Cancer

Pharmacology of Chemoprevention

Carcinogenesis is a chronic and multistep process that results in malignancy. Malignant cells acquire the ability to invade or metastasize. Metastasis is often the first evidence of malignant disease. During the continuum of carcinogenesis, therapeutic interventions can be used to arrest or reverse this process. This is known as cancer chemoprevention. Effective cancer chemoprevention should suppress or block the clinical manifestation of malignancies by treating lesions before clinical signs or symptoms arise. A strong rationale for cancer chemoprevention as an attractive therapeutic strategy stems from considerable preclinical, clinical, and epidemiological findings. Cancer chemoprevention has been shown to be a valid clinical approach in the use of tamoxifen in a randomized phase III trial that demonstrated how this selective estrogen receptor modulator (SERM) reduced the risk of breast cancer in high-risk women. Other clinical trials have demonstrated favorable activity in the treatment of certain premalignant diseases, as will be discussed. Effective cancer chemoprevention should involve lifestyle, dietary, nutritional, pharmacological, and other interventions. This article focuses on the pharmacological basis for clinical cancer prevention by emphasizing candidate classes of agents that are promising for use in cancer chemoprevention.

I. INTRODUCTION

Cancers arise as a result of carcinogenesis, a chronic and multistep process. This stems from mutagenic damage to growth-regulating genes and their products that alters gene expression and ultimately confers changes that lead to the development of invasive or metastatic malignant disease. This process leads to progressive changes of cells that result in premalignancy and eventually overt malignancy. The steps defined in this process include (1) initiation, where DNA damage occurs; (2) promotion, where genetic or epigenetic alterations confer additional genomic damage; and (3) progression to invasive or metastatic disease. Conceivably, each of these steps would be attractive pharmacological targets for cancer chemoprevention.
It has been recognized that carcinogenic exposure leads to fields of altered cells that exist before malignancies are clinically evident. This concept was first proposed by Slaughter in 1953 and provided a basis for understanding how carcinogen-exposed cells that result from failure to repair genomic damage are of clonal origin. Some of these genetically altered cells may progress to a malignant phenotype. It is possible that markers of these carcinogenic changes at affected tissues will determine clones of genetically altered cells that are at especially high risk for malignant conversion. It is not known what precise cassette of carcinogenic or genetic changes are required for the maintenance or progression of premalignant lesions.       
Distinct changes may be required for each affected tissue or carcinogenic agent. These alterations likely involve dominant genetic changes through the activation of oncogenes and recessive genetic events through the inactivation of tumor suppressor genes. The underlying genetic properties of a cell may promote susceptibility for malignant transformation such as in inherited cancer-prone syndromes. Tumor- matrix interactions and neoangiogenesis also have important roles in the maintenance or progression of premalignant cells. Because carcinogenesis is of a chronic nature, pharmacological interventions are attractive to arrest or reverse these progressive changes evident following genomic damage. The concept of cancer chemoprevention, coined by Sporn, stresses therapeutic interventions at the earliest steps of carcinogenesis, as this would avoid the clinical consequences of malignancies.       
A clinical validation of the concept of cancer chemoprevention was shown through a randomized trial using the SERM tamoxifen to reduce breast cancer risk in high-risk women. Risk reduction was found only for hormone-sensitive breast cancers and provided a basis for Federal Drug Administration (FDA) approval. Undesirable estrogenic effects, such as those involving the endometrium, have resulted in an active search for a SERM that would preserve the chemopreventive effects of tamoxifen in the breast while retaining desirable estrogenic effects, such as those in preventing osteoporosis and promoting hypocholesterolemia. This clinical finding in reduction of hormone sensitive breast cancer risk has underscored the need to identify other pharmacological agents that would be effective in preventing hormone-resistant breast cancers.       
Many interventions could reduce specific cancer risk, including lifestyle changes, dietary interventions, and effective screening of high-risk individuals, as will be discussed elsewhere in this encyclopedia. This article focuses on chemopreventive agents that exert their actions through specific pharmacological mechanisms. Representative classes of chemoprevention agents will be discussed, including specific agonists or antagonists for members of the steroid receptor superfamily of nuclear receptors, selective cyclooxygenase-2 (COX-2) or inducible nitric oxide synthase (iNOS) inhibitors, and other agents.

II. PHARMACOLOGY

Pharmacological cancer chemoprevention strategies could target multiple steps during carcinogenesis. Agents can act by blocking DNA damage that occurs as an initiating step in carcinogenesis or by blocking or reversing the progression of premalignant cells that have already acquired genomic damage. Agents may also act at the promotion or progression steps of carcinogenesis. In targeting these steps of carcinogenesis, cell-stromal interactions, as well as neoangiogenesis, would play important roles in the development of invasive malignancy. An empirical approach to cancer chemoprevention has been replaced by clinical strategies that emphasize the mechanisms of action of candidate chemopreventive agents in the design of clinical trials. These strategies build on basic scientific insights into pathways involved in cancer chemoprevention.       
Several examples illustrate this point. Inducible COX-2 is involved in the synthesis of prostaglandins from arachidonic acid and is often activated during inflammation. Evidence exists for COX-2 as a therapeutic target for cancer chemoprevention. Genetic findings implicate a role for COX-2 in preventing colon carcinogenesis. Genetically modified mice have been engineered to harbor defects of the adenomatous polyposis coli (APC) gene, which results in intestinal adenomatous polyps as well as an increase in COX-2. The relevancy of this genetically modified mouse model to disease in humans is shown by the finding that COX-2 overexpression relative to adjacent normal tissues was found as frequent in clinical colon cancers.       
The role of COX-2 in colon carcinogenesis was examined further by engineering mice with defects in COX-2 and APC. In contrast to the increase in intestinal polyps observed in APC-deficient mice, a reduction in polyp formation occurred in mice deficient in COX-2. These and other studies provided a basis for clinical trials that target inducible COX-2 with pharmacological inhibitors in selected patients at risk for colon carcinogenesis. These findings were extended to the clinical setting. Beneficial clinical effects have been reported with selective COX-2 inhibitors in the treatment of patients with the familial adenomatous polyps (FAP) syndrome.


FIGURE 1 Structures of candidate cancer chemoprevention agents. These are representative pharmacological agents that target distinct pathways and exhibit specific mechanisms of action. Several interact with nuclear receptors, others affect enzymatic pathways, and some have mechanisms of actions that are currently under active investigation.

These clinical findings obtained in patients with an inherited risk for colon carcinogenesis set the stage for use of COX-2 inhibitors (Fig. 1) in other individuals with high risk for colon cancer. A further rationale for targeting COX-2 in colon carcinogenesis comes from epidemiological findings in individuals who chronically received nonsteroidal anti-inflammatory drugs (NSAIDs) and have reduced incidence of carcinogenesis. While NSAIDs do not selectively inhibit COX-2 rather than COX-1, these and other findings support the view that a selective COX-2 inhibitor would have a beneficial impact in colon cancer prevention. Selective COX-2 inhibitors would be expected to have reduced side effects than a nonselective inhibitor, as the constitutively expressed COX-1 is not targeted. This could favor chronic clinical administration as would be needed for the prevention of colonic or other tumors.      
Another example of successful targeting of a pathway involved in the suppression of carcinogenesis is found in the use of SERMs to reduce breast cancer risk. Administration of the SERM tamoxifen (Fig. 1) has been shown to reduce breast cancer risk in women at high risk for hormone-sensitive breast cancer. A positive proof of a principle randomized trial has provided a basis for additional clinical breast cancer prevention trials that test other SERMs that would exert the desired tissue estrogenic actions, such as in preventing osteoporosis, while antagonizing undesirable estrogenic effects that promote carcinogenesis in breast, uterine, or ovarian tissues. This possibility is under clinical study with the SERM raloxifene. Distinct raloxifene response elements exist and could contribute to different pharmacological effects of this SERM relative to others. In the future, other SERMs will be examined in relevant clinical trials. These could exploit the fact that a second estrogen receptor (ER ) has been identified that may exert distinct biological effects.
An analogous chemopreventive approach could target the androgen receptor in prostate cancer. Prostatic intraepithelial neoplasia (PIN) has been identified as a precursor lesion in prostatic carcinogenesis. Whether antiandrogen-based strategies will prevent the progression of PIN to prostate cancer is the subject of future work in this field. The development of transgenic prostate cancer models should be of assistance in evaluating the efficacy and activities of pharmacological agents that target the androgen receptor in prostate carcinogenesis.      
Antiproliferative, differentiating-inducing as well as proapoptotic agents can target carcinogenesis. The retinoids, derivatives of vitamin A, are a class of prevention agents that could exert desired clinical chemoprevention effects by targeting these and other biological pathways. The retinoids are natural and synthetic derivatives of vitamin A that have diverse chemical structures, pharmacological properties, nuclear receptor affinities, and associated toxicity profiles.      
A strong rationale for a role of retinoids in cancer therapy or prevention stems from results obtained from experimental animal models, epidemiological studies, and clinical trials. Wolbach and Howe focused initial attention on vitamin A-dependent pathways as important in epithelial cell homeostasis in 1925. These investigators found that vitamin A deficiency in rodents caused squamous metaplasia in the trachea as well as at other epithelial sites. Notably, correction of this deficiency by vitamin A treatment reverses these metaplastic changes. These metaplastic changes are similar to those that arise in smokers, implicating a role for vitamin A-dependent signals in suppressing lung carcinogenesis. Further evidence for an association between vitamin A and cancer incidence stems from epidemiological data demonstrating an inverse relationship between vitamin A levels and incidence of cancer at specific epithelial sites.      
These and other findings provided a basis for use of retinoids in cancer prevention. Additional support for a retinoid role in cancer prevention derived from clinical trials conducted using retinoids that resulted in the successful treatment of certain premalignant conditions such as oral leukoplakia, cervical dysplasia, and xeroderma pigmentosum. Other clinical trials revealed retinoid activity in reducing some second primary cancers. These include independent retinoid trials that demonstrated a reduction in second aerodigestive tract cancers in patients having prior head and neck, lung, or hepatocellular carcinomas. In contrast to these promising trials, a randomized intergroup trial conducted in subjects treated with 13- cis retinoic acid following resection of stage I lung cancers did not show clinical benefit in smokers, although a reduction in second cancers was observed in subjects who never smoked. These findings, when coupled with those reported in large randomized trials using β-carotene in primary lung cancer prevention in high-risk individuals, indicate that a negative clinical interaction can exist when a chemopreventive agent is administered to active smokers. There is a need to combine lung cancer prevention agents with smoking cessation.      
Mechanisms responsible for the reported reduction of second cancers by retinoid treatment of nonsmokers need to be determined. A better understanding of relevant mechanisms should prove useful in the selection of the optimal retinoid for use in cancer chemoprevention. Two classes of retinoid nuclear receptors exist. These are the retinoid acid receptors (RARs) and the retinoid X receptors (RXRs). These share homology with other members of the steroid receptor superfamily of nuclear receptors, which include the glucocorticoid receptor, vitamin D receptor, and estrogen receptor, among others. There are three subtypes of RARs (RARα, RARβ, and RARγ) and RXRs (RXRα, RXRβ, and RXRγ) and several isoforms exist. Orphan nuclear receptors have been identified where the physiological ligands remain to be discovered.      
The ligand-binding domain of individual retinoid nuclear receptors is where specific retinoids bind. These nuclear receptors also contain DNA-binding domains that recognize defined responsive elements in genomic DNA. Following these ligand-receptor and receptor-DNA interactions, direct target genes that signal retinoid biological effects are activated or repressed. Retinoid nuclear receptors can heterodimerize or homodimerize and associate with two classes of coregulator proteins known as inhibitory corepressors and stimulating coactivators. Protein- protein interactions between retinoid receptors and their coregulators provide another level of regulation to the retinoid signaling pathway, as these can affect the basal transcriptional machinery through chromatin remodeling via changes in the state of acetylation. Coregulators represent additional pharmacological targets in cancer prevention.      
Pharmacological agonists and antagonists have been engineered to affect specific components of the retinoid signaling pathway. For instance, all-trans retinoic acid is an agonist for the RAR but not the RXR pathway, whereas the ligand 9-cis retinoic acid is bifunctional, activating the RAR and RXR pathways. An RXR agonist, known as a rexinoid, has been approved for clinical use by the FDA. Other retinoids target the AP-1 transcription factor. Some retinoids, such as N-(4-hydroxyphenyl)retinamide (4HPR) act through receptor-independent mechanisms (Fig. 1) and preferentially signal apoptosis in responsive cells.      
Randomized cancer chemoprevention clinical trials have emphasized the use of classical retinoids that activate the RAR pathway. Because repressed expression of RARβ is frequent in several epithelial cancers, including lung cancers, this could contribute to the clinical chemopreventive effects observed in subjects entered into trials to reduce primary or second lung tumors. The mechanisms responsible for RARβ repression are under active study. Preclinical evidence points to a role for methylation-induced silencing of this nuclear receptor. Perhaps demethylation agents that target RARβ sequences could be used in conjunction with the optimal retinoid to overcome RARβ repression and elicit the desired clinical chemoprevention effects. Clinical cancer chemoprevention studies could consider the use of retinoids that do not activate the classical retinoid signaling pathway. This approach might bypass a common defect observed in aerodigestive tract tumors, the suppression of RARβ.      
There is a need for additional candidate cancer chemoprevention agents that target specific cellular pathways. A partial list of candidate cancer chemoprevention agents appears in Fig. 1. In addition to pharmacological agents designated as retinoids, rexinoids, or SERMs, agents that act through other nuclear receptors include those affecting the vitamin D receptor (known as deltanoids) and those acting through PPAR-γ. One promising class of potential chemoprevention agents is synthetic triterpenoids, which are derivatives of natural products, known as cyclosqualenoids. Triterpenoids exhibit potent differentiation-inducing, antiproliferative, and antiinflammatory activities. Pertinent to their potential role in cancer chemoprevention, one of the synthesized triterpenoids known as CDDO (Fig. 1) suppresses induction of the inflammatory enzymes iNOS and COX-2. Whether these findings will be extended into the setting of clinical cancer chemoprevention is the subject of ongoing work.

