pairs of hematopoietic organs located along the dorsal heart vessel, behind the brain hemispheres. In several mutant cases, undifferentiated hemocytes are released in large amount into the hemolymph where they behave aggressively invading all larval tissues and causing their destruction, whereas in other mutants, the hemocytes remain confined to the hematopoietic organs which expand massively.

7 When the control is lacking in cancer development Hyperplastic growth Hyperplasia of the imaginal discs are characterized by a massive proliferation of the imaginal cells which maintain a columnar shape with a normal apical-basal polarity. During the prolonged larval life the hyperplastic discs can grow to several times their normal size and, keep a folding pattern, albeit abnormal. In addition, the discs appear often to be duplicated. Furthermore, the hyperplastic disc cells retain some capacity to differentiate, displaying upon differentiation a restricted pattern of adult cuticular structures. In some of these mutants, other proliferating tissues such as the brain hemispheres, the imaginal rings of the foregut, hindgut, and salivary glands, exhibit variable overgrowth. However, when transplanted into adult hosts, the hyperplastics tissues are unable to grow. Imaginal discs are the only tissues which have been shown to give rise to hyperplastic growth although it is not excluded that some other tumourous tissues may fall into this category. Similar to human tumours, the classification of Drosophila tumours suffers from a lack of invariant properties. Presently, we can only define Drosophila tumours on the basis of in vivo structural and behavioural characteristics of the overgrown cells. We are still lacking good markers, molecular or biochemical, for characterizing the various forms of tumorous growth. An ultimate goal of our investigations is to elucidate the molecular function of the genes which control cell proliferation and to determine whether a common mechanism may underlie each categories of tissue overgrowth. 5. Tumour suppressor genes in Drosophila In Drosophila, mutagenesis screens and the analysis of spontaneously occurring mutations have allowed the identification and genetic mapping of a series of tumour suppressor genes and genes controlling tissue overgrowth. Known genes described in the literature are listed in table 2. Tissue overgrowth appears at different developmental stages and in distinct tissues: two mutations give rise to potential neoplasia during embryogenesis, thirty-three produce visible malignancies during the larval development in either the brain hemispheres, the imaginal discs, the hematopoietic organs or combination of these tissues, eight affect the germ line. Furthermore, five mutations cause well-defined hyperplasia of the imaginal discs. Capacity for tissue overgrowth in mitotic clones has been found in the case of eight mutations. All these genes act as recessive determinants of tissue overgrowth and are classified as tumour suppressor genes, for their normal function is to control cell proliferation and/or differentiation. In Drosophila convincing demonstration of tumour suppression can be obtained by introducing an intact allele of a cloned tumour suppressor gene into the genome of homozygous mutated animals and by showing restoration of a normal growth and reversion to a normal phenotype (Opper et al 1987). Such type of experiment has proved successful in the case of the lethal(2)giant larvae gene. Introduction of a normal copy of the lethal(2)giant larvae tumour suppressor gene into the genome of animals deficient for this gene can prevent the occurrence of tumorigenesis and restore a complete development of these animals. This result shows that the l(2)gl gene behaves as a tumour suppressor (Opper et al 1987).

8 544 Bernard M Mechler Table 2. Neoplastic and hyperplastic mutants in D. melanogaster.

9 When the control is lacking in cancer development 545 a The mutant insects develop neoplasia during the embryonic development and die during embryogenesis. b The mutant insects develop neoplasia during the larval development and die as third-instar larvae or pseudopupae. c Dominant temperature-sensitive mutant (Hanratty and Ryerse 1981). Thus, 56 genes causing tissue overgrowth are presently known in Drosophila. Their identification has occurred erratically over the last 25 years and, until recently, no systematic search for mutations causing either neoplastic transformation of embryonic tissues, or overgrowth of imaginal tissues has been undertaken. However, the recent development of derivatives of transposable P-element allowing the tagging of specific genes and their direct molecular cloning can provide a valuable approach for identifying new tumour suppressor genes and estimating their numbers. It would be interesting to know whether new mutagenesis experiments will constantly result in the isolation of already known genes and will indicate that the Drosophila genome is saturated for mutations affecting tumour suppressor genes, or may reveal new genes and, in this case, how many? Answer to this question may also help us in estimating the number of genes controlling cell growth and tumorigenesis in Drosophila. Recent P-element mediated mutagenesis experiments of the X-chromosome of D. melanogaster have conducted to the isolation of nine complemention groups producing overgrowth of the haematopoietic organs (Watson et al 1991). In this experiment 1,100 X-linked lethal mutations have been examined and all the new genes are only represented by single mutant allele. This indicates that the X-chromosome has not yet been saturated for mutations affecting tumour suppressor genes. Furthermore,

10 546 Bernard M Mechler a P-mediated mutagenesis experiment of the second chromosome has led to the isolation of 16 new mutations causing overgrowth of imaginal tissues has been isolated (Torok et al 1993b). These mutations are distributed into 14 genes, 13 of which are represented by single mutant alleles. Cytogenetic and genetic analyses have shown that these mutations affect genes which are distinct from the five tumour suppressor genes which are known on the second chromosome. Since only one half of the collection of 2,700 second chromosome lethals has been screened, one can expect the discovery of a series of additional mutants affecting tumour suppressor genes in the other half of the collection. Furthermore it is also reasonable to envisage that some of the future mutants will correspond to unidentified genes. The results of this analysis show again that the second chromosome of Drosophila has not yet been fully saturated for P-element mutations causing overgrowth of imaginal tissues. All together, eleven tumour suppressor genes controlling hyperplasia and neoplasia of imaginal tissues are presently known on the X-chromosome and 19 on the second chromosome. Since the X-chromosome represents approximately 20% of the size of the genome of Drosophila, and the second chromosome 40%, it is reasonable to envisage that at least 55 tumour suppressor genes may be present in Drosophila, but this number can certainly be larger. It would not be surprising if the actual number of tumour suppressor genes in Drosophila amounts to 100 to 200. Analysis of the mutants isolated by I Kiss and coworkers at Szeged revealed a series of new phenotypes which have not yet been described, such as simultaneous overgrowth of three distinct tissues: brain hemispheres, imaginal discs and haematopoietic organs, or overgrowth limited to only some imaginal discs, such as the labial, clypeo-labral and antennal discs but surprisingly not the eye disc which is intimately associated with the antennal disc (Torok et al 1993b). Mutations in tumour suppressor genes cause complex biological effects that not only affect the tissues with potential tumorous growth but may also exert deleterious effect on other tissues. In these respects, the work of Hadorn's group has shown that the l(2)gl mutation affects the development of numerous tissues (i.e., atrophy of the male germ line, underdevelopment of the imaginal cells of the salivary glands and the gut) before the appearance of the tumorous growth (for review see Hadorn 1961). Therefore, these mutations produce pleiotropic effects, neoplasia or hyperplasia being the most striking feature. However, the other damages should not fall into oblivion because they may help us to understand the normal function of these genes. In view of the complex biological effects caused by these mutations, the molecular cloning of tumour suppressor genes is an essential step for understanding the primary genetic changes that lead to these abnormalities and particularly to tumorigenesis. At the molecular level the first tumour suppressor gene, the lethal(2)giant larvae gene of D. melanogaster, was isolated in 1985 and ten more genes controlling cell overgrowth have since been characterized (see table 3). Four of these genes (l(2)gl, fat, dlg and air ' ) display sequence similarity to mouse or human genes (Mahoney et al 1991; Bryant and Woods 1992; Watson et al 1992; Stewart and Denell 1993; Tomotsune et al 1993). However, none of these human homologues to Drosophila genes have so far been recognized as tumour suppressor genes, although there are already hints that they may be deregulated in certain human cancers. The fat gene encodes an enormous cell-adhesion molecule of the cadherin family

11 When the control is lacking in cancer development 547 Table 3. Cloned tumour suppressor genes in Drosonhila. (Mahoney et al 1991). This transmembrane protein contains more than 5,000 amino acids with 34 tandem cadherin domains, and four EGF-like repeats. In Drosophila absence of the fat protein leads to hyperplastic growth of the imaginal discs. In higher vertebrates several cadherin genes have been isolated but so far none has been directly implicated in tumorigenesis. However, recent cloning of one of the candidate colorectal carcinoma tumour suppressor genes, the DCC (deleted in colorectal carcinoma) gene, has revealed that this gene encodes a large cell adhesion molecule, distinct from cadherin. The DDC protein contains four immunoglobulin-like domains and a fibronectin domain related to the domains present in N-CAM, LI, and other members of this family of CAM (Fearon et al 1990). The dlg gene encodes a homologue to a guanylate kinase gene (Woods and Bryant 1991; Bryant and Woods 1992; Cho et al 1992), first identified in yeast, of which two distinct forms have been recently found in human (Koonin et al 1992). The dig protein is localized at septate junctions between epithelial cells of

12 548 Bernard M Mechler the imaginal discs and other tissues. In Drosophila mutations in dig produce neoplasia in the imaginal discs and overgrowth of the brain. Guanylate kinases are enzyme which transfer phosphate from ATP to GMP and convert it into GDP, the penultimate step before the conversion of GDP to GTP. Thus, by regulating the availability of GDP in the cells, these enzymes may influence signalling processes involving guanine nucleotides. In these respects, the gene responsible for Von Recklinghausen's neurofibromatosis (NF1) has recently been cloned (Cawthon et al 1990; Wallace et al 1990; Xu et al 1990) and shown to encode a GTPase activating protein. In humans inactivation of NF1 leads to abnormal proliferation of cells of the neural crest origin. Thus, the decrease of GTP production in dig animals may cause similar tissue overgrowth as the absence of a GTPase activating protein in human. Mutations in the air 8 gene reduce drastically the level of expression of a gene which was found to encode the ribosomal protein S6. The expression of the human homologue is differentially regulated and appears to be over expressed in human colon carcinomas (Staniunas et al 1990). By contrast, in Drosophila a strong reduction of S6 protein expression leads to neoplasia of the hematopoietic organs (Watson et al 1992; Stewart and Denell 1993). Recently, a mouse homologue to the l(2)gl gene, designated as mgl-i, has been isolated by virtue of its binding to the Hox-C8 protein in native chromatine (Tomotsune et al 1993) indicating that mouse nzgl-1 expression may be regulated by homeobox proteins. These four examples show that study of tumour suppressor genes in Drosophila offers interesting possibilities for understanding the regulatory circuitry that governs cell proliferation and differentiation and thus the molecular mechanisms of cancer pathogenesis. In addition they indicate also that these mechanisms may not be so different between the fruit fly and human. 6. The lethal(2)giant larvae gene 6.1 Genetics and developmental biology of l(2)gl The best known tumour suppressor gene of Drosophila is the lethal(2)giant larvae l(2)gl gene. Homozygous mutations in the l(2)gl gene lead to neoplastic transformation of the neuroblasts and ganglion mother cells of the adult optic centres in the larval brain and neoplasia of the imaginal discs (Gateff and Schneiderman 1967, 1969, 1974). In the mutant animals these neoplasms first become visible in the third larval instar. The brain hemispheres and imaginal discs grow to several times their normal size during the extended life of the l(2)gl larvae. The l(2)gl gene was discovered by Bridges in 1933 (Bridges and Brehme 1944) and intensively studied by Hadorn and his collaborators (for review, see Hadorn 1961). The l(2)gl mutation was first recognized as a lethal factor producing pleiotropic effects with degeneration of the imaginal discs of the head, thorax and genitalia, aplasia of germ cells in the testes but not of the ovaries and underdevelopment of the salivary glands, fat bodies and prothoracic cells of the ring gland, which is the source of ecdysone. Lack of ecdysone leads to considerable delay of pupation, and the mutant animals die as giant larvae or pseudopupae.

