Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Abstracts of Talks GENEâENVIRONMENT INTERACTIONS RELATED TO COLON CANCER2 David Alberts, M.D. and M. Elena Martinez, M.D. Rates of Colorectal Cancer and Adenomatous Polyps. Colorectal cancer remains the third leading cause of cancer deaths in each sex and second overall in the United States (Landis et al., 1999), despite the fact that it is largely a preventable disease. Approximately half of diagnosed individuals will die of this malignancy. Because adenomatous polyps are precursors to colorectal cancer, assessing the effect of environmental and genetic factors in adenoma occurrence and recurrence instead of cancer might help identify relatively asymptomatic individuals who are at increased risk of cancer and who would benefit most from an overall public health intervention. Additionally, the identification of risk fac- tors for recurrence may help define follow-up screening protocols. Although we have obtained some clues regarding risk factors for newly diagnosed adenomas (Neugut et al., 1993), few data exist on predictors of adenoma recurrence among individuals with resected adenomas (Davidow et al., 1996; Tseng et al., 1997; Baron et al., 1998; Hyman et al., 1998; Whelan et al., 1999). Genetic Basis for Colorectal Neoplasia. The genetic basis for the develop- ment of colorectal cancer involves the accumulation of specific somatic muta- tions in proto-oncogenes and tumor suppressor genes with increasing age 2Supported by a Cancer Research Foundation of America grant, Public Health Service grants CA- 41108 (Colon Cancer Prevention Program Project) and CA-23074 (Arizona Cancer Center Core Grant) from the National Cancer Institute, and a Career Development Award (KO1 CA-79069-10) grant to M.E. Martinez from the National Cancer Institute 70
ABSTRACTS OF TALKS 71 (Kinzler and Vogelstein, 1996). However, only a small proportion of colorectal cancers are attributable to inheritance of these rare, highly penetrant mutated genes. Epigenetic changes, such as alterations in DNA methylation (e.g., CpG island methylator phenotype, or CIMP) and gene expression, also may play a critical role in the development of this malignancy (Baylin et al., 1998). It is evident too that variability in carcinogen-metabolizing genes influences the risk of colorectal neoplasia in humans (Gertig and Hunter, 1998; Hussain and Harris, 1998a; Pereira, 1998). It is clear that susceptibility to colorectal cancer is related to interindividual variability in biotransformation of endogenous and exogenous substances, as well as in DNA repair and cell cycle control. Common genetic variation may enhance susceptibility to environmental carcinogens by altering the rates of acti- vation and detoxification of carcinogens. The interactions of environmental fac- tors with metabolic polymorphisms may act via a model in which the exposure alone, but not the variant genotype alone, increases disease risk; however, expo- sure interacts with the variant genotype to further increase risk in the exposed individuals (Vineis, 1997). The same interaction can also modulate disease patho- genesis in that exposure and susceptibility factors may alter the effects of other risk factors, such as folate intake, methylenetetrahydrofolate reductase (MTHFR), CIMP, selenium, or celecoxib intervention, and cyclooxygenase (COX) upregulation, on adenoma recurrence. An example of such an interaction is the relationship between alcohol intake, folate, and MTHFR. Classification of sub- groups of the population into those who may be more vulnerable to the effects of certain carcinogens may also have important implications beyond risk assess- ment. Through the identification of an increased risk in certain subgroups, dis- ease risk factors may be better defined. However, to date, sample sizes for most studies attempting to uncover geneâenvironment interactions have been small, limiting the potential for detecting significant findings. CIMP as a Marker of Gene Methylation. As stated previously, a genetic basis for cancer has been established with the assumption that the age (and mutagen exposure) related accumulation of somatic mutations accounts for the increased incidence of cancer with age (Ames et al., 1993). The actual rate of mutation accumulation in aged tissues is more uncertain, with some investiga- tors finding lower than expected mutation rates (Warner and Price, 1989; Bohr and Anson, 1995), possibly reflecting the presence of additional mechanisms for activation and/or inactivation of genes important in the carcinogenesis process. In the past few years, there has been renewed interest in epigenetic mechanisms in carcinogenesis (Jones, 1996; Baylin et al., 1998). Epigenetics refers to the study of changes in gene expression that can be mitotically inherited, without associated changes in the coding sequence of the affected genes. Aging and transformed cells show profound changes in gene expression, many of which cannot be accounted for genetically (Sager, 1997). Methylation of DNA within
72 CANCER AND THE ENVIRONMENT promoter-associated CpG islands can be a powerful molecular mechanism for gene silencing (Razin and Riggs, 1980; Adams and Burdon, 1982). In mammals, 5-cytosine methylation of CpG dinucleotides is the only natu- rally occurring modification of DNA. DNA methylation patterns form early in development with the establishment of gene and tissue-specific patterns of me- thylation, which are relatively stable (Razin and Riggs, 1980). In humans, ap- proximately 70% of CpG dinucleotides are methylated in adult cells (Adams and Burdon, 1982; Bird, 1992). The function of normal DNA methylation remains controversial, as suggested by the fact that highly expressed genes tend to be hypomethylated, and that silent genes tend to be hypermethylated (Cedar, 1988; Bird, 1992). Issa and coworkers (Toyota et al., 1999a, 1999b) have developed a defini- tion for a methylated phenotype referred to as CIMP. This phenotype was devel- oped after analysis of a number of colon cancers, colorectal adenomas, and nor- mal colonic mucosa. The research group found that CIMP-positive colorectal cancers averaged 5.1 methylated loci (out of 7) versus 0.3 for CIMP-negative tumors. Additionally, some genes were methylated in an age-related manner, while others were more clearly associated with cancer. Based on this work, CIMP-positive adenomas were more likely to have Ki-ras mutations, while CIMP-negative adenomas were more likely to have mutations in p53. Further- more, CIMP-positive adenomas were found to have lower levels of COX2 ex- pression because of promoter methylation, and their dependence on expression of DNA methyltransferase to maintain tumor suppressor gene silencing via hypermethylation. Folate, MTHFR, and Gene Methylation. An increasing epidemiologic body of evidence from case-control (Ferraroni et al., 1994) and cohort studies (Giovannucci et al., 1995, 1998; Glynn et al., 1996) supports the important role of folate in reducing the risk of colorectal cancer. Another study (Ma et al., 1997), which did not have comprehensive dietary data, showed an inverse asso- ciation between plasma folate and risk of colon cancer. Folate intake and blood levels have also been consistently associated with lower risk of colon adenomas (Giovannucci et al., 1993; Tseng et al., 1996). Recent results indicate that in- creased consumption of folic acid from supplements, after a period of 15 or more years, may decrease the risk of colon cancer by about 75% (Giovannucci et al., 1998). Giovannucci et al. (1993) have proposed that the increased risk associated with low folate levels is related to intracellular methylation defects. Addition- ally, these investigators proposed that alcohol consumption increases the risk of colorectal neoplasia by acting as a folate antagonist; this hypothesis is based on data demonstrating the modifying effect of folate and methionine on the alcohol and colorectal neoplasia relationship (Giovannucci et al., 1993, 1995; Ma et al., 1997). Given the epidemiologic evidence for the proposed protective effect of folate on colorectal neoplasia, some studies have explored the mechanisms involved in
ABSTRACTS OF TALKS 73 this association. Because folate is the primary methyl donor in cellular metabo- lism (Hoffman, 1985), markers of folate status are important factors to address in the etiology of gene methylation. A critical role of folate is in synthesizing methionine from homocysteine (Hoffman, 1985). Methionine, in turn, is con- verted to S-adenosylmethionine (SAM), the primary methyl donor in the reaction transferring a methyl group to the enzyme 5â-cytosine-DNA methyltransferase. Transfer of a methyl group from SAM to methyltransferase produces S-adeno- sylhomocysteine (SAH), which is then hydrolyzed to homocysteine. Folate is also essential for nucleotide biosynthesis (Eto and Krumdieck, 1986). In folate deficiency, thymidylate shortages cause an imbalance in the thymidylateâdeoxy- uridylate pool and a resultant incorporation of uridylate into DNA. Excess uridylate incorporation into DNA results in unstable chromosomes, decreased DNA repair, and increased chromosome breaks (Barclay et al., 1982; Reidy, 1987; Everson et al., 1988; Dianov et al., 1991; James et al., 1992). Genetic Polymorphisms of MTHFR. A genetic factor that modifies the effects of folate status has recently been identified that includes the inherited variation in the activity of MTHFR, a critical enzyme involved in the production of the form of folate that supplies the methyl group for methionine synthesis (Kutzbach and Stokstad, 1971). Different endogenous forms of folate, 5-methyl- tetrahydrofolate, and 10-methylenetetrahydrofolate, are essential for DNA me- thylation and DNA synthesis, respectively. A common thermolabile polymor- phism in the MTHFR gene (C677âT, alanineâvaline) has been shown to be protective against colon cancer in some (Chen et al., 1996; Ma et al., 1997; Slattery et al., 1999; Ulrich et al., 1999) but not all (Chen et al, 1998) studies. Low MTHFR activity is thought to protect against colorectal cancer since less tetrahydrofolate is converted to 5-methyltetrahydrofolate, allowing more folate to be shunted toward DNA synthesis and repair. In these studies, an inverse association was shown for the presence of the val/val genotype and colorectal cancer among individuals with adequate folate intake, whereas this effect was not seen among those with low folate intake (Chen et al., 1996; Ma et al., 1997). Since MTHFR is required to convert 5,10-methylenetetrahydrofolate to 5- methyltetrahydrofolate, individuals with low MTHFR levels (val/val homozy- gotes) would be expected to have relatively high levels of 5,10-methylene- tetrahydrofolate resulting from the low levels of the 5-methyltetrahydrofolate (Frosst et al., 1995). Therefore, in a low-folate environment where there are inadequate quantities of methyl groups for these pathways, individuals with the val/val genotype do not divert as much folate from the thymidylate pathway to the methylation pathway, resulting in lower SAM levels and high levels of ho- mocysteine. Under conditions in which folic acid levels are insufficient to meet metabolic needs, val/val homozygotes would be less efficient at diverting folate metabolites into the 5-methyltetrahydrofolate product, resulting in a shortage of methyl groups. When levels of 5,10-methylenetetrahydrofolate (which is required
74 CANCER AND THE ENVIRONMENT to convert deoxyuridylate to thymidylate) are low, misincorporation of uracil for thymidine may occur during DNA synthesis (Wickramasinghe and Fida, 1994), possibly increasing spontaneous mutation rates (Weinberg et al., 1981), sensitiv- ity to DNA-damaging agents (Meuth, 1981), frequency of chromosomal aberra- tions (Sutherland, 1988; Fenech and Rinaldi, 1994), or errors in DNA replication (Hunting and Dresler, 1985; Fenech and Rinaldi, 1994; James et al., 1994). Folate Deficiency and CpG Island Gene Methylation. The proposed mechanism for the above-reviewed studies relates to dietary factors that influ- ence methyl group availability, which can in turn affect DNA methylation. DNA methylation is an essential mechanism of gene regulation, and disturbances may cause differential gene expression (Cedar, 1988). In animal models, folate defi- ciency can cause imbalances in DNA methylation (Wainfan and Poirier, 1992; Kim et al., 1996). Furthermore, folate deficiency in rats has been shown to induce DNA strand breaks and altered methylation within the p53 tumor sup- pressor gene (Kim et al., 1997) and to result in deoxynucleotide pool distur- bances (James et al., 1992). Given this literature, support for the potential effect of folate status and methyl group availability in the etiology of CIMP status exists. Additional supporting evidence for this association derives from our own study (Martinez et al., 2001), in which a low intake of folate was associated with a significantly higher risk of Ki-ras mutations in adenomatous polyps. Previous work by Toyota et al. (1999a) indicates that CIMP-positive adenomas are more likely to harbor Ki-ras mutations than CIMP-negative adenomas. Thus, the po- tential role of folate in the etiology of Ki-ras, mutations along with data support- ing the high rate of these mutations in CIMP-positive adenomas, suggests that folate may be involved in the etiology of CIMP-positive adenomas. Dietary Folate Intake, MTHFR Status, and Colorectal Cancer. Published data on the interaction between folate and MTHFR in the etiology of colorectal neoplasia are inconsistent, suggesting that this interaction is more complex than originally proposed. In a recent report of more than 3,000 case and control par- ticipants (Slattery et al., 1999), a lower risk of colorectal cancer associated with higher intake of folate was shown among individuals with the val/val genotype as compared to those with the ala/ala genotype who had low folate intake (odds ratio (OR) = 0.6; 95% confidence interval = 0.4â1.0). Of particular interest, the effects of folate among the val/val genotypes appeared to be stronger for the proximal (OR = 0.5) compared to the distal (OR = 0.8) colon. Based on Issaâs work, CIMP-positive adenomas were more prevalent in the proximal colon, which may be related to factors affecting folate metabolism and methyl group availability. Plasma Homocysteine, MTHFR Status, and Polyp Recurrence. We pro- spectively examined whether plasma levels of homocysteine were associated