III. CLINICAL CONSIDERATIONS

Clinical cancer chemoprevention trials have features distinct from therapeutic trials. To exert the desired clinical effects, chemoprevention agents are often administered on a chronic basis and should have few, if any, associated clinical toxicities. For individuals who are at increased risk for cancer, primary cancer prevention with chemopreventive agents, when coupled with lifestyle or dietary changes, would be an attractive approach to reduce cancer risk. Even in individuals at high risk for a primary cancer, a cancer chemoprevention agent would not be clinically adopted when clinical side effects are evident when used in cancer chemoprevention. In contrast, subjects who have already had a cured primary cancer may accept some side effects of chemoprevention agents if this would reduce the risk of a second primary cancer. Because candidate cancer chemopreventive agents are often selected for use based on mechanisms of action, one way to limit clinical toxicities of cancer chemopreventive agents would be through combination therapy. Agents targeting different chemopreventive pathways would each be administered at dosages lower than when these are used as single agents. This could yield more than additive chemopreventive activities while retaining acceptable clinical toxicity profiles. Perhaps synergistic clinical actions would be exerted by pharmacological agents that affect distinct chemopreventive pathways.      
Cancer chemoprevention trials are of a large size and require long clinical follow-up. If clinical outcome is the sole end point for the assessment of chemopreventive activity, then progress in this field will not be rapid. For this reason, biomarkers or intermediate end points have been proposed as ways to assess chemoprevention responses even before the clinical outcome is known. Biomarkers and intermediate end points are indicative of changes that increase the risk of cancer development in affected cells or tissues. Examples could indicate genomic instability that leads to additional chromosomal abnormalities (such as aneuploidy or loss of heterozygosity), cell cycle deregulation that alters the proliferative state, or changes in transcription due to the basal methylation or acetylation status of the genome, among other changes. Specific genetic alterations that occur in carcinogenesis include those affecting oncogenes (ras family, myc family, epidermal growth factor receptors, and others) or tumor suppressor genes, such as p53. These changes might be targets for cancer chemoprevention or surrogate markers for response to cancer chemopreventive agents.

IV. CONCLUSIONS

Pharmacological interventions can be used to reverse or arrest the progression of carcinogenesis at specific cell or tissue sites. Cancer chemoprevention is an attractive approach to reduce the societal burden of cancer by treating carcinogenesis before lesions become clinically evident. Given the chronic nature of interventions for cancer chemoprevention, pharmacological agents should be administered with few, if any, associated clinical toxicities. Biomarkers or intermediate end points could prove useful to identify chemopreventive targets as well as highlight those changes that would place cells or tissues at high risk for malignant transformation. Changes in these markers represent potential surrogate end points for clinical cancer chemoprevention trials. In the near term, as the clinical cancer chemoprevention field advances, it will be important to understand how preventive agents act and when they should be administered for primary or secondary cancer chemoprevention.

Acknowledgments
This work was supported in part by NIH RO-1-CA8756 (E.D.), RO1-CA62275 (E.D.), RO1-CA78814 (M.B.S.), the Department of Defense Grants DAMD17-99-1-9168 and DAMD17- 98-1-8604, the American Cancer Society Grant RPG-90-019- 10-DDC (E.D.), the National Foundation for Cancer Research (M.B.S.), and the Oliver and Jennie Donaldson Trust. M.B.S. is the Oscar M. Cohn Professor. We thank Dr. Nanjoo Suh, Dartmouth Medical School, for helpful consultation and Ms. Ann Frost for expert editorial assistance.

Ethan Dmitrovsky
Michael B. Sporn
Dartmouth Medical School

See Also
CELLULAR RESPONSES TO DNA DAMAGE ; CHEMOPREVENTION TRIALS ; ESTROGENS AND ANTIESTROGENS ; MOLECULAR EPIDEMIOLOGY AND CANCER RISK ; MOLECULAR MECHANISMS OF CANCER INVASION ; MULTISTAGE CARCINOGENESIS ; NUTRITIONAL SUPPLEMENTS AND DIET AS CHEMOPREVENTIVE AGENTS ; TUMOR CELL MOTILITY AND INVASION

GLOSSARY
biomarkers and intermediate end points Markers of the carcinogenesis process that highlight cells or tissues at risk for malignant conversion; these may serve as surrogates of response in clinical cancer chemoprevention trials.

cancer chemoprevention Use of dietary, nutritional, or pharmacological interventions to inhibit development of invasive cancer by blocking DNA damage that initiates carcinogenesis or by arresting or reversing the progression of premalignant cells that have already acquired genomic damage.

premalignancy Cells or tissues that are at an intermediate step in the carcinogenesis process and have acquired some, but not all, features of transformation; these cells are diagnosed based on histopathologic features and often exhibit genetic changes.

primary prevention Therapeutic interventions to prevent primary cancers from arising in high-risk individuals.

secondary prevention Therapeutic interventions to prevent second cancers from arising in patients cured of a primary cancer.

Bibliography
Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study Group. (1994). The effect of vitamin E and beta carotene on the incidence of lung cancer and other cancers in male smokers. N. Engl. J. Med. 330, 1029-1035.
Fisher, B., Costantino, J. P., Wickerham, D. L., Redmond, C. K., Kavanah, M., Cronin, W. M., Vogel, V., Robidoux, A., Dimitrov, N., Atkins, J., Daly, M., Wieand, S., Tan- Chiu, E., Ford, L., and Wolmark, N. (1998). Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J. Natl. Cancer Inst. 90, 1371-1388.
Hennekens, C. H., Buring, J. E., Manson, J. E., Stampfer, M., Rosner, B., Cook, N. R., Belanger, C., LaMotte, F., Gaziano, J. M., Ridker, P. M., Willett, W., and Peto, R. (1996). Lack of effect of long-term supplementation with beta carotene on the incidence of malignant neoplasms and cardiovascular disease. N. Engl. J. Med. 334, 1145-1149.
Hong, W. K., Endicott, J., Itri, L. M., Doos, W., Batsakis, J. G., Bell, R., Fofonoff, S., Byers, R., Atkinson, E. N., Vaughan, C., Toth, B. B., Kramer, A., Dimery, I. W., Skipper, P., and Strong, S. (1986). 13-cis-retinoic acid in the treatment of oral leukoplakia. N. Engl. J. Med. 315, 1501-1505.
Hong, W. K., Lippman, S. M., Itri, L. M., Karp, D. D., Lee, J. S., Byers, R. M., Schantz, S. P., Kramer, A. M., Lotan, R., Peters, L. J., Dimery, I. W., Brown, B. W., and Goepfert, H. (1990). Prevention of second primary tumors with isotretinoin in squamous-cell carcinoma of the head and neck. N. Engl. J. Med. 323, 795-801.
Kraemer, K. H., DiGiovanna, J. J., Moshell, A. N., Tarone, R. E., and Peck, G. L. (1988). Prevention of skin cancer in xeroderma pigmentosum with the use of oral isotretinoin. N. Engl. J. Med. 318, 1633-1637.
Lippman, S. M., Lee, J. J., Karp, D. D., Vokes, E. E., Brenner, S. E., Goodman, G. E., Khuri, F. R., Marks, R., Winn, R. J., Fry, W., Graziano, S. L., Gandara, D. R., Okawara, G., Woodhouse, C. L., Williams, B., Perez, C., Kim, H. W., Lotan, R., Roth, J. A., and Hong, W. K. (2001). Randomized phase III intergroup trial of isotretinoin to prevent second primary tumors in stage I non-small-cell lung cancer. J. Natl. Cancer Inst. 93, 605-618.
Meyskens, F. L., Jr., Surwit, E., Moon, T. E., Childers, J. M., Davis, J. R., Dorr, R. T., Johnson, C. S., and Alberts, D. S. (1994). Enhancement of regression of cervical intraepithelial neoplasia II (moderate dysplasia) with topically applied all-trans-retinoic acid: A randomized trial. J. Natl. Cancer Inst. 86, 539-543.
Muto, Y., Moriwaki, H., Ninomiya, M., Adachi, S., Saito, A., Takasaki, K. T., Tanaka, T., Tsurumi, K., Okuno, M., Tomita, E., Nakamura, T., and Kojima, T. (1996). Prevention of second primary tumors by an acyclic retinoid, polyprenoic acid, in patients with hepatocellular carcinoma. Hepatoma Prevention Study Group. N. Engl. J. Med. 334, 1561-1567.
Nason-Burchenal, K., and Dmitrovsky, E. (1999). The retinoids: Cancer therapy and prevention mechanisms. In "Handbook of Experimental Pharmacology" (H. Nau and W. Blaner, eds.), Vol. 139. pp. 301-322.
Springer, Berlin. Omenn, G. S., Goodman, G. E., Thornquist, M. D., Balmes, J., Cullen, M. R., Glass, A., Keogh, J. P., Meyskens, F. L., Valanis, B., Williams, J. H., Barnhart, S., and Hammar, S. (1996). Effects of a combination of beta carotene and vitamin A on lung cancer and cardiovascular disease. N. Engl. J. Med. 334, 1150-1155.
Pastorino, U., Infante, M., Maioli, M., Chiesa, G., Buyse, M., Firket, P., Rosmentz, N., Clerici, M., Soresi, E., Valente, M., Belloni, P. A., and Ravasi, G. (1993). Adjuvant treatment of stage I lung cancer with high-dose vitamin A. J. Clin. Oncol. 11, 1216-1222.
Slaughter, D. P., Southwick, H. W., and Smejkal, W. P. (1953). "Field cancerization" in oral stratified squamous epithelium: clinical implications for multicentric origin. Cancer 6, 963-968.
Sporn, M. B., Dunlop, N. M., Newton, D. L., and Smith, J. M. (1976). Prevention of chemical carcinogenesis by vitamin A and its synthetic analogs (retinoids). Fed. Proc. 35, 1332-1338.
Steinbach, G., Lynch, P. M., Phillips, R. K. S., Wallace, M. H., Hawk, E., Gordon, G. B., Wakabayashi, N., Saunders, B., Shen, Y., Fujimura, T., Su, L.-K., and Levin, B. (2000). The effect of celecoxib, a cyclooxygenase-2 inhibitor, in familial adenomatous polyposis. N. Engl. J. Med. 342, 1946-1952.
Suh, N., Wang, Y., Honda, T., Gribble, G. W., Dmitrovsky, E., Hickey, W. F., Maue, R. A., Place, A. E., Porter, D. M., Spinella, M. J., Williams, C. R., Wu, G., Dannenberg, A. J., Flanders, K. C., Letterio, J. J., Mangelsdorf, D. J., Nathan, C. F., Nguyen, L., Porter, W. W., Ren, R. F., Roberts, A. B., Roche, N. S., Subbaramaiah, K., and Sporn, M. B. (1999). A novel synthetic oleanane triterpenoid, 2-cyano-3,12- dioxoolean-1,9-dien-28-oic acid, with potent differentiating, antiproliferative, and anti-inflammatory activity. Cancer Res. 59, 336-341.
Wolbach, S. B., and Howe, P. R. (1925). Tissue changes following deprivation of fat-soluble vitamin A. J. Exp. Med. 42, 753-777.

Chemoprevention Trials

Cancer prevention efforts have been a long and arduous process. As the biological basis for carcinogenesis continues to be elucidated, different strategies for prevention have emerged. The success of recent clinical trials designed to prevent cancer in patients who are at increased risk of cancer (cancer "chemoprevention" trials) suggests that chemoprevention is a rational and appealing treatment strategy. Success in the prevention of epithelial cancers suggests that chemopreventive agents can interrupt the carcinogenic process. This article focuses on current chemoprevention research both laboratory and clinical.      
Chemopreventive agents, how they might interrupt the biological effects of genetic changes to inhibit abnormal epithelial cell growth and differentiation, and the genetic events associated with epithelial carcinogenesis are discussed and the usefulness of intermediate biomarkers as markers of premalignancy is reviewed. Finally, chemoprevention trials are analyzed with emphasis on strategies of trial design and clinical outcome and future directions in chemoprevention are proposed that are based on recently acquired mechanistic insight into carcinogenesis and chemoprevention.

I. INTRODUCTION

Epithelial cancers have historically been a cause for frustration by both patients and practitioners. Since the mid-1970s, the mortality rate from epithelial malignancies has improved only slightly despite advancements in cancer therapeutics. Although some patients will present with early stage epithelial malignancies, the majority of these patients will present with locally advanced or metastatic disease. Surgical intervention with adjuvant treatment, including radiotherapy and/or chemotherapy, has offered some improvements in long-term survival rates. However, local recurrence and especially the development of second primary tumors have impacted both morbidity and mortality in this patient population. Thus, a novel approach to these epithelial cancers is needed. Epithelial carcinogenesis is a multistep process in which an accumulation of genetic events leads to a progressively dysplastic cellular appearance, deregulated cell growth, and finally carcinoma.

II. CHEMOPREVENTION

Chemoprevention can be defined as the use of specific natural or synthetic chemical agents to reverse, suppress, or prevent carcinogenic progression to invasive cancer. Sporn first defined the term chemoprevention in 1976. Cancer chemoprevention is a rapidly developing field that approaches carcinogenesis from a different perspective. Previously, early detection techniques were employed to reduce morbidity and mortality with respect to cancer treatment. In lung cancer, this included early chest X-ray and sputum cytology analysis in individuals at high risk. Despite these early detection techniques, overall mortality did not improve. Chemoprevention bridges basic biologic research with clinical chemical intervention and attempts to halt the process of carcinogenesis. Its principles are based on the concepts of field cancerization and multistep carcinogenesis. In field cancerization, diffuse epithelial injury results from carcinogen exposure in the aerodigestive tract, genetic changes, and premalignant and malignant lesions in one region of the field translate into an increased risk of cancer developing in the entire field. Multistep carcinogenesis describes a stepwise accumulation of alterations, both genotypic and phenotypic. Arresting one or several of the steps may impede or delay the development of cancer. This has been described particularly well in studies involving precancerous and cancerous lesions of the head and neck, which focus on oral premalignant lesions (leukoplakia and erythroplakia) and their associated increased risk of progression to cancer. As the new millennium begins, exciting new techniques to fight cancer are being devised, including biologic interventions and genetic manipulations. Intermediate markers of response are needed to assess the validity of these approaches in a timely and cost-efficient manner.

III. BIOLOGICAL BASIS OF EPITHELIAL CARCINOGENESIS

A. Field Cancerization
Field cancerization was originally described by Auerbach in 1953. In the upper aerodigestive tract (UADT), the surface epithelium, or field, is exposed in large amounts to environmental carcinogens, predominantly tobacco smoke. Pathologic evaluation of the epithelial mucosa of the UADT located adjacent to carcinomas frequently reveals hyperplastic and dysplastic changes. These premalignant changes found in areas of carcinogen-exposed epithelium adjacent to tumors are termed "field cancerization" and suggest that these multiple foci of premalignancy could progress concurrently to form multiple primary cancers.       
Second primary tumors (SPTs) have become the leading cause of mortality in head and neck cancer and this best illustrates the concept of field cancerization. Multiple genetic abnormalities have been detected in normal and premalignant epithelium of the lung and UADT in high-risk patients. In limited studies, when primary tumors and SPTs are analyzed for p53 mutations, evidence supports the independent origin of these tumors. Mutations of p53 can occur in only one of the tumors or distinct mutations occur between the primary and SPT. Continued work in analyzing molecular characteristics of primary and second primary cancers is needed.

B. Multistep Carcinogenesis
Pathological observations in field cancerization gave rise to the hypothesis of multistep carcinogenesis, which proposes that neoplastic changes evolve over a period of time progressing from normal, to hyperplastic, to dysplastic, and finally to fully malignant phenotypes. Elucidating the mechanism of multistep carcinogenesis has awaited the integration of molecular biological techniques into pathologic evaluation of epithelial lesions. This has led to the discovery of genetic abnormalities in premalignant and malignant epithelial cells. Studies have identified different carcinogenic stages, including initiation, promotion, and progression. Initiation involves direct DNA binding and damage by carcinogens and is rapid and irreversible.      
Promotion leads to premalignancy and is generally irreversible involving epigenetic mechanisms. Progression is the period between premalignancy and the cancer and is generally irreversible involving genetic mechanisms. The stages of promotion and progression are prolonged. Genetic damage appears to accumulate during neoplastic transformation, and specific genes have been discovered that, when altered, may play a role in epithelial carcinogenesis.       
These include both tumor suppressor genes and protooncogenes, which encode proteins that are involved in cell cycle control, signal transduction, and transcriptional regulation. Tumor suppressor genes inhibit clonal expansion by suppressing cell growth and genomic mutability. Some tumor suppressors that have been linked to epithelial carcinogenesis include p53, retinoblastoma (Rb), DCC (deleted in colorectal carcinoma), MCC (mutated in colorectal carcinoma), and APC (adenomatous polyposis coli) genes. Over 500 genes that play vital roles in cell signaling and growth control are altered in cancer cells and, as protooncogenes, may be involved in the process of neoplasia.       
These include ras, the myc (C-, N-, and Lmyc), erbB [erbB1 (epidermal growth factor receptor), erbB2 (her2/neu), erbB3 (her3), and erbB4/her4]. Chromosomes, also extensively damaged during epithelial carcinogenesis, are detected in the form of nuclear DNA adducts, cytoplasmic DNA fragments, or micronuclei and chromosomal abnormalities, which include aneuploidy as well as intrachromosomal deletions and amplifications. Future studies designed to discover genes disrupted by these chromosomal lesions may reveal both known and novel tumor suppressor genes and oncogenes, which play a role in epithelial carcinogenesis.

IV. BIOLOGICAL BASIS OF CHEMOPREVENTION

Chemoprevention trials are based on the hypothesis that interruption of the biological processes involved in carcinogenesis will inhibit this process and, in turn, reduce cancer incidence. This hypothesis provides a framework for the design and evaluation of chemoprevention trials, including the rationale for the selection of agents that are likely to inhibit biological processes and the development of intermediate markers associated with carcinogenesis. DNA damage associated with epithelial malignancies is thought to occur partly through oxidative processes induced by free radicals. Aberrant epithelial proliferation and differentiation are hallmarks of premalignant lesions.      
Treatment approaches include interrupting any of these processes. Development of intermediate markers for chemoprevention trials is crucial. Because treatment-induced improvements in cancer incidence require years to evaluate, monitoring intermediate markers that are both modulated by chemoprevention treatment and correlate with a reduction in cancer incidence would allow a more expeditious evaluation of potentially active chemopreventive agents.      
Premalignant lesions are a potential source of intermediate markers, and if disappearance of these lesions correlates with a reduction in cancer incidence, then markers of premalignancy will serve as intermediate end points for chemoprevention trials. Future studies in chemoprevention will continue to test this hypothesis.

V. CHEMOPREVENTION TRIALS

Cancer chemoprevention is still investigational, although its role in oncologic practice continues to expand. Prevention in cancer has become more prominent as frustration over the failures of current therapeutic modalities has grown. A variety of chemopreventive agents have been studied in over 40 randomized trials since 1990. Some clinical activity has been demonstrated proving the potential utility of this method of cancer prevention. Still, large randomized trials are needed before chemoprevention agents can be fully integrated into standard oncologic practice. Major trials are listed in Table I.

TABLE I Completed Randomized Chemoprevention Trials

Author Study setting Design Number Intervention Outcome
Head and neck          
Hong (1986) Oral leukoplakia Phase IIb 44 Isotretinoin (1-2 mg/kg/day) Positive
Stich (1988) Oral leukoplakia Phase IIb 65 Vitamin A (200,000 IU/week) Positive
Han (1990) Oral leukoplakia Phase IIb 61 Retinamide (40 mg/day) Positive
Lippman (1993) Oral leukoplakia Phase IIb (maintenance) 70 Isotretinoin (0.5 mg/kg/day) Positive

Chiesa (1993)

 Oral leukoplakia Phase IIb (maintenance) 80 Fenretinide Positive

Hong (1990)

Prior SCC Phase III 103 Isotretinoin (50-100 mg/m2/day) Positive
Lung          
Heimberger (1988) Metaplasia (sputum) Phase IIb 73 Vitamin B12 (500 μg/day); folic acid (10 mg/day) Positive
Arnold (1992) Metaplasia (sputum) Phase IIb 150 Etretinate (25 mg/day) Negative
Van Poppel (1992) Micronuclei (sputum) Phase IIb 114 β-Carotene (20 mg/day) Positive
Lee (1993) Metaplasia (biopsy) Phase IIb 87 Isotretinoin (1 mg/kg/day) Negative
Pastorino (1993) Prior NSCLC Phase III 307 Retinyl palmitate (300,000 IU/day) Positive (SPT)
Kurie (2000) Metaplasia/dysplasia Phase IIb 82 4-HPR Negative
Lippman (2000) Prior NSCLC Phase III 1166 Isotretinoin (30 mg/day) Negative
Colorectal          
Bussey (1982) FAP Phase IIb 36 Vitamin C (3 g/day) Positive (polyp)
McKeown-Eyssen (1988) Resected adenoma Phase IIb 137 Vitamins C (400 mg/day) and E (400 mg/day) Negative
Decosse (1989) FAP Phase IIb 58 Vitamins C (4 g/day) and E (400 mg/day); fiber (22.5 g/day) Positive (polyp)
Gregoire (1989) Prior colon cancer Phase IIb 30 Calcium (1200 mg/day) Negative (LI)
Stern (1990) Prior FAP Phase IIb 31 Calcium (1200 mg/day) Negative (LI)
Labayle (1991) FAP Phase IIb 10 Sulindac (100 mg three times/day) Positive (polyp)
Wargovich (1992) Resected adenoma Phase IIb 20 Calcium (2000 mg/day) Positive (LI)
Paganelli (1992) Resected adenoma Phase IIb 41 Vitamins A (30,000 IU/day), E (70 mg/d), and C (1 g/day) Negative (LI)
Alberts (1992) Resected adenoma Phase IIb 100 WBF (2.0 or 13.5 g/day); calcium (250 or 1500 mg/day) Negative (LI)
Nugent (1993) FAP Phase IIb 14 Sulindac (200 mg twice a day) Positive (polyp)
Giardiello (1993) FAP Phase IIb 22 Sulindac (150 mg twice a day) Positive (polyp)
MacLennan (1993) Resected adenoma Phase IIb 378 Fat (<25% of calories); WBF (11 g/day); β-carotene (20 mg/day) Negative (polyps)
Bostick (1993) Resected adenoma Phase IIb 21 Calcium (1200 mg/day) Negative (LI)
Roncucci (1993) Resected adenoma Phase IIb 209 Vitamins A (30,000 IU/day), C (1 g/day), and E (70 mg/day); lactulose (20 g/day) Positive (vitamins> lactulose)
Gann (1993) U.S. male physicians Phase III 22,071 Aspirin (325 mg qod) Negative
Steinbach (2000) FAP Phase IIb 77 Celecoxib 100 mg twice a day Negative
        Celecoxib 400 mg twice a day Positive
Schatzkin (2000) Resected adenoma Phase IIb 2079 High fiber, low fat, counseling Negative
Alberts (2000) Resected adenoma Phase IIb 1429 Wheat bran, 2.0 or 13.5 g daily Negative
Skin          
Moriarty (1982) Actinic keratoses Phase IIb 50 Etretinate (75 mg/day) Positive
Watson (1986) Actinic keratoses Phase IIb 15 Etretinate (75 mg/day) Positive
Kligman (1991) Actinic keratoses Phase IIb 527 Topical tretinoin (0.05%) Negative
Kligman (1991) Actinic keratoses Phase IIb 455 Topical tretinoin (0.10%) Positive
Moon (1993) Prior BCC/SCC Phase III 524 Isotretinoin (5-10 mg/day), retinol (25,000 IU/day) Negative
Greenberg (1990) Prior BCC/SCC Phase III 1805 β-Carotene (50 mg/day) Negative
Tangrea (1992) Prior BCC Phase III 981 Isotretinoin (10 mg/day) Negative
Moon (1993) Prior actinic keratoses Phase III 2298 Retinol (25,000 IU/day) Positive
Clark (1996) Prior BCC/SCC Phase III 1312 Selenium (200 mcg) Negative
Esophagus/stomach          
Munoz (1985) Geographic high risk (Huixan) Phase IIb 610 Retinol (50,000 IU/week); riboflavin (200 mg/week); zinc (50 mg/week) Negative (dysplasia)
          Positive (micronuclei)
Zaridze (1993) Geographic high risk (Uzbekistan) (oral leukoplakia and/or chronic esophagitis) Phase IIb 532 Riboflavin (80 mg/week); vitamins A (100,000 IU/week) and E (80 mg/week); β-carotene (40 mg/day) Negative
Blot (1993) Geographic high risk (Linxian) Phase III 29,584 Multiple vitamins/minerals Positive (stomach)
Li (1993) Geographic high risk (Linxian, dysplasia) Phase III 3318 Multiple vitamins/minerals Negative
Bladder          
Alfthan (1983) Superficial tumors (resected) Phase IIb 32 Etretinate (25-50 mg/day) Positive
Pederson (1984) Superficial tumors (resected) Phase IIb 73 Etretinate (50 mg/day) Negative
Studer (1984) Superficial tumors (resected) Phase IIb 86 Etretinate (25-50 mg/day) Positive
Cervix          
Byrne (1986) Dysplasia (CIN 2,3) Phase IIb 26 HLI (0.8 × 106 IU/week) Negative
Yliskoski (1990) Dysplasia (CIN 1,2) Phase IIb 20 HLI (9 × 106 IU/day) Negative
Frost (1990) Dysplasia (CIN 2) Phase IIb 10 IFN-α2b (4 × 106 IU/day) Negative
de Vet (1991) Dysplasia (CIN 1-3) Phase IIb 278 β-Carotene (10 mg/day) Negative
Butterworth (1992) Dysplasia (CIN 1,2) Phase IIb 235 Folic acid (10 mg/day) Negative
Chu (1993) Dysplasia (CIN 1,2) Phase IIb 298 Folic acid (5 mg/day) Negative
Meyskens (1993) Dysplasia (CIN 2,3) Phase IIb 301 Topical tretinoin (0.372%) Positive (CIN 2)
Breast          
BCPT LCIS, high-risk factors Phase III 13,388 Tamoxifen 20 mg daily Positive-early stop
Powles (1998) Family history Phase III 2494 Tamoxifen 20 mg for 8 years Negative
Veronesi (1999) Previous hysterectomy Phase III 5408 Tamoxifen 20 mg for 5 years Negative

 

A. Trial Design
Chemoprevention trials are conducted in three phases similar to other clinical trials. Phase I trials determine the toxicity of an agent alone or in combination with other agents. However, in latter phases, the ideal end point of chemoprevention trials is a reduction in cancer incidence as patients who are enrolled in study are free of cancer as opposed to traditional phase II studies. Because cancer incidence is extremely low among the general population with no known cancer risk factors, demonstrating a treatment-induced reduction in cancer incidence among the general population requires large randomized studies, which are expensive and time-consuming.      
The time and resources required for chemoprevention trials can be significantly reduced by targeting high-risk populations and utilizing potential intermediate biomarkers. This point is best illustrated by ongoing trials of chemoprevention of upper aerodigestive tract cancer. In these studies, patients are definitively treated for a stage I or II cancer of the UADT. Although recurrence is a concern, development of second primary tumors is the leading cause of cancer-related death following treatment in early stage disease as reported by Vikram. Therefore, the end points in these trials are the occurrence of SPTs and survival. SPTs occur at a rate of 3-7% per year in these patients. To test whether a given chemopreventive agent has an effect, if a cancer incidence rate of 6% is used, it is necessary to follow 1000 patients for a period of 5 years. As of September 1, 1999, the Retinoid Head and Neck Second Primary Trial had completed accrual with 1384 registered patients and 1191 patients randomized and eligible. Interim analysis performed in May 2000 indicated significantly higher recurrence rates in active smokers vs former smokers and significantly higher smoking related SPT rates in active smokers vs never smokers, with intermediate rates for former smokers. This trial will be unblinded and definitively analyzed in September 2002, when all randomized patients have completed active therapy.     
The risk of SPTs is much lower in most other epithelial cancers. For patients who have undergone resection of a breast cancer, the risk of a SPT is 0.8% per year. When compared to UADT cancer chemoprevention trials, breast cancer chemoprevention trials require up to 5 to 10 times more patients with longer clinical follow-up. As 5000-10,000 patients are required for these trials, this is still less than the 20,000 or more required to conduct a study in the general population.      

Intermediate end points can reduce the duration of studies as far as cost and resources. By utilizing bio-markers that highly predict the development of cancer rather than assessing the actual development of cancer, chemoprevention trials can be designed to test the effects of potential agents in a smaller population in a shorter period of time.

B. Chemopreventive Agents
Approximately 2000 natural and synthetic agents have been shown in experimental systems to have chemopreventive activity. Agents that have been studied in clinical trials include retinoids, N-acetylcysteine, β-carotene, calcium, α-tocopherol, selenium, tamoxifen, finasteride, and nonsteroidal antiinflammatory drugs (NSAIDs.) Of the group, the retinoids have been studied most extensively as chemopreventive agents. Vitamin A was first noted to be an essential nutrient in 1913, and its deficiency was associated with changes in epithelial histology in 1925. Since that time, it has been shown that vitamin A deficiency is associated with bronchial metaplasia and an increased incidence of cancer. Vitamin A exists as preformed vitamin A (retinol esters, retinol, and retinal) and provitamin A carotenoids (β-carotene and metabolic precursors of retinol).      
Retinoids occur in natural (all-trans retinoic acid or ATRA, 13-cis retinoic acid or 13-cRA, 9-cis retinoic acid or 9-cRA, retinyl palmitate) and synthetic [fenretinide or N- (4-hydroxyphenyl)retinamide (4- HPR)] forms. They are important for normal cell growth and differentiation, as well as for the regulation of apoptosis. In preclinical models, they have been shown to suppress or reverse epithelial carcinogenesis in lung, oral cavity, esophagus, bladder, skin, mammary gland, prostate, and liver tissues. Clinical trials have studied naturally occurring retinoids, including tretinoin or ATRA, its stereoisomer isotretinoin or 13-cRA, and the retinoid-related molecule β-carotene. Synthetic retinoids that have demonstrated clinical activity in chemoprevention trials include retinyl palmitate, fenretinide or 4-HPR, and etretinate. A retinoid that will be studied in future clinical trials is the naturally occurring ATRA stereoisomer 9-cRA. Of all chemopreventive agents, retinoids have the best-defined mechanism of action. Retinoids function as ligands for intracellular receptors. Cytosolic receptors for retinol (CRBP) and retinoic acid (CRABP) bind retinoids and appear to regulate the transfer of retinoids into the nucleus. Nuclear retinoid receptors were discovered in 1987 and are believed to mediate the effects of retinoids and are members of a larger steroid superfamily of nuclear receptors that includes glucocorticoid, thyroid, vitamin D, progesterone, and estrogen receptors, among others. There are two families of retinoid nuclear receptors, RAR and RXR, and each family consists of three members: α, β, and γ forms.      
These receptors can exist as heterodimers or homodimers. RARs bind ATRA and 9-cRA, and RXRs bind 9-cRA and bexarotene (Targretin). Liganded nuclear receptors function as transcription factors that bind DNA and regulate the expression of genes that mediate retinoid cell functions, including growth, differentiation, and apoptosis. Retinoid nuclear receptors associate with inhibitory corepressors or stimulatory coactivators that after their transcriptional activities.      
Some retinoids, including 4-HPR, which has potent apoptosis-inducing activity, do not bind retinoid receptors. Other retinoids have not demonstrated any ability to bind either family of nuclear receptors (carotenoids), whereas their metabolites do (13-cRA, and retinol). Whether these retinoids are metabolized to another form that can bind to known nuclear receptors or bind to as yet undiscovered receptors is unknown. Studies have shown intracellular interconversion of 13-cRA to ATRA, demonstrating the importance of retinoid metabolism in determining the biological effects of retinoid treatment.      
The nuclear retinoid receptor RAR-β seems to have a major role in UADT carcinogenesis. Important observations of RAR-β made in various studies include its absence in many head and neck carcinomas and lung cancer cell lines and its ability to suppress tumorigenesis. When upregulated with isotretinoin treatment, increased RAR-β expression and clinical response correlate in 40 to 90% of cases. Thus, RAR-β is the best indicator to date of retinoid chemoprevention efficacy in human head and neck carcinogenesis and a good biomarker for continued research. Surprisingly though, strong intratumoral RAR-β expression has been associated with a poorer outcome in early stage lung cancer. Thus, the biological mechanism and clinical outcome still need further investigation.

C. Trial Results
1. UADT Trials
The upper aerodigestive tract has served as a good model for chemoprevention and its utility. Premalignant oral lesions include erythroplakia and dysplastic leukoplakia. Retinoids, β-carotene, vitamin E, and selenium have shown activity in the reversal of oral premalignancy, but only retinoids have demonstrated positive results in randomized trials. Isotretinoin, retinal (a synthetic retinamide), and fenretinide were used in these randomized trials. Of these agents, the best characterized is isotretinoin, which has been studied in two related randomized trials. In 1986, Hong reported a placebo-controlled, double-blind study of 44 patients randomized to receive 13-cRA (1-2 mg/kg/day) or placebo for 3 months and followed for 6 months. The clinical and histologic major response rates to 3 months' treatment with high-dose isotretinoin were 67 and 54%, respectively (P = 0.0002 and 0.01), after 6 months of follow-up; clinical and histologic rates of response to placebo were 10%. Toxicity was severe with the highdose regimen, and more than 50% of treated patients relapsed within 2 to 3 months after discontinuing therapy.      
A second trial was instituted to examine the effects of low-dose maintenance isotretinoin or β-carotene for 9 months following 3 months of high-dose induction therapy with isotretinoin. Sixty-six patients completed the first phase of the study, and the premalignant disease progression rates were 8% in patients who received low-dose isotretinoin and 55% in patients who received β-carotene (P < 0.001). The percentage of patients whose lesions decreased in size was 33% in patients treated with isotretinoin and 10% in patients treated with β-carotene. Carcinoma developed in seven patients who received β-carotene and only one who received isotretinoin. This study not only confirmed the activity of isotretinoin demonstrated in the first study, but also showed that isotretinoin is superior to β-carotene in this setting and revealed that low-dose maintenance therapy with isotretinoin may yield better long-term effects than a short course of high-dose therapy. However, a 10-year update of the study revealed no differences in cancer rates between the two groups. These studies established the rationale for the treatment of premalignant disease with chemopreventive agents.       
These findings were further elaborated by studies that examined the effects of retinoid treatment in head and neck cancer patients who had been cured of their primary disease. A placebo-controlled, randomized trial of 103 patients who received high-dose isotretinoin (50-100 mg/m2 for 12 months) following surgery and/or radiotherapy for early stage disease revealed that isotretinoin reduced the rate of SPT development after 32 months (P = 0.005), but this effect decreased after a median follow-up of approximately 5 years (P = 0.04). The treatment had no effect on local, regional, or distant recurrence or overall survival rates. In addition to reversing oral premalignancy, treatment with retinoids reduced the incidence of SPTs, the major cause of mortality in head and neck cancer patients. This effect diminished with time after concluding treatment. Many patients had side effects to the retinoid, which points to the need for developing better tolerated, more efficacious therapies. To further evaluate the retinoid effect on SPTs, a randomized, double-blind, multiinstitutional trial was launched examining the effect of lowdose 13-cRA in previously definitively treated patients with early stage head and neck cancer. Patients received 30 mg of 13-cRA per day for 3 years and are followed for 4 years documenting any end points, recurrences, or SPTs. This trial is planned to be unblinded in 2002. These results, positive or negative, will help establish a new standard in head and neck cancer chemoprevention.      
Another study examined the effect of a second generation retinoid, etretinate, in patients with early stage head and neck squamous cell carcinoma treated with surgery or radiation. This randomized study followed 316 patients who were randomized to receive etretinate, 50 mg/day for the first month, followed by 25 mg/day in the following months, or placebo for 24 months. There were no significant differences regarding local, regional, and distant metastases at 5 years. Treatment was discontinued in 33% of patients taking etretinate and 23% taking placebo (P < 0.05) because of toxicity. This study failed to show any effect in prevention of SPTs with etretinate.       
Adjuvant therapy in advanced head and neck cancer has also been examined. Trials used a combination of 13cRA (50 mg/m2/day), interferon-α (3 MU/m2/t.i.w.), and α-tocopherol (1200 IU/day) given for 12 months. In a phase II study of 45 evaluable patients with definitively treated stage III/IV head and neck cancer, 39 have no evidence of disease with only 6 recurrences at 21 months median follow-up. The 2- year disease-free survival of the study is 84%. This suggests that perhaps in patients with a higher risk of developing recurrence or SPTs, more aggressive therapies for prevention are warranted.      
EUROSCAN, a large phase III study encompassing 2592 patients, reported no benefit of chemopreventive agents in patients with head and neck or lung cancer in terms of survival, event-free survival, or SPT. In the EUROSCAN study, patients were randomized to receive supplementation with retinyl palmitate, N-acetylcysteine, both drugs in combination, or placebo for 2 years. Sixty percent of patients had a history of head and neck cancer and 40% had a history of lung cancer. Patients were grouped as current/ former (93.5%) smokers and never (6.5%) smokers. However, data regarding smoking status, verification thereof (cotinine levels), and its impact relative to SPTs and recurrence were not presented. Further research and studies are needed to formulate a risk model in head and neck cancer, which can help identify that section of the general population highest at risk for development of cancer.

2. Lung Trials
The rationale for prevention of lung cancer is similar to that in head and neck cancer. In both diseases, chronic exposure to tobacco is the major risk factor and dysplastic epithelial lesions are thought to represent a premalignant stage. Preclinical data indicate that retinoids reverse dysplastic bronchial epithelial lesions. Despite these data, placebo-controlled, randomized trials in smokers have revealed that retinoid treatment adds no significant benefit to the effects of smoking cessation and reversal of bronchial metaplasia. In light of results demonstrating that retinoids reduce SPTs in patients who have had a lung cancer resected (see later), bronchial metaplasia may not accurately reflect the chemopreventive effects of retinoids on bronchial epithelium. Research is underway to identify intermediate markers that predict retinoid chemopreventive effects on bronchial epithelial cells.      
In resected NSCLC patients, SPTs occur at the rate of 2-4% per year. Similar to its effects in head and neck cancer patients, retinoid treatment reduces the incidence of SPTs in lung cancer patients who have undergone resection. In the only completed trial addressing this question, 307 patients whose stage I NSCLC were completely resected were randomized to receive 12 months of treatment with retinol palmitate (300,000 IU a day) or no treatment. At a median of 46 months follow-up, patients who received retinol palmitate had a 35% lower incidence of SPTs than the control group (3.1% vs 4.8%). As in studies of head and neck cancer patients, retinoid treatment had no observed effect on survival duration or the rate of primary disease recurrence. EUROSCAN, as mentioned earlier, also showed no chemopreventive benefit in lung cancer patients. These trials in NSCLC patients point out the need to further investigate the effects of different retinoids in this setting and to extend these trials to include patients whose small cell lung cancer (SCLC) has been cured, who have SPTs rates twofold higher than in patients who have been treated for early stage head and neck cancer.      
Trials in lung cancer chemoprevention have demonstrated the importance of smoking status and use of these agents. The Alpha-Tocopherol, Beta Carotene (ATBC) Cancer Prevention study was a randomized, double-blind, placebo-controlled primaryprevention trial in which 29,133 Finnish male smokers received α-tocopherol 50 mg a day alone, β- carotene 20 mg a day alone, both α-tocopherol and β-carotene, or placebo. These men were between 50 and 69 years of age and all smoked five or more cigarettes a day. Patients were followed for 5-8 years.      
Lung cancer incidence, the primary end point, did not change with the addition of β-tocopherol alone, nor did overall mortality. However, both groups who received β-carotene supplementation (alone or with α-tocopherol) had an 18% increase in the incidence of lung cancer. There appeared to be a stronger adverse effect from β-carotene in those men who smoked more than 20 cigarettes a day. This trial raised the serious issue that pharmacologic doses of β-carotene could potentially be harmful in active smokers. The β-Carotene and Retinol Efficacy Trial (CARET) confirmed results of the Finnish trial. This randomized, double-blinded, placebo-controlled trial tested the combination of 30 mg β-carotene and 25,000 IU retinyl palmitate against placebo in 18,314 men and women age 50-69 years at high risk for lung cancer; 14,254 had at least a 20 pack-year smoking history and were either current or recent former smokers.       
Four thousand and sixty men had extensive occupational exposure to asbestos. This trial was stopped after 21 months because no benefit and even possibly harm were found. Lung cancer incidence, the primary end point, increased 28% in the active intervention group. Overall mortality also increased 17% in this group. Given these results, as well as those of the ATBC trial, high-dose β-carotene is not recommended for high-risk patients who continue to smoke. The Physicians Health Study, a randomized, doubleblind, placebo-controlled trial studied 22,071 healthy male physicians: 11,036 received 50 mg of β-carotene on alternate days and 11,035 received placebo. The use of supplemental β-carotene showed virtually no adverse or beneficial effects on cancer incidence or overall mortality during a 12-year follow-up.      
In China, a study evaluating β-carotene, α- tocopherol, and selenium in the prevention of gastric and esophageal cancer showed a nonsignificant decrease in the risk of lung cancer in a small cohort of patients. Subgroup analysis of the aforementioned studies, especially ATBC and CARET, have provided few explanations for the increase in lung cancer incidence. It seems β-carotene has a harmful effect only in high-risk heavy smokers or those with previous exposure to asbestos. Current recommendations are for these people to avoid supplemental β-carotene in large doses. Including the results of EUROSCAN mentioned previously, much work is needed before chemoprevention agents can be instituted in lung cancer. Currently, an ECOG trial is studying patients with stage I lung cancer and the effect of daily selenium supplementation.

3. Colorectal Trials
Colorectal cancer is associated with premalignant lesions, including polyps and dysplastic epithelium. Large polyps with villous elements and dysplastic epithelia are more likely to progress to carcinoma than small, tubular polyps. The premalignant stages of colon carcinogenesis present an evaluable process for chemoprevention trials.      
The agents most widely used in colorectal cancer chemoprevention trials are NSAIDs and calcium salts. In phase II trials, the NSAID sulindac achieved a statistically significant decrease in both mean polyp number and diameter compared to placebo. Aspirin also shows promise in colon cancer prevention. Epidemiologic studies suggest that aspirin inhibits colon carcinogenesis. The effects of low-dose aspirin on colon cancer incidence were examined in the large-scale U.S. Physicians Health Study. This study revealed that aspirin had no effect on polyp or cancer incidence, but the trial's premature closing because of the finding of a statistically significant effect on myocardial infarction incidence most likely limited the power of this study to detect a modest effect of aspirin in preventing colon cancer. Despite evidence of colorectal cancer prevention, NSAIDs still require more investigation.      
Agent specificity, mechanism of action, and dose and duration of treatment, in addition to adverse side effects, have made recommendations for prevention difficult. Better understanding of the role of aspirin and other NSAIDs in colon cancer chemoprevention awaits further randomized, placebo-controlled trials. Cyclooxygenase-2 (COX-2) inhibitors have shown promise. Forms of the COX enzyme demonstrated include COX-1, which is constitutively expressed, and COX-2, which is overexpressed in inflammatory cells and sites of inflammation. Most NSAIDs inhibit both enzymes. Inhibition of COX-2 specifically reduces untoward side affects such as ulcers and gastritis. In familial adenomatous polyposis (FAP), patients develop hundreds of polyps in their colon due to mutation of the adenomatous polyposis coli (APC) gene. Strong evidence for the activity of COX-2 inhibitors in the treatment and prevention of FAP has been demonstrated in mouse models and, more recently, human patients. In a recent double-blind, placebo-controlled study, 77 patients with FAP received Celebrex, a COX-2 inhibitor, 100 mg or 400 mg twice daily or placebo for 6 months. Patients who received 400 mg twice daily had a 28% reduction in the mean number of colorectal polyps and a 31% reduction in polyp burden, or the sum of polyp diameters. Based on these studies, trials to assess the prevention of adenoma development in adolescents with preclinical FAP will be needed.      
Fiber in the prevention of colon cancer has been examined in several studies. The Polyp Prevention Trial studied 2079 patients with a history of colorectal adenomas randomized to receive counseling, a low-fat, high-fiber diet rich in fruits and vegetables, or to continue their current diet without counseling. Colonoscopy after 1 and 4 years found no difference in the incidence of recurrent adenomas. Another study by the Phoenix Colon Cancer Prevention Physician's Network studied 1429 patients with a history of colorectal adenoma. These patients were randomized to receive supplemental wheat bran, 2.0 or 13.5 g a day. Again, no difference in the incidence of recurrent adenomas was found between the two groups. Currently, there is no prospective evidence that fiber supplementation is effective for colorectal cancer prevention.       
Calcium salts are available in several forms (gluconate, citrate, and carbonate). Randomized trials testing the effects of calcium as a chemopreventive agent have used cellular proliferation rates within colonic mucosal crypts as an index of response in resected colon cancer patients. Three of the four randomized trials performed to date have shown an increase in cellular proliferation rates within colonic mucosal crypts as an index of response in resected colon cancer patients.      
Three of the four randomized trials performed to date have shown an increase in cellular proliferation with calcium treatment. A randomized, doubleblind trial tested the effect of dietary supplementation with calcium carbonate in 930 patients with a recent history of colorectal adenomas. The calcium carbonatetreated group was found to have a lower risk of recurrent adenoma. The results of this study, which were modestly significant, continue to support the investigation of calcium carbonate as a possible chemopreventive agent. If calcium is found to be an effective treatment in colorectal cancer chemoprevention, then intermediate end points other than cellular proliferation will be needed for future trials.

4. Skin Trials
Chemoprevention of skin cancer has been reported in two large trials using selenium and retinols. The objective of the selenium study was to determine whether supplemental selenium would decrease the incidence of cancer, specifically basal cell and squamous cell carcinomas of the skin. This multicenter, double-blind, randomized trial evaluated 1312 patients ages 18-80 years having a history of skin cancer who were given 200 μg of selenium (0.5 g high selenium brewer's yeast tablet) or placebo daily. Selenium had no effect on skin cancer incidence. However, secondary end point analyses revealed that selenium supplementation was associated with significantly lower incidences of total nonskin cancer and total nonskin and overall cancer mortality rates. In addition, lung cancer, prostate cancer, and colon cancer incidences were significantly reduced. Future trials will test the efficacy of selenium in lung and with α-tocopherol in prostate cancer.      
Retinoids have demonstrated activity in the reversal of actinic keratoses, a premalignant skin disorder. Trials have tested either topical tretinoin or various retinoids given systemically. Small-scale trials in patients at high risk for skin cancers such as xeroderma pigmentosa or in renal transplant patients have shown that prolonged treatment with high-dose isotretinoin or etretinate reduced the incidence of invasive skin cancers. In a phase III trial, 2297 patients who were at a lower risk for skin cancer, including patients with actinic keratoses, received oral retinol (25,000 IU) or placebo daily for 5 years. Retinol treatment was effective in reducing the incidence of squamous cell skin cancer, but not basal cell carcinoma. However, in patients with a prior skin cancer, three phase III trials that tested the effects of retinol, β-carotene, and low-dose isotretinoin demonstrated that these retinoids had no effect on the incidence of SPTs. Further studies are needed to validate the effects of these agents in skin and other cancers.

5. Breast Trials
Like UADT and lung cancers, breast cancer provides a model to study the effects of chemopreventive agents. Chemoprevention strategies in breast cancer target the development of SPTs, which occur at a rate of 0.8% per year, as well as preventing breast cancer in high risk patients. Pooled data from over 40 randomized adjuvant trials revealed that tamoxifen, given over prolonged periods, reduces the incidence of SPTs by 39% in postmenopausal women who have undergone resection. Observing that tamoxifen reduced the incidence of contralateral breast cancer when used as adjuvant therapy, investigators hypothesized that a possible benefit could exist in the prevention of breast cancer in high-risk patients. This initiated several large studies, including the U.S. Breast Cancer Prevention Trial (BCPT) or NSABP P-1 (National Surgical Adjuvant Breast and Bowel Project) trial that was launched in 1992. They studied 13,388 patients who were at increased risk for breast cancer: greater than age 60 years, elevated Gail assessment in those aged 35-59 years, and patients with a history of lobular carcinoma in situ.      
Patients were randomized to take tamoxifen or placebo for 5 years. This trial was stopped early and unblinded after interim analysis results demonstrated a 50% reduction in new tumors. Two other studies, the Royal Marsden Hospital (RMH) Tamoxifen Chemoprevention Trial with 2494 patients and the Italian Tamoxifen Prevention Trial with 5408 patients, are still blinded because preliminary results did not indicate a reduction in breast cancer incidence. The Italian study, a multicenter, double-blind, placebo-controlled trial, evaluated the effect of taking tamoxifen for 5 years in healthy women. The RMH study included women age 30 to 70 years with a family history of breast cancer.      
Another trial, MORE, is studying 3 years of treatment with raloxifene to prevent osteoporosis. It too has reported a reduction in breast cancer incidence in women taking raloxifene. Currently, the NSABP-P2 or STAR (Study of Tamoxifen and Raloxifene) is ongoing in the United States. It will enroll tens of thousands of patients to help determine whether a difference exists in treatment between the two drugs. Raloxifene is another type of SERM, which binds with high affinity to estrogen receptors and preliminarily has shown a marked effect on estrogen receptor-positive tumors. The final results of these trials are still pending.      
Fenretinide, a synthetic vitamin A derivative, has been studied in breast cancer since 1979. A phase III trial initiated in 1987 randomized 2972 women with a history of stage I breast cancer to receive 200 mg of fenretinide daily or no intervention for 5 years in an attempt to reduce contralateral breast cancer. Although no significant difference was found between the two groups after a median of 8 years, there was a trend for benefit in premenopausal women. This has led to a current trial studying the effect of tamoxifen and fenretinide in premenopausal women at increased risk for breast cancer.      
Other strategies for chemoprevention trials in breast cancer include studying the effects of lutenizing hormone-releasing hormone (LHRH) agonists in high-risk premenopausal women as well as aromatase inhibitors in postmenopausal women.

6. Esophageal/Gastric Trials
Esophageal carcinoma has been associated with tobacco and alcohol exposure in the United States and with nutrient deficiencies and exposure to N-nitrous compounds in China. Randomized chemoprevention studies in China, which tested the effects of combinations of agents such as retinol, riboflavin, zinc, and vitamin E, have revealed a reduction in micronuclei frequency but not in the incidence of premalignant lesions in the esophagus. A cooperative study with the United States in Linxian, China, studied the effects of one of four combinations of vitamins/minerals for 5 years (retinol and zinc, riboflavin and niacin, vitamin C, molybdenum and β-carotene, vitamin E and selenium). No statistically significant relationships were correlated with the intervention. However, in secondary analysis, selenium, β-carotene, and vitamin E use was associated with a statistically significant lower mortality rate in all cancers, predominantly gastric cancer. No effect was seen with esophageal cancer. The interpretation of these studies is made difficult by the use of readily available vitamin supplements in the control arm, blurring potential differences from the treatment arm. This represents an inherent flaw in trials utilizing vitamin supplements and dietary intervention. Other trials include examining the effects of vitamin/ mineral supplementation in patients with esophageal dysplasia; 3,318 patients with defined histologic evidence received treatment for 6 years and demonstrated lowered mortality as well as a reduced risk of esophageal or gastric dysplasia. Other trials currently ongoing are comparing gastric dysplasia and treatment of Helicobacter pylori and chronic atrophic gastritis with oltipraz treatment.

7. Bladder Trials
Previous studies using retinoids to prevent recurrence and SPTs in bladder cancer patients have been small and limited by high toxicity. Two trials with prolonged low-dose etretinate appeared to have some effect. One double-blind, placebo-controlled trial studied 30 patients with superficial bladder tumors and the preventive effect of etretinate. A reduction in recurrence was seen in the treated group. Another trial, which was multicenter and randomized, evaluated 79 patients with superficial papillary bladder tumors with 25 mg of etretinate or placebo daily. Time to first recurrence was similar in the two groups; however, the interval to subsequent tumor recurrence was significantly longer in the treated group. Some mild toxicity was experienced in these trials and, although small, indicate beneficial effects of retinoids in prevention. Better tolerated retinoids such as fenretinide are being used in ongoing bladder cancer trials, and combinations of these with other treatment modalities, such as intravesical chemotherapy, NSAIDs, and oltipraz, are under investigation.

8. Cervical Trials
Cervical cancer has well-recognized premalignant dysplastic stages that are accessible for study. Interferon, folic acid, β-carotene, and retinoids have been used in the treatment of cervical dysplasia, and only retinoids have yielded positive results. Nutritional studies have helped define micronutrients of interest (folate, carotenoids, vitamin C, vitamin E). Other interesting medications under evaluation include retinoids [4-hydroxyphenylretinamide (4-HPR), retinyl acetate gel, topical all-trans retinoic acid], polyamine synthesis inhibitors [α-difluoromethylornithine (DFMO)], and nonsteroidal anti-inflammatory drugs (ibuprofen). Phase I chemoprevention studies of the cervix have tested retinyl acetate gel and all-trans retinoic acid. Topical all-trans retinoic acid has been shown to increase the histological regression rate in women with moderate cervical intraepithelial hyperplasia (CIN 2). Phase II trials of all-trans retinoic acid, β-carotene, and folic acid have been and are being carried out, whereas phase III trials of all-trans retinoic acid have been completed and have shown significant regression of CIN 2 but not CIN 3. Clearly, further studies are warranted to formulate chemoprevention strategies in cervical cancer.

9. Prostate Trials
Chemoprevention in prostate cancer is a relatively new area of research. The Prostate Cancer Prevention Trial (PCPT) was launched in 1993 with its principle end point being a reduction of biopsy-proven prostate cancer incidence. Patients included 18,882 males age 55 years and older who are received finasteride, 5-α- reductase inhibitor, 5 mg or placebo daily for 7 years. The trial is still blinded and will achieve its primary end point in 2004. Selenium and vitamin E have been observed to decrease prostate cancer incidence in several trials. Therefore, the National Cancer Institute will launch a trial entitled "SELECT." This phase III trial will enroll more than 32,000 men and assess prostate cancer prevention with selenium, vitamin E, selenium and vitamin E, and placebo. Results of this trial, which will last over 10 years, will help elucidate new strategies in prostate cancer chemoprevention.

10. Other Cancers
Studies of retinoids in other cancers such as hepatocellular carcinoma have been described as well. A prospective randomized study reported in 1996 studied 89 patients who were definitively treated for primary hepatoma. Patients received polyprenoic acid (600 mg daily) or placebo for 12 months. There was a significant decrease in recurrence as well as second primary hepatomas in the retinoid-treated group. At 38 months, 27% of treated patients had recurrence or SPT versus 49% in the placebo group (P = 0.04). Further studies are needed to verify these encouraging results.

VI. FUTURE DIRECTIONS

A. Risk Models
Treatment of epithelial malignancies is moving toward prevention. Since the early 1980s, many patients at high risk for epithelial cancers, mostly those with premalignant lesions or a history of epithelial malignancy, have been enrolled in chemoprevention trials. While this accounts for a substantial number of patients, the vast majority of epithelial malignancies arise in patients with no history of either of these risk factors. This population is, at present, unrecognizable and is therefore not entering chemoprevention trials. While this accounts for a substantial number of patients, the vast majority of epithelial malignancies arise in patients with no history of either of these risk factors. This population is, at present, unrecognizable and is therefore not entering chemoprevention trials. This points out the need to develop a risk model to use in identifying high-risk people from the general population.      
Developing a risk model will require the identification of markers that will predict the likelihood of cancer development. Ongoing research will determine the cancer risks associated with the presence of premalignant epithelial lesions and the genetic abnormalities they contain. These studies will be advanced by improved techniques in the diagnosis of premalignant epithelial foci. Spectroscopic analysis of epithelial tissues performed at endoscopy can differentiate normal, dysplastic, and malignant areas. This technique is based on fluorescent emission spectra of epithelial cells following laser excitation and this reflects differences between normal and malignant cells in their endogenous flavins, riboflavins, and other fluorophors.      
These endoscopic techniques will aid in the identification and acquisition of premalignant tissue for histologic and genotypic analyses. By using this technique, genetic events that are associated with malignant progression can be analyzed, including changes in chromosomal ploidy, micronuclei, proliferation antigens, point mutations in ras and p53, amplification of the myc or erbB-2 genes, and deletions incorporating these markers into ongoing chemoprevention trials in patients with prior malignancies, clinical correlation will reveal the power of these markers to predict the likelihood of an SPT. To determine whether these variables can be used in a model to predict cancer risk among the general population with no history of cancer, prospective studies must be performed. Because cancer risk among the general population is low, population sample size for these studies will be extremely large and will require years of clinical follow-up. Analysis of biopsy material and patient examinations will be extremely expensive to carry out. However, these studies are necessary to the construction of a framework on which chemoprevention trials can be built. Not until high-risk patients are identified can potential chemopreventive agents be tested.

B. Intermediate End Points
Work is underway to identify cellular and molecular markers that change during chemopreventive treatment and correlate with a treatment-induced reduction in cancer incidence. Intracellular pathways activated or inhibited by chemopreventive agents offer potential intermediate markers of response. Of the pathways known to be important in mediating the effects of chemopreventive agents, the retinoid pathway is the best characterized. Cytoplasmic and nuclear retinoid receptors are key elements of retinoid signal transduction. In vitro studies have shown that the expression of RARs is activated within hours of retinoid treatment. A growing body of evidence reveals that the activated expression of specific RARs is crucial to the process of retinoid-induced tumor differentiation. For example, in P19 murine teratocarcinoma cells, retinoid refractoriness correlates with aberrant RAR-α expression. The mechanism by which retinoids induce tumor differentiation may be different from that of their chemopreventive effects.       
Growth inhibition, apoptosis, reversal of dysplasia, and stabilization of the DNA damage process are potential mechanisms of retinoid actions. In acute promyelocytic leukemia patients, tretinoin induces differentiation of promyelocytic cells within weeks. The minimum duration of retinoid treatment necessary to prevent epithelial cancers is not known. If months of treatment are required, then events other than receptor activation may be necessary for retinoidinduced chemoprevention. In addition to retinoid receptors, specific growth factors, their receptors, and carbohydrate antigens have been shown to be modulated by retinoid treatment in vitro and may be important in mediating retinoid effects in epithelial cells. Previous studies have shown that retinoid treatment downregulates the expression of transforming growth factor α, epidermal growth factor receptor, and fibroblast growth factor-4. To examine the roles of these and other biological events as potential intermediate markers for chemoprevention trials, prospective clinical studies that correlate treatmentinduced changes in these cellular components with clinical outcome are needed.      
Premalignant lesions are another source of intermediate end points. Phenotypic analysis of hyperplastic and dysplastic changes can be performed by light microscopy and immunostaining for proliferation antigens such as proliferating cell nuclear antigen and nuclear protein that binds the Ki-67 antibody. Analysis of genetic changes associated with premalignancy offers a potentially more objective means of diagnosis. For example, mutations in p53 have been found in preneoplastic lesions associated with most epithelial cancers. Because mutations can occur over a large region of the p53 gene, a mutation at a specific site, involving an alteration to a specific nucleotide, may be considered a clonal marker. Thus, disappearance of a clonal marker during treatment with a chemopreventive agent could represent regression of a premalignant lesion and may have prognostic value. Work is underway to develop the ability to detect the presence of such clonal markers in sputum, urine, and stool. This will pave the way to testing their usefulness as intermediate markers in chemoprevention trials.

C. Chemopreventive Agents
There is a long list of potentially effective chemopreventive agents that await clinical trials. As shown in Table I, the list includes a broad range of compounds, including vitamins, minerals, antioxidants, antiinflammatory agents, and steroid hormone antagonists. As single agents alone, these compounds will provide material for important clinical trials over at least the next decade.     
Retinoids have undergone extensive development as chemopreventive agents. In addition to 13-cRA, another retinoic acid stereoisomer, 9-cRA, has been developed for clinical trials. 9-cRA is able to bind and activate both RAR and RXR receptor families. Because of its novel nuclear receptor affinities, 9-cis RA may have biochemical effects that other retinoic acid stereoisomers do not have. In addition, synthetic retinoids have been developed that have receptorspecific activity. These agents may have greater efficacy than natural retinoids if particular receptors are implicated in epithelial carcinogenesis. One site in which this may prove true is UADT. RAR-β expression is lower in cancers in this region than in adjacent histologically normal mucosa. In UADT cancer patients, treatment with retinoic acid leads to a chemopreventive response that occurs in UADT cancer patients who have undergone resection. The mechanisms by which retinoids induce a chemopreventive effect in the UADT and lung are unknown, but insight was provided by in vitro experiments that showed that RAR-β appears to suppress cell growth.      
In a lung cancer cell line, overexpression of RAR-β through stable transfection led to a suppression of growth and tumorigenicity in nude mice. Thus, novel retinoids that target RAR-β may be more efficacious than other retinoids in UADT and lung cancer chemoprevention trials. Recently though, several studies have suggested an adverse effect of retinoids in current smokers producing a higher incidence of lung cancer. If RAR-β regulates growth in epithelial cells of the lung and UADT, then the RAR-β gene might be considered as a therapeutic agent in gene therapy trials in the future. The settings in which RAR-β might be most efficacious are unknown, but trials in lung cancer prevention should be considered. The rationale for its use in prevention trials rests on the observation that benign bronchial epithelial cells are more responsive to the growth suppressive effects of retinoids than lung cancer cells, suggesting that retinoid receptors may have greater biological activity in benign bronchial epithelium than in lung cancer cells. Trials such as these await the development of vectors that can safely and effectively deliver genes of interest to the target tissues.       
Specific gene mutations found in premalignant lesions, such as ras and p53 point mutations, may offer additional opportunities for intervention at the molecular level. For example, posttranslational farnesylation of mutant ras is necessary for activation of its transforming properties. FTIs block the enzyme farnesyltransferase or its substrate, farnesyl. Although the technology to deliver efficiently peptides or proteins in clinical trials has not yet been developed, this is the goal of several ongoing studies. Similarly, in vitro studies suggest that changes in mutant p53 phosphorylation can induce wild-type p53 properties. Such an approach might be considered to restore p53 tumor suppressor activities within tumor cells.      
Another direction in chemoprevention trials that may prove fruitful is combination therapy. Combining agents that mediate their effects through different pathways might enhance the ultimate chemopreventive effect. For example, agents that decrease the accumulation of intracellular-free radicals, such as oltipraz, might be combined with retinoids that activate a separate pathway, i.e., the modulation of cell growth and differentiation. Other agents have enhanced the effects of retinoids. In tumor differentiation models, for example, the effects of retinoids are augmented by agents such as phorbol esters or cyclic AMP, which activate protein kinase C (PKC) and protein kinase A, respectively. In a human teratocarcinoma cell line, activation of these kinases enhanced the effects of retinoic acid on RAR-β activation, demonstrating coupling of these kinase pathways with retinoid receptors. In addition to retinoid receptors, another mechanism through which retinoids and kinases might couple is transforming growth factor-β (TGF-β). Intracellular TGF-β production is increased by either retinoic acid treatment or PKC activation.      
Thus, retinoid and protein kinase pathways converge on TGF-β. Transforming growth factor-β has been shown to induce profound growth suppression in many cell types. Studies that illuminate the convergence of different intracellular pathways by molecules that mediate growth suppression, such as retinoid receptors and TGF-β, provide a basis for combination therapy. Combination therapy might be used in the design of future chemoprevention trials. These trials might await the development of agents that interact with specific kinase pathways. One drug that should be considered in future chemoprevention trials is bryostatin, a PKC activator that is now entering phase I trials. Other agents that could potentially be effective preventive agents in high-risk patients because of their favorable side effect profiles include farnesyltransferase inhibitors, which inhibit ras, monoclonal antibodies and small molecules such as tyrosine kinase inhibitors to epidermal growth factor or vascular endothelial growth factor. Cyclooxygenase of lipoxygenase inhibitors and other NSAIDs need to be further studied as chemopreventive agents. Immunotherapy with interferon-α combinations is also under investigation. These agents all have promise in high risk patients as primary prevention agents or in the adjuvant setting.

VII. SUMMARY

Since the mid-1980s, the goal of treatment for epithelial cancers has begun to shift from the eradication of metastatic disease to the prevention of cancer. Along these lines, advancements have been made in the prevention of UADT cancer by treatment with retinoids as well as the use of tamoxifen in breast cancer and targeting COX-2 in FAP colon cancer. Further advances in epithelial cancer prevention await the development of cancer risk models and intermediate markers that can be incorporated into the design of chemoprevention trials. These needs may be met through advancements in our understanding of genetic events that occur during epithelial carcinogenesis, such as point mutations of ras and p53. Pathways activated or inhibited by chemopreventive agents may offer additional intermediate markers of response, including retinoid receptor expression, which increases following retinoid treatment, epidermal growth factor, and cyclooxygenase. Prospective phase III clinical trials are necessary to examine these possibilities and to establish eventually chemoprevention strategies as standard of care in health policy.

Edward S. Kim
Fadlo R. Khuri
Waun Ki Hong
University of Texas M. D. Anderson Cancer Center, Houston

See Also
CHEMOPREVENTION, PRINCIPLES OF ; COLORECTAL CANCER: MOLECULAR AND CELLULAR ABNORMALITIES ; HEAD AND NECK CANCER ; LUNG CANCER: MOLECULAR AND CELLULAR ABNORMALITIES ; MOLECULAR EPIDEMIOLOGY AND CANCER RISK ; MULTISTAGE CARCINOGENESIS

GLOSSARY
chemoprevention Treatment with natural or synthetic agents that prevent the development of cancer in cancernaive patients (primary chemoprevention) or in patients who have been cured of a prior cancer (secondary chemoprevention).

DNA damage Genomic damage resulting from carcinogen exposure that leads to abnormalities in coding sequence (point mutations, amplifications, deletions, additions) or in chromosomal content (aneuploidy).

epithelial cancer Cancer that originates in the epithelial lining of any organ.

field cancerization Hypothesis in which diffuse epithelial injury results from chronic carcinogen exposure; genetic changes and the presence of premalignant and malignant lesions in one region of the field are associated with an increased risk of cancer developing throughout the entire field.

intermediate end point Any biological or genetic variable that changes during chemopreventive treatment and that can be used as a marker to predict a reduction in cancer incidence in chemoprevention trials.

multistep carcinogenesis Hypothesis stating that a cancer cell evolves from a normal cell as a result of a sequence of genomic injuries, which leads to a progressively dysplastic appearance and abnormal biological behavior.

oncogene A gene that, when expressed at abnormal levels or in a mutated state, contributes to the carcinogenic process through dysregulation of cell growth and differentiation.

premalignancy An epithelial region that is histologically abnormal but not fully malignant in appearance or behavior and that is at risk for neoplastic progression in the future.

retinoid receptor A cytoplasmic or nuclear protein that binds retinoids and functions in retinoid signal transduction.

retinoids A chemically diverse group of natural and synthetic compounds that are related to vitamin A and bind a specific receptor or set of receptors.

second primary tumor A tumor that occurs in a patient who has been cured of a prior cancer and that is of a different histology and/or occurs in a separate location from the original cancer.

tumor suppressor gene A gene that normally functions as a physiologic inhibitor of abnormal clonal expansion, preventing the outgrowth of cells that have undergone oncogenic mutations.

Bibliography
Bollag, G., and McCormick, F. (1992). GTPase activating proteins. Semin. Cancer Biol. 3, 199-208.
Brison, O. (1993). Gene amplification and tumor progression. Biochim. Biophys. Acta 1155, 25-41.
Cho, K. R., and Vogelstein, B. (1992). Suppressor gene alterations in the colorectal adenoma-carcinoma sequence. J. Cell. Biochem. 16G, 137-141.
Decensi, A., and Costa, A. (2000). Recent advances in cancer chemoprevention, with emphasis on breast and colorectal cancer. Eur. J. Cancer 36, 694-709.
Fisher, B., Costantino, J. P., Wickerham, D. L., et al. (1998). Tamoxifen for prevention of breast cancer: Report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J. Natl. Cancer Inst. 90, 1371-1388.
Harris, C. C., and Hollstein, M. (1993). Clinical implications of the p53 tumor-suppressor gene. N. Engl. J. Med. 329, 1318-1326.
Hong, W. K. (2000). Chemoprevention in the 21st century: Genetics, risk modeling, and molecular targets. David A. Karnofsky Memorial Award Lecture, J. Clin. Oncol. 18(21 suppl.), 95-185.
Hong, W. K., and Sporn, M. B. (1997). Recent advances in chemoprevention of cancer. Science 278(5340), 1073-1077.
Hong, W. K., Lippman, S. M., Itri, L. M., et al. (1990). Prevention of second primary tumors with isotretinoin in squamous-cell carcinoma of the head and neck. N. Engl. J. Med. 323(12), 795-801.
Kim, E. S., Hong, W. K., and Khuri, F. R. (2000). Prevention of lung cancer: The new millennium. Chest Surg. Clin. North Am. 10(4), 663-690.
Lippman, S. M., Brenner, S. E., and Hong, W. K. (1994). Cancer chemoprevention. J. Clin. Oncol. 12, 851-873.
Lippman, S. M., Lee, J. J., Karp, D. D., et al. (2001). Phase-III intergroup trial of isotretinoin to prevent second primary tumors in stage-I non-small-cell lung cancer. J. Natl. Cancer Inst. 93(8), 605-618.
Lippman, S. M., Lee, J. J., and Sabichi A. L. (1998). Cancer chemoprevention: Progress and promise. J. Natl. Cancer Inst. 90(20), 1514-1528.
Sabichi, A. L., Lerner, S. P., Grossman H. B., and Lippman, S. M. (1998). Retinoids in chemoprevention of bladder cancer. Curr. Opin. Oncol. 10, 479-484.
van Zandwijk. N., Dalesio, O., Pastorino, U., et al. (2000). EUROSCAN, a randomized trial of vitamin A and N-acetylcysteine in patients with head and neck cancer or lung cancer. For the European Organization for Research and Treatment of Cancer Head and Neck and Lung Cancer Cooperative Groups. J. Natl. Cancer Inst. 92(12); 977-986.
Weinberg, R. (1991). Tumor suppressor genes. Science 254, 1138-1146.

Cancer Risk Reduction

Cancer Risk Reduction (Diet/Smoking Cessation/ Lifestyle Changes)

Many cases of cancer can be prevented. Generally, people can reduce their risks for developing cancer by making wise lifestyle choices such as eating low-fat, high-fiber diets that include a variety of vegetables and fruits, avoiding tobacco use, being physically active, and minimizing sun exposure. Specific genetic susceptibilities, however, can influence cancer risk associated with certain lifestyle factors, and variation in risk exists among individuals. Guidelines for implementing lifestyle choices to reduce cancer risk have been formulated to help people adopt cancer-protective behaviors.

I. INTRODUCTION

The risk of developing cancer, a disease that affects many millions of people worldwide, can be reduced markedly by approaches that encourage primary prevention. A landmark report by Doll and Peto in 1981 summarized the evidence relating lifestyle choices, including diet, tobacco use, and sun exposure, as well as other environmental factors (e.g., occupation, ionizing radiation), to cancer risk. The report suggested that 75-80% of cancer cases were potentially avoidable and that dietary factors, tobacco use, and sun exposure were associated with approximately 35, 30, and 1-2%, respectively, of all cancers. Scientists now recognize that the effects of environmental factors, including lifestyle choices, on cancer risk can be influenced by a person's genetic susceptibility.

II. DIET

A. The Diet-Cancer Relationship
A considerable body of evidence--experimental, epidemiologic, and clinical--indicates that dietary factors, both individual food constituents and dietary patterns, play a major role in determining cancer risk. Generally, the evidence supports inverse associations between cancer risk and intakes of vegetables, fruits, whole grains, dietary fiber, certain micronutrients, and certain types of fat (e.g., n-3 fatty acids, particularly n- 3/n-6 fatty acid ratios), as well as direct associations between cancer risk and intakes of excessive calories, alcohol, total fat, and certain types of fat (e.g., saturated fat). To illustrate, epidemiologic studies have provided consistent and convincing evidence for increased cancer risk when migrants from countries with a low-fat, high-fiber diet adopt the high-fat, low-fiber diet characteristic of Western countries. The Western countries also have higher rates of obesity. Of interest is the possible cancer-protective Mediterranean diet, which emphasizes a high intake of vegetables, fruits, whole grains, fish (high in n-3 fatty acids), and olive oil (high in monosaturated fatty acids). Interactions likely occur among many dietary constituents. However, neither these interactions nor their influence on cancer risk is well understood. Thus, at present it is difficult to tease out the specific effects of individual dietary components from diet and cancer research data.      
Randomized, controlled trials (RCTs) offer one of the best means for testing diet and cancer hypotheses developed from the insights provided by epidemiologic and experimental studies. Among specific dietary constituents investigated in RCTs, vitamins A, C, and E, folic acid, calcium, and selenium have shown promise in reducing cancer risk at certain sites (e.g., vitamin E and selenium for prostate cancer). Some diet and cancer trials include efforts to modify certain lifestyle choices other than diet, such as smoking and level of physical activity, that also could influence cancer risk and thus influence study results.

B. Gene-Nutrient Interactions
Cancer risk reduction by dietary modification likely will depend, in part, on increased understanding of both gene-nutrient interactions and the role of genetic differences (polymorphisms) among individuals. Genes involved in carcinogenesis influence metabolic activation/detoxification, DNA repair, chromosome stability, activity of oncogenes or tumor suppressors, cell cycle control, signal transduction, hormonal pathways, vitamin metabolism pathways, immune function, and receptor or neurotransmitter action. Exposure to the same quantitative level of dietary factor(s) or dietary carcinogens can increase cancer risk in one individual but not in another, depending on specific susceptibilities to gene-nutrient interactions. For example, an intake of heterocyclic amines (HAs), produced by grilling of meat, increases the risk of colorectal cancer. However, people with polymorphic forms of the enzymes N-acetyltransferase type 2 (NAT2) and cytochrome P4501A2 (CYP1A2), which cause rapid activation of HAs, have a greater risk of colorectal cancer than people with polymorphic forms that cause a slow activation of HAs. Folate, found in green, leafy vegetables, is also associated with polymorphisms and cancer risk. A specific polymorphism in the gene that codes for methylenetetrahydrofolate reductase--an enzyme critical to DNA methylation and synthesis--can reduce colorectal cancer risk by altering cellular responses to dietary folate and methionine.     
Understanding gene-nutrient interactions and individual differences in genetic susceptibilities may lead to future dietary intervention strategies to reduce cancer risk. Focusing on polymorphisms in intervention studies allows investigators to develop study designs, stratify participants, and analyze results based on genetic differences within the study population. For example, polymorphisms in genes affecting the use of vitamin D confer different levels of risk for prostate cancer--a fourfold increase in risk among individuals with one polymorphism and a 60% reduction in risk among individuals with another. This kind of information is important for researchers designing trials to investigate the role of vitamin D, or its analogs, in prostate cancer risk. As more knowledge is gained about how genetic polymorphisms influence cancer risk through specific dietary constituents or patterns, researchers can refine recommendations for healthful eating regimens to reduce cancer risk.

TABLE I National Cancer Institute Dietary Guidelines

Reduce fat intake to 30% or less of calories
Increase fiber intake to 20 to 30 g daily, with an upper limit of 35 g
Include a variety of vegetables and fruits in the daily diet
Avoid obesity
Consume alcoholic beverages in moderation, if at all
Minimize consumption of salt-cured, salt-pickled, and smoked foods

C. Dietary Guidelines
As research continues into the role of specific dietary constituents in cancer risk, it is important that clinicians and the public be advised about the importance of modifying diets to reduce cancer risk. Since the mid-1970s, various scientific organizations around the world, including the World Cancer Research Fund (WCRF) in association with the American Institute for Cancer Research (AICR), the American Cancer Society (ACS), and the U.S. National Cancer Institute (NCI), have developed dietary guidelines to promote cancer risk reduction as a population strategy.      
The guidelines of the NCI, outlined in Table I, are in agreement with dietary guidelines developed by other organizations. The average American, according to the latest national health assessment, consumes too much fat (too much saturated fat, too little monosaturated fat), too little fiber, and too few vegetables, fruits, and whole grains. This pattern is seen in many developed countries worldwide. People should be encouraged to adhere to dietary guidelines that promote good health. It is incumbent that physicians, nutritionists, and registered dietitians play a key role in advocating healthful diets that have the potential to reduce cancer risk and are beneficial overall.

III. SMOKING CESSATION

A. Smoking and Cancer Risk
Lung cancer is the most common cancer in the world, accounting for almost 13% of all cancers (18% for men, 7% for women). Smoking is the single most important lifestyle factor contributing to lung cancer incidence worldwide. In fact, at least 20 carcinogens that are components of tobacco smoke cause lung tumors in either animals or humans. In addition to lung cancer, smoking substantially increases the risk for cancers of the larynx, oral cavity, and esophagus and contributes to risk for cancers of the pancreas, uterine cervix, kidney, and bladder. People may have different susceptibilities to developing smoking-related cancers because of genetic variations in their metabolism of tobacco smoke carcinogens. Furthermore, the adverse effects of smoking on cancer risk can be enhanced by alcohol consumption and by the presence of other environmental carcinogens, particularly radon and asbestos.      
In many areas of the world, particularly in third-world countries, tobacco use is rising. Although it is estimated that as many as 30% of cancers in developed counties are tobacco related, tobacco use in these countries remains high. In the United States, about one-fourth of adults currently smoke, and tobacco use among high school students has increased steadily since the early 1990s. Results of a national 1999 survey show that almost 35% of high school students smoke.      
Nonsmokers exposed to environmental tobacco smoke (ETS), as a result of either household or occupational exposure, inhale and metabolize components of the smoke and thus also may be at increased risk for lung cancer. An analysis of 37 case-control and cohort studies indicated that, for men and women combined, the risk of lung cancer increased by 23% for a nonsmoker who lived with a smoker; also, risk appeared to be directly related to the number of cigarettes smoked by the spouse and the duration of exposure.      
Educational strategies to prevent the start of smoking by adolescents and smoking cessation approaches to stop tobacco use by smokers permanently are among the most effective ways to reduce cancer risk. Although 70% of smokers claim to be interested in quitting smoking, most of them have no immediate plans to quit. When smokers try to quit on their own, relapse in the majority occurs within days, and only about 7% achieve long-term success. However, longterm success rates can be increased to 15-30% if interventions such as pharmacotherapies and intensive counseling are used. A combination of pharmacotherapy and counseling should be used for adult smokers trying to quit, except for light smokers (< 10 cigarettes/day), women who are either pregnant or breastfeeding, or when pharmacotherapy is medically contraindicated.

B. Pharmacotherapies for Smoking Cessation
Nicotine, present in all tobacco products, is an addictive substance. For most users, tobacco use results in true drug dependence. Not all smokers, however, have the same level of nicotine dependence; a high number of cigarettes smoked each day, frequent smoking in the morning, and smoking while ill indicate a high dependence level. Evidence demonstrating that pharmacotherapies can help people addicted to nicotine quit smoking is strong and consistent, and several pharmacotherapies--primarily nicotine replacement therapies (NRTs)--have been approved for smoking cessation by the U.S. Food and Drug Administration (Table II). All currently approved pharmacotherapies approximately double the chances that an attempt to quit smoking will be successful. Abundant evidence confirms that both nicotine gum and nicotine patches are safe and effective aids to smoking cessation. Nicotine nasal spray, which delivers larger doses of nicotine more rapidly than nicotine gum and patches, is also an effective smoking cessation aid. The nicotine inhaler, a plastic rod containing a plug impregnated with nicotine, is designed to combine pharmacologic and behavioral substitution strategies; it results in blood levels of nicotine similar to those from use of nicotine gum. Bupropion, an antidepressant and the only approved nonnicotine pharmacotherapy, appears to be an effective aid to smoking cessation and is safe even when used jointly with NRT. Although bupropion likely does not influence smoking cessation via its antidepressant effect, its mechanism of action is unclear.

TABLE II Pharmacotherapies for Smoking Cessation

Pharmacotherapy Availability Side effects
Nicotine gum Over-the-counter only Mouth soreness Indigestion
Nicotine patch Over-the-counter and prescription Skin reaction at patch site
Insomnia
Nicotine nasal spray Prescription only Nasal/throat irritation
Sneezing, coughing
Nicotine inhaler Prescription only Mouth/throat irritation
Coughing
Bupropion Prescription only Dry mouth
Insomnia

C. Behavioral Interventions
Behavioral interventions can range from minimal intervention, such as a 3-min intervention during a routine visit to a physician, to intensive group or one-on-one counseling using multiple sessions. In the United States, more than 70% of smokers visit a healthcare setting every year, providing an excellent opportunity for brief advice and counseling. In fact, patients who smoke expect their physician to inquire into their smoking habits and advise them regarding cessation techniques. Studies show that brief physician advice to quit smoking can produce cessation rates of 5-10% each year. Simple intervention strategies that can be used by physicians to help their patients stop smoking are outlined in Table III. Effective counseling is essential to help a smoker who is willing to quit achieve long-term success in smoking cessation. Practical counseling that provides coping strategies and basic information about smoking/quitting, social support offered as part of treatment, and social support arranged outside of treatment (e.g., community resources) all are important components of effective counseling.

TABLE III Synopsis for Physicians: How to Help Your Patients Stop Smokinga

Ask about smoking at every opportunity a. "Do you smoke?" If so, "How much?"
b. "How soon after waking do you have your first cigarette?"
c. "Are you interested in stopping smoking?"
d. "Have you ever tried to stop before?" If so, "What happened?"
Advise all smokers to stop a. State advice clearly, e.g., "As your physician, I must advise you to stop smoking now."
b. Personalize the message to quit. Refer to the patient's clinical condition, smoking history, family history, personal interests, or social roles.
Assist the patient in stopping a. Set a quit date. Help the patient pick a date within the next 4 weeks.
b. Provide self-help materials. Review the materials with the patient, if desired.
c. Consider prescribing nicotine replacement therapy, especially for highly addicted patients (smoke one pack or more a day or smoke within 30 min of waking).
d. Consider signing a stop-smoking contract with the patient.
e. If the patient is not willing to quit now, provide motivating literature and ask again at the next visit.
Arrange follow-up visits a. Set a followup visit within 1 to 2 weeks after the quit date.
b. Contact the patient within 7 days after the initial visit; reinforce the decision to stop and remind the patient of the quit date and the follow-up visit.
c. At the first follow-up visit, discuss the patient's smoking status to provide support and help prevent relapse. Relapse is common; if it happens, encourage the patient to try again immediately.
d. Set a second follow-up visit in 1 to 2 months. For patients who have relapsed, discuss the circumstances of the relapse and other special concerns.

aAdapted from Glynn, T., and Manley, M. (1993). "How to Help Your Patients Stop Smoking: A National Cancer Institute Manual for Physicians." NIH Publication No. 93-3064. Public Health Service.

IV. LIFESTYLE CHANGES

A. Physical Activity
Regular physical activity is associated with reduced all-cancer mortality through various mechanisms, including altering hormone levels and increasing immune system activity, energy expenditure, and antioxidant activity. Of particular interest is the role of physical activity in reducing weight and body fat and the possible subsequent effect on cancer risk reduction. For example, being overweight or obese is associated with an increased risk of hormone-related cancers.      
Losing weight and body fat reduces circulating levels of estrogen and progesterone, hormones related to breast and colorectal cancer. Studies suggest that moderate physical activity can reduce breast cancer risk in both premenopausal and postmenopausal women. For colorectal cancer, leanness and regular physical activity have been consistently associated with reduced risk in both men and women. Numerous studies have reported that when caloric intake exceeds energy output, there is an increased risk of cancer of the colon, rectum, prostate, endometrium, breast, and kidney.      
The amount of physical activity needed to maintain a healthy weight, lose weight, and promote good health, including reducing cancer risk, is recommended by various organizations in the United States, including the NCI and ACS, to be 30 min of moderate physical activity on most days of the week. This level of activity might include walking briskly (3-4 miles per hour) for about 2 miles, gardening and yard work, jogging, or swimming. The activity does not need to be continuous; the key is to exercise on a regular basis.

B. Sun Exposure
Exposure to sunlight, the main source of ultraviolet (UV) radiation, is implicated as a causative factor in the development of skin cancer--the most common form of cancer, with about 1.3 million new cases each year in the United States. People with red or blond hair and fair skin that freckles or burns easily are at especially high risk for skin cancer. Basal cell and squamous cell carcinomas, both highly curable, account for the majority of skin cancers. However, the incidence of malignant melanoma, the most dangerous form of skin cancer, is increasing. Total worldwide incidence increased about 15% between 1985 and 1990 estimates. In the United States, the lifetime risk of developing melanoma was 1 in 1500 in the 1930s. Now, the risk is 1 in 74, and there will be an estimated 47,700 new cases in the year 2000.      
Abundant evidence has established that skin cancer risk can be reduced by limiting exposure to sunlight and, thus, to UV radiation. Generally, effective protective behaviors include avoiding the sun during midday (especially between 10 AM and 2 PM), wearing protective clothing and broad-brimmed hats, wearing sunglasses, avoiding tanning beds and sun lamps (these are also sources of UV radiation), and using sunscreen that has a sun protection factor (SPF) of 15 or higher, even on hazy or cloudy days.      
The relationship between sunscreen use and melanoma risk is somewhat controversial. Many epidemiologic studies investigating the association between melanoma risk and sunscreen use have found either reduced risk or no clear association. However, findings in others have suggested that sunscreen use is associated with an increased risk for melanoma, leading some to question the advisability of widespread recommendations for its use. It has been hypothesized that people who use sunscreen primarily to avoid sunburn during intentional sun exposure, such as sunbathing, might increase their sun exposure time when using sunscreen, thus increasing their exposure to UV radiation and risk for melanoma. At present, accumulated evidence on sun exposure and skin cancer still warrants using sunscreen as part of an overall sun protection strategy, during both intentional and unintentional exposures (e.g., in gardening or hiking).  

Peter Greenwald
National Cancer Institute, National Institutes of Health, Bethesda, Maryland

Darrell E. Anderson
Sharon S. McDonald
Scientific Consulting Group, Inc., Gaithersburg, Maryland

See Also
CHEMOPREVENTION, PRINCIPLES OF ; CHEMOPREVENTION TRIALS ; NUTRITIONAL SUPPLEMENTS AND DIET AS CHEMOPREVENTIVE AGENTS ; TOBACCO CARCINOGENESIS

GLOSSARY
enzyme A protein molecule produced by living organisms that catalyzes chemical reactions of other substances without itself being destroyed or altered upon completion of the reactions.

epidemiology The science concerned with the study of factors determining and influencing the frequency and distribution of disease and other health-related events and their causes in a defined human population.

gene-nutrient interaction The influence exerted by genes and nutrients on each other. The action may be unilateral or reciprocal and usually involves the alteration of metabolic pathways or products.

micronutrient A vitamin or mineral that the body must obtain from outside sources. Micronutrients are essential to the body in small amounts because they are either components of enzymes (the minerals) or act as coenzymes in managing chemical reactions.

oncogene Mutated and/or overexpressed version of a normal gene that in a dominant fashion can release the cell from normal restraints on growth and thus, alone or in concert with other changes, convert a cell into a tumor cell.

pharmacotherapy The treatment of diseases or conditions by medicines.

polymorphism The occurrence of different forms (alleles) of a gene (typically greater than 1%) in individual organisms or in organisms of the same species, independent of sexual variations.

primary prevention The identification, control, and avoidance of environmental factors related to cancer development.

randomized, controlled trials A clinical trial that uses a control group of people given an inactive substance (placebo) and an intervention group given the substance or action under study.

sunscreen A substance applied to the skin to protect it from the effects of the sun's rays; sunscreens act by either absorbing ultraviolet (UV) radiation or reflecting incident light. Their effectiveness is rated by their sun protection factor (SPF); e.g., a sunscreen with an SPF of 15 allows only 1/15 of the incident UV radiation or light to reach the skin.

Bibliography
Doll, R., and Peto, R. (1981). The causes of cancer: Quantitative estimates of avoidable risks of cancer in the United States today. J. Natl. Cancer Inst. 66, 1191.
Fiore, M. C. (2000). A clinical practice guideline for treating tobacco use and dependence: A U.S. Public Health Service report. J. Am. Med. Assoc. 283, 3244.
Hackshaw, A. K., Law, M. R., and Wald, N. J. (1997). The accumulated evidence on lung cancer and environmental tobacco smoke. Br. Med. J. 315, 980.
Heber, D., Blackburn, G. L., and Go, V. L. W. (eds.) (1999). "Nutritional Oncology." Academic Press, New York.
Hein, D. W., Doll, M. A., Fretland, A. J., Leff, M. A., Webb, S. J., Xiao, G. H., Devanaboyina, U. S., Nangju, N. A., and Feng, Y. (2000). Molecular genetics and epidemiology of the NAT1 and NAT2 acetylation polymorphisms. Cancer Epidemiol. Biomark. Prev. 9, 29.
Pandey, M., Mathew, A., and Nair, M. K. (1999). Global perspective of tobacco habits and lung cancer: A lesson for third world countries. Eur. J. Cancer Prev. 8, 271.
Perera, F. P. (1996) Molecular epidemiology: Insights into cancer susceptibility, risk assessment, and prevention. J. Natl. Cancer Inst. 88, 496.
Reif, A. E., and Heeren, T. (1999). Consensus on synergism between cigarette smoke and other environmental carcinogens in the causation of lung cancer. Adv. Cancer Res. 76, 161.
Rigel, D. S., and Carucci, J. A. (2000). Malignant melanoma: Prevention, early detection, and treatment in the 21st century. CA Cancer J. Clin. 50, 215.
Sinha, R., and Caporaso, N. (1999) Diet, genetic susceptibility and human cancer etiology. J. Nutr. 129, 556S.
U.S. Department of Health and Human Services (1996). "Physical Activity and Health: A Report of the Surgeon General." U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Centers for Chronic Disease Prevention and Health Promotion, Atlanta, GA.
U.S. Department of Health and Human Services (2000). "Reducing Tobacco Use: A Report of the Surgeon General." U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health, Atlanta, GA.
Weinstock, M. A. (1999). Do sunscreens increase or decrease melanoma risk: An epidemiologic evaluation. J. Invest. Dermatol. Symp. Proc. 4, 97.
World Cancer Research Fund (1997). "Food, Nutrition and the Prevention of Cancer: A Global Perspective." American Institute for Cancer Research, Washington, DC.

Antioxidants: Carcinogenic and Chemopreventive Properties

Chemical carcinogens, which are present widely in our environment, are considered to play an important role in the causation of most human cancers. They may be classified into genotoxic and nongenotoxic types, the former including nitrosamines, aromatic hydrocarbons, aromatic amines, and nitrofurans, and the latter being exemplified by peroxisome proliferators, antioxidants, chlorinated pesticides, uracil, and D-limonene. For primary prevention of human cancer, it is essential that we eliminate carcinogens as far as possible and ingest possible chemopreventors. One problem with eliminating carcinogens is the existence of nongenotoxic compounds which escape detection in short-term assays. The development of chemopreventors therefore assumes particular importance in approaches to the prevention of cancer. Since antioxidants possess both carcinogenic and chemopreventive properties, a detailed analysis of this group of agents should contribute to their introduction.

I. INTRODUCTION

Antioxidants have been used widely in the food industry to prevent or retard autooxidation of fats, oils, and fat-soluble compounds. Some water-soluble antioxidants prevent oxidation in aqueous material.      
Apart from foodstuffs, they are contained in various cosmetics, medicines, plastics, and rubbers. They may be synthetic or naturally occurring; the latter type of antioxidants are widely distributed in plants. Humans may thus be exposed to mixtures of synthetic and naturally occurring antioxidants orally from foodstuffs or through the skin from cosmetics. Generally, antioxidants can be classified into five types:
1. Primary antioxidants which terminate free radical chain reactions (phenolic compounds, tocopherols, tertiary amines, flavonoids)
2. Oxygen scavengers which react with oxygen and remove it (ascorbic acid and its derivatives)
3. Secondary antioxidants which decompose lipid peroxides into stable products (sulfur compounds, seleno-compounds)
4. Enzymic antioxidants which remove highly oxidative species (superoxide dismutase, catalase, glutathione peroxidase)
5. Chelating agents which chelate metallic ions such as copper and iron (citric acid, phytic acid)      
Of these antioxidants, ascorbic acid and chelating agents are able to enhance the antioxidant action of tocopherols or phenolic compounds, and are therefore called synergists. Antioxidants are generally not mutagenic as evaluated by the Ames test and may even inhibit the mutagenic activity of mutagens. Moreover, they inhibit chemical carcinogenesis in various organs of rodents when they are given prior to and/or simultaneously with certain carcinogens.       
Therefore, antioxidants may have potential applications as potent chemopreventors in man. However, some of them have been shown to enhance second stage chemical carcinogenesis in rodents when administered after exposure to carcinogens. In addition, the synthetic antioxidant BHA, which is commonly used throughout the world as a food additive, was demonstrated to induce forestomach carcinomas in F344 rats of both sexes and in male Syrian golden hamsters. Subsequently, modifying effects of antioxidants and the mechanism(s) underlying antioxidant induction of tumors have been the focus of extensive investigations.      
In addition to their antioxidant effects, many antioxidants possess various kinds of biological activities such as enzyme induction, interference with the immune response, anti-viral activity, anti-inflammatory activity, interference with prostaglandin synthesis, inhibition of platelet aggregation, and protection against reperfusion injury. Some of these properties are closely linked to chemical carcinogenesis.      
The present review article covers the latest results of research on antioxidants, especially in relation to neoplastic and chemopreventive properties.

II. ANTIOXIDANT CARCINOGENICITY

Carcinogenicity studies of antioxidants have been extensive since the demonstration of forestomach carcinogenicity for the synthetic antioxidant BHA in 1983. Before then, BHA had been used widely in the world as a safe food antioxidant and was generally regarded to be beneficial to humans due to its lack of mutagenicity and its inhibition of rodent carcinogenesis induced by several carcinogens.      
In the first examination of long-term BHA exposure, male and female F344 rats were continuously administered 2% (maximum tolerable dose) or 0.4% BHA in the diet for 2 years. Histopathological examination showed that the high dose of BHA induced forestomach squamous cell carcinomas at incidences of 34.6% in males and 29.4% in females. No neoplastic lesions which could be attributed to BHA administration were observed in any other organs. This was the first report that antioxidants could induce tumors in rodents. Subsequently, BHA was also shown to be carcinogenic to make Syrian golden hamster forestomach at doses at 1 and 2% in the diet, although the results were equivocal when male B6C3F1 mice received doses of 0.5 and 1%. In a dose-response study using F344 male rats at levels of 0.125 to 2%, only the highest dose caused squamous cell carcinomas while 1% induced only a low incidence of squamous cell papillomas; no tumors were observed at doses lower than 0.5%.      
Since the carcinogenicity of BHA appeared limited to the forestomach which humans do not possess, it was necessary to examine its effects on other animals without a forestomach for the evaluation of human hazard potential. Continuous oral treatment with BHA at doses lower than 1% of 20 months in guinea pigs, daily intragastric doses of 500 mg/kg body weight of BHA for 20 days and then 250 mg/kg body weight 5 times/week until 84 days in cynomolgus monkeys, or 0-100 mg/kg/days or dietary 0.25-1% BHA for up to 1 year in beagle dogs did not cause any histopathological changes in any tissue. Therefore it has been concluded that the carcinogenic potential of BHA is limited to forestomach epithelium.      
The carcinogenicity of butylated hydroxytoluene (BHT), which is as potent as BHA as an antioxidant and is commonly used in cosmetics or as a food additive, has also been extensively studied. Wistar rats and B6C3F1 mice of both sexes were treated with BHT at dose levels of 0.25 or 1.0%, and 0.02, 0.1, or 0.5% in the diet, respectively, for up to 104 weeks. However, no tumors were induced that could be attributed to the treatment. On the other hand, administration of 1 or 2% BHT in the diet to male B6C3F1 mice for 104 weeks was associated with an increased incidence of hepatocellular adenomas or foci of alteration in a clear dose-dependent manner.       
Although there was an inverse dose relationship, male C3H mice fed diets containing 0.5 or 0.05% of BHT for 10 months also had significantly increased incidences of liver tumors as compared to those kept on basal diet alone. Olsen et al. also reported weak tumorigenicity for BHT in the rat liver in a two generation study. However, none of the evidence to date is unequivocal and the question of BHT carcinogenicity in rodents remains open.      
Although no standard carcinogenic bioassays have been reported for tert-butylhydroquinone (TBHQ), continuous feeding of this antioxidant for <20 months at doses of 0-0.5% did not result in any compound-related gross or microscopic lesions. Propyl gallate was also studied in F344 rats and B6C3F1 mice by feeding at dietary levels of 5000-20,000 mg/kg, but no dose-related increases in incidences of tumors or differences in survival were found. The potential carcinogenicity of other gallates has not been fully evaluated.      
Since BHA was found to induce pronounced forestomach hyperplasia in a short period and forestomach squamous cell carcinoma in rats and hamsters in the long term, many other structurally related phenolic antioxidants were examined for associated proliferative activity in the forestomach epithelium as an aid to prediction of potential to induce forestomach tumors. Among these are TBHQ, 4-methoxyphenol, 1,4-dimethoxybenzene, catechol, resorcinol, hydroquinone, 3-methoxyphenol, 2-methoxyphenol, anisole, 4-cresol, phenol, 4-hydroxybenzoic acid n-alkyl esters (alkyl parabenes), 4-hydroxybenzoic acid esters, gallic acid, caffeic acid, sesamol, chlorogenic acid, syringic acid, ferulic acid, eugenol, esculin, 4-methylphenol, 4-tert-buthylphenol, pyrogallor, methylhydroquinone, 2-tert-butyl-4-methylphenol, and BHT tested in short-term feeding studies in rats or hamsters at a dose of 0.7-2% in diet; 4-methoxyphenol and sesamol were found to be as active as BHA in the induction of forestomach hyperplasia and, in addition, they caused a circular deep ulceration parallel to the limiting ridge. Caffeic acid, 2-tert-butyl-4-methylphenol, and 4-tert-butylphenol also induced pronounced hyperplasia in the forestomach epithelium. The labeling index in the glandular stomach was also significantly increased in animals fed catechol and 4-methoxyphenol. This prompted investigation in long-term experiments. Chemical structures and the results of carcinogenicity studies of antioxidants are presented in Fig. 1 and Table I, respectively.


FIGURE 1 Chemical structures of carcinogenic antioxidants.

TABLE I Incidences of Tumors in the Stomach and Kidney

Chemicals Sex No. of rats No. of rats with (%) Squamous cell carcinoma (forestomach) No. of rats with (%) Adenocarcinoma (G1 stomach) No. of rats with (%) Adenoma (kidney)
BHA M 52 18 (35)*** 0 0
  F 51 15 (29)*** 0 0
4-Methoxyphenol M 30 23 (77)*** 0
  F 30 4 (13) 0 0
Caffeic acid M 30 17 (57)*** 0 4 (13)
  F 30 15 (50)*** 0 6 (20)*
Sesamol M 29 9 (31) *** 0 0
  F 30 3 (10) 0 0
Catechol M 28 0 15 (54)*** 0
  F 28 0 12 (43)*** 0
4-Methylcatechol M 30 17 (57)*** 17 (57)*** 0
  F 30 12 (40) 14 (47)*** 0
Hydroquinone M 30 0 0 14 (47)***
  F 30 0 0 0
Control M 30 0 0 0
  F 30 0 0 0

*P < 0.02.
***P < 0.001 vs control group value.

When male and female F344 rats were treated with caffeic acid, sesamol, 4-methoxyphenol, or 4- methylcatechol at a dose of 2% or catechol at a dose of 0.8% in the diet for 2 years, the agents except catechol induced significantly increased incidences of forestomach squamous cell carcinomas. Females were less sensitive than males. In addition, 4-methylcatechol induced carcinomas in glandular stomach epithelium. Although catechol did not induce tumors in the forestomach, it did induce glandular stomach carcinomas. In a dose-response study, even 0.16% catechol in the diet induced adenomas at a low incidence.       
Since this demonstration of catechol (1,2-dihydroxybenzene) carcinogenicity, hydroquinone (1,4-dihydroxybenzene), one of its isomers, was further examined in male and female rats at a dose of 0.8% in diet. Whereas it was not found to be carcinogenic for either the forestomach or glandular stomach, it induced a 46.6% incidence of kidney adenomas, predominantly in male rats. 1,2,4-Benzenetriol, protocatechuic acid, protocatechualdehyde, dopamine, and DLdopa, all dihydroxybenzene derivatives like caffeic acid and catechol, were examined for their potency to induce cell proliferation in the rat forestomach and glandular stomach epithelium. However, only 1,2,4- benzenetriol and dopamine, each at 1.5% in the diet, were effective in increasing the BrdU-labeling index in the forestomach epithelium. 2-Methoxyphenol, with one hydroxy substituent replaced by a methoxy substituent, also lacked any effects on cell proliferation in either forestomach or glandular stomach epithelium. Therefore, the ortho-dihydroxy structure appears important, but substituents actually determine cell proliferation on stimulus.      
Many phenolic compounds that do not show stim- ulation activity per se on the forestomach epithelium cause very strong cell proliferation or tumorigenicity when they are combined with sodium nitrite (NaNO2). For example, continuous oral treatment with 0.8% catechol alone in the diet for 51 weeks induced mild forestomach hyperplasia. The grade of forestomach hyperplasia considerably increased and papillomas were also found in 4 of 15 rats with a simultaneous administration of catechol and NaNO2.      
When the combined effects of various phenolic compounds and NaNO2 on rat forestomach cell proliferation were further examined in a 4-week experiment, the cell proliferative response to known forestomach carcinogens such as sesamol, 4-methoxyphenol, and 4-methylcatechol was further enhanced by simultaneous treatment with NaNO2. In addition, markedly increased cell proliferation was found when 2% hydroquinone, 2% pyrogallol, 2% gallic acid, or 2% TBHQ in the diet was combined with 0.3% NaNO2 in the drinking water, although individual phenolic compounds or NaNO2 did not stimulate any proliferating activity or only mild hyperplasia. It has been known that mutagenic diazocompounds are formed by the reaction of phenols and NaNO2 under acidic conditions.
Such compounds could be responsible for the cell proliferation. Cell proliferation induced by these phenolic compounds and NaNO2 in combination was much more pronounced than with the known forestomach carcinogens caffeic acid, sesamol, 4- methoxyphenol, and 4-methycatechol. Similar effects were observed when the nonphenolic antioxidant sodium ascorbate (NaASA) or ascorbic acid (ASA) was given to rats simultaneously with NaNO2. Therefore, many phenolic compounds as well as NaASA and ASA may possