The genetic analysis of ovarian cancer.

Ovarian cancer represents the fifth most significant cause of cancer-related death for women and the most frequent cause of death from gynaecological neoplasia in the Western world. The incidence of ovarian cancer in the UK is over 5000 new cases every year, accounting for 4275 deaths per year (Chang et al., 1994). A recent meta-analysis of all randomised trials of patients with epithelial ovarian cancer after surgery demonstrated an overall 5 year survival of 30% (Advanced Ovarian Cancer Trialists Group, 1991). Five year survival rates are as follows: stage I, 70%; stage II, 45%; stage III, 17%; and stage IV, 5% (Chang et al., 1994). The high overall mortality is due to the majority of patients presenting with stage III and IV disease. Clearly, any methods that enable the early detection of ovarian cancer would lead to a significant decrease in mortality. Ovarian cancer encompasses a broad spectrum of lesions, ranging from localised benign tumours and tumours of borderline malignant potential, through to invasive malignant adenocarcinomas. Histologically, the common epithelial ovarian cancers, which account for 90% of all ovarian cancer, are classified into several types, that is serous, mucinous, endometrioid, clear cell, Brenner, mixed epithelial and undifferentiated tumours. The different histological subtypes reflect the considerable differentiation potential of the ovarian surface epithelium. The aetiology of ovarian cancer is not completely understood, although both epidemiological and genetic associations have been recorded. Epidemiological factors related to ovulation seem to be important (Fathalla, 1971), whereby ovarian epithelial cells undergo several rounds of division and proliferative growth to heal the wound in the epithelial surface, thereby increasing the chance of a genetic accident during the repair process, such as the activation of an oncogene or the inactivation of a tumour-suppressor gene (Berek et al., 1993). The genetic changes occurring in epithelial ovarian cancer are also poorly understood and, except for the analysis of the p53 gene, the majority have not yet been defined. This review focuses on the current understanding of cytogenetic abnormalities, linkage and allele loss studies that signpost chromosomal regions which may contain relevant genes. The emphasis of this review is on recessively acting rather than dominant genes (reviewed recently in Berchuck et al., 1992) as the isolation of tumoursuppressor genes will lay the foundation for an improved understanding of the mechanisms involved in tumorigenesis.

Ovarian cancer represents the fifth most significant cause of cancer-related death for women and the most frequent cause of death from gynaecological neoplasia in the Western world. The incidence of ovarian cancer in the UK is over 5000 new cases every year, accounting for 4275 deaths per year (Chang et al., 1994). A recent meta-analysis of all randomised trials of patients with epithelial ovarian cancer after surgery demonstrated an overall 5 year survival of 30% (Advanced Ovarian Cancer Trialists Group, 1991). Five year survival rates are as follows: stage I, 70%; stage II, 45%; stage III, 17%;and stage IV, 5% (Chang et al., 1994). The high overall mortality is due to the majority of patients presenting with stage III and IV disease. Clearly, any methods that enable the early detection of ovarian cancer would lead to a significant decrease in mortality.
Ovarian cancer encompasses a broad spectrum of lesions, ranging from localised benign tumours and tumours of borderline malignant potential, through to invasive malignant adenocarcinomas. Histologically, the common epithelial ovarian cancers, which account for 90% of all ovarian cancer, are classified into several types, that is serous, mucinous, endometrioid, clear cell, Brenner, mixed epithelial and undifferentiated tumours. The different histological subtypes reflect the considerable differentiation potential of the ovarian surface epithelium.
The aetiology of ovarian cancer is not completely understood, although both epidemiological and genetic associations have been recorded. Epidemiological factors related to ovulation seem to be important (Fathalla, 1971), whereby ovarian epithelial cells undergo several rounds of division and proliferative growth to heal the wound in the epithelial surface, thereby increasing the chance of a genetic accident during the repair process, such as the activation of an oncogene or the inactivation of a tumour-suppressor gene (Berek et al., 1993). The genetic changes occurring in epithelial ovarian cancer are also poorly understood and, except for the analysis of the p53 gene, the majority have not yet been defined. This review focuses on the current understanding of cytogenetic abnormalities, linkage and allele loss studies that signpost chromosomal regions which may contain relevant genes. The emphasis of this review is on recessively acting rather than dominant genes (reviewed recently in Berchuck et al., 1992) as the isolation of tumoursuppressor genes will lay the foundation for an improved understanding of the mechanisms involved in tumorigenesis.