ABSTRACTS OF TALKS 75 with the risk of recurrence of adenomatous polyps (Martinez et al., 2001). Analy- ses were conducted among 1,014 men and women, 40 to 80 years of age, en- rolled in a Phase III trial testing the effects of a wheat bran fiber intervention on adenoma recurrence. We also examined whether the association between plasma homocysteine and adenoma recurrence was modified by the MTHFR genotype among 961 participants with genotype data. Homocysteine in plasma was ana- lyzed at baseline by high-performance liquid chromatography (HPLC). MTHFR genotyping was performed by high-throughput microarray technology. Com- pared to participants with lower plasma homocysteine levels, those with higher levels were older, were more likely to be male, had lower intakes of total (dietary plus supplemental) folate, had higher alcohol intakes, and had lower plasma folate levels. After adjustment for age, gender, number of colonoscopies, and a history of previous polyps, the odds ratio for adenoma recurrence for individuals with homocysteine levels >11.6 mmol/l was 1.45 (95% CI = 0.98â2.14; P-trend = 0.02) compared to those with levels <7.8. When we assessed the relationship between homocysteine levels and adenoma recurrence according to MTHFR sta- tus, individuals with the TT genotype and homocysteine levels above the median (>9.4) had a higher risk of recurrence (relative risk = 1.96) compared to those with the CC genotype and homocysteine levels below the median. The results of these analyses suggest a modest effect of plasma homocysteine levels on ad- enoma recurrence and a risk-enhancing effect of high homocysteine levels on adenoma recurrence among individuals with the TT MTHFR genotype. DIET AND RISK OF CHILDHOOD CANCER Greta Bunin, Ph.D. Epidemiologists often broadly define environment to encompass anything that is not genetics. Diet is an integral part of the environment. All solids includ- ing fish, foul, meat, grains, vegetables, and foods may contain trace contami- nants (e.g., pesticides, heavy metals, polychlorinated biphenyls [PCBs]). The evidence is just beginning to emerge on the role of diet and cancer in children and, more specifically, the role of diet in the generation of cancer. Fewer than 20 studies have looked for a link between childhood cancer and diet. Most commonly, researchers have investigated the link between brain can- cer and diet because of a hypothesis based on animal data. The hypothesis postu- lates that children with greater exposure to N-nitroso compounds (NOCs) and their precursors are more likely to develop a brain tumor compared to other children. In many species of animals, NOCs are highly potent carcinogens. Some NOCs induce nervous system tumors, and for a few NOCs, the risk of tumor development is multiplied when the exposure occurs in utero. Human exposure to NOCs is widespread, and they have been detected in many common products, including cigarette smoke, automobile interiors, and
76 CANCER AND THE ENVIRONMENT cosmetics. Additionally, we are also exposed to precursors that combine to form NOCs in our stomachs and elsewhere in our bodies. In fact, most human expo- sure is thought to occur via synthesis in the body from precursors. Substances such as vitamins C and E inhibit the formation of NOCs and, thus, may be important for the prevention of brain tumors. Diet is a major source of NOCs, NOC precursors, and NOC inhibitors. Meats cured with nitrites, such as hot dogs and lunch meat, contain NOCs and NOC precursors, while fruits, vegetables, and vitamin supplements contain NOC inhibitors. Therefore, researchers hypothesize that a motherâs frequent eating of cured meats and infrequent eating of fruits and vegetables increase the risk of brain tumors in children. Of the eight studies investigating the NOC hypothesis, four found a signifi- cant doubling of risk of brain cancer when the mother frequently consumed cured meats during pregnancy. In a fifth study, a similar association was ob- served but it was not significant. Two studies had small numbers of children with brain tumors, which may explain why they failed to detect a difference. The last study looked at a less common type of brain tumor and observed no associa- tion with cured meat. This tumor type has a different sex and age distribution than the most common type and therefore might have a different etiology as well. Overall, the data are fairly consistent for an association between frequent consumption of cured meats during gestation and childhood brain tumors. Fur- ther research will be needed to determine if the brain tumors are due to the NOCs in cured meats or to a nutrient such as high fat or low folate in the diet. SIMILARITIES OF PROSTATE AND BREAST CANCER: EVOLUTION, DIET, AND ESTROGENS Donald S. Coffey, Ph.D. The risk of both prostate and breast cancer is similar and primarily deter- mined by the environment in which one lives, and this risk can vary more than tenfold between countries. In contrast, no risk exists for human seminal vesicle cancer, thus demonstrating tissue specificity for cancer in the human. There is also species specificity because there is no risk for prostate cancer in any other of the thousand of aging mammal species except the dog. Evolution indicates that the prostate and breast appeared at the same time 65 million years ago with the development of mammals. All male mammals have a prostate; however, the presence of seminal vesicles is variable and is determined by the diet so that species primarily eating meat do not have seminal vesicles. The exception is the human, who has seminal vesicles and consumes meat, although this is a recent dietary change. Human lineage departed from other higher primates 8 million years ago. The closest existing primate to humans is the bonobo (pigmy chim- panzee), which does not eat meat but exists primarily on a high-fruit and fresh
ABSTRACTS OF TALKS 77 vegetable diet. Homo sapiens evolved only about 150,000 years ago, and only in the last 10% of that time (10,000â15,000 years ago) did humans and dogs dra- matically alter their diets. This is the time when humans domesticated the dog, bred animals, grew crops, and cooked, processed, and stored meats and veg- etables. Current epidemiologic evidence and suggestions for preventing prostate and breast cancer in humans indicate that we should return to the original type of diets under which our ancestors evolved. The recent development of the West- ern-type diet is associated with breast and prostate cancer throughout the world. It is believed that the exposure to and metabolism of estrogen, and the dietary intake of phytoestrogens, combined with fat intake, obesity, and burned food processing, may all be related to hormonal carcinogenesis and oxidative DNA damage. An explanatory model is proposed. (For details see Coffey, 2001.) EFFECT OF HERBAL THERAPIES ON PROSTATE CANCER Robert S. DiPaola, M.D. Background. Herbal mixtures are popular alternatives to demonstrated therapies. PC-SPES, a commercially available combination of eight herbs, is used as a nonestrogenic treatment for cancer of the prostate. Since other herbal medicines have estrogenic effects in vitro, we tested the estrogenic activity of PC-SPES in yeast and mice and in men with prostate cancer. Methods. We measured the estrogenic activity of PC-SPES with transcrip- tional activation assays in yeast and biologic assay in mice. We assessed the clinical activity of PC-SPES in eight patients with hormone-sensitive prostate cancer by measuring serum prostate-specific antigen and testosterone concentra- tions during and after treatment. Results. PC-SPES had estrogenic activity similar to that of 1 nM estradiol, and in ovariectomized CD-1 mice, the herbal mixture increased uterine weights substantially. In six of six men with prostate cancer, PC-SPES decreased serum testosterone concentrations (P < 0.005), and in eight of eight patients, it de- creased serum concentrations of PSA. All eight patients had breast tenderness and loss of libido, and one had venous thrombosis. HPLC, gas chromatography, and mass spectometry showed that PC-SPES contains estrogenic organic com- pounds that are distinct from diethylstilbestrol, estrone, and estradiol. Conclusions. PC-SPES has potent estrogenic activity. The use of this un- regulated mixture of herbs may confound the results of standard or experimental therapies and may produce clinically significant adverse effects. Further studies to identify the estrogen(s) responsible for this activity are warranted.
78 CANCER AND THE ENVIRONMENT COLORECTAL CANCER AND ENVIRONMENTAL RISK FACTORS Raymond N. DuBois, M.D., Ph.D. Risk factors for colorectal cancer include a positive family history, meat consumption, smoking, and alcohol consumption. A reduction in risk for the disease is associated with vegetable intake, use of nonsteroidal anti-inflamma- tory drugs (NSAIDs), hormone replacement therapy, and physical activity. There are several genetic and epigenetic alterations that are known to be involved in the development of colorectal cancer. These alterations are important in both inherited syndromes such as familial adenomatous polyposis (FAP) or hereditary nonpolyposis colorectal cancer (HNPCC) and in sporadic tumors. It will be im- portant to understand the roles of environmental exposure and host susceptibility to develop better screening, prevention, and treatment strategies. Population-based studies indicate a 40â50% reduction in mortality from colorectal cancer in persons using NSAIDs on a regular basis (Smalley and DuBois, 1997). Colorectal cancer is a major cause of death from cancer in West- ern civilizations, claiming more than 55,000 lives in the United States each year. Environmental and dietary factors play an important role in the etiology of this disease as well as the known genetic components. Research efforts have been focused on understanding the molecular basis for the chemoprotective effects associated with use of aspirin and other NSAIDs. NSAIDs inhibit both cyclooxygenase-1 (COX-1) and cyclooxygenase-2 (COX-2) activity (Williams and DuBois, 1996). Since COX-2 levels are increased in a number of solid tumors, this enzyme may serve as a molecular target for cancer prevention (Sheng et al., 1997). Recent clinical studies indicate that the presence of COX-2 in human lung and colon cancers is associated with a negative clinical prognosis (Achiwa et al., 1999; Sheehan et al., 1999). Therefore, COX-2 inhibitors are currently being evaluated for the prevention and/or treatment of cancer in hu- mans (Steinbach et al., 1999). GENES AND THE ENVIRONMENT IN CANCER ETIOLOGY Joseph F. Fraumeni, Jr., M.D. The importance of environmental factors in human cancer has long been evident from the striking international variation reported in cancer incidence, resulting in estimates that perhaps 80% of all cancer in the United States is potentially preventable or avoidable. Further indications of environmental can- cer come from the shifts in the cancer experience of migrant populations whose rates tend to approximate those of the host country, the geographic patterns of cancer within the United States, the changing incidence of certain cancers over time, ethnic and socioeconomic differentials, and the abundant epidemiologic evidence linking carcinogenic risks to a variety of lifestyle and other environ-
ABSTRACTS OF TALKS 79 mental exposures. Not long ago, the role of inherited susceptibility in human cancer was considered to be quite small given the rarity of hereditary cancer syndromes, but recent progress in identifying and characterizing highly pen- etrant but relatively rare susceptibility genes in familial cancer has revolution- ized our understanding of genetic mechanisms and their critical importance in cancer etiology. Of special significance to the public health burden of cancer, however, are the common polymorphic susceptibility or modifier genes that con- fer low relative and absolute risks, but high population-attributable risks in the presence of relevant environmental exposures. The two classes of genes repre- sent parts of a continuum because even the highly penetrant genes responsible for hereditary cancer may involve environmental exposures for expression, as illustrated by the susceptibility to carcinogens in hereditary retinoblastoma and LiâFraumeni syndrome. Especially exciting is the opportunity to parlay discoveries of polymorphic genes and their functions into a better understanding of environmental carcino- genesis. By incorporating careful exposure assessment and mechanistically plau- sible candidate susceptibility genes into epidemiological study designs, it should be possible to identify the more subtle risks due to specific dietary and nutri- tional factors, metabolic alterations, environmental pollutants, and other com- mon exposures that have eluded traditional epidemiologic approaches. Although the molecular and statistical tools to examine complex geneâenvironment inter- actions are still in development, opportunities now exist for population and fam- ily-based studies using biomarkers that integrate the search for susceptibility genes and the exogenous and endogenous exposures that cause cancer. While the methodologic challenges of âmolecular epidemiologyâ are formidable, this inter- disciplinary approach to cancer etiology should provide unprecedented opportu- nities to enlarge our understanding of environmental and genetic risk factors and their biological pathways, and set the stage for new clinical and public health strategies aimed at preventing and controlling cancer. CANCER DISPARITIES IN APPALACHIA Gilbert H. Friedell, M.D. Despite recent good news about decreasing U.S. cancer mortality rates, not all population subgroups are sharing in this success story. Progress toward meet- ing the cancer-related Healthy People 2010 goals will be hampered by the nationâs inability to deal effectively with the greater cancer burden borne by certain vulnerable populations. These âspecial populations,â defined as popula- tion groups at higher-than-average risk of death, disease, and disability, include people with low incomes, African Americans, Hispanics, American Indians, and other ethnic minorities.