13 When the control is lacking in cancer development 549 However, atrophy of the ring gland is apparently not the primary cause of the mutation but appears to result from the invasive growth of malignant neuroblasts within the brain hemispheres that disrupts essential functions required for the maturation of the ring gland. This is supported by the observations that transplantation of a normal ring gland l(2)gl-deficient larvae (Hadorn 1937) or the injection of ecdysone (Karlson and Hauser 1952) can induce pupation but cannot fully rescue the development of the mutant larvae. Furthermore, ecdysone deficiency is not sufficient to cause neoplastic growth (Garen et al 1977). Therefore, the l(2)gl phenotype does not seem to result from a deficiency in ecdysone or its receptor (Richards 1976). l(2)gl was first mapped distally to position 21C at the left end of the second chromosome (Lewis 1945). Molecular cloning has shown that l(2)gl is located at position 21A and is the first gene at this end of the chromosome (Mechler et al 1985). On its distal side l(2)gl is flanked by telomeric repetitive sequences. Numerous spontaneous mutant alleles have been isolated from wild Drosophila populations in the Soviet Union (Golubovsky 1978, 1980) and in California (Green and Shepherd 1979). Molecular analysis of the chromosomal structure of these mutant alleles has revealed that almost all consist of terminal deletions that have removed the l(2)gl gene or part of it (Mechler et al 1985). Examination of naturally occurring variants has shown that approximately 1% of flies in these widely separated populations had mutations at the l(2)gl locus (Golubovsky 1980). The high frequency of l(2)gl mutations and the preponderance of deletions suggest that the l(2)gl gene is associated with an unstable chromosomal region, possible linked to telomeric variation (Mechler et al 1985). 6.2 Isolation of the l(2)gl gene The l(2)gl gene was isolated by molecular cloning in (Mechler 1984; Mechler et al 1985) and shown to encompass 13 kb of DNA at the extremity of the left arm of the second chromosome. The initial identification depended on detection of deletions involving large segments at the left end of chromosome 2 which were visualized by in situ hybridization on polytene chromosomes of salivary glands or by means of Southern blotting analysis. The limits of the gene were determined by mapping 20 distinct chromosomal rearrangements and by relating the structural disruptions to one of the transcription units contained within the cloned region. This procedure identified a 13-kilobase (kb) transcription unit encoding two classes of overlapping transcripts of about 4 5 and 6 kb, respectively. In all examined l(2)gl mutants, this transcription unit was structurally altered (Mechler et al 1985; Lutzelschwab et al 1986). Positive identification of the biologically active l(2)gl gene was obtained by reintegrating a kb DNA fragment containing the putative l(2)gl gene into the genome of l(2)gl-deficient animals by P-mediated germline transformation and showing that this sequence fully rescued the development of the insects that would otherwise have succumbed with brain and imaginal disc neoplasia (Opper et al 1987; Jacob et al 1987).

14 550 Bernard M Mechler 6.3 The l(2)gl protein The entire 13 kb genomic DNA segment and several cdnas have been sequenced (Jacob et al 1987). This analysis showed that the l(2)gl gene encodes a polypeptide of amino acids with a molecular weight of about 127 kda, designated as p127. Polyclonal and monoclonal antibodies prepared against l(2)gl fusion proteins or synthetic peptides recognize a polypeptide of 130 kd in size (Lutzelschwab et al 1987; Merz et al 1990) which can be metabolically labelled with 32 P (Strand et al 1991). No other post-translational modifications have been so far detected (Strand et al 1991). Although l(2)gl does not appear to encode a membrane protein, having no likely secretory signal sequence or membrane-spanning domain, others have found similarities in the p127 sequence with extracellular domains of the L-CAM family (Lutzelschwab et al 1987) or the cadherin family of cell adhesion molecules (Klambt et al 1989). Recent biochemical investigations and cell fractionation studies have revealed, however, that l(2)gl encodes an intracellular protein which is recovered either free in the cytoplasm and tightly attached to the inner face of the plasma membrane at junctional sites between cells (Merz et al 1990; Strand et al 1991; Strand, Raska and Mechler, submitted). Amino acid sequence analysis and comparison with protein sequences show the presence of several amino acid motifs, such as reiterated heptad units of hydrophobic amino acid residues and repeated motifs showing homologies with motifs present in ß-subunits of G-proteins (Dalrymple et al 1989; Mechler et al 1991). Such motifs indicate that p127 may interact with other proteins and/or form multimeric complexes with itself. Moreover, the regular spacing of the hydrophobic amino acid residues as well as the other motifs playing a potential role in protein-protein interaction are strongly conserved during evolution and can be found in l(2)gl homologues of other dipteran species (Torok et al 1993a). The presence of potential motifs for protein-protein interaction has instigated studies to determine whether any domain of p127 can promote its multimerization or its interaction with other proteins. Recent analysis has shown that several segments of p127 can indeed induce multimerization of a fused protein A, which by itself remains in a monomeric form (R Jakobs, D Strand and B Mechler, in preparation). Furthermore, gel filtration studies have revealed that p127 is always recovered in large oligomeric complexes, ranging between 500 and 1,000 kda (D Strand and B Mechler, personal communication). Analysis of these complexes has shown that p127 is the predominant component in these complexes suggesting further that p127 can by itself form quaternary structures. In addition to selfaggregation of p127, these studies have also revealed the presence of additional proteins in the p127 complexes. One of these components was found to be a serine protein kinase which specifically phosphorylates p127 at multiple sites and displays similar characteristics to those of protein kinase A (A Kalmes, D Strand and B Mechler, personal communication). Cell fractionation studies and immunohistochemistry have revealed that a significant proportion of p127 can bind to the plasma membrane. Examination of the membrane association of p127 was performed by using selective detergent extraction procedures and solubilizing agents which have been previously used to study interactions of proteins with the cytoskeleton (Geiger 1983; Garziani et al 1989; McCrea and Gumbiner 1991). The studies showed that, following treatment with the non-ionic

15 When the control is lacking in cancer development 551 detergent Nonidet P40 which solubilizes the membrane lipids, a large proportion of the membrane associated p12.7 was recovered as large insoluble aggregates which were found to contain elements of the membrane cytoskeleton such as actin. Treatments with various solubilizing agents such as high salt, low ph or chaotropic agents (6M urea) released the same proportion of p127 as the non-ionic detergent, leaving the majority of p127 in an insoluble form, whereas strong ionic detergents or high ph solubilized p127 effectively (Strand, Raska and Mechler, submitted). All together these data indicate that p127 forms or participates to a cytoskeletal network and may be a component of a signal transduction pathway involved in intercellular communication which directs cell differentiation. 6.4 l(2)gl gene expression The spatio-temporal pattern of l(2)gl transcription and protein expression has been studied by in situ hybridization and immunostaining. Both analyses have shown ubiquitous gene expression during early embryogenesis. The l(2)gl transcripts are first found in all embryonic cells from the blastoderm stage (2. 5 hours of development) up to completion of the germ band extension (about 8. 5 hours). During this period l(2)gl expression is uniform and relatively intense over all embryonic cells. At the time of the dorsal closure (about 10 hours) the l(2)gl transcripts become gradually restricted to the epithelial cells of the midgut, where they persist until the end of embryogenesis (Mechler et al 1989; Merz et al 1990). Immunostaining with anti-p 127 antibodies has shown that the pattern of protein expression generally follows the pattern of gene transcription. During early embryogenesis, p127 is ubiquitously expressed in all embryonic cells from the syncytial embryogenesis, p127 is ubiquitously expressed in all embryonic cells from the syncytial blastoderm stage up to germ band retraction and then becomes gradually restricted to the midgut epithelium by the end of embryogenesis. In the early syncytial embryo, p127 is first concentrated in the apical periplasm between the nuclei and the egg surface. During cellularization, p127 extends downward along the growing plasma membranes, forming nods linked by microfibrilles. During gastrulation p127 becomes diffusely distributed in the whole cytoplasm as well as localized along the plasma membranes. This pattern of diffuse distribution of p127 in the cytoplasm remains constant during germ band elongation and retraction at which point the p127 becomes more localized to the lateral faces of epithelial cells such as the midgut cells. The expression in the digestive tract is then maintained during all the larval development up to the mid third larval instar at which time p127 is intensively expressed in all imaginal discs but not in the brain hemispheres. The absence of p127 expression in larval brain tissues was further confirmed by dissecting brains and imaginal discs of late third instar larvae and analysing p127 by immunoprecipitation and Western blotting procedures. No p127 could be detected in brain extracts although p127 could be easily identified in extracts of imaginal discs. The absence of p127 expression in brain tissues indicates that the larval expression of p127 in imaginal discs bears no direct contribution to the control of cell proliferation in these tissues and points out that the critical period of p127 expression for preventing tumorigenesis occurs at a much earlier developmental, stage, namely, during embryogenesis.

