Heritable Factors In Gynecologic Cancer

Ovarian cancer encompasses ovarian epithelial neoplasia, derived from the germinal epithelium located on the surface of the ovary; germ cell neoplasia; and stromal neoplasia. Germ cell tumors are relatively more common during childhood. By the sixth decade of life, epithelial ovarian tumors are far more frequent⁷; their mean age at occurrence is about 56 years.

Ovarian epithelial cancers include tumors with serous, mucinous, clear cell, and endometrioid cell types, as well as mixed lesions. Although the incidence of ovarian epithelial cancer is lower than that of either cervical or endometrial cancer, the frequency of deaths is higher. The overall 5-year disease-free survival rate of 20% to 35% is largely a reflection of the difficulty in diagnosing the disease at an early stage; approximately 70% of epithelial ovarian carcinomas are in stages II to IV at diagnosis.⁸

Many of the established risk factors for epithelial ovarian carcinoma are related to the general phenomenon of “incessant ovulation,” including nulliparity, increased number of uninterrupted ovulations, early menarche, and late menopause.⁹ Oral contraceptive use and multiparity seem to lessen the risk.¹⁰ However, the published data for such risk factors is inconclusive and controversial. It is suggested that only a small fraction of ovarian cancer risk is attributable to these established risk factors.

Incidence rates for ovarian cancer are highest in the United States and other western nations, suggesting an importance of environmental and lifestyle factors (e.g. diet) in the etiology of this tumor. This is supported by the observation that Japanese women residing in the United States show a prevalence approximating that of other women in the United States, whereas the incidence of ovarian epithelial cancer is lower in the female population of Japan and other Asian countries than in white women in the United States.¹¹

A family history of ovarian cancer confers the greatest risk of all known factors, other than age, for developing the disease.^12,13,14 Consistent with this observation are estimates that at least 5% to 10% of all epithelial ovarian carcinoma cases result from hereditary predisposition,^15,16,17,18 with the germline inheritance of a mutant gene conferring autosomal dominant susceptibility with high penetrance.^{13,16,19,20,21}

There are two discrete manifestations of hereditary ovarian carcinoma: the breast and ovarian cancer syndrome and the hereditary nonpolyposis colorectal cancer (HNPCC) syndrome.^20,22,23,24 Genetic linkage analysis reveals that the majority of breast and ovarian cancer families are linked to the BRCA1 gene; however, some cases of hereditary ovarian cancer are linked to the BRCA2 gene. The majority of HNPCC families are linked to one of four genes encoding a family of DNA mismatch repair (MMR) proteins.^24A–24D

Breast and ovarian cancer syndrome accounts for 65% to 75% of all hereditary ovarian cancer cases.^17,19,22 The genetic relationship of these two malignancies in a hereditary context was first demonstrated in population-based, case-control epidemiologic studies.^18,25 Later, BRCA1 was demonstrated to be responsible for these hereditary malignancies.^5,26 The breast and ovarian cancer families not linked to BRCA1 are linked to the newly described BRCA2 locus on chromosome 13q12–13, especially those with cases of male breast cancer.^6,27,28

HNPCC, known previously as cancer family syndrome and Lynch syndrome II, is an autosomal dominant genetic syndrome characterized by three or more first-degree relatives with colon or endometrial cancer, at least two of whom are diagnosed with colon cancer at age 50 years or younger.²⁹ HNPCC accounts for approximately 10% to 15% of all hereditary ovarian cancer cases.²² The risk of ovarian cancer in these families is increased 3.5-fold over that expected in the general population.²⁴ In addition to cancers of the colon, endometrium, and ovary, HNPCC family members are at increased risk for cancers of other gastrointestinal sites and the upper urologic tract.²⁴ Significant diversity in ovarian cancer frequency is seen among HNPCC families, suggesting genetic heterogeneity.²⁴ Consistent with this observation are linkage data indicative of at least four genetic loci that contribute to the HNPCC phenotype.^24A–24D The recent cloning and characterization of the genes responsible for HNPCC have provided significant insights into the etiology of HNPCC-associated tumors and the potential for genetic screening for this disorder.^30,31

BRCA1

The BRCA1 gene consists of 22 coding exons distributed over approximately 100 kb of genomic DNA on chromosome 17q21.⁵ The 7.8-kb mRNA transcript is expressed most abundantly in the testis and thymus and, at lower levels, in the breast and ovary. Young women with germline alterations in BRCA1 develop breast cancer at rates 100-fold higher than that in the general population, and BRCA1-null mice die before day 8 of development. Functional studies support the classification of BRCA1 as a tumor-suppressor gene, because retroviral transfer of wild-type BRCA1 suppresses ovarian and breast carcinoma cell growth in vitro and tumorigenicity in vivo.³² Disruption of the Brca1 gene in mouse mammary epithelial cells causes increased apoptosis and abnormal ductal development. Mammary tumor formation occurs after long latency and is associated with genetic instability characterized by aneuploidy, chromosomal rearrangements and triggers further alterations, including the inactivation of p53, that lead to tumor formation.^32A The BRCA1 protein localizes in discrete nuclear foci in a variety of epithelial cell lines, including those derived from breast malignancies. Immunohistochemical staining of human breast specimens also revealed BRCA1 nuclear foci in benign breast, invasive lobular cancers and low-grade ductal carcinomas. Conversely, BRCA1 expression was reduced or undetectable in the majority of high-grade, ductal carcinomas, suggesting that absence of BRCA1 may contribute to the pathogenesis of a significant percentage of sporadic breast cancers.^32B The nuclear protein BRCA1 is a coactivator of the P53 protein³³ and has the properties of a transcription factor. It can interact with the recombination and repair protein RAD50.³⁴ It also plays an important role in cellular responses to DNA damage that are mediated by the hRad50-hMre11-p95 complex.^34A The expression pattern of BRCA1 during mouse development indicates that the gene is expressed in rapidly proliferating tissues undergoing differentiation, including the mammary gland, and that expression is regulated by ovarian steroid hormones.³⁵ Bidirectional promoters between as well as within the exons of both genes are critical for the optimal and coordinated expression of both genes. BRCA1 has been found to share a bidirectional promoter with a newly cloned LAI3B gene indicating its involvement with gene regulation.³⁶ The mechanisms of BRCA1-mediated growth regulation and tumor suppression remain unknown. Recently, one mechanism by which BRCA1 contributes to cell cycle arrest and growth suppression has been demonstrated to be through the induction of P21.³⁴ Mutational inactivation of BRCA1 would thus be expected to affect the expression of other genes, involved presumably in the regulation of growth or differentiation in breast and ovarian epithelium. It has also been suggested that alternative mechanisms, such as disruption of transcription, may also be involved in the suppression of BRCA1 gene expression/function in breast cancers.³⁷ A 36-bp region was determined to be important for the positive regulation of BRCA1 transcription utilizing systematic promoter deletions and transient transfection assays. Deletion of this positive regulatory region resulted in a significant loss of promoter activity. Disruption of the DNA-protein complexes could affect normal BRCA1 transcription and may contribute to breast and ovarian cancer susceptibility.³⁷ Recently BRCA1 was found to inhibit signaling by the ligand-activated estrogen receptor (ER-alpha) through the estrogen-responsive enhancer element and to block the transcriptional activation function AF-2 of ER-alpha. These results raise the possibility that wild-type BRCA1 suppresses estrogen-dependent transcriptional pathways related to mammary epithelial cell proliferation and that loss of this ability contributes to tumorigenesis. This may partly explain why mutations of the BRCA1 gene confer increased risk for breast, ovarian, and prostate cancers.^37A

Mutations of BRCA1 are located throughout the gene with no evidence of clustering. Approximately one third of the distinct mutations are recurrent with three mutations found relatively frequently.³⁸ About 80% of the mutations are loss-of-function nonsense or frameshift alterations. Other studies showed that allelic deletions affecting the 17q21 region in BRCA1-linked breast or ovarian cancers involve the wild-type chromosome,^39,40 as would be expected if BRCA1 were behaving as a typical tumor-suppressor gene. It is important to understand how specific mutations of BRCA1 may preferentially contribute to breast or ovarian cancer and how other genetic and/or environmental factors may affect penetrance and tissue specificity.