80 CANCER AND THE ENVIRONMENT In addition, the National Cancer Institute (NCI) has stated that it considers rural residents to constitute âa special population.â Rural Americans tend to be older, poorer, less educated, and more likely to be uninsured than their urban counterparts. Rural communities have higher rates of chronic illness and disabil- ity, and report poorer overall health status than their urban neighbors. Residents in rural areas generally have less contact and fewer visits with physicians and, in general, lower levels of preventive care. In addition to factors related to rural health status and practices, there are systemic factors related to rural life in general (e.g., lack of public transportation and lower levels of other community services) that may also contribute to less than optimal cancer control. All of these factors are evident in the largely rural and predominantly white population of Appalachia, particularly in the Central Highlands, including the Appalachian counties of Ohio, West Virginia, Kentucky, Tennessee, and Vir- ginia. Lung cancer is a leading cause of male cancer deaths in central Appala- chia, with the highest rate in Appalachian Kentucky, the geographic area where the Behavioral Risk Factor Surveillance Survey data (BRFSS) indicate the high- est rates of cigarette smoking in the state. Cervical cancer mortality rates are also higher in central Appalachia than in the U.S. population as a whole. Data from the Kentucky cancer registry showed that the incidence of inva- sive cervical cancer and lung cancer in eastern Kentucky is higher than the incidence of these cancers in the overall Kentucky population and in the popula- tion covered by the Surveillance, Epidemiology, and End Results (SEER) pro- gram. It is, however, quite similar to the incidence of lung cancer and cervical cancer in the predominantly urban, African American population of Kentucky. Poverty is a common characteristic in much of this region. Some of the counties in Appalachian Kentucky, for example, are among the poorest in the country. In the same geographic areas, the level of literacyâindicated by the highest grade of formal schooling completedâis also lower than in most of the country. Problems associated with poverty are similar to those confronting poor populations in other parts of the country, but the latter are often characterized by race or ethnicity rather than by socioeconomic status (SES). This use of race and ethnicity as surrogates for poverty has obscured the fact that the problems related to cancer in the poor white population are comparable in many ways to those seen in recognizable minority populations. Individuals living in poverty often do not receive quality health care, includ- ing cancer prevention, diagnosis, treatment, and appropriate follow-up care, be- cause services are not available, accessible, and/or utilized. Behavioral risk fac- tors, such as tobacco use, poor nutrition, obesity, and underutilization of cancer screening examinations are more evident in impoverished populations. The so- cial environment in which poor people live also prevents the development of healthy behaviors. Freeman has pointed out that poverty âis a proxy for other elements of living, including lack of education, unemployment, substandard housing, poor nutrition, risk-promoting lifestyles and behaviors, and a dimin-
ABSTRACTS OF TALKS 81 ished access to health care,â all of which affect individual chances of developing cancer and surviving it. However, until cancer surveillance incorporates socio- economic status into its database, the relationship between poverty and cancer in population groups will be difficult to sort out. Data concerning income or other elements of SES generally are not col- lected by either hospital- or population-based cancer registries. It is therefore difficult to identify individuals whose income is below the poverty line. In Ap- palachia, however, and specifically in the defined geographic area of central Appalachia, a high proportion of the almost entirely white, largely rural popula- tion is poor. Research is hampered by the lack of access to necessary data. Currently in eastern Kentucky, the average levels of income, education, and other elements of SES can be determined, but obtaining this information on an individual basis for patients at the present time is quite difficult. The necessary data are generally not available in the medical records of cancer patients. Moreover, to have mean- ingful, population-based data for the purposes of comparison, this information would necessarily have to be a part of the medical record for all patients at both in-patient and ambulatory facilities. Some barriers to increased participation in cancer control programs exist at all socioeconomic levels, (e.g., lack of information about cancer and about both the availability and the benefit of cancer screening). Fear of what might be found during such an examination mitigates against women either gaining information about cancer or doing something with the information once it is obtained. Addi- tionally, health literacy continues to be a problem. The average reading level in this region is approximately at the fifth or sixth grade, making it difficult for individuals to understand and correctly respond to higher-level printed materials. Addressing issues such as these at the community level will be necessary. TUMOR SUPPRESSOR GENES: AT THE CROSSROADS OF MOLECULAR CARCINOGENESIS, MOLECULAR EPIDEMIOLOGY, AND HUMAN RISK ASSESSMENT Curtis C. Harris, M.D. Environmental, occupational, and recreational exposures to carcinogens con- tribute to cancer risk in humans. Cancer formation is a multistage process in- volving the activation of proto-oncogenes and the inactivation of tumor suppres- sor genes. Carcinogens can interact during any of these stages through genetic and epigenetic mechanisms. Mutational spectra of cancer-related genes (e.g., p53, BRCA-1, and p16INK4) may provide a molecular link between etiological agents and human cancer. Mutations in the evolutionarily conserved codons of the p53 tumor suppressor
82 CANCER AND THE ENVIRONMENT gene are common in diverse types of human cancer (Hollstein et al., 1994), and the mutational spectra differ among cancers of the colon, lung, esophagus, breast, liver, brain, reticuloendothelial tissues, and hemopoietic tissues. Analysis of these mutations may provide clues to the mutagenic mechanisms and the function of specific regions of p53 and generate hypotheses for investigation (Hussain and Harris, 1998a). Most transversions in lung, breast, and esophageal carcinomas are dispersed among numerous evolutionarily conserved codons within the p53 domain responsible for sequence-specific DNA binding and transcriptional ac- tivity. Transitions predominate in colon, brain, and lymphoid malignancies. Mu- tational hotspots at CpG dinucleotides in codons 175, 248, 273 of the p53 gene and condon 282 may reflect an endogenous mutagenic mechanism, (e.g., the deamination of 5-methylcytosine to thymidine). Oxyradicals and nitrogen oxyradicals may enhance the rate of deamination. For example, we have ob- served that (1) an increased production of nitric oxide (NO) by nitric oxide synthase-2 (NOS2) is associated with p53 cytosine to thymidine (C to T) transi- tions during colon carcinogenesis (Ambs et al., 1999); (2) p53 transrepresses basal and cytokine-induced NOS2 expression in vitro (Forrester et al., 1996) and in vivo (Ambs et al., 1998a); and (3) NO increases both the expression of the vascular endothelial growth factor and angiogenesis (Ambs et al., 1998b). p53 G:C to T:A (where G = guanosine and A = adenosine) transversions are the most frequent substitutions observed in cancers of the lung, breast, stomach, and liver, and are more likely to be due to bulky carcinogenâDNA adducts. G:C to T:A transversions also are more common in lung cancers from smokers compared to never smokers (Takeshima et al., 1993; Hussain and Harris, 1998b) and are more frequent in lung cancers from women compared to men (Guinee et al., 1995). The high frequency of G to T p53 mutations in the nontranscribed DNA strand is a reflection of strand-specific repair of the transcribed strand (Evans et al., 1993). The p53 gene may also contribute to DNA repair and apoptosis by proteinâ protein interactions with transcriptionârepair factors, XPB (ERCC3) and XPD (ERCC2), and in TFIIH (Wang et al., 1994, 1995a, 1996). A p53 mutation, allelic deletion and/or posttranslationally modified protein can be an early event in bronchial, mammary, or esophageal carcinogenesis (Bartek et al., 1990; Davidoff et al., 1991; Bennett et al., 1992; Sozzi et al., 1992; Sundaresan et al., 1992; Vahakangas et al., 1992; Nuorva et al., 1993) and may prove useful in the early diagnosis of cancer. In liver tumors from persons living in geographic areas where aflatoxin B1 (AFB) and hepatitis B virus (HBV) are cancer risk factors, most p53 mutations are at the third nucleotide pair of codon 249 (Hsu et al., 1991). A dose-dependent relationship between dietary AFB intake and codon 249ser p53 mutations is ob- served in hepatocellular carcinoma. Exposure of AFB to human liver cells in vitro produces 249ser (AGG to AGT) p53 mutants (Aguilar et al., 1993; Mace et al., 1997). The mutation load of 249ser mutant cells in nontumorous liver also is positively correlated with dietary AFB exposure (Aguilar et al., 1994). These
ABSTRACTS OF TALKS 83 results indicate that the expression of the 249ser mutant p53 protein provides a specific growth and/or survival advantage to liver cells. Because cellular context may influence the pathobiological effects of specific mutants of p53, the 249ser mutant may be especially potent in hepatocytes by the enhanced growth rate of p53-null HEP-3B cells by transfected 249ser mutant p53, and indicates a gain of oncogenic function (Ponchel et al., 1994). The 249ser mutant p53 is more effec- tive than other p53 mutants (143ala, 175his, 248trp, and 282his) in inhibiting wild- type p53 transcriptional activity in human liver cells (Forrester et al., 1995). One model concerning the generation of liver cancers with the 49ser mutation is the following: (1) AFB is metabolically activated to form the promutagenic N7dG adduct; and (2) enhanced cell proliferation due to chronic active viral hepatitis allows both the fixation of the G:C to T:A transversion in codon 249 of the p53 gene and selective clonal expansion of the cells containing this mutant p53 gene. HBV also has significant pathobiological effects. For example, the HBVX gene is frequently integrated and expressed in human hepatocellular carcinomas from high-risk geographic areas (Unsal et al., 1994; Paterlini et al., 1995). Hepatitis B viral gene products may form complexes with cellular transcription factors (e.g., ATF2; Maguire et al., 1991), upregulate transcription of cellular and viral genes (Twu and Schloemer, 1987; Spandau and Lee, 1988; Shirakata et al., 1989; Caselmann et al., 1990; Kekule et al., 1990) including NOS2 (Elmore et al., 1997), or activate the ras-raf-MAP kinase signaling cascade (Benn and Schneider, 1994). Inactivation of the p53 tumor suppressor gene functions, including DNA repair and apoptosis, may be another consequence of the cellular proteinâHBV oncoprotein complex formation. The HBVX protein binds to p53 (Pirisi et al., 1987; Wang et al., 1994; Ueda et al., 1995), sequesters it in the cytoplasm (Elmore et al., 1997), and inhibits its sequence-specific DNA binding and tran- scriptional activity (Wang et al., 1994). The HBVX protein also inhibits p53- dependent apoptosis (Wang et al., 1995b). In nucleotide excision DNA repair, the HBVX protein may modulate p53 function (Wang et al., 1995a; Jia et al., 1999), including the repair of AFB1âDNA adducts. HBV integration also could increase genomic instability, including abnormal chromosomal segregation, and increase the rates of DNA recombination (Hino et al., 1989, 1991). Three other associations between the p53 mutational spectra and carcinogen exposure have been observed. The induction of skin carcinoma by ultraviolet light is indicated by the occurrence of p53 mutations at dipyrimidine sites, in- cluding CC to TT double-base changes (Brash et al., 1991; Ziegler et al., 1994). The p53 mutational spectrum in radon-associated lung cancer from uranium min- ers also differs from lung cancer caused by tobacco smoking alone (Vahakangas et al., 1992; Taylor et al., 1994). Hepatic angiosarcomas induced by occupa- tional exposure to vinyl chloride have a high frequency of A:T to T:A p53 mutations when compared with sporatic angiosarcoma (Hollstein et al., 1994; unpublished results). In summary, these differences in mutational frequency and spectra among human cancer types indicate the following: (1) the etiological