16 552 Bernard M Mechler In the adults, p127 expression occurs essentially in two different organs: the digestive tract and the reproductive organs. In the ovaries, p127 is strongly expressed in the germanium and in previtellogenic egg chambers and more moderately in vitellogenic egg chambers. In previtellogenic egg chambers, p127 is uniformly distributed in the cytoplasm of all nurse cells, oocytes and follicle cells. On the plasma membranes, p127 is essentially located along the lateral sides of the cells and is virtually absent from the basal and apical membrane domains of these cells. The absence of membrane localization is particularly noticeable in the regions of contact between the germ line cells and the follicle cells. In the last previtellogenic stage, the egg chambers begin to elongate and the follicle cells become enlarged and columnar. As soon as yolk formation begins at stage 8, p127 staining disappears completely from the oocyte and fades gradually in the nurse cells when it becomes undetectable from stage 10. However, during all these stages, p127 is intensively expressed in the follicle cells until they become thinner and squamous at stage 11. Our studies demonstrate that in adult female ovaries both the germ line and mesodermally derived cells express high level of p127. These results corroborate previous observations made by Hadorn and Gloor (1942) indicating that l(2)gl gene activity was required for terminal differentiation of oogenesis. These researches showed that transplanted l(2)gl ovaries into normal third instar larvae develop synchronously with the metamorphosing host and, at the time of hatching, apparently reach the same stage as normal ovaries. However, the development of the egg chambers is profoundly perturbed with variable number of nurse cells and an incomplete layer of follicle cells grouped together in a disorganized mass resembling tumours (Gloor 1943). In addition, recent analyses involving pole cell transplantation and genetic mosaics have demonstrated that the absence of l(2)gl function in either the germ line or the follicle cells prevents egg development (Szabad et al 1991) In the testis, p127 is essentially detected at the extreme apex of the sperm tube in a region where the spermatogonial cells are located. This finding shows also that the l(2)gl function may be required for the differentiation of the male germ line and is supported by previous observations showing that the male gonads of hatching l(2)gl larvae are degenerated (Gloor 1943). 6.5 Biology of l(2)gl function The results of our molecular analyses indicate that the critical period of gene expression for preventing tumorigenesis can be associated with the embryonic period of l(2)gl transcription when the gene is ubiquitously expressed in all cells, and especially in the progenitors of the cells that will become neoplastically transformed in the mutant animals. Furthermore, explants of l(2)gl embryonic tissues were shown to grow malignantly in adult hosts (Gateff and Schneiderman 1974), suggesting further that the critical period for commitment to tumorous growth takes place during embryogenesis. However, as mentioned above, 1(2)gl gene expression occurs during two phases of Drosophila development. The respective contributions of each period of l(2)gl expression were analysed by inducing the appearance of l(2)gl clones in otherwise heterozygous animals and by studying the fate of the clones during development.

17 When the control is lacking in cancer development 553 Analysis of the l(2)gl mosaic animals (Mechler et al 1990; A Mishra-Sinha and B Mechler unpublished results) showed that neoplastic growth occurs in clones of cells that have lost the l(2)gl gene in the preblastoderm syncytial embryos before any expression of the l(2)gl gene. Clones produced at later embryonic stages, but still during the embryonic phase of l(2)gl expression, do not display the neoplastic phenotype but are unable to complete differentiation. Finally, when l(2)gl - clones arise during larval stages, the l(2)gl deficient tissues show a nearly normal or normal development. With the indication that the l(2)gl gene produces an intracellular protein and that the l(2)gl activity is cell autonomous, these data show that the critical period for the establishment of l(2)gl neoplasia takes place during embryogenesis, when intense l(2)gl gene activity occurs in all embryonic cells. Immunocytochemistry as well as biochemical investigations and cell fractionation studies have shown that the l(2)gl protein participates in a cytoskeletal network which can either be diffused in the entire cytoplasm or bound to the inner face of lateral cell membranes in regions of cell junctions. This finding indicates that p127 may directly contribute to the structure and maintenance of cellular architecture, in particular, cell polarity, and thus to the structure of tissue organization. The structural alterations induced by l(2)gl inactivation may, in turn, reduce the potential of a cell to receive and process inter- and intra-cellular signals, and can be therefore critical for differentiation. The binding of p127 to the innerface of the plasma membrane and its tight association to a serine protein kinase suggest that p127 may act as a transducer in a signal pathway linking plasma membrane bound receptors to intracellular effector proteins. By blocking or altering the reception of signals, the disruptions induced by l(2)gl inactivation prevent the cells from progression in their normal programme of development and keep them as undifferentiated stem cells which continue to grow and divide. Present investigations are concerned with the identification of the nature and specificity of the kinase associated with p127 as well as the other components participating to the large p127 complexes. Moreover, studies are also carried out to determine by which mechanism the l(2)gl protein becomes attached to the plasma membrane and the relevance of this binding with respect to l(2)gl function. Acknowledgements The studies reported here were supported by grants of the Deutsche Forschungsgemeinschaft, the Swiss Cancer League, the European Commission, Contract No. CI1*-CT and ERB-SC1-CT and general funds of the Deutsches Krebsforschungszentrum. References Becker H J 1976 Mitotic recombination; in The genetics and biology of Drosophila (eds) M Ashburner and E Novitski (London: Academic Press) Vol. I C, pp Boedigheimer M and Laughon A I993 expanded: a gene involved in the control of cell proliferation in imaginal discs; Development Bridges C B and Brehme K F 1944 The mutation of Drosophila melanogaster (Carnegie Institute of Washington) Publication No. 522 Bryant P Experimental and genetic analysis of growth and cell proliferation in Drosophila