The data linking BRCA1 to most hereditary ovarian cancers are now unequivocal. It has been thought that this gene would be found to play a major role in sporadic ovarian carcinomas as well. This speculation based on the observation that loss of heterozygosity, the genetic hallmark of tumor-suppressor gene inactivation, is observed on chromosome 17q in up to 80% of sporadic ovarian carcinomas.^41,42,43 However, somatic mutations of BRCA1 are rare.^{26,44,45,46,47} The questions whether somatic mutation of the P53 gene is required for BRCA-linked ovarian tumorigenesis and whether the spectrum of additional somatic molecular genetic alterations present in these tumors differs from that known to exist in sporadic ovarian cancers were recently studied in forty tumors, 29 linked to BRCA1 and 11 linked to BRCA2, for mutational alterations in P53, KRAS, ERBB2, CMYC, and AKT2.⁴⁸ The presence of a P53 mutation in 80% of these cancers indicates that P53 mutation is common but not required for BRCA-linked ovarian tumorigenesis; however, a significantly higher proportion of the P53 mutations in BRCA2-linked cancers were deletions or insertions compared with the more typical spectrum of missense mutations seen in BRCA1-linked cancers. Additionally, BRCA-linked ovarian carcinomas seem to develop through a unique pathway of tumorigenesis that does not involve mutation of KRAS or amplification of ERBB2, CMYC, or AKT2.⁴⁸ Fine deletion mapping studies implicate that one and perhaps two regions distal to BRCA1 may harbor an additional tumor-suppressor gene or genes.^49,50,51 Although further studies are required to fully determine the frequency of somatic BRCA1 mutations in sporadic ovarian cancers compared with the frequency of germline mutations in hereditary tumors, the data published to date indicate that mutational inactivation of BRCA1 is necessary at a relatively early stage of embryologic development to be able to contribute to ovarian tumorigenesis. If that was true, somatic mutation of BRCA1 in the adult ovarian epithelium does not provide a significant selective advantage in a developing malignancy leading to clonal selection and manifestation as a somatic mutation in the ovarian cancer.

BRCA2

The BRCA2 gene consists of 26 coding exons distributed over approximately 70 kb of genomic DNA, encoding a transcript of 11 to 12 kb.^6,27 As for BRCA1, BRCA2 mRNA is most expressed in testis and thymus, with lower levels in breast and ovary.^6,27 BRCA1 and BRCA2 share a number of additional structural and functional similarities: unusually large genes in terms of the number of exons and size of the encoded message, a large exon 11 which contains approximately half of the entire coding region, translation start sites in exon 2, relatively A-T rich, and most remarkably, the only sequence similarities noted in database searches are to each other, specifically in the region containing a consensus granin motif suggesting that both genes may encode secreted proteins.⁵² Expressions of both BRCA2 and BRCA1 during the progression of the cell cycle are very similar.⁵³ BRCA2 RNA was low in G₀ and early G₁ phases, and was then upregulated at the G₁/S phase junction. It was maintained at relatively high levels when cells progressed through S and G₂/M phases. The level of BRCA2 transcript decreased as cells were released from nocodazole-mediated metaphase arrest.⁵³ These results suggest important functions for both BRCA2 and BRCA1 in regulation of cell growth.

Mutations of BRCA2 are dispersed throughout the gene with no evidence for hot spots.^{6,27,54,55,56,57,58} The majority of mutations are frameshift, with microdeletions being most common; microinsertions, nonsense, and missense mutations are rare. Loss of heterozygosity at the BRCA2 locus in tumors from linked individuals always involves the wild-type allele, consistent with the classification of BRCA2 as a tumor-suppressor gene.^59,60 As for BRCA1, it was believed that somatic mutations of BRCA2 might be involved in a significant fraction of sporadic ovarian cancers, based on the high frequency of allelic loss on chromosome 13q observed in these tumors.^61,62 Although the analysis of a large sample of unselected ovarian cancers indicates that loss of heterozygosity of the BRCA2 locus is seen in more than half of the tumors, somatic mutations of the gene are rare.⁵⁸ Future studies on the function and expression of BRCA1 and BRCA2 should provide insights into the mechanism through which germline but not somatic mutations in these genes contribute to ovarian tumorigenesis.

Mismatch Repair Genes

The HNPCC syndrome arises from an inherited defect in any one of four known genes: hMSH2 (chromosome 2p), hMLHI (chromosome 3p), hPMSI (chromosome 2q), or hPMS2 (chromosome 7p).^{63,64,65,66,67} The proteins encoded by these genes participate in the same DNA MMR pathway, and loss-of-function mutations are associated with genetic instability in the tumors of affected family members.^68,69,70 Although mutations have been described in all four genes, the majority of HNPCC appears to be linked to either hMSH2 or hMLHI with greater than 90% of all reported mutations affecting one of these two genes. Direct evidence for the association of genetic instability and mutated MMR genes is derived from the following observations: nuclear extracts from human tumor cell lines harboring mutations in one or another of these genes are unable to efficiently repair heteroduplex DNA fragments when compared with extracts from normal cell lines^71,72; purified hMLHI and hPMS2 proteins are able to restore mismatch repair activity to defective human cells⁷³; and transfer of a normal human chromosome 3 into tumor cells with mutant hMLHI restores microsatellite stability.⁷⁴

The defective MMR gene is most readily observed through somatic length alterations in simple repeat sequences (e.g. CA) which are located throughout the genome and are known as micro-sat-ellites. Replication errors in these repeat sequences are common, and their inefficient repair results in the microsatellite instability phenotype. A large number of studies have documented microsatellite instability in many sporadic tumor types, including HNPCC syndrome.^75,76 Although mutations of the MMR genes have been readily identified in numerous HNPCC kindreds,^{63,65,66,67,77,78} MMR gene mutations in sporadic tumors with the microsatellite instability phenotype are rarely detected.^79,80 It is likely that this type of genetic instability arises in sporadic tumors through a molecular genetic pathway distinct from that in HNPCC-associated cancers.