84 CANCER AND THE ENVIRONMENT contributions of both exogenous and endogenous factors to human carcinogen- esis; (2) specific proliferative effects conferred by different mutant p53 genes in different human cell types; and (3) hypotheses for investigation (Hussain and Harris, 1998b). These genetic changes in the tumor suppressor genes also have implications for cancer diagnosis, prognosis, and therapy (Harris and Hollstein, 1993; Harris, 1996). The association of a suspected carcinogenic exposure and cancer risk can be studied in populations by classic epidemiologic techniques. However, these tech- niques are not applicable to the assessment of risk in individuals. A goal of molecular epidemiology is to integrate molecular biology, in vitro and in vivo laboratory models, biochemistry, and epidemiology to infer individual cancer risk (Harris, 1991; Shields and Harris, 1991; Perera and Santella, 1993; Perela, 1997; Ponder, 1997). Carcinogenâmacromolecular adduct levels and somatic cell mutations can be measured to determine the biologically effective dose of a carcinogen. Molecular epidemiology also explores host cancer susceptibilities, such as carcinogen metabolic activation, DNA repair, endogenous mutation rates, and inheritance of mutated tumor suppressor genes. Substantial interindividual variation for each of these biological end points has been shown (Harris, 1991) and highlights the need for assessing cancer risk on an individual basis. Given the pace of the past decade, it is feasible that future advances will allow molecu- lar epidemiologists to develop a cancer risk profile for an individual that in- cludes assessment of a number of exposure and host factors. This will help focus preventive strategies and strengthen quantitative risk assessments. GENETIC EPIDEMIOLOGY AS A TOOL FOR GENEâ ENVIRONMENT INTERACTIONS Kari Hemminki, M.D., Ph.D. Age-incidence relationships and experimental evidence suggest that cancer is a polygenic multifactorial disease (Armitage and Doll, 1954; Kinzler and Vogelstain, 1996). Tumors are monoclonal, implying that multiple hits need to affect a single clone of cells. Multifactorial diseases include an environmental component, which has been assumed to be the main cause in most types of cancer. The support for multistage carcinogenesis in vivo is limited. Almost all the known cancer syndromes are monogenic, and they conform to a two-stage model in requiring inactivation of the two copies of the tumor suppressor gene (Vogelstein and Kinzler, 1998). In this presentation, I consider the effects of multistage carcinogenesis for study of familial cancer, based on Swedish popula- tion and cancer registries. These sources of data allow estimations of the envi- ronmental and heritable contributions in the causation of cancers. It is estimated that some 1% of cancer is caused by the currently known cancer syndromes and up to 5% by highly penetrant single-gene mutations
ABSTRACTS OF TALKS 85 (Lynch et al., 1995; Vogelstein and Kinzler, 1998). These data apply to domi- nant Mendelian conditions, which can be assessed in family studies covering two or more generations. However, such studies provide no data on recessive Mendelian conditions and have a limited resolving power on polygenic condi- tions (Fearon, 1997). Consequently, apart from highly penetrant single-gene mutations, the estimation of the total hereditary contribution is extremely diffi- cult. The risks posed by low-penetrance single-gene mutations, polygenes, and recessive genes are poorly understood. Twin Model: Nonparametric Approach. Studies among twins are tradi- tional tools for dissecting questions about disease etiology, genes, and environ- ment. The twin model is particularly valuable because the mode of inheritance need not to postulated, (i.e., the approach is nonparametric, and the results are informative of the overall genetic effects). Polygenic effects are diluted among dizygotic twins but not among monozygotic twins. Twin studies invite genetic interpretation because monozygotic twins are genetically identical and dizygotic twins share half of their segregating genes. Thus, if monozygotic twins are more similar for a trait than dizygotic twins, genetic effects are likely to be involved. If there is twin similarity not accounted for by genetic effects, this indicates that shared environmental effects, (e.g., shared childhood experiences such as diet), contribute to variance in the trait. Yet the rareness of twinning has limited this approach to a few publications addressing the relative importance of genetic and environmental effects in cancer, the largest of these studies originating from the Nordic countries. A cancer study was carried out by pooling data from the Swedish, Finnish, and Danish twin registries for joint analysis (Lichtenstein et al., 2000). The aim of this study was to provide reliable estimates of genetic and environmental effects for the most common cancer sites and to assess the modification of such estimates by age at diagnosis. Data from 90,000 twins were combined to assess the cancer risks at 28 sites for co-twins of twins with cancer. A structural equa- tion model, MX, was applied in estimating the proportions of variation in can- cers due to environmental and inherited causes. The nonshared random environ- mental effect was the largest factor for all cancers, accounting for 58â82% of the total variation. Statistically significant heritability estimates were detected for cancers of the colorectum (35%), breast (27%), and prostate (42%). Estimates for the shared environmental effects ranged from 0 to 20%, but none were statis- tically significant. There were no significant differences between sexes at any of the sites. The Family-Cancer Database. The Swedish Family-Cancer Database now contains data on 10 million people, organized in families, and their 1 million cancers, retrieved from the Swedish Cancer Registry (Hemminki and Vaittinen, 1999). The Family-Cancer Database is the largest population-based data set ever
86 CANCER AND THE ENVIRONMENT used for studies on familial cancer. The database has been used in some 80 studies characterizing familial risks of various cancers. It has been used to model for cancer causation, using the same MX program employed in the above- mentioned twin study (Hemminki et al., 2001). Because of the overwhelming size of the data, most estimates for environmental and heritable causes were statistically significant. However, the results were not different from the twin studies. Environmental causes explained most of the total variation for all neo- plasms except thyroid cancer, for which heritable causes were largest. For ex- ample, the estimate for heritable causes of 25% in breast cancer was in line with the estimate from twins. Multiple Cancers. An increased occurrence of second primary cancers can result from intensive medical surveillance after first diagnosis, therapy-induced exposure to X-rays and carcinogens, and shared environmental and hereditary causes between the first and second cancer. I will take up only the aspect relating to genetic epidemiology, the hereditary risks possibly revealed in second can- cers. The risks for second cancer tend to be much higher than those for the first cancer. Very high risks for second cancer were noted for nose, skin (squamous cell carcinoma), connective tissue, and leukemia. We have analyzed the effect of family history on some cancers, such as breast cancer, and it is an important factor but affects a relatively small proportion of patients with second breast cancer (Dong and Hemminki, 2001; Vaittinen and Hemminki, 2000). The data suggest that patients with second cancer include a subgroup with a strong genetic predisposition to cancer, which often cannot be predicted by a family history. Such risks would be typical of polygenic diseases, and our tenet in the recent work on second cancers has been that they may serve as a unique population model for polygenic cancers (Dong and Hemminki, 2001; Hemminki and Mutanen, 2001). Interpretations. There is a consensus on the predominant importance of environmental factors and somatic events in human cancer, and the present re- sults on the twin and family sets quantified the effect of nonshared environment to range from 40 to 90% for different cancers. Nonshared environment encom- passes anything that is not hereditary and not shared between the relatives (spo- radic causes of cancer). It is of interest to note that this effect was large for some cancers of identified environmental causes, such as lung and cervical cancers. Shared environment, summing up common family experiences and habits of family member, accounted for 0â30% of etiology. The structural equation mod- eling carried out can accommodate both dominant and recessive Mendelian modes and polygenic modes of inheritance. Thus, the results on heritability sum- marize the total genetic effects. The recent data identified significant heritable effects for colorectal, breast, and prostate cancer. Heritabilities for these cancers
ABSTRACTS OF TALKS 87 were estimated to be between 15 and 40%, challenging previous estimates of the magnitude of genetic effects, based mainly on high-penetrance dominant condi- tions. For all cancer the genetic effect was 26%. Moreover, we found evidence for heritability of all main cancers, ranging from 1 to 50%. The frequencies of mutations in the known high-risk susceptibility genes BRCA1 and BRCA2 in breast cancer and DNA mismatch repair genes in HNPCC are so low that they explain at most 10% of the genetic effects noted (Peto et al., 1999; Salovaara et al., 2000). For prostate cancer, candidate genes have been mapped but not identi- fied. These findings suggest that other genes are yet to be identified, but because they are likely to be relatively common and of moderate risk only, the incrimina- tion will be difficult. We conclude that the overwhelming cause of cancer in these twin popula- tions was nonshared environment, accounting for some 70% of all cancer. Ana- lytical and molecular epidemiological studies provide tools to identify and quan- tify risk factors contributing to environmental effects. Etiologic clues may even be found in childhood environment or in long-lasting family habits, because the shared environment appeared to contribute to some forms of cancer in which common environmental risk factors have not been identified, including those of the gallbladder and corpus uteri. The most challenging result, however, was the large heritability of all cancer and of colorectal, breast, and prostate cancer in particular. The twin method probably detected recessive and polygenic cancers that are difficult to detect in other types of family studies. If these proportions were reliable, they would reveal major gaps in our understanding of the molecu- lar basis of heritable cancer. Polygenic Cancer and Future Study Designs. The data presented on twins, families, and second cancers provide additional support for the multistage theory of carcinogenesis. However, overall, direct support for the theory is limited even though a common belief is that it has been proven long since. If most cancers are indeed polygenic, this should be adequately considered in study designs for gene- mapping approaches. Linkage analysis in families of multiple affected individu- als does not suit polygenic diseases well, but it is still the main approach in attempts to identify cancer-related genes. Instead, large case-control studies on well-defined patient series may be more powerful (Risch and Merikangas, 1996). The studies can be carried out on patients without regard to family history be- cause otherwise the selection causes bias toward monogenic or oligogenic domi- nant cancers. Monozygotic twins and patients with multiple cancers would be very suitable study population, but identification of such patients may be cum- bersome.
88 CANCER AND THE ENVIRONMENT EPIDEMIOLOGY AND GENETIC SUSCEPTIBILITY TO BREAST CANCER3 Brian E. Henderson, M.D. and Kenneth T. Norris, Jr. In 2000, it is estimated that 184,200 women in the United States will be diagnosed with breast cancer, reflecting the high incidence rates now experi- enced by many racialâethnic groups including African American, Japanese, and white women. Much lower rates of breast cancer are found in multiethnic cohort (MEC) Latina women, due especially to its extremely low incidence in first- generation migrants. The increase in breast cancer rates for Asian women born in the United States compared to traditional Asian women (Parkin et al., 1992; Zeigler et al., 1993) is striking. While much of this increase may be explained by their transition to a more Western lifestyle (e.g., decrease in age of menarche and use of hormone replacement therapy), the fact that their cancer rate is as high as other racialâethnic groups despite their much lower postmenopausal weight (Probst-Hensch et al., 2000) suggests that both genetic and environmental differ- ences may be important. We have had a long-standing interest in the role of sex steroids in the etiol- ogy of breast cancer, especially related to estrogen and, more recently, progestin stimulation of the breast. This interest has been driven by the premise that estro- gen and progestin are the primary determinants of cell proliferation in breast epithelium and that cell proliferation is a prerequisite for many of the genetic changes necessary for a cell to transform to a malignant phenotype. The strong, consistent association between a womanâs menstrual history and breast cancer risk implicates lifetime exposure to sex steroid hormones as a major factor in the causation of breast cancer. A recent meta-analysis of epidemiological studies implicated estrogens more directly, by showing that circulating levels of the biologically most potent estrogen, estradiol (E2), are significantly higher in breast cancer patients compared to controls (Thomas et al., 1997). Moreover, we have found that plasma estrogen levels differ by racialâethnic group and that these differences appear to contribute to racialâethnic variation in breast cancer rates (Probst-Hensch et al., 1999). We (and others) have recently found that exog- enous exposure to these steroids, as combined estrogen and progestin replace- ment therapy, also substantially increases the risk of breast cancer (Ross et al., 2000). Androgens, as a precursor of estrogen biosynthesis or by direct action, also may contribute to breast carcinogenesis (Dorgan et al., 1997; Hankinson et al., 1998b). Three biosynthesis genes (CYP17, CYP19, HSD17B1) form the foundation of our current activities. These genes are important rate-limiting factors in estro- 3This work is support by NCI grant numbers CA63464 and CA54281.