Clonality
At surgery, tumours are frequently found in both ovaries and at other locations in the abdomen and pelvis, raising the possibility of a multifocal origin. However, it appears that, like most other neoplasms, ovarian cancer is clonal in origin (Bello and Rey, 1990;Boltz et al., 1990;Pejovic et al., 1991;Jacobs et al., 1992;Mok et al., 1992;Tsao et al., 1993 Evidence for clonality is provided when the loss of genetic material, abnormalities of karyotype and/or point mutations which have contributed to the initial malignant transformation are still present in the malignant cells of metastatic deposits. Several studies (for example Tsao et al., 1993) have shown that patterns of allelic deletion and chromosome methylation were identical in both the primary lesion and associated metastatic tumour within a given patient, thus providing support for the unifocal origin of ovarian tumours.
The genetic model for multistep tumour progression of colorectal tumours (Fearon and Vogelstein, 1990) has features which may be relevant for ovarian cancer, though the progression from benign to malignant in ovarian tumours is controversial (for example Powell et al., 1992). There is currently no definite evidence to show whether ovarian carcinomas develop by multistep progression or whether they arise de novo, that is each disease stage represents a distinct entity. At the recent Helene Harris Memorial Trust Meeting (Blackett and Sharp, 1994) it was concluded that at least a small proportion of ovarian cancers appear to arise from pre-existent benign tumours. The uncertainty of the origins of ovarian cancer may be resolved by the detailed molecular analysis of tumours.

Tumour-suppressor genes
Recent evidence indicates that a normal cell is converted to a malignant counterpart following the accumulation of a critical number of mutations within regulatory genes. These genes fall into two classes: oncogenes (or proto-oncogenes), which promote cell growth, and tumour-suppressor genes, which inhibit cell growth. Proto-oncogenes are necessary for normal growth and differentiation, but when altered by such events as mutation, translocation or amplification they function as transforming oncogenes. The activation of several proto-oncogenes (such as c-erbB-2, c-fms, c-myc and Ki-ras) occurs relatively frequently but appears to be unrelated to prognosis.
Tumour-suppressor genes, like oncogenes,-are involved in the regulation of cellular growth and differentiation. However, tumour-suppressor genes act recessively, that is it is the loss or inactivation of both copies of a tumour-suppressor gene that removes normal constraints to cell proliferation. In this model of carcinogenesis, loss or inactivation of a tumour-suppressor gene can be due to one of several mechanisms, such as point mutation, deletions, mitotic recombination and/or chromosomal loss.
Many chromosomal regions have been implicated to contain tumour-suppressor genes and are thought to be involved in ovarian tumour progression when analysed by a variety of approaches.

Cytogenetic abnormalities
In most solid tumours, cytogenetic abnormalities are complex and it is difficult to identify specific karyotypic changes which Genetic anlysis of woaan cancer AN Shelling et al 522 are consistently present for a particular type of cancer. The majority of epithelial ovarian cancers appear to be aneuploid and contain a vaanetv of structural chromosomal abnormalities. However. some non-random chromosomal alterations have been identified in ovarian cell lines and tumours. including chromosomes 1.3. 6.9.11.12.17.19 and X (Wake et al.. 1980: Whang-Peng et al.. 1984: Atkin and Baker. 1987: Jenkyn and McCartney. 1987: Sheer et al.. 1987: Smith et al.. 1987Pejovic et al.. 1989Pejovic et al.. . 1990Pejovic et al.. . 1991Pejovic et al.. . 1992Tanaka et al.. 1989: Bello and Rey. 1990: Roberts and Tattersall. 1990: Islam et al.. 1993: Thompson et al.. 1994. The cytogenetic data have allowed investigators to evaluate the role of some of these chromosomal alterations using more sensitive and precise methods. that is using highly polymorphic markers for linkage analysis of familial cancer and loss of heterozygosity studies in sporadic tumours.

Linkage
The majority of ovarian cancers are sporadic. but a predisposition to tumour development can be inherited as an autosomal dominant trait. Female members of ovarian cancer families may have a lifetime risk for ovarian cancer 2or 3-fold greater than the general female population. and are often found clustered with stomach. breast and colon cancer (Blackett and Sharp. 1994). Recently. a large international consortium has used polymorphic DNA markers to link more than 200 families with breast and ovarian cancer to a susceptibility gene at chromosome 17q21. known as BRCA1. leading to the recent identification of the BRCA1 gene (Miki et al.. 1994). The combined data have demonstrated that in almost all families with breast and ovarian cancer. and about half of those with only breast cancer. the disease can be linked to the BRCA1 gene (Black and Solomon. 1993). Loss of heterozygosity studies on tumours from patients within ovarian cancer families have also consistently shown chromosome 17 loss within the region which contains the wild-type BRCA 1 gene (Smith et al.. 1992). leaving the mutant BRCA1 gene on the remaining chromosome 17. suggesting that it is a tumour-suppressor gene. Overall. germline BRCA 1 mutations may account for as many as 10% of ovarian cancers (Blackett and Sharp. 1994). however high loss of heterozygosity in the BRCA 1 region of 60% in sporadic ovanran tumours suggests that somatic alterations in BRCA 1 (not observed by Futreal et al.. 1994) or a nearby gene may be important in a larger proportion of these cancers.  (33)  9p 49 157 (31) Data summan'sed from Eccles et al. (1990Eccles et al. ( . 1992b. Lee et al. (1990). Okamoto et al. (1991). Sato et al. (1991). Tsao et al. (1993). Viel et al. (1991Viel et al. ( . 1992. Zheng et al. (1991. 1993). CheneVix-Trench et al. (1992. Gallion et al. (1992). Jacobs et al. (1992. 1993). Jones and Nakamura (1992). Saito et al. (1992. 1993). horizontal line (33%) represents the average LOH (taken as total number of chromosome arms lost total number of tumours). The location of some known candidate genes is indicated.

Loss of heterozygositv
The search for loss of heterozygosity is now widely accepted as a means of identifying recessive genes involved in the aetiology of hereditary and sporadic tumours. Frequent allele loss at specific loci suggests that these loci may contain tumour-suppressor genes. Some authors  have suggested that loss of heterozygosity occurring more frequently than a baseline level of 35% is more likely to represent important. potentially causative, genetic events than a secondary phenomenon associated with generalised genomic instability. Some loss of heterozygosity studies have shown quite variable results, making it often difficult to identify clearly regions of interest. These differences may be due to insufficient numbers of tumours being tested, uninformative loci on the particular chromosome arm being tested or the inability of the researcher to dissect tumour material away from normal tissue. Other causes may be more significant. such as inherent genetic differences in the study population or differences in the tumour subtype. stage. grade or incidence of prior treatment in the tumour being evaluated. In an attempt to adjust for some of these variables. results of chromosome arm loss from a number of loss of heterozygosity studies have been pooled (Table I and Figure 1). An attempt has been made to avoid duplicating data from different studies and, where possible. only results from malignant tumours have been included. This approach may not be totally valid, as it would not expose all potential tumour-suppressor genes mutated in more subtle ways. such as by small deletions or point mutations, however it does provided a useful indicator of generalised allele loss. This is especially significant in ovarian cancer. in which loss of heterozygosity for a single marker may frequently equate with loss of heterozygosity of all informative markers on a chromosome arm (Foulkes et al.. 1993a). Similar regions of allele losses are seen in a variety of solid tumours. for example 17p is lost not only in ovarian cancer (63%). but also in osteosarcoma (71%). non-small-cell lung cancer (62%). oesophageal (62%). breast (61%) and hepatocellular (54%) cancer (Yamaguchi et al.. 1992;Tsuchiya et al.. 1992: Aoki et al.. 1994Devilee et al.. 1991: Fujimori et al.. 1991 respectively. Several chromosomal regions identified as containing potential tumour-suppressor genes implicated in ovarian cancer are discussed in detail below.

Chromosome 6
Allelic losses of up to 50% involving 6q have been frequently reported (38%. Table I). Several studies have shown by either loss of heterozygosity (Sato et al.. 1991Saito et al.. 1993 or cytogenetic abnormalities (Wake et al., 1980). that these changes occur more frequently in serous adenocarcinomas, implying that 6q may be important in the pathogenesis of the more common serous adenocarcinomas (Sato et al.. 1991). Evidence for a critical region on chromosome 6 (6q26-6q27) has been provided by allele loss studies using cosmids derived from chromosome 6 (Saito et al., 1992) on a panel of ovarian tumours. Two cosmids delineated the region of minimal loss in the tumour from one patient to chromosome 6q27. The potential distance between the two cosmids has been estimated to be 2 megabases based on the CEPH genetic map (Saito et al., 1992). Using cosmids mapped to chromosome 6q by Nakamura and co-workers (Saito et al.. 1992), six cell lines have been studied in detail using fluorescence in situ hybridisation (Lastowska et al.. 1994). Three of the six cell lines show abnormalities in this region, which suggests that a gene (or genes) localised to 6q26-27, and also a region proximal to 6q24. may play a role in the development of ovarian cancer. Recently, Wan et al. (1994) have identified three regions on chromosome 6 which show increased levels of allele loss: at 6q27, at a more proximal site (6q21-25) and at a region on the short arm that includes the WAF-1 /Cip-l gene (6p21).
Chromosome transfection studies have shown that chromosome 6. and especially 6q, contains gene(s) that cause senescence (Hubbard-Smith et al., 1992;Gualandi et al.. 1994;Sandhu et al.. 1994) and/or reverse the tumorigenic or metastatic features of various tumour cell lines (Trent et al.. 1990;Yamada et al., 1990;Negrini et al., 1994;Welch et al.. 1994). It remains to be seen whether these regions contain a gene or genes involved in ovarian cancer.

Chromosome 11
In epithelial ovarian cancer, loss of heterozygosity of 33% on lIp has been reported (Table I). This may be a late event in tumour progression (Vandamme et al., 1992). The important sites of deletion have been mapped to llpl3 between loci Dl1S16 and catalase, corresponding to the position of the Wilms tumour gene (W71), although no abnormalities in the WT1 gene have been found (Viel et al., 1994), and to llpl5.5, telomeric to the P-globin gene (Vandamme et al., 1992;Viel et al., 1992). In some tumours there was concomitant deletion in both regions, suggesting that they may act synergistically. Recently, it has been shown that introduction of normal human chromosome 11 altered the trans-formed phenotype of an ovarian cell line (Cao et al.. 1993). Foulkes et al. (1 993b) analysed 11 q in response to the numerous cytogenetic abnormalities including translocations and deletions involving 1 lql3-qter in epithelial ovarian cancer. They found a minimal region of loss at 1 Iq23.3-qter. thus suggesting that there may be a third tumour-suppressor gene on chromosome 11.

Chromosome 13
The overall loss of heterozygosity of chromosome 13q alleles is 41% (Table I). Initially the retinoblastoma (RB) gene locus  was a candidate tumour-suppressor gene for ovarian cancer, however inactivation of the RB gene leading to abnormal RB protein expression is extremely rare (Dodson et al., 1994;. This would suggest that another tumour-suppressor gene(s) other than RB must be involved on chromosome 13 in the progression of ovarian cancer. Recently, a gene predisposing for familial breast cancer, BRCA2. has been mapped to 13ql2-13 (Wooster et al.. 1994). Loss of chromosome 13 appears to be specific for high-grade tumours (Kim et al.. 1994). which suggests that allelic loss of 13 either causes or occurs soon after the development of invasive or metastatic abilities.
p53 Mutations in the p53 tumour-suppressor gene. which is located on 17p, occur in up to 50% of all human cancers. and are found in both inherited and sporadic tumours. Two biochemical features are clearly important in the normal role of p53 for growth suppression. First, p53 binds to and thereby suppresses various transcription factors, including those that bind to TATA elements and, second, it transcriptionally activates the expression of a number of genes which encode proteins that can suppress cell division. Tumour data have shown two types of mutational events in p53 are required to cause a phenotypic effect on cell growth. First, the loss of the wild-type allele, which is frequently observed when high loss of heterozygosity is seen on chromosome 17p (p53 is located on 17pl3.l). Second, many studies have shown a high frequency of mutations in p53 (546/1125; 49%) (Table II). Point mutations within the p53 gene frequently t~~~~~~~~~~~~~~~~S -e 524 cause conformational changes w-hich stabilhse and extend the half-life of the mutant p53 proteins. causina them to accumulate in the nucleus and allowxing them to be detected immunohistochemicallx. ser-ing as a rapid and effectix-e means of screening for p53 mutations. Clearl. p53 mutation is not a common feature of benign (O 122 tumours) or borderline tumours (2 866 2'0( (Table II>. Furthermore. p53 mutations appear to be less common in localised tumours. occurring in 105 284 (3-'~o stage I and II tumours as compared w-ith 351 608 (580o( late-stage tumours (stages III and IV) (Table III). This would suggest that pS3 mutations occur as a later ex-ent in tumour progression. Although p53 ox-erexpression occurs more frequentlxin late-stage tumours. oN-erexpression has not been showxn to has-e a correlation w-ith surx-ixal (Hartmann et al.. 1994).
As in other tumours. the analysis of the spectrum of mutations in the p53 gene may prox-ide information about the origins of the mutations that gix-e rise to the tumours. Prex-ious studies (Hollstein et al.. 1991 hax-e shown that 98°o of mutations fall in exons 5 -. w-hich are highlxex-olutionanlx conserx-ed. In the analysis of ox-an'an cancer (Kohler et  follow-ing o-ulation. is a fax-ourable molecular mechanism to explain Fathalla's hxpothesis  since no enx-ironmental carcinogens hax-e been conx-incinglIassociated w-ith ox-arian cancer.
Other chromosome 1-tiumrour-suppressor genes Allelic loss on lVq may relon the loss of txwo or more genes. The familial ox-arian breast cancer locus (BRCA4 1 ) on chromosome IVq2l is a likely candidate. howex-er. it does not appear to be important in sporadic cancer (Futreal et al.. 1994(. Sex-eral inx-estiaators hax-e found loss at more distal Iq regions to the BRC.41 gene (Eccles et al.. 1990: Russell et al.. 1990: Foulkes et al.. 1991 1993 . It appears that 1 7q loss occurs before 1 I7p loss. as loss of heteroz-xgosity at 1 q has been reported in benign and borderline oxarian tumours (Russell et al.. 1990: Gallion er al.. 1992. MlanNstudies hax-e shoxwn that a great majority of ox-arian tumours hax-e probably lost one copx of an entire chromosome 1-. thus deleting p53. BRC.1I and other potential tumour-suppressor genes in a single ex-ent. In most cases. the loss appears to in-olx-e the w-hole chromosome. probably due to non-d-sjunction. xwith or wxithout reduplication (Foulkes et al.. 1993a(. Chromosome 1J? The DCC locus (deleted in colon cancer) on chromosome 18 appeared to be a good candidate gene for ox-arian cancer.
particularly as both colon and oxarian carcinomas arise from normal epithelia. xwhich suggests that similar genetic events may be required. Overall. 420o of tumours show-ed loss of heterozygositv on 18q (Table I). w-hereas 18p only show-ed 140o loss. The DCC locus and alleles surrounding it have been analvsed in detail . High loss of heteroz-zosity wvas found at one or more loci in approximately 60°o of the tumours studied. and tended to occur more frequently in advanced staze tumours. The smallest region of overlap of allele loss unexpectedly did not include the DCC locus. This suggests that another locus exists on 18q near the DCC gene.

Chromosome X'
As oxvarian cancer is a female cancer. there might be a specific role for the X chromosome. Ox-erall. both Xp (380o) and Xq (2900o have a high lexel of loss of heterozxvositV (Table I>. This appears to be highest around the OTC locus (Xp1l.l) (530o: 9 17) (Yang-Feng ei al.. 1992). Loss of heterozxgosity on Xp maybe specific for ovarian cancer (Yani-Feng et al.. 1993). howxexer other tumours have not yet been tested with X and chromosome markers. Cvtozenetic analysis of the X chromosome in oxarian patients frequently identifies the loss of the X chromosome often at quite high lexels. for example Tanaka et al. (1989( found loss of in 8 9 oxarian carcinomas. It has been suggested that loss of X may be a primary or early ev-ent in ovarian tumour dexelopment (Thompson et al.. 1994(. In addition to allele loss. the selectixe inactix-ation of X chromosome genes by hy permethylation may contribute to the inactivation of a tumour-suppressor gene. how-ever this form of allele inactixation is thought to be a secondarx exent in tumour progression (Laird and Jaenisch. 1994.

Conclusion
The positional cloning of putatix-e tumour-suppressor genes identified from allele loss studies *-ill lay the foundation for a better understanding of the pathogenesis of ox-arian cancer. The identification of BRC41 and BRC.4' u-ould be of direct clinical benefit to probands in breastoarian cancer families. The isolation and characterisation of oncogenes and tumour-suppressor genes has sexveral clinical applications. First. persons at high risk of oxarian cancer (such as oxarian cancer families ) can be screened by molecular approaches and offered prophylactic oophorectomy if they carry the defectixe gene. Second. it is also conceixable that such genes or their products may be the basis of a general screening approach for ox-arian cancer. Diaanosis could be made relatixely simply by the identification of mutant gene products in the blood. or by the detection of antibodies made by the patient against the mutant gene product. Third. neu-er therapeutic approaches designed to inactix ate mutant gene products (e.g. c-erbB-2' or mimic or restore the normal biological function of genes like p53 will be possible. Finally.
gene therapy x-ould be an appealingz xvay to restore function in patients x-ho hax-e oxarian cancer once it is possible to  BRCA1 (Hosking et al.. 1995: Merajver et al.. 1995.