It is not clear how microsatellite instability contributes to tumorigenesis in the ovary or in other organs affected by the HNPCC syndrome. Microsatellites exist throughout the genome in predominantly noncoding regions of DNA. Simple repeat sequences are known to occur in the coding regions of genes, however, their somatic mutation may result in loss of function for genes critical to the regulation of proliferation, invasion, or metastasis. Alternatively, microsatellite instability may affect global patterns of gene expression through alterations in chromatin architecture. The organization of DNA within the nucleus is cell and tissue specific, and the interaction of DNA with the nuclear matrix forms fixed organizing sites for several nuclear functions including gene transcription.⁸³ Matrix attachment regions of DNA are known to be characterized by repetitive elements.⁸⁴ Thus, mutations affecting repeat length may lead to changes in chromatin structure and gene transcription. However, more evidence is needed to support this hypothesis.

The cloning and characterization of genes responsible for hereditary predisposition to ovarian cancer have profound implications for the presymptomatic identification of susceptible individuals and families, the clinical management of these patients, and ultimately, the saving of lives from a generally fatal disease. These technological advances are accompanied by an equally significant body of scientific, ethical, legal, and psychosocial questions. Although most published consensus opinions have generally urged caution to one degree or another in the widespread implementation of genetic screening for cancer predisposition, it is evident that tests for mutations in some of the genes discussed in this article have recently become commercially available and that commercial availability will become more widespread in the very near future. Therefore, some issues regarding mutation detection, risk assessment and clinical management need to be addressed.

Few scientific or medical advances have been associated with implications as complex and controversial as that of genetic testing for cancer predisposition. The major concern are those related to scientific, medical, and technical matters such as the reliability of genetic screening tests, the interpretability and predictive value of positive test results, and the clinical ability to prevent cancers in presymptomatic individuals who test positive. The more detailed discussions of concerns relating to genetic testing for cancer predisposition in general,^85,86 genetic counseling for HNPCC,^87,88 genetic counseling for breast and ovarian cancer,^89,90 BRCA1 testing in families with hereditary breast and ovarian cancer,⁹¹ the behavioral and psychosocial effects of BRCA1 testing,⁹² genetic testing in minors,⁹³ and insurance issues related to genetic testing⁹⁴ are provided by recently published reviews. Even though there are a lot of different options about the genetic test, one issue is common and is clear: it is critical that we create safeguards to ensure that the benefits of testing exceed the risks.

Mutation analysis of the genes involved in hereditary ovarian cancer is technically arduous. For example, in the individuals from breast and ovarian cancer families and from site-specific ovarian cancer families, the BRCA1 and BRCA2 genes with enormous coding regions of 5.7 and 10.2 kb, respectively, must be examined. To screen individuals from HNPCC families, up to four genes with coding regions ranging in size from 2.2 to 2.8 kb must be tested. Although several recurrent mutations have been observed for BRCA1, BRCA2, hMSH2, and hMLHI, the spectrum of reported mutations for each of these genes generally spans the entire coding region. Genetic linkage analysis may be used to indicate which gene may be involved in a particular cancer family, but this procedure requires DNA samples from multiple family members in multiple generations, and the results are not always reliable.

Several screening procedures commonly used in research studies to search for evidence of mutations in a particular gene prior to sequence analysis are based on the amplification of gene or cDNA fragments by the polymerase chain reaction (PCR), followed by gel electrophoresis to visualize products with an altered conformation and gel mobility indicative of a mutation, such as single-strand conformation polymorphism (SSCP) analysis. However, none is completely reliable in terms of detecting all mutations present. Direct sequence analysis must be performed on altered products following screening by the above techniques to identify a particular mutation. The direct sequencing of entire coding regions with automated technology is the most straightforward and reliable procedure for mutation analysis of these genes, but is technically and economically prohibitive for all but a small number of centers. Thus, all of the currently available screening procedures are less than completely reliable for mutation detection, and even direct sequence analysis protocols do not typically include promoter and intronic regions which may contain functionally significant mutations. However, after a mutation has been detected in an individual, the identification of other affected family members is relatively simple, inexpensive, and accurate, using either the direct sequencing of a single PCR product containing the mutation or a sequence-specific screening procedure such as allele-specific oligonucleotide hybridization. In this circumstance, other family members may be categorized as positive or negative for a specific mutation with high accuracy.

The translation of knowledge regarding a specific mutation into clinically useful information is currently one of the most formidable obstacles in cancer genetic testing. A negative test result has little predictive value for an individual from a family with no mutation. Although most of the mutations described for the BRCA1, BRCA2, and MMR genes are truncating in nature and may be reasonably predicted to have an inactivating effect on the encoded proteins without guarantee, missense and other noncoding mutations must be examined for biologic relevance and no straightforward interpretation can be made. The penetrance of BRCA1 mutations for either breast or ovarian cancer is estimated at about 95%,^95,96 with a cumulative risk of about 63% for ovarian cancer in particular by age 70 years. The penetrance of mutant BRCA2 for ovarian cancer is much lower than that of BRCA1. The penetrance of MMR gene mutations for cancer at any site is estimated at about 90%^24,29 with a particularly lower penetrance for ovarian cancer. Little is known regarding factors responsible for the variable penetrance and tissue-specific expressivity of mutations in any of these genes.

A related issue is the prevalence of specific mutations in particular populations, the presence of which facilitates genetic screening and the accumulation of data bearing on risk assessment. The current estimate of BRCA1 mutation carriers in the general population is approximately 1 in 800 women; for BRCA2 mutations, the risk for ovarian cancer is probably too low to have a measurable effect on general population-based prevalence studies.⁹⁷ Specific mutations in both of these genes have been found to occur in significantly higher frequency in particular populations: frameshift mutation in BRCA1, 185delAG, appears to occur in approximately 1% of the Ashkenazi Jewish population and in about 20% of Jewish women with early onset breast cancer^98,99; a frameshift mutation in BRCA2, 6174delT, is seen in about 8% of Ashkenazi Jewish women with early onset breast cancer and about 1% in this entire ethnic population.^100,101 The BRCA1 185delAG mutation may account for 35% to 40% of early onset ovarian cancer in Jewish women,^97,102 whereas few data yet exist on the possible ovarian cancer risk conferred by BRCA2 6174delT. Other so-called “founder mutations” have been reported in several populations, including recurrent BRCA1 mutations in Canadian¹⁰³ and Swedish¹⁰⁴ breast and ovarian cancer families and a common BRCA2 mutation in Icelandic cancer families.⁵⁵

Although the genes involved in hereditary ovarian cancers have only recently been identified, the prior recognition of distinct hereditary ovarian cancer syndromes with characteristic clinical and histopathologic manifestations was observed. Mucinous tumors are exceedingly underrepresented among all forms of hereditary ovarian carcinoma.^{15,17,19,22,105,106} The prevalence of serous histology in ovarian cancers associated with germline BRCA1 mutations has been documented through linkage analysis. Thirty-four of 36 BRCA1-linked patients were of serous histology¹⁰⁶ in one study, and 43 of 53 ovarian cancers from individuals with germline BRCA1 mutations had serous histology in another.¹⁰⁷ Borderline ovarian tumors also appear to be underrepresented among cases of hereditary ovarian cancer.^{14,17,19,22,107}

Data regarding age of onset and clinical outcome at hereditary ovarian cancer also indicate a distinct underlying biology. The average age of onset for hereditary ovarian cancer cases is approximately 10 years younger than that of the general population.^15,20,21,22 In five studies regarding age data for hereditary ovarian cancer cases associated with germline BRCA1 mutations, the average age at diagnosis for a combined total of 144 patients was significantly lower than an average age of 60 years for the diagnosis of epithelial ovarian carcinoma in the general population.^{102,107,108,109,110}

There is no confirmed evidence that patients with “familial” ovarian cancer exhibit a higher 5-year survival than control cases.^22,111 Interestingly, a study reported that patients with germline mutations of the BRCA1 gene, had a considerably more favorable clinical course compared with sporadic ovarian cancer patients matched for age, stage, grade, and histologic subtype.¹⁰⁷ A similar trend of longer survival has been observed in familial breast cancer patients with linkage to the BRCA1 locus compared with matched control cases.¹¹² More studies are necessary to expand on and confirm this observation. With the accumulation of mutation data the ability to correlate clinical and histopathologic features with hereditary ovarian cancers should improve dramatically.

Clinical Management

It is clear that development of the most effective guidelines for the clinical management of patients with an inherited predisposition to ovarian cancer will require further research, especially regarding the ability to translate genetic status into clinical risk because of the previously mentioned complexities. It is well accepted that clinical management will require the interaction of educated physicians and genetic counselors to develop strategies most appropriate for the unique needs and circumstances of individuals and families, based on rapidly evolving research developments.

Although specific risk figures concerning the likelihood of developing ovarian malignancy cannot be provided, available data indicate that prophylactic oophorectomy should be considered if multiple family members (e.g. siblings, mother) develop serous adenocarcinoma of the ovary. Indeed, hyperplastic changes have been observed on the surface (epithelium) of ovaries removed solely because of a family history of ovarian neoplasia. Multiple papillary projections have been described on ovaries of an asymptomatic woman who showed abnormal cells in a cul-de-sac aspirate.¹¹³ Similar changes were observed in three of eight women undergoing prophylactic oophorectomy.¹¹⁴ However, such findings were not confirmed by others.¹¹⁵

The issue of the effectiveness of prophylactic surgery in preventing cancer is probably of major concern. For ovarian cancer risk, the major concern is the development of primary peritoneal carcinoma indistinguishable from primary ovarian carcinoma, subsequent to prophylactic oophorectomy. It was reported that in 16 families with strong histories of ovarian cancer, 3 of 28 women who had undergone prophylactic oophorectomy developed peritoneal carcinomatosis 1, 5, and 11 years later.¹¹⁶ Two additional case reports of this phenomenon were subsequently published.^21,117 Based on these case reports, Piver and colleagues examined the Gilda Radner Familial Ovarian Cancer Registry over a 10-year period and identified 931 families containing 2221 cases of familial ovarian cancer.¹¹⁸ Of 324 women in these families who underwent prophylactic oophorectomy, 6 (1.8%) developed peritoneal carcinoma indistinguishable from ovarian carcinoma 1 to 27 years after prophylactic surgery. Thus, oophorectomy does not necessarily abolish totally the risk of “ovarian cancer.”¹¹⁹

Although it is clear that there is a risk of developing peritoneal cancer following prophylactic oophorectomy in women with genetic predisposition to ovarian cancer, it is also certain that cancer may be prevented by prophylactic surgery in a significant fraction of these women. The fact is that the lifetime risk of ovarian cancer for a BRCA1 mutation carrier exceeds 60%. More controlled studies are needed to allow a more precise evaluation of the risk reduction by this procedure. Prophylactic oophorectomy has long been a relatively well-accepted standard of care for women with a family history of ovarian cancer. Forty-four percent of British gynecologists would perform prophylactic oophorectomy in women with a strong family history of ovarian cancer.¹²⁰ There is no doubt that oophorectomy will have a great impact on the quality of life of those women if that procedure is performed at the wrong time. In addition, the general clinical course and prognosis of so-called “familial” ovarian cancer are at least as good as control cases if not better. So oophorectomy should not be performed in those women at high risk too early. Currently, there are no accurate data available regarding the timing of this procedure; however, a time point 10 years before the youngest age at which onset of disease was observed in the family.

Beyond routine gynecologic examination, the most conservative clinical intervention for asymptomatic women at increased risk for ovarian cancer would involve measurement of serum CA125 levels and/or transabdominal or transvaginal sonography. These procedures have been tested for their utility in detecting early stage ovarian cancers in the general population and are generally considered ineffective as screening modalities.^{121,122,123,124} One might also consider the prophylactic use of oral contraceptives for any time where pregnancy is not desired.

Clinical management of high-risk members of HNPCC families is even more difficult than that encountered with patients at risk primarily for ovarian cancer or breast and ovarian cancer, as a result of the numerous cancers for which these patients are at risk. The recommendations of the International Collaborative Group on HNPCC include the option of prophylactic hysterectomy and bilateral salpingo-oophorectomy in women who are undergoing subtotal colectomy following the diagnosis of colon cancer or in asymptomatic, high-risk members (i.e. MMR gene mutation carriers) of HNPCC families with severe cancer phobias.^87,88 Otherwise, surveillance guidelines for high-risk individuals include, in addition to colonoscopy procedures, endometrial biopsy, transvaginal ultrasound, Doppler color blood flow imaging of the ovaries, and serum CA125 measurements annually after the age of 30 years. Formal recommendations by a collaborative or consensus group have not been made with regard to surveillance procedures for members of BRCA1 or BRCA2 families. The ultrasound imaging and serum tests would also seem reasonable for women at high risk for ovarian cancer outside the context of HNPCC.

In females, germ cell neoplasia most often takes the form of benign cystic teratomas (dermoids). By definition, at least two of the three germ cell layers must be present. Malignant germ cell tumors include immature teratomas, endodermal sinus tumors, and dysgerminomas.

First- and second-degree relatives of probands with ovarian germ cell malignancies do not have an increased risk for similar tumors.²³³ In this study, 78 mothers, 87 sisters, 135 aunts, and 156 grandmothers of 78 presumptive 46,XX patients (ages ranging from newborn to 20 years) with malignant ovarian germ cell tumors, excluding cases of dysgerminoma and gonadoblastoma were reviewed. None had a malignant ovarian germ cell neoplasm or other malignant ovarian neoplasm.

That benign cystic teratomas are heritable is evident on the basis of frequent bilaterality (20%) and early age at onset. Both characteristics suggest genetic tendencies. Indeed, teratomas have been reported in twins, in nontwin siblings, in triplets, in a mother and her two adult daughters, and in three generations.^{125,126,127,128,129,130}

The pathogenesis of cystic teratomas involves parthenogenesis. That is, these tumors arise from a single germ cell that reproduces itself. Studies using polymorphic chromosomal variants initially suggested that the error was restricted to meiosis II¹³¹; however, more recent data suggest that the error may also involve meiosis I.¹³² Coupled with observations of inherited tendencies for teratomas, these data imply genetic control over meiosis.

No unusual characteristics have been uncovered in familial cases, but few data exist. Neither HLA studies nor comparative frequencies of parthenogenesis are available. Too few cases of familial teratomas exist to compare ages at onset between familial and nonfamilial cases.

Dysgerminoma has been reported in 46,XX individuals in two and perhaps three generations of a Jamaican kindred.¹³³ Familial dysgerminomas are otherwise not reported in 46,XX persons. However, the tumor is relatively common among individuals with 46,XY or 45,X/46,XY gonadal dysgenesis.

Ovarian neoplasms may arise from ovarian stromal cells—granulosa cells, Sertoli/Leydig (interstitial) cells, or fibrous elements (fibromas). All of these tumors are rare, but familial aggregates of each are reported. Epidemiologic studies specific for stromal tumors are not available, but several lines of evidence suggest that genetic factors may be involved.

Granulosa cell tumors occur in both young girls as well as in postmenopausal women, suggesting different pathogenetic mechanisms. Granulosa cell tumors are observed frequently in the Peutz-Jeghers syndrome, which is an autosomal dominant disorder characterized by circumoral melanin deposits and colonic polyposis. In one series, 16 of 115 patients with the Peutz-Jeghers syndrome had a coexisting ovarian tumor, typically a granulosa cell tumor.¹³⁴ On the other hand, familial aggregates in otherwise normal persons seem rare.

Reported familial aggregates of arrhenoblastoma involve siblings, a mother and daughter, and paternal cousins.^{135,136,137,138,139} Too few cases exist to permit comparisons between familial and nonfamilial cases. A syndrome of familial arrhenoblastomas and thyroid disease has been claimed by some^140,141 but doubted by others.¹⁴²

Ovarian fibromas are rare and usually not familial. An autosomal dominant gene appears to have caused ovarian fibromas in at least one family.¹⁴³ Ovarian fibromas are also associated with the basal cell nevus syndrome,¹⁴⁴ the Gorlin syndrome,¹⁴⁵ and ataxia telangiectasia.¹⁴⁶ Ovarian fibrosarcoma has also been reported in each of dizygotic twins.¹⁴⁷

Gestational trophoblastic diseases constitute a spectrum of interrelated disorders that include hydatidiform mole, invasive mole, and choriocarcinoma.^148,149 Trophoblastic neoplasias arise in association with pregnancy. Hydropic avascular villi, trophoblastic hyperplasia, and absence of a fetus characterize benign trophoblastic disease (hydatidiform mole). Hydatidiform moles per se are benign, but 3% develop into choriocarcinoma. In turn, half of choriocarcinomas follow from moles, one fourth follow a spontaneous abortion, and one fourth follow normal pregnancy.

The incidence of trophoblastic disease is 0.5% to 1.0% in North America but considerably higher in Taiwan, India, the Philippines, and elsewhere in Asia.^148,150 Because Asian women residing in the United States show incidences lower than in their native lands, genetic factors are not considered paramount. Nutritional or other nongenetic factors are assumed to be more relevant; currently, a relation with dietary vitamin A intake is a popular hypothesis.

Because gestational trophoblastic disease is usually successfully treated, the possibility of recurrence in the same patient arises. Indeed, the likelihood of gestational trophoblastic disease recurring in subsequent pregnancies in the same patient is increased over the background risk.^148,149,151 Pooled data indicate that the frequency of recurrent trophoblastic disease is 1.3% (67 of 5002 subjects in 18 reports).¹⁵¹ The actual risk is probably lower because some patients in the 18 reports were surely referred to university centers because of repetitive neoplasia. For example, only 3 of 449 (0.67%) women treated for trophoblastic neoplasia at Northwestern University manifested a subsequent tumor.¹⁵² After two gestational trophoblastic disease events, the recurrence risk rises to 28% in future pregnancies. Most recurrences have involved complete moles, but Honore has reported recurrent partial mole in the same subject.^153,154

Trophoblastic disease also occurs in more than one family member. Affected family members include twins, cousins, and a mother and daughter.^155,156 Because population-based data are not available, the significance of these aggregates remains unclear, especially when they are observed in regions where the incidence is high. Furthermore, the nutritional factors cited previously would be expected to result in certain familial aggregates. No study has compared familial and sporadic cases.

Pivotal for future studies is the realization that hydatidiform moles can be subcategorized into two pathologic types. The classic mole is more likely to become malignant and is not associated with coexisting fetal parts.^{157,158,159,160,161} Classic moles, or complete moles, are diploid with both haploid complements being of paternal origin (androgenetic). Approximately 90% are 46,XX; the remainder are 46,XY but still purely paternal in origin. In 46,XY moles, fertilization could involve either one XY diploid sperm or two haploid sperm. In 46,XX moles, the two paternal complements are typically identical (homozygous). Androgenetic origin implies either a failure of the female pronucleus to participate in syngamy or expulsion of the maternal haploid complement following syngamy.

The nonclassic mole, or partial mole, is rarely malignant and is often associated with coexisting fetal parts. Partial moles are usually triploid, the result of two paternal and one maternal haploid complements. The most likely explanation is dispermy. The most common complement is 69,XXY, followed by 69,XXX; other complements are rarer. Although prognosis seems more favorable for triploid moles than for diploid moles, clinical management is at present identical for the two groups.

In gestational trophoblastic disease, a husband and wife often share HLA antigens (22 of 35 pooled cases).¹⁶⁰ In such cases, the paternal haplotype shared with the mother almost always (21 of 22 cases) segregates into the mole.¹⁶² Increased malignancy associated with exclusively paternal histocompatibility loci would be consistent with the ostensibly paradoxical observation that disparity between maternal and paternal HLA genes enhances implantation.¹⁶³ No particular HLA antigen has been identified as a predisposing factor in molar disease.

Epithelial changes in the cervix occur very often, constituting the basis for periodic cytologic testing (Papanicolaou smear). Often the cytologic changes (altered nuclei) merely reflect readily reversible inflammatory changes. Other changes, proved by biopsies, indicate preinvasive or invasive neoplasia.

Most (85%) cervical cancer is squamous in origin. Adenocarcinoma of the cervix accounts for the remaining 15% of cervical cancer. Epidemiologic studies have not attempted to distinguish between the two histologic types. Thus, adenocarcinoma may or may not share similar etiologies with squamous cervical cancer.

Squamous cell carcinoma of the cervix shows striking epidemiologic features consistent with data suggesting that heritable factors are not of paramount importance.¹⁶⁴ Incidence is inversely correlated with age at first intercourse and positively correlated with number of sex partners. The disorder is relatively more common among prostitutes and extraordinarily rare among celibates. All these observations suggest an infectious etiology. Indeed, strong associations between cervical cancer and condyloma have been observed.¹⁶⁵ Protein products of human papillomaviruses (HPV types 11, 16, and 18) are observed in cervical carcinoma.¹⁶⁵ Inactivation of P53 through mutation or binding with the oncoprotein of human papillomaviruses in cooperation with specific oncogene (MYC) activation are thought to cause malignant transformation of host cells.^166,167

Heritable factors are not common in squamous cervical cancer. Genetic studies have failed to detect familial tendencies.^165,168 The few reported familial aggregates^169,234 are surely no more frequent than expected on the basis of either siblings sharing similar socioeconomic status or coincidental development of a relatively common neoplasia in multiple family members.

Recently, activating mutations of fibroblast growth factor receptor 3 (FGFR3) were found in 25% of primary cervix carcinomas (3/12) and 35% of bladder carcinomas (9/26).^165A Matched constitutional DNA, available in the nine bladder tumor cases with mutations, contained wild-type sequences, demonstrating the somatic nature of these mutations in the bladder cancer. As corresponding normal DNAs were not available for cervix tumors and all three cervix carcinomas have the same point mutation: S249C, TCC to TCG, Ser to Cys,^165A it would be interesting to see whether any germline mutations of FGFR3 could be found in cervix cancer.

Like cancers of the cervix, uterine cancers consist of several different histologic types. By far the most common involves the endometrial glands as an endometrial adenocarcinoma. Rarer lesions arise from the myometrium (leiomyosarcomas) or the endometrial stroma (stromal sarcomas and mixed müllerian tumors). Women with endometrial cancer are characterized by relatively late age of marriage, relatively few pregnancies, and relative infertility.¹⁷⁰ The common pathogenesis is believed to involve intervals in which estrogens are not counteracted by progesterone. Thus, ovulatory cycles and intermittent pregnancies (progesterone-dominated milieu) protect against neoplasia.

Adenocarcinoma of the endometrium is also associated with obesity, hypertension, and diabetes mellitus.¹⁷⁰ Inasmuch as each trait shows genetic susceptibilities, one would predict a priori the existence of genetic susceptibilities in adenocarcinoma of the endometrium. Similar reasoning applies to a neoplastic susceptibility in polycystic ovarian disease, a heritable disorder characterized by anovulation and, hence, predisposing to adenocarcinoma of the endometrium.

Probably independent of the above susceptibilities, however, familial tendencies are documented in endometrial cancer. Albert and Child¹⁶⁸ observed family patterns in their epidemiology study, and clinicians often recite their clinical impression of familial aggregates. One often-cited study is that of Lynch and colleagues¹⁷¹ who investigated 154 probands with endometrial carcinoma. Their mean age was 63 years; obesity was present in 80%, hypertension in 65%, and diabetes mellitus in 65%. Of the 154, 11 had multiple primary neoplasias. Adenocarcinomas of either the endometrium or colon were present in first-degree relatives of 16% of the 154 probands. Unfortunately, frequency among first-degree relatives was not stated, and the origin of the sample from which the 154 probands were drawn was also not clear. Further problems with the study of Lynch and colleagues¹⁷¹ include determining the relationship between familial adenocarcinoma of the endometrium and colon and the “cancer family syndrome.” Some familial aggregates of endometrial cancer may represent the cancer family syndrome, whereas others need not. Moreover, the genetic mechanism in these two conditions may or may not be different.

In contrast to the relatively high proportion of familial cases observed by Lynch and colleagues,¹⁷¹ most gynecologic oncologists seem to observe far fewer familial cases. In most clinical reports there is little or no mention of familial aggregates. Of course, such studies are rarely conducted by persons whose discipline is genetics. Overall, the authors would counsel a recurrence risk not greater than 5% for first-degree relatives, adding the caveat that definitive statements are not possible.

Recently, the MSH6 gene, one of the DNA mismatch repair genes (MMR), has been found to be associated with familial endometrial cancer.^168A Among 214 HNPCC kindreds, 71 complying with the clinical Amsterdam criteria (Ams+ ),²⁹ and 143 kindreds not fully fulfilling the Amsterdam criteria (Ams-), 9 different MSH6 pathogenic germline mutations were found in 10 kindreds. These mutations were scattered along the coding sequence of MSH6 and predicted the truncation of its protein product. Seven of 10 MSH mutations were found in atypical HNPCC families not fulfilling the Amsterdam criteria. These kindreds displayed a very high frequency of atypical hyperplastic lesions and carcinomas of the endometrium: 73% in female MSH6 mutation carriers compared with 29% in women with MSH2 mutations and 31% in patients with structional defects in MLH1. The characteristic clinical features of the MSH6 mutation carriers are delayed age of cancer onset and incomplete penetrance. In view of this finding, MSH6 mutation analysis should be recommended in suspected familial gynecologic cancer syndromes, particularly in those families with an excess of endometrial cancers.

No gene predisposing to early development of sporadic endometrial cancer (not complying with the new modified Amsterdam criteria)^168B has been found to date.

Other gynecologic cancers include carcinomas of the vagina, vulva, and fallopian tubes. Carcinomas of the vagina and vulva are usually squamous in type and typically occur in older women. Prenatal diethylstilbestrol exposure is associated with clear cell carcinomas of the vagina, a distinctly different histologic type. Carcinomas of the fallopian tube are even rarer. No familial aggregates have been reported for any of these neoplasias.

Although the genetic basis of familial breast and ovarian cancer has been intensively investigated and has recently been further elucidated,⁵ very few genetic markers have so far been available for the 90% to 95% of patients about to contract these malignancies in the absence of a positive family history. Recently, we have reported the inheritance pattern of a genetic polymorphic 320-bp insertion in intron G of the human progesterone receptor gene and named it PROGINS.^172,173 The PROGINS restriction fragment length polymorphism (RFLP) marking a 320-bp insertion of Alu sequence in intron G was combined with point mutations in exon 4 and exon 5 of the HPR gene. The exon 5 alteration was a silent third base change, whereas the point mutation in exon 4 lead to an amino acid change of valine to leucine causing the mutated HPR protein. PROGINS was shown to co-segregate with increased risk of ovarian cancer in different ethnic groups. Higher frequency of PROGINS was associated with higher incidence of sporadic ovarian cancer but not endometrial and cervix cancer. The underlying mechanism is not clear. It was suggested that increased protein stability of the mutant protein which leads to increased transcriptional activity may contribute to the increased risk of developing ovarian cancer.^172,173 Another genetic marker, P53/PIN-3, has also been shown to be associated with an increased risk of developing sporadic ovarian cancer. PIN-3 is a 16-bp insertion in the intron 3 of the P53 gene. Homozygous status of PIN-3 was associated with an eightfold increased risk of ovarian cancer. No increased risk of ovarian cancer was observed in patients with heterozygous PIN-3.¹⁷⁴ The relationship of this genetic background and predisposition to ovarian cancer is not clear.

With the rapid progress in molecular biology and its use in medicine, it is to be expected that more and more reliable genetic markers both for familial and sporadic gynecologic cancer will be found in the near future and ultimately may make prophylactic approaches to gynecologic cancer development possible.

To better understand the complexity of gynecologic cancer development, including the heritable factors and rudimentary molecular genetics associated with these diseases, a basic background in the molecular mechanisms of carcinogenesis is valuable.

Cancer is generally viewed as a genetic disease.^175,176,177 However, there are important differences between cancer and most other genetic diseases. Unlike genetic disorders in general, cancer does not arise from a single mutation but as a result of a series of genetic changes successively transforming the cell from a normal to a malignant state.^175,176,177 These genetic alterations may result in activation of cellular oncogenes through chromosome rearrangements, gene amplification, point mutations, or viral insertion, or in inactivation of tumorsuppressor genes through point mutation, deletion, or partial or whole chromosome loss.^{1,61,176,178,179} It is well accepted that a concerted effort of these genetic changes is required to bring about successful transformation.

In most solid tumors, genetic abnormalities are complex^{180,181,182,183} and it is difficult to identify specific genetic changes, which are consistently present for a particular type of cancer. Attempts to demonstrate an association between a certain structural change such as point mutations in these genes and development of a certain cancer have generally been unsuccessful. For example, the BRCA1 and BRCA2 genes localized on chromosomes 17q and 13q, which are responsible for hereditary breast and ovarian cancer, are rarely mutated in sporadic breast and ovarian cancer.^{44,45,54,55,56,58,184} The RB gene on chromosome 13q,^185,186,187 which plays an important role in retinoblastoma, the P16 (CDKN2) gene in chromosome 9p,¹⁸⁸ and the P21 gene on chromosome 6p^189,190 are other examples. Among the known tumor-suppressor genes, P53, located on chromosome 17p, is the one for which a role in controlling tumorigenesis was most clearly demonstrated.^178,191,192 It is, however, not cancer type specific.^178,193 So far no genetic changes have been reported to be present only in one given cancer.

For most cancers, complex genetic alterations including chromosomal abnormalities, oncogene activation of genes such as MYC1, ERBB2, AKT2 and tumor-suppressor gene defects such as P53 are found at the same time, even though there is evidence suggesting that oncogene activation is an early event and the alteration of tumor-suppressor genes occurs later. The temporal relation of these mutations (i.e. early or late events in carcinogenesis) remains to be determined.

The complexity of genetic abnormalities in cancer is also reflected by the levels of expression of specific proteins involved in signal transduction,^{189,194,195,196,197,198,199} hormonal regulation,^200,201,202 and other cellular pathways such as extracellular matrix degradation, which are generally altered to a certain extent in malignant tumors.^{203,204,205,206,207,208} Although those changes are undoubtedly determined by underlying genetic alterations, the nature of the specific signal transduction pathways altered in a given tumor is currently not predictable. For example, neither increased levels of CA125 in ovarian cancer nor elevated levels of alpha 1-fetoprotein in liver cancer are always present.

A better understanding of the molecular events leading to the acquisition of allelic losses, oncogene activation, tumor-suppressor gene inactivation and elucidation of the links between a specific molecular genetic defect, a specific signal transduction mechanism or other cellular pathways as well as the development of technical approaches for the rapid identification of activated or altered pathways in individual tumors would undoubtedly provide important clues regarding the initial molecular changes associated with malignant transformation and lead to significant changes in our approach to cancer therapy.

Knudson and coworkers²⁰⁹ proposed a useful model to explain certain observations in heritable development of cancer. Cancer is proposed to arise in a two-step fashion. Each step requires a mutation. The first mutation (step 1) may be either germinal (inherited) or somatic (acquired). The second mutation (step 2) is almost always somatic. The Knudson model explains observations that Mendelian tumors (e.g. retinoblastoma) are not only usually bilateral but also show earlier ages at onset than do nonmendelian tumors. All cells in individuals having a gene for a mendelian tumor are assumed to have accomplished step 1 as a result of having been born with the mutant gene. Conversely, an individual born without the germinal mutation could develop a histologically similar tumor only as result of oncogenic agents sequentially producing two somatic mutations. Naturally, this would arise less often because not a single event but two independent events are necessary. This model is very useful to understand the key issue of carcinogenesis, but it cannot explain the evidence observed in sporadic human cancer and knockout mouse experiments. Mice without a certain gene which was demonstrated to be involved in the process of carcinogenesis developed normally and did not have an increased risk of developing cancer.^{210,211,212,213} Alterations of many tumor-suppressor genes and oncogenes are not observed in a variety of sporadic cancers.

Gene Network of the Cell

The cell is the level at which cancer develops. It is estimated that the genome of a human cell contains more than 100,000 genes. At a given time in a given cell only a small proportion of genes are actively involved in cell life. These actively transcribed genes work in a well-controlled system (upregulate or downregulate the expression of each other), determining the behavior of the cell in response to external signals and are called the gene network.

Each living cell has its own gene network. Different cells in the same biologic object have the same genome but different gene networks. This is supported by “chromosome walking,” which demonstrated that different cell types contain the same one-dimensional genomic DNA whereas different proteins are expressed in individual cells. Two classes of genes, namely oncogenes and tumor-suppressor genes implicated in cell proliferation and differentiation, embryogenesis and oncogenesis, are the key genes in the gene network controlling the behaviors of the cell.

Cellular oncogenes are the driving force in the gene network. They are not only found in viruses but also in normal human cells as well as in malignant human cells. Oncogene sequences in normal cells are known as protooncogenes. When protooncogenes are altered they become functional oncogenes known as cellular oncogenes. Presence of oncogenes in all biologic species indicates a very important function. These genes are highly conserved during evolution. Their products belong to the regulatory network relaying external signals from the membranes toward the nucleus and allowing cells to adapt their division rate to the demand of the organism. Genes directing cells toward rapid cell division could be integral to successful embryonic development, especially in the initial weeks after conception. After differentiation, nonspecific cell growth would no longer be desirable. However, oncogenes could persist to play a role in cell division. Activation of oncogenes can occur by a variety of mechanisms (i.e. activation by point mutation; structural aberration such as translocation, insertion, inversion or gene amplification).^{175,176,210,211,212,213,214,215}

Oncogenes are considered dominant transforming genes because activation due to alterations in only a single allele can transform cells. There is strong evidence that activation of oncogenes is involved in carcinogenesis. A considerable amount of data has accumulated showing that many oncogenes are altered in their structural organization or expression in tumor cells. However, among all the oncogenes described so far, only a few altered in their structural organization or expression have been demonstrated in a given cancer. For example, mutations of RAS genes are found in a significant proportion of various human neoplasms.²¹⁶ RAS overexpression has been demonstrated in ovarian cancer. However, mutations of the RAS genes are rare in ovarian cancer.^217,218,219 The most common RAS gene alterations in ovarian cancers involve KRAS. Mutation in the KRAS gene has been observed in malignant ovarian tumors but occurs with a particularly high frequency (50%) in borderline tumors.^219,220 No HRAS mutations have been reported and only a few cases have had KRAS or NRAS mutations.

Amplification of RAS genes is also infrequent in ovarian cancer.^221,222,223 The MYC1 oncogene, a transcription factor coding for a protein that is involved in cell transition from the G₀ to the G₁ phase of the cell cycle, is another example. Abnormalities of MYC1 are a common finding in leukemia and a number of solid tumors (e.g. carcinoma of the breast and cervix and small-cell lung cancer) and are often associated with more aggressive tumors. Increased expression of MYC protein due to amplification of the gene was reported in approximately 20% to 50% of ovarian cancers but not in benign or borderline ovarian tumors.²²⁴ However, major structural rearrangements of the MYC gene do not occur in ovarian cancer.^225,226 A number of other oncogenes (e.g. INT2, FMS, MDD, AKT2) which have also occasionally been shown to be amplified in ovarian cancer^227,228,229 are not considered to play a important role in the carcinogenesis of ovarian cancer. All these observations indicate that only a small number of oncogenes are involved in the carcinogenesis.

Tumor-suppressor genes, like oncogenes, are involved in the regulation of cellular growth and differentiation. Unlike oncogenes, these genes behave in a recessive manner and are the major force for keeping the gene network in a stable state. Loss or inactivation of both copies of a tumor-suppressor gene removes normal constraints of cell proliferation, although there are now examples of inherited cancers (e.g. familial adenomatous polyposis, multiple endocrine neoplasia 2a and 2b) where the germline alteration of a single allele of the susceptibility gene is sufficient to cause an altered phenotype. Loss or inactivation of a tumor-suppressor gene can be due to one of several mechanisms, such as point mutation, deletion, mitotic recombination, chromosomal loss and loss of imprinting in certain genes. In solid tumors, recessive genes are more contributory to pathogenesis than dominant genes. Many chromosomal regions have been implicated in harboring tumor-suppressor genes and are thought to be involved in tumor progression when analyzed by a variety of approaches. It is this area of research that has received a lot of attention as it is a direct way of identifying genes that are important in pathogenesis.

Many tumor-suppressor genes have been identified and are found to be altered in certain tumors such as RB in retinoblastoma and DCC in colon cancer. However, not all tumor-suppressor genes are involved in the carcinogenesis of a given cell. For example, tumor-suppressor genes such as RB, DCC, FAP, and WAF1 are rarely altered in sporadic ovarian cancer, and only five mutations in BRCA1^44,45 and four mutations (180 cases) in the BRCA2 have been reported in sporadic ovarian malignancy.⁵⁸ It would be very important to know which tumor-suppressor genes are actually incorporated into the gene network of a given cell.

In a well differentiated cell the gene network is quite stable. The internal state of the gene network is reflected by its products, proteins, which carry out the business of each living cell. The behavior of the cell under different circumstances (external signals) is controlled by adjusting the gene network through regulating the expression of certain genes without reconstructing the gene network. This is supported by the observation that in a well differentiated cell only the proportions of individual proteins are changing in response to external signals. No new proteins which are the products of other genes are produced.

In an undifferentiated cell the gene network is unstable. The response to external signals is usually through reconstructing the gene network. The external signals can be produced by the cell itself, by other cells, or by other stimuli. The reconstruction of the gene network is characterized by systematic controlled incorporation of certain genes into the gene network and a shutdown of other genes already existing in the gene network. In another words, alterations of the genome are observed in three dimensions and in one dimension. This is supported by the observation that the expression of many proteins existed before getting lost during cell differentiation while many other new proteins are synthesized.

Pattern formation is a central feature in development; it is the process whereby states are assigned to the cells according to their position. It is fascinating to see how a single cell develops into a complicated biologic object with well-differentiated tissues and organs while each individual cell still contains the same genome and interacts with each other by means of different signals. We consider the development and differentiation as a process of reconstruction and reorganization of the gene network. This process involves activation and inactivation of certain genes. The extracellular signals are the directing force for this process. In a physiologic state, activation and inactivation of genes is performed by binding of the promoter of target genes to protein or protein-protein complexes or hypermethylation or through other ways. The process of activation and inactivation of genes is reversible. Although alterations of three-dimensional genetic structures of the genome happen, no one-dimensional genetic alterations occur. In contrast, one-dimensional genetic alterations are the key characteristic of pathologic activation and inactivation of genes. Pathologic activation and inactivation of genes is due to one of several mechanisms such as point mutation, deletion, insertion, mitotic recombination, chromosomal loss and loss of imprinting in certain genes. This process is irreversible.

Tumor development and progression can be considered as one of many unstable gene networks under uncontrolled reconstruction due to pathologic activation and inactivation of key genes (i.e. tumor-suppressor gene and oncogene) of the gene network. The driving forces of selection inherent in malignancy can be considered as the sum of intact or altered tumor-suppressor genes and oncogenes incorporated into the new gene network. Tumor development and progression is associated with a simplified gene network characterized by overall increased loss of protein types which are present in a well differentiated cell. Inactivation of tumor-suppressor genes in the gene network is the key step of tumor development. This is supported by the observation that inactivation of tumor-suppressor genes is the most common event in solid tumors and is associated with the more complicated pictures of genetic alterations found in more advanced cancers, whereas oncogene activations were more common in benign or less advanced tumors, which have less-complicated genetic alterations.²³⁰ This is probably because of more uncontrolled reconstruction or reorganization of the gene network due to impaired function of tumor-suppressor genes leading to more chaos of the gene network and the whole genome in the tumor cell.

Although the neoplastic phenotype is derived partially from epigenetic alterations in gene expression, the driving force of malignant transformation appears to be the sequential mutation of protooncogenes and tumor-suppressor genes, with their subsequent selection and accumulation.¹⁷⁸ Inactivation of tumor-suppressor genes in the gene network decreases the force of stabilizing the gene network while activation of oncogenes drives the cell through the cell cycle under less control. A series of genetic alterations is needed to change the well established gene network in a given cell to a malignant gene network characterized by giving the cell reproductive immortality and a remarkable equivalence with our unicellular ancestors. The data to support the multistep, multigenic paradigm are extensive,^{61,175,177,231} but perhaps the most compelling evidence is that the age-adjusted incidence for most human epithelial tumors increases at roughly the fifth to sixth power of elapsed time, suggesting that a series of five to six genetic alterations are rate limiting for carcinoma development.²³²

Experimental data suggest that the gene network is nonobjective. A gene in the gene network can contribute to the emergence of more than just one phenotypic trait or a phenotypic trait can be determined by the expression of several genes. This implies nonlinearity (i.e. lack of the proportional relation between input and outcome), complexity (i.e. emergence of the hierarchical network of multiple cross-interacting elements that is sensitive to initial conditions, possesses multiple equilibria, organizes spontaneously into different morphologic patterns, and is controlled in a dispersed rather than a centralized manner), and quasi-determinism (i.e. coexistence of deterministic and nondeterministic events) of the gene network. Because of its complexity, the same phenotype can be associated with a number of alternative sequences of genetic events. Moreover, the primary cause initiating phenotypic evolution of cells, such as malignant transformation, can be determined to be probable but cannot be identified unequivocally. It is well known that morphologically similar tumors presenting in any assigned stage may behave in radically different fashions, which seriously hampers the physician's ability to accurately predict clinical behavior in a given case. It is hypothesized that genetic mutations inside the gene network alter both the molecular and informational structure of the gene network which make the already complicated story even more complicated. It is impossible to accurately predict the behavior of a given cell without knowing more about the gene network inside the cell.

In a given cell the gene network consists of only a small proportion of tumor-suppressor genes, oncogenes and other structure genes of the whole gene pool. Only those tumor-suppressor genes and oncogenes which are incorporated into the established gene network are involved in the process of carcinogenesis. Inactivation of tumor-suppressor genes or activation of oncogenes outside the gene network may have no or weak impact on the process of tumor development. Definition of individual gene networks in a given cell would be the key step to give us the ability to accurately predict the behavior of a cell. However, definition of the individual gene network in a given cell will be very difficult because of its complexity, even though it consists only of a small proportion of genes of the whole gene pool. With the application of advanced genetic techniques such as fluorescent in situ hybridization, comparative genome hybridization, PCR, the construction of whole genome mapped polymorphic markers and genetic maps of individual chromosomes, radiation hybrid cell maps, developments in physical mapping consisting of genomic maps, random sequencing of cDNA clones from tissue specific libraries and their mapping, we may be able to accurately define the gene network in a given cell in the foreseeable future.

If the outlined model were correct, pivotal to the question of whether genetic factors play roles in neoplasia is whether those genetic factors have an impact on the stability of the gene network. Identifying the crucial genes in each individual gene network may narrow our focus to the identification of the first mutation in the cascade of carcinogenetic events. If this concept proves realistic, site-specific screening programs that identify individuals at high risk for certain cancers could be envisioned.