ABSTRACTS OF TALKS 89 gen production. While our current focus is on estrogen biosynthesis, variants in other candidate genes in estrogen and progesterone biosynthesis, transport, me- tabolism, and binding are being explored jointly with our colleagues at the White- head InstituteâMassachusetts Institute of Technology (MIT). The CYP17 C allele is associated with elevated endogenous estrogen levels in both pre- and postmenopausal women, and based on data from the MEC we found that women carrying the C allele were at an increased risk of advanced breast cancer (Feigelson et al., 1997). Several polymorphisms have been identi- fied in HSD17B1, including a common polymorphism in exon 6 that results in an amino acid change from serine (allele S) to glycine (allele G) at position 312 (Normand et al., 1993; Mannermaa et al., 1994). Although current evidence indicates that this amino acid change may not affect the catalytic or immuno- logic properties of the enzyme (Puranen et al., 1994), an early report suggested that individuals who are homozygous for A are at increased risk of breast cancer (Mannermaa et al., 1994). Among MEC participants, in a preliminary analysis, women homozygous for allele A have been found to be 38% more likely to develop advanced-stage breast cancer than women homozygous for G. CYP17 and HSD17B1. In a combined analysis of CYP17 and HSD17B1, we found that carrying one or more âhigh-riskâ allele(s) increases the risk of advanced breast cancer in a doseâresponse fashion and that this relationship could be explained by endogenous serum estrogen levels (Siegelmann-Danieli and Buetow, 1999). The relative risk among women carrying four high-risk alle- les for CYP17 and HSD17B1 (C/C/S/S) was 2.76 compared to those carrying no high-risk alleles. Using an alternative model in which the risk of breast cancer was estimated per unit change in serum estrogen as predicted by genotype (using other data sources), we found an RR of 1.76 for the C/C/S/S genotype with comparable adequacy of fit to our data. We plan to extend these studies in order to evaluate this relationship within each racialâethnic group and in relationship to circulating levels of sex steroid hormones and binding proteins (E1[estrone], E2, bioavailable E2, SHBG, estrone sulfate, androstenedione, dehydroepiandrosterone [DHEA], DHEA sulfate, 5-androstene-, and androstenediol) in prediagnostic bloods collected in both Los Angeles and Ha- waii. CYP19. The CYP19 gene codes for the aromatase enzyme. Aromatase cata- lyzes the irreversible conversion of androstenedione to E1 and testosterone to E2. Tissue-specific aromatase expression is determined in part by tissue-specific pro- moters, which give rise to transcripts with unique 5â-noncoding termini that correspond to untranslated sections of exon I. These specific 5â-termini sequences are all spliced onto exon II at the common 3â-splice junction. The association of obesity with postmenopausal blood estrogen levels as well as breast cancer risk may at least in part be mediated by an increased aromatization of androstenedi-
90 CANCER AND THE ENVIRONMENT one in the adipose tissue of obese women (Huang et al., 1997; Siegelmann- Danieli and Buetow, 1999). Several polymorphisms have been described in the CYP19 gene (Sourdaine et al., 1994; Probst-Hensch et al., 1999; Healy et al., 2000). The functional relevance is unknown for all of the polymorphisms identi- fied so far. The only polymorphism in the coding region of CYP19 leading to a nonconservative amino acid substitution (exon VII, Arg264Cys) has no apparent effect on aromatase activity or response to aromatase inhibitors (Watanabe et al., 1997; Kristensen et al., 1998), nor has it been associated with breast cancer risk (Kristensen et al., 1998; Probst-Hensch et al., 1999). However, a possible role of a tetranucleotide repeat polymorphism of unknown functional relevance in breast cancer has been suggested by several studies (Rasmussen and Cullen, 1998; Haiman et al., 1999; Siegelmann-Danieli and Buetow, 1999), but we and others could not confirm this (Sourdaine et al., 1994; Probst-Hensch et al., 1999). The search for functionally relevant polymorphisms in CYP19 is currently focused on the regulatory regions of the gene, and we are studying these regions through collaborations with Dr. Nicole Probst at the University of Basel and Dr. Sue Ingles at the University of Southern California. GHâIGF Pathway. Insulin-like growth factors (e.g., IGF-1) also cause epi- thelial breast cells to divide, and it is hypothesized that the mitogenic effect of estrogen may be mediated through an IGF-1 signaling pathway (Holly, 1998). Members of the growth hormoneâIGF axis (GHâIGF) may exert a direct effect on breast cancer risk or potentiate the effects of estrogens on breast epithelium (Oh, 1998). IGFs have been recognized as major regulators of mammary epithe- lium and breast cancer cell growth (Clarke et al., 1997) and act as mitogens, as well as potent survival factors. In experimental studies, E2 has been found to stimulate breast epithelial cell proliferation in normal human breast tissue xeno- grafted into athymic mice via a paracrine mechanism involving IGFs (Hankinson et al., 1998a). Epidemiologic studies support a strong positive association be- tween IGF-1, IGF-binding protein-3 (IGFBP-3), and breast cancer risk (Bohlke et al., 1998; Byrne et al., 2000), with increased mammographic density being an important breast cancer risk factor (Rosen et al., 1998). Initial findings have been limited to premenopausal breast cancer. Among postmenopausal control women in the MEC, we found an association between circulating levels of IGF- 1 and breast cancer incidence by racialâethnic group. In the MEC, we will ex- plore the relationship between breast cancer risk, plasma IGF (free IGF-1, IGF- 1, IGF-2, IGFBP-3) levels in prediagnostic bloods, and genetic variants in the GHâIGF axis. A microsatellite repeat (CA) in the IGF-1 gene approximately 1 kilobase (kb) upstream from the transcription start site has been associated with lower mean levels of plasma IGF-1 (Rosen et al., 1998), and we are currently exploring the association between plasma IGF-1, the microsatellite repeat, and breast cancer risk among women in the MEC. We also will genotype these same women for a G to C transversion (exon I, position 2132) in the IGFBP-3 gene, of
ABSTRACTS OF TALKS 91 yet unknown functional relevance, that may serve as a marker for IGFBP-3 plasma levels. Moreover, we will expand our investigation of candidate genes in the GHâIGF pathway through our collaboration with the Whitehead Instituteâ MIT. While we have focused our major research effort on estrogen biosynthesis, genetic variation, and breast cancer risk, we have concomitantly evaluated germline missense variants in genes involved in breast cancer progression. We have confirmed a previous observation that a missense variant (Ile 655 Val) in Her-2/neu is associated with breast cancer risk and that this variant affects stages at diagnosis (McKean-Cowdin et al., in progress). We are pursuing similar mis- sense variants in BRCA1, BRCA2, and AT. MIGRANT FARMWORKER CHILDREN AT HIGH RISK FOR PESTICIDE EXPOSURE MarÃa A. HernÃ¡ndez-Valero, Dr.P.H. There are approximately 3â6 million migrant and/or seasonal farmworkers (MSFs) in the United States, appoximately 85â90% are from ethnic and racial minorities (e.g., Hispanics, African Americans). Hispanics of Mexican descent constitute the majority (approximately 92%) of the MSF population, with chil- dren and adolescents comprising 20â25% of the total number. The majority of the agricultural studies have been conducted among farm owners and operators, while cancer research among MSF is almost nonexistent. The lack of research is often attributed to the perceived difficulty in conducting epidemiological studies among this underrepresented population. This is true even though MSFs are chronically exposed to pesticides and other agricultural exposures, sometimes starting at a very young age when susceptibility may be of great importance. MSF children are chronically exposed to pesticides via their parentsâ occu- pation, including application drift; overspray; carry-home exposures from par- ents; exposure in utero; breast-feeding; going to or working in the fields with their parents; and the foods they eat. A pilot feasibility study was conducted to measure 21 organochlorine pesti- cides (OCPs) and correlate the measured levels with reported exposures in 62 Mexican American MSFs (36 children and 26 adults) home-based in the Hous- ton metropolitan area. The pilot sutdy objectives were to (1) provide quantitative measures of OCP levels in the serum of MSF adults and children; (2) obtain epidemiological data on work histories, field exposures, pregnancy, lactation, medical histories, and other factors from the standardized bilingual Migrant Farmworker Questionnaire developed by the National Cancer Institute; and (3) correlate the epidemiological data with the measured OCP levels. The referent laboratoryâs detection limit for OCPs was 0.3 ng/ml (parts per billion [ppb]).
92 CANCER AND THE ENVIRONMENT The study population was composed of 42% males and 78% females. The majority of the children or adolescents in the study were born in the United States (89%), with an average age of 12.4 Â± 7.5. The opposite was observed among the adults who for the most part were born in Mexico (73%), with an average age of 45.4 Â± 12.1. Two OCPs and/or metabolites (DDE and mirex) were detected in most of the samples tested. Five other OCPs were also detected in adults only (DDT, b-HBC, d-HBC, g-chlordane and oxychlordane). The aver- age DDE serum level for the children was low (1.6 Â± 1.6 ppb); the adultsâ level was higher than expected (15.4 Â± 17.2 ppb), almost five times higher than the referent laboratory populationâs average (3.2 Â± 1.8). The average mirex levels among both adults and children were almost identical (adults 1.8 Â± 0.6, children 1.7 Â± 0.7) and were also higher than expected, approximately eight times higher than the mean levels measured in the referent population (<0.3, nondetectable). In conclusion, traces of OCPs that had been banned in the United States for many decades are still measured in the serum of MSFs. OCPs, although decreas- ing in use, may still pose a threat to MSF children through their continued use and their persistence in the environment and body tissues. Since MSF children are potentially exposed to pesticides, both occupationally and environmentally through several pathways, there is a need to monitor this high-risk population. It may be necessary to legislate stricter public health measures aimed at reducing pesticide exposure among MSF children, and restricting these children from working or entering the fields with their parents. In regards to future research, (1) MSF children need to be included in prospective cohort studies to prove scientifically whether or not chronic exposure to pesticides from childhood into adulthood places humans at risk of developing deleterious health outcomes, in- cluding cancer during childhood and later in life, (2) the measurement of OCP levels should be included in ongoing and future studies; and also (3) the dietary intake of MSF children should be studied to evaluate the possible synergistic action of diet and pesticide exposure. CHEMICALS AND CHROMOSOMES, CHILDREN AND CANCER, CLUSTERS AND COHORTS IN A NEW CENTURY Richard Jackson, M.D., M.P.H. Eighty-six percent of the U.S. population says that environmental factors are either important or very important in causing disease. Thus, it shouldnât be sur- prising to researchers and policymakers that a cancer cluster is assumed (by the general public) to be environmental until proven otherwise. Cancer clusters can be the bane of existence for many state and local health officers and have been called the epidemiologists foolâs goldâthe idea being that you shouldnât spend your time chasing them because they never pan out. However, if one uses good
ABSTRACTS OF TALKS 93 communication skills, science, medicine, and policy, there are tremendous op- portunities to meet public health needs and achieve good outcomes. I studied many disease clusters during my time in the Public Health Depart- ment in California, a state that uses many types of pesticides. In fact, California uses about a quarter of the nationâs pesticides, and at least 5% of all the pesti- cides in the world. The uses of these pesticides become important as we discuss water pollution, air pollution, agriculture, economics, and workersâ health. We as researchers and government officials have had and continue to have many problems that hamper our ability to handle clusters effectively, including: lack of obligatory reporting of pesticides, unless they are Category Oneâextremely toxic; data gaps on particular chemicalsâchemicals that were thought to be safe but turned out to have health effects; and lack of disease registries to ascertain the baseline levels of a disease in a given region. Without this information, it is hard to be able to answer fundamental questions. Clusters also require a commu- nity relations specialists who not only speaks their language, but is able to under- stand their issues. In the future, approaching disease clusters will require collection of disease rates in registries because if we donât know the baseline level of any disease, it is impossible to know if the reported cluster is a statistical aberration or not. We also need better tracking of exposures, because the more accurately you docu- ment peopleâs precise level of exposures, the more accurately you can calibrate risks. The Centers for Disease Control and Prevention (CDC) has just started to release a report on a series of chemicals in the American people. We will con- tinue adding 25 chemicals each year until we have a total of 100 in four years. We are moving beyond the traditional two-by-two table investigations where we look at exposed, unexposed, disease, no disease. The era of looking at exposures and not looking at the genetic makeup of the individual in our epidemic studies is over. Trying to tease out nature versus nurture is a recipe for disaster. We need to be able to look at both of these at the same time. HEALTH DISPARITIES: DO GENEâENVIRONMENT INTERACTIONS PLAY A ROLE? Lovell Jones, Ph.D. When we discuss health disparities, I often remember the saying âif you always do what you have always done, we will always get what we already got.â When we approach the efforts to address the lack of measurable progress in tackling health disparities, we tend to fall back on what we did before. It may be under a different name or it may be packaged in a different box, but ultimately it is the same strategy. In a recent study, people questioned whether health disparities are âreal.â
94 CANCER AND THE ENVIRONMENT Consider the fact that breast cancer in Hispanic females has tripled over the last 15 years. African American females under the age of 35 have a 50% higher incidence of breast cancer than white females. These are just two examples. We need to understand that these are not access issuesâaccess to health care does not play a role in incidence. It is not an issue of poverty because it crosses all socioeconomic sectors. These disparities are growing and are becoming a major problem in this nation. If these trends and incidences were reported for white females, every alarm in this nation would be going off. â¢ What about other ethnic groups and other forms of cancer? Here are a few examples: â¢ Vietnamese women have cervical cancers at nearly five times the rate of white females. â¢ African American men under age 65 have nearly twice the rate of pros- tate cancer of white males. â¢ Hispanics, Native Americans, and Alaskan Natives are nearly twice as likely as Caucasians to have diabetes. â¢ In Pacific Islanders, certain cancers are 60 times more prevalent than in Caucasians. As we begin to address these disparities in terms of genetics and geneâ environment interactions, we will have to answer what role race and ethnicity will play. We need to understand that racial classifications are a social construct and not a biological construct. Yet racial classification still has an impact on the health of this nation. In the United States, we use the one-drop rule. If you have one drop of blood or if someone could trace one drop of African American bloods, you are African Americanâno matter your phenotype. Just as not all African Americans, Hispanics, and Asians may be similar culturally (within their ethnic groups), they may not have the same phenotypes. Most importantly, as we continue our research into geneâenvironment interactions, we need to remember that one size does not fit all. DIET AND OTHER ENVIRONMENTAL FACTORS AS MODIFIERS OF CANCER RISK John A. Milner, Ph.D. The last decade has witnessed important advances in the understanding of factors that influence cancer risk. Several environmental factors continue to sur- face as potentially instrumental in explaining the wide global variation in the incidence and biological behavior of tumors. Undeniably, exposure to environ- mental agents and/or ultraviolet radiation may contribute to oxidative stress or other biochemical events that foster uncontrolled cell proliferation. Clearly, diet
ABSTRACTS OF TALKS 95 is a significant environmental factor to which an individual is continually ex- posed. Although an individualâs diet may serve as a protector against the poten- tially lethal effects of reactive oxygen species and toxic environmental chemi- cals, under some circumstances it may also be a significant source of deleterious compounds. Thus, an individualâs diet may increase or decrease cancer risk de- pending on its composition. A variety of linkages between diet habits and cancer risk have surfaced in both epidemiological and preclinical studies. Some of the most compelling evi- dence linking diet and cancer comes from the epidemiological observation that increased vegetable and fruit consumption is associated with a reduction in the risk for cancers of the mouth and pharynx, esophagus, lung, stomach, colon, and rectum. Likewise considerable preclinical evidence points to a host of essential and nonessential nutrients as modifiers of cancer risk at a variety of sites. Part of this variation in cancer risk may arise from variation in the intake of one or more essential nutrients supplied by either plant or animal food sources. Vegetables derived from various parts of plants including roots (e.g., carrots, parsnips), leaves (e.g., spinach, lettuce), flowers (e.g., artichoke, broccoli), stalks (e.g., celery, rhubarb), and seeds (e.g., corn, peas), as well as a host of fruits, provide thousands of chemically diverse phytonutrients that may contribute to these ob- servations. Some of these phytonutrients, including flavonoids, carotenoids, organosulfides, and isothiocyanates, have been the focus of recent research to determine both their effects on risk and their mechanism(s) of action. While the risk of developing several cancers has been linked with dietary patterns, frequent inconsistencies are noted. These inconsistencies may reflect the multifactorial and complex nature of cancer, the specificity that individual dietary constituents have in modifying specific genetic pathways, and the tempo- ral relationship between dietary intervention and phenotypic changes in tumor incidence or behavior. Again, the complexity of defining the precise role of diet is magnified by the numerous and diverse essential (i.e., folate, selenium, vita- min E, n-3 fatty acids, and calcium) and nonessential (i.e., oligofructose, allyl sulfurs, carotenoids, flavonoids, and isothiocyanates) components that may alter one or more phases of the cancer process and the temporal and compensatory responses to these dietary constituents. Because of the chemical and biological diversity of dietary components and the range of molecular targets, the elucida- tion of the importance of diet is proving to be not only an exciting undertaking, but also an immense challenge. The past decade has witnessed great strides in understanding the biological basis of cancer. Discoveries that both essential and nonessential dietary nutrients can markedly influence several key biological events including cell cycle regula- tion, processes involved with replication/or transcription, immunocompetence, and factors involved with apoptosis have strengthened convictions that specific foods and/or components may markedly influence cancer risk. Unraveling which dietary component is most important in establishing cancer risk is a daunting
96 CANCER AND THE ENVIRONMENT task but should be easier with new gene and protein technologies. For instance, limonene (a monoterpene found particularly in citrus fruits) addition to tumor cells has been reported to enhance 42 genes and suppress another 58 genes. Since several of the identified genes are involved in the mitoinhibitory trans- forming growth factor (TGF) signal transduction pathway, this provides support for the hypothesis that monoterpenes may initiate mitoinhibitory and apoptotic signaling through a signal transduction-related mechanism. Similarly, studies have been undertaken with a variety of other nutrients including selenium, isothiocyanates and allyl sulfide. Dietary isothiocyanates; have been reported to modify at least 20 different gene products associated with cancer prevention. Knockout and transgenic animals are beginning to provide clues to the specific site of action of specific dietary components and should be used more exten- sively as tools for probing mechanisms of action. It should be noted that dietary allyl sulfur compounds can lead to a number of changes that may influence overall cancer risk including blocking nitrosamine formation, retarding carcino- gen bioactivation, blocking cell division, promoting apoptosis, and retarding an- giogenesis. Recent studies from our laboratory found that diallyl disulfide com- pounds lead to marked changes in more than 20 genes associated with the cancer process. Collectively these studies reveal that individual dietary components are capable of bringing about a host of intracellular changes that may influence cancer risk. The utilization of genomic technologies to evaluate the effects of nutrients offers hope in determining which cellular change is most important in bringing about a change in the incidence or behavior of a tumor. It should be noted that preclinical evidence suggests that diverse dietary constituents includ- ing selenium, allyl sulfur, genistein, and resveratrol can influence the same ge- netic pathways. Such common effects raise concerns about potential interactive and cumulative effects among nutrients. Thus, a reductionist approach to under- standing the role of diet in cancer prevention may produce oversimplifications and confusion. Preclinical evidence exists that such diverse dietary components as folate, allyl sulfur, genistein, and resveratrol can alter genes and pathways associated with tumor cell proliferation and apoptosis. Part of this protection may relate to their ability to prevent oxidative damage. Compounds suppressing oxidative stress have been reported to produce changes in c-fos, c-jun, and c-myc and the tumor suppressor gene p53, as well as genes coding for the syntheses of protec- tive molecules such as metallothioneins, glutathione, and stress proteins. Aston- ishing strides have been made in understanding how molecules and genetic path- ways differ in precancerous and malignant cells and from their normal counterparts. Capitalizing on the differences in cellular signatures that are char- acterized by active and inactive genes and cellular products should assist in determining who should and should not benefit from intervention strategies. Clearly such information will help clarify the reason for discrepancies among preclinical, epidemiological, and intervention studies.
ABSTRACTS OF TALKS 97 At least part of this variation in response to dietary components may relate to the consumerâs genetic profile. It is now becoming apparent that the preva- lence of polymorphisms is variable among studied populations and these differ- ences could influence the response to diet. For example, in a random sample of participants in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC Study), there was a low prevalence of polymorphisms in genes coding for activation (phase I) enzymes CYP1A1 (0.07) and CYP2E1 (0.02) and a high prevalence in genes coding for detoxification (phase II) enzymes GSTM1 (0.40) and NQO1 (0.20). Evidence exists that several genetic polymorphisms may modulate cancer risk through their influence on folate metabolism, including two polymorphisms of the MTHFR gene C677 CâT (alanine âvaline) and 1298 AâC (glutamateâalanine) and a polymorphism of MTR, the gene that codes for methionine synthase C2756 AâG (aspartateâglycine); all of these polymorphisms reduce enzyme activity. Epidemiologic studies have reported that when folate intake was adequate, colorectal cancer risk was reduced (abou 50%) in individu- als with the MTHFR 677TT genotype compared with the MTHFR 677CC geno- type, and the risk of adult acute lymphocytic leukemia (ALL) was reduced by 77%. Variation in the response to folate metabolism is not unique since other studies suggest that variation in receptors for vitamin D may also be linked to cancer risk. Considerably more information is needed about how genetic poly- morphisms influence the response to dietary components and ultimately cancer risk. Unquestionably cancer is intertwined with environmental factors including diet. Strategies to prevent cancer through modification of either diet or specific dietary patterns, although intriguing and likely a low-cost health care strategy, will probably not be uniformly effective for all individuals. A better understand- ing of geneânutrient interactions will be needed to unravel who might benefit most from dietary intervention and who might be placed at risk. Future research in nutrition and cancer prevention must give top priority to studies that seek to understand the basic molecular and genetic mechanisms by which nutrients in- fluence the various steps in carcinogenesis. While the challenges to researchers will be enormous, the potential rewards in terms of reducing cancer morbidity and mortality will be of an equally great magnitude. MOLECULAR PATHOGENESIS OF LUNG CANCER John D. Minna, M.D. We and others have hypothesized that clinically evident lung cancers have accumulated 10â20 different genetic abnormalities in dominant oncogenes and/ or tumor suppressor genes (TSGs) (Sekido et al., 1998; Fong et al., 1999). If true, this hypothesis has important ramifications for the clinic. For example, it should be possible to discover carcinogen-exposed respiratory epithelial cells
98 CANCER AND THE ENVIRONMENT with only a subset of these changes and intervene with very early treatment and/ or chemoprevention. A related hypothesis is that these changes are recurrent and common between different tumors. If true, this has implications for directing the search for specific diagnostic and therapeutic targets and indicating the likeli- hood that all of the changes are required for the malignant phenotype. There have been many studies published on searching for genetic abnormalities in lung cancer (Sekido et al., 1998; Fong et al., 1999). However, with few exceptions, these studies have not been global in nature either in testing for genome-wide abnormalities or in testing for multiple abnormalities in the same individual lung cancer. To approach these two hypothesis in a global and quantitative fashion, we have performed a high resolution (10 cM) genome-wide search for loss of heterozygosity (LOH, allele loss) in 36 lung cancer cell lines using 399 polymor- phic markers. Individual tumors averaged 17â22 chromosomal regions involved in frequent, recurrent allele loss (âhot spotsâ), and these regions were signifi- cantly different between small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC) (Girard et al., 2000). On average, 35% of the markers showed allele loss in individual tumors, with an average size of subchromosomal regions of loss of 50â60 cM. We found 22 different regions with more than 60% LOH, 13 with a preference for SCLC, 7 for NSCLC, and 2 affecting both histologic types. This provides clear evidence on a genome-wide scale that SCLC and NSCLC differ significantly in the TSGs that are inactivated during their patho- genesis. However, in all other aspects (e.g. fractional allele loss, number of breakpoints, number of microsatellite alterations) SCLC and NSCLC were not significantly different. The chromosomal arms with the most frequent LOH were 1p, 3p, 4p,4q, 5q, 8p, 9p (p16), 9q, 10p, 10q, 13q (Rb), 15q, 17p (p53), 18q, 19p, Xp, and Xq. We next conducted detailed high-density allelotyping (~5-cm level) on chro- mosome arms 3p, 4p, 4q, and 8p (average of 25 markers per chromosome arm) in 66 microdissected primary archival lung cancers (22 SCLC, 21 squamous, 22 adenocarcinomas), as well as microdissected respiratory preneoplastic lesions from patients with lung cancers and from cigarette smokers (Shivapurkar et al., 1999; Wistuba et al., 1999b, 2000a). Allelic losses of 3p were found in 96% of lung cancers and in 78% of the preneoplastic/or preinvasive lesions. The 3p allele losses were often multiple and discontinuous, with areas of LOH inter- spersed with areas of retention of heterozygosity. There was progressive in- crease in the frequency and size of 3p allele loss regions with increasing severity of histopathological preneoplastic/preinvasive changes. Analysis of all of the data indicated multiple regions of localized 3p allele loss. A panel of six markers in the 600-kb 3p21.3 homozygous deletion region showed loss in 77% of lung cancers (100% SCLC, 100% squamous, 90% adenocarcinomas), 70% of normal or preneoplastic/preinvasive lesions associated with lung cancer, and 49% of 47 normal, mildly abnormal, or preneoplastic/preinvasive lesions found in smokers without lung cancer (Wistuba et al., 2000a). This was in contrast to 0% loss in 18
ABSTRACTS OF TALKS 99 epithelial samples from seven never smokers. We have identified all of the genes in this completely sequenced region, and several of them appear to suppress the tumorigenic phenotype when introduced back into lung cancers with multiple other genetic lesions (Lerman and Minna, 2000). Of these, the best studied is the RASSF1A mRNA isoform of the RASSF1 locus (Dammann et al., 2000; Burbee et al., 2001). This gene is rarely mutated in lung cancer, but its expression is lost by promoter-acquired hypermethylation in ~90% of SCLCs and 30â40% of NSCLCs. Methylation in NSCLCs is associated with adverse survival and treat- ment of lung cancer cells with 5-azacytidine reactivates RASSF1A expression. This isoform contains a RAS binding domain and a putative diacylglycerol bind- ing domain and suppresses the tumorigenic phenotype of lung cancer cells (Dammann et al., 2000; Burbee et al., 2001). This 600-kb region and the 3p14.2 (FHIT/FRA3B) and 3p12 (U2020/DUTT1) regions were common, independent sites of breakpoints (Wistuba et al, 2000a). We conclude that 3p allele loss is nearly universal in lung cancer pathogenesis; involves multiple, discrete, 3p LOH sites that often show a âdiscontinuous LOHâ pattern in individual tumors; occurs in preneoplastic/preinvasive lesions of smokers with and without lung cancer; and frequently involves breakpoints in at least three very small, defined genomic regions. These findings are consistent with previously reported LOH studies in a variety of tumors showing allele loss occurring by mitotic recombination and induced by oxidative damage. Similarly, high frequencies of LOH (86% SCLC, 100% squamous, 81% adenocarcinomas) for the 8p21â23 regions were detected in primary lung cancers (Wistuba et al., 1999b). The LOH commenced early during the multistage development of lung cancer at the hyperplasia/metaplasia stage in cancer patients and in smokers without cancer. Of interest, 8p21â23 allelic losses always followed 3p and usually followed 9p allele loss. In contrast to 3p LOH, no 8p LOH was found in histologically normal epithelium; however, 15% of mildly abnormal, 50% of dysplastic, and 92% of carcinoma in situ le- sions had 8p21â23 LOH. Allelic loss occurred in 65% of smokers without can- cer and persisted for up to 48 years after smoking cessation. Frequent LOH of 4p and 4q markers was seen in SCLC and mesotheliomas (Region 1 4q33â34 >80%, Region 2 4q25â26 >60%, and Region 3 4p15.1â15.3 >50%) but was much less frequent (20â30%) in NSCLC where the most frequent pattern was loss of Re- gion 3 alone (Shivapurkar et al., 1999). For 3p, the regions of loss in SCLC and squamous cancers were usually quite large, often involving multiple markers, whereas those in adenocarcinomas were much smaller and usually involved only one or two markers. SCLC had significantly higher frequency of 4p and 4q allele loss (two separate regions on each arm) than did NSCLC (Shivapurkar et al., 1999). The converse was seen with the 8p allelotype (Wistuba et al., 1999b). Tumor-acquired promoter methylation is a new, important mutational mechanism for inactivating TSGs. We found tumor-acquired aberrant promoter methylation in nine genes in 107 resected primary NSCLCs (RAR beta 40%, TIMP-3 26%, p16ink4a 25%, MGMT 21%, FHIT 37%, DAPK 19%, ECAD 18%,
100 CANCER AND THE ENVIRONMENT p14ARF 8%, and GSTP1 7%) (Park et al., 1999; Virmani et al., 2000; Zochbauer- Muller, et al., 2001a, 2001b). At least one gene was methylated in 86% of NSCLCs, whereas normal lung from these same patients was methylated in only a few patients. In addition, we found 63/87 (72%) of SCLCs to exhibit RAR promoter hypermethylation. The methylation events occurred independently of one another. However, about 13% of the NSCLCs exhibited more frequent pro- moter hypermethylation and thus are candidates for having a âglobal CpG island methylator phenotype.â In nonmalignant bronchial epithelium 218 foci (195 of histologically normal or slightly abnormal epithelium and 23 of dysplastic epithelium) were studied from 19 surgically resected lobectomy specimens (Park et al., 1999). Thirteen (68%) of the nineteen specimens had at least one focus of bronchial epithelium with molecular changes. At least one molecular abnormality was detected in 32% of the 195 histologically normal or slightly abnormal foci and in 52% of the 23 dysplastic foci. Extrapolating from a two-dimensional analysis, we estimate that most clonal patches contain approximately 90,000 cells. Although, in a given individual, tumors appeared homogeneous with respect to molecular changes, the clonally altered patches of mildly abnormal epithelium were heterogeneous. Our findings indicate that multiple small clonal or subclonal patches containing molecular abnormalities are present in normal or slightly abnormal bronchial epithelium of patients with lung cancer. In detailed studies of bronchial epithe- lium and bronchial biopsies from current or former smokers without lung cancer, we also find thousands of clonal patches showing allele loss in histologically normal-appearing respiratory epithelium. In fact, these patches can be detected more than 30 years after cessation of cigarette smoking. This would suggest the potential for damaged stem cells to repopulate. We also investigated the relationship between the amount of smoking and the degree of methylation of the p16, RASSF1A, RAR, APC, and HCAD(CDH13) genes in more than 200 resected NSCLCs from Japan with known smoking history. We found that p16 and RASSF1A developed promoter hypermethylation related to the amount of cigarette smoking, while RAR, APC, and HCAD(CDH13) methylation occurred independently of the amount of cigarette smoking (ZÃ¶chbauer-MÃ¼ller et al., in preparation). To investigate whether methylation of genes such as RAR, HCAD(CDH13), p16, RASSF1A, and FHIT occurs in smoking-damaged epithelium before lung cancer develops, we analyzed oropharyngeal brushes, sputum samples, bronchial brushes, and bronchoalveolar lavage (BAL) samples from more than 100 heavy smokers without evidence of cancer (Zochbauer-Muller et al., 2001a). Methyla- tion of at least one gene was present in one or more specimens from nearly 50% of the smokers. However, the frequency of methylation of these genes found in the epithelial samples from heavy smokers was lower than the frequency found in primary lung cancers. These findings indicate that promoter-acquired methy-
ABSTRACTS OF TALKS 101 lation should be tested as an intermediate marker of risk assessment and re- sponse to chemoprevention. Small cell lung cancer has many distinct morphologic and biochemical fea- tures (such as neuroendocrine phenotype) distinguishing it from the non-small cell lung cancer histologic types (Sekido et al., 1998, 2001; Fong et al., 1999). These distinctions are of diagnostic importance and commit patients with differ- ent histologic types to quite different initial treatment regimens. With the excep- tion of bronchoalveolar lung cancer, all histologic types have smoking and to- bacco carcinogens as the major underlying etiologic factor. Clearly, SCLC etiology is strongly tied to cigarette smoking. Therefore, we have sought to answer the following major questions: Are there differences in the number or type of acquired molecular abnormalities between SCLC and NSLC; what are the specific genes involved; and what is the nature of the molecular changes that are found in smoking-damaged bronchial epithelium accompanying SCLC and NSLC? Finally, are there gene expression profiles that distinguish these two major histologic types? There is a wealth of information concerning molecular abnormalities in SCLC (Fong et al., 1999; Sekido et al., 2001). Ras mutations represent an obvi- ous difference, they are found in ~30% of NSCLCs (predominantly adenocarci- nomas) but, to our knowledge, have never been found in SCLCs (with more than 100 tumors analyzed). In fact, introducing a mutant ras allele into SCLCs in vitro has led to an alteration of the cellular phenotype to one more like NSCLC. A related component in the same signal transduction pathway is Her2/neu, which is overexpressed in ~30% of NSCLCs but rarely overexpressed in SCLCs. We have recently found a related signal pathway activation difference for the ERK/ MAP kinase pathway. While ERK1 and ERK2 proteins are expressed in all histologic types of lung cancer, we find constitutive activation (detection of the âactiveâ phosphorylated forms of ERK1 and ERK2 using specific antibodies) in 80% of NSCLCs but <5% of SCLCs. Autocrine growth factors involving neu- roendocrine regulatory peptides (e.g., bombesinâgastrin-releasing peptide) were first described for SCLC. However, recently it has become clear that both SCLC and NSCLC can express these peptides and their specific receptors (Sekido et al., 1998, 2001; Fong et al., 1999). While there are some differences related to histology (e.g., expression of neuromedin B in NSCLCs), it appears that both histologic types use this mechanisms. Myc oncogenes are overexpressed fre- quently in both SCLC and NSCLC. C-myc is overexpressed in both SCLC and NSCLC, but the overexpression of myc family members L-myc and N-myc is usually only found in SCLC. The p53 gene is frequently mutated in both SCLC and NSCLC, but this occurs in >90% of SCLCs and ~50% of NSCLCs. The other components of the p53 pathway (such as MDM2 and p14ARF) need to be studied. In comparing the type of mutations occurring in p53, there appear to be no differences (e.g., in nucleotide-type change or location in the p53 open reading frame) between SCLC
102 CANCER AND THE ENVIRONMENT and NSCLC, providing evidence that the carcinogenic insult was similar. An- other major difference is seen in the RB/p16 signaling pathway. This pathway is inactivated in the vast majority of all histologic types of lung cancer. However, the target mutated various dramatically between histologic types. In SCLC, Rb is inactivated in >90% of cases, with loss of protein expression usually occurring with truncating mutations. It is very rare for SCLCs to have mutations inactivat- ing the expression of p16. In contrast, NSCLCs inactivate p16 expression in ~50% of cases, while loss of expression of Rb protein occurs in <20% (Sekido et al., 1998). There appear to be no differences in mutational frequencies for inacti- vation of FHIT occurring in 50â70% of all lung cancers (Fong et al., 1999; Sekido et al., 2001). Finally, we have looked at the bronchial epithelium accompanying SCLC and NSCLC for the occurrence of clonal alterations using precise laser capture microdissection with subsequent allelotyping for polymorphic markers (Wistuba et al., 2000b). In NSCLC, we frequently find clones of cells with molecular abnormalities in histologically affected epithelium (e.g., carcinoma in situ, dys- plasia, hyperplasia) and occasionally in normal-appearing epithelium in the case of current or former smokers. In SCLC, these histologic preneoplastic changes were minimal. However, in studies of histologically normal respiratory epithe- lium, we found a severalfold increased rate of allele loss in SCLC compared to NSCLC patients. Thus, the smoking-damaged histologically normal epithelium associated with SCLC appeared âgenetically scrambledâ and had incurred sig- nificantly more damage than the epithelium accompanying NSCLCs. We con- clude that SCLCs and NSCLCs do not differ significantly in the number of genetic alterations that occur, however SCLCs do differ significantly from NSCLCs in the specific genetic alterations that occur. In addition, smoking- damaged bronchial epithelium accompanying SCLCs appears to have undergone significantly more acquired genetic damage than that accompanying NSCLCs. Future studies need to identify the specific genes involved at these multiple sites and determine whether these provide new tools for early molecular detection, for monitoring of chemoprevention efforts, and as specific targets for developing new therapies. We conclude from our global and quantitative studies that clinically evident lung cancers have acquired 20 or more clonal genetic alterations; SCLC and NSCLC have acquired different genetic lesions; alterations in 3p TSGs appear especially early, followed by changes in 9p, 8p, and then multiple other sites; tumor-acquired promoter hypermethylation is a frequent mutational mechanism in lung cancer; changes consistent with oxidative damage leading to mitotic recombination are frequently seen; smoking-damaged histologically normal epi- thelium, as well as epithelium with preneoplastic/preinvasive changes, has thou- sands of clonal patches containing genetic alterations; and correcting even single genetic abnormalities can reverse the malignant phenotype. All of these observa-
ABSTRACTS OF TALKS 103 tions are ready for translation into the clinic for new methods of diagnosis, risk assessment, prevention, and treatment. BREAST CANCER GENETICS: BRCA1 AND BRCA2 GENES Olufunmilayo Olopade, M.D. Breast cancer is a genetic disease, caused by spontaneous mutations in so- matic cells or by germline inheritance of mutations in breast cancer susceptibil- ity genes. Germline mutations in the BRCA1 or BRCA2 susceptibility genes result in breast cancers characterized by young age of onset, bilaterality, associa- tion with ovarian cancer and other tumor types, vertical transmission, and dis- tinct tumor phenotypes. Because breast cancer develops in 37â85% of women that carry BRCA1 or BRCA2 mutations, genetic testing of individuals with a high risk of familial breast cancer is an important part of a cancer control effort. Somatic genomic rearrangements that cause breast cancer include amplification of the HER-2/neu gene, which is associated with a poor prognosis, relative resis- tance to chemotherapy and tamoxifen, and sensitivity to Herceptin. Detection of HER-2/neu amplification in tumors is therefore an important factor in prognosis and choice of therapies. These examples reveal the clinical value of addressing the genetic basis of cancer and illustrate the importance of understanding genetic mechanisms in developing methods of cancer prevention, early detection, and targeted therapies. TYPES AND TRENDS OF CHILDHOOD CANCER; CANCER IN CHILDHOOD CANCER SURVIVORS Leslie Robison, Ph.D. In the United States, cancer is the leading cause of death due to disease among individuals between the ages of 1 and 20. The annual incidence rate is 15/ 100,000, which translates into a cumulative risk of cancer equivalent to 1 in 300 by the age of 20 years. The types of malignancies in children and adolescents differ from adults and include (annual rate per million and proportion): leukemia (37, 24.8%); lymphoma (24, 16.1%); brain and central nervous system (CNS) (25, 16.8%); neuroblastomoa (7, 4.7%); retinoblastoma (3, 2.0%); kidney, pre- dominantly Wilmsâ tumor (6, 4.0%); liver, predominaently hepatoblastoma (2, 1.3%); bone, primarily osteoscarcoma; and Ewingâs sarcoma (9, 6.0%); soft tis- sue, predominaently rhabdomyosarcoma (11, 7.4%); germ cell (10, 6.7%); and others (15, 10.0%). There are distinct age-specific patterns of incidence; most notable are the peaks in incidence of acute lymphoblastic leukemia that occur between the age
104 CANCER AND THE ENVIRONMENT of 3 and 6 years; the aggregation of neuroblastoma, retinoblastoma, and Wilmsâ tumor in children below the age of 5; the increasing incidence with age of lym- phoma; and the relatively constant incidence of brain and CNS malignancies. Overall, males have a higher rate of malignancies than females, which is attribut- able primarily to a higher incidence among males of lymphomas and acute lympohoblastic leukemia. In the 15â20 years of age group, females have a higher incidence of cancer than males. Further, the annual incidence (per million popu- lation) of childhood and adolescent cancers differs by race: Caucasian (161.7), black (124.6), Hispanic (145.6), Asian/Pacific Islander (136.8), and Native American (79.6). Observations during the past several decades have identified a modest, but consistent, increase in the incidence of childhood cancers. Secular trends have varied with specific categories, but the most consistent increases have been seen in acute leukemia and tumors of the central nervious system. The survival rate for childhood and adolescent cancer has increased dra- matically during the past three decades. Currently, more than 70% of individuals diagnosed with cancer before age 15 will survive five or more years from diag- nosis, with the majority being cured of their original malignancy. With these improvements in treatment and survival, it is estimated that approximately 1 in every 900 individuals in the United States between the age of 15 and 45 is now a survivor of childhood cancer. These survivors are, however, at increased risk for long-term complications of their initial cancer and subsequent therapy. Late se- quelae of childhood cancer can include an increased risk of second and subse- quent malignancies, as well as serious organ dysfunction and psychosocial ef- fects. As more patients survive and the length of follow-up grows, patterns of second and subsequent malignancies are being identified in survivors, including increased rates of breast cancer, thyroid malignancies, CNS tumors, and leuke- mia. CANCER TREATMENT BASED ON IMMUNE STIMULATION Steven Rosenberg, M.D., Ph.D. Tumor infiltrating lymphocytes (TILs) obtained from patients with mela- noma have been used to clone the genes encoding the antigens recognized by these TILs. TILs have been identified that can recognize unique cancer antigens on murine and human cancers including melanoma, breast cancer, colon cancer, and lymphoma. The major histocompatibility complex MHC restricted recogni- tion of human cancer antigens was detected by assaying panels of human leuko- cyte antigen (HLA)-typed target cells and by transfection into target cells of genes encoding the appropriate HLA specificities. In clinical trials of TILs ad- ministration, 36% of patients with metastatic melanoma underwent objective cancer remission. TILs trafficked to and accumulated in cancer deposits.
ABSTRACTS OF TALKS 105 GENETIC SUSCEPTIBILITY TO LUNG CANCER4 Margaret R. Spitz, M.D. Less than 20% of long-term smokers develop lung cancer by age 75. Geneti- cally determined factors that abrogate the effects of environmental carcinogens may explain differences in susceptibility. The challenge in quantitative risk as- sessment is to account for this interindividual variation in susceptibility to car- cinogens. Evidence of familial aggregation of lung cancer provides indirect sup- port for the role of genetic predisposition to lung cancer. These patterns of inheritance studies suggest that a small proportion of lung cancer is due to âlung cancer genesâ that are probably of low frequency, but high penetrance. However, the study of low-penetrance, high-frequency genes is likely to be more useful in elucidating the causal pathways for the vast majority of lung cancers. Lung cancer risk is dependent on the dose of tobacco carcinogens, which in turn is modulated by genetic polymorphisms in the enzymes responsible for carcinogen activation (e.g., myeloperoxidase) and detoxification (e.g., glu- tathione s-transferases), as well as by the efficiency of the host cells in monitor- ing and repairing tobacco carcinogen DNA damage. Individuals with susceptible genotypes (or adverse phenotypes) tend to develop lung cancer at earlier ages and with lower levels of tobacco exposure. On the other hand, the genetic com- ponent in risk tends to be lower at high dose levels, when environmental influ- ences overpower genetic predisposition. We have applied phenotypic assays to measure DNA repair capacity (DRC) by means of the in vitro host cell reactivation assay using plasmids damaged with benzo[a]pyrene. DRC was significantly lower in cases than controls, lower in women compared to men, and lower in younger compared to older cases. There was a statistically significant trend for increasing risk with decreasing DRC and an odds ratio (OR) of 2.54 (P < 0.05) for lung cancer in the least efficient repair stratum. The mutagen sensitivity assay, in which in vitro mu- tagen-induced breaks are quantitated as a measure of carcinogen sensitivity, has also been identified as a significant risk factor for lung cancer. Mutagens used are bleomycin and benzo[a]pyrene. Higher risk estimates are evident for current compared to former smokers and lighter smokers (less than one pack per day) compared to heavier smokers. There was a doseâresponse relationship with ad- justed ORs for increasing quartiles of induced chromatid breaks for both bleomycin sensitivity and benzo[a]pyrene sensitivity (trend P < 0.0001). On stratified analyses, the risk for both adverse phenotypes (suboptimal DRC and mutagen sensitivity) was fivefold. 4Supported by National Cancer Institute grants CA55769 (to M.R.S.) and CA 68437 (to W.K.H.).
106 CANCER AND THE ENVIRONMENT Genotypeâphenotype and dietâgene interactions are also being studied in- tensively. For example, while the GSTM1 null genotype was not an apparent independent risk factor for lung cancer, in the presence of low isothiocyanate intake, the OR for the GSTM1 null genotype was 2.33. There was no increased risk in any stratum for former smokers. It is most likely that multiple susceptibility factors must be accounted for to represent the true dimensions of geneâenvironment interactions. In the near fu- ture, microarray technology will facilitate the performance of large-scale, low- cost genotyping. The ethical, educational, social, and informatics considerations that will result are challenging. However, the ability to identify smokers with the highest risks of developing cancer has substantial preventive implications for intensive screening and smoking cessation interventions and for enrollment into chemoprevention trials. ENVIRONMENT AND BREAST CANCER Mary S. Wolff, Ph.D. Wide variations are seen in the incidence of breast cancer, both internation- ally and nationally among ethnic groups. In the United States, Hispanic women have lower rates of breast cancer than black and white women, and lower rates than Japanese Americans but higher than other women of Asian ancestry. Genet- ics, diet, and reproductive factors do not explain all of the differences. Indeed, women with inherited genetic predisposition may never suffer from breast can- cer, just as not all smokers get lung cancer. Environmental exposures have been implicated, but few specific agents are strongly related to breast cancer. How- ever, individuals respond differently to exposures in terms of their innate ability to metabolize chemicals. Therefore, diet, lifestyle, and adverse exposures must collaborate with susceptibility factors to incur risk for breast cancer. Breast cancer etiology is complex because tumorigenesis can arise from a combination of many different mechanisms over a very long time. Because breast cancer risk is strongly associated with reproductive hormones, any role for envi- ronmental exposures must act in concert with endogenous hormones. Environ- ment, genes, and hormones must work together at specific end pointsâextend- ing from perinatal mammary cell development, to onset of puberty, through birth, lactation, and menopauseâfollowing the course of tumor progression and, after diagnosis, prognosis for recurrence. In this age of generally low exposures, relative risks for main effects are low, and many environmental exposures and genetic variants alone are not strong risk factors for breast cancer. Combinations of exposures may obscure expo- sureâdisease relationships. Crucial exposures as well as critical reproductive end points may have happened years before a tumor is found. The second generation of studies will address whether complex genetic factors, hormonal milieu, or
ABSTRACTS OF TALKS 107 dietary intakes alter environmental risk factors. Such effects may be responsible for differences in breast cancer among racial and ethnic groups who may have risks related to genetic polymorphisms and excessive exposures. PRIORITIES AND SPECIAL POPULATIONS: TIES THAT BIND Armin Weinberg, Ph.D. It becomes clear as we examine charts and data on health such as the those in Healthy People 2000 that socioeconomic status (SES) plays a role in cancer and many other health care issues. Individuals and groups with a higher SES (1) can obtain better housing, (2) can live in better neighborhoods, (3) have opportu- nities to engage in healthy behaviors, (4) have better access to health care, and (5) can more readily participate in clinical trials. Thus, in order to understand clinical data as it relates to geneâenvironment interactions, we need to include analysis of the communities and SES. Studies today use descriptive phrases such as âspecial populations,â âprior- ity populations,â and âvulnerable populations.â These terms are used extensively in research and discussions, but there is a great diversity of opinion as to their definition. As we continue to use these phrases in our communities, research will have to refine the definitions to better describeânot labelâthese groups. Addi- tionally, they will have to provide a certain degree of flexibility to accommodate the subgroups that emerge. More than 2,500 individuals immigrate to the United States each dayâa trend that complicates geneâenvironment research. Many of these individuals are from Latin America and Mexico, as well as other countries throughout the world. As we start to include these foreign-born residents in our studies, we must pay attention to the fact that these individuals are mostly young and have had different exposures in their country of origin at a time when they were most vulnerable. Additionally, migratory patterns have shifted in the United States as individuals from different countries or regions of a country immigrate distinctly to geographic regions in the United States. The NCIâs national special population networks have been formed to con- sider factors related to these and other issues. The steering committee for one network Radas En Acum, which has research sites in California, Illinois, New York, Florida, and Texas, was formed to help establish both a research agenda and research priorities. In addition to the genetics and the geneâenvironment issues, language and health literacy will require special attention. As we try to close the gap in enrollment in clinical trials, these other issues will become more important. We will have to be sensitive to cultural differences in these communi- ties and ensure that material is available in many languages. Further, as we talk to communities about geneâenvironment interactions, we must communicate what this means, what we want, and what we hope to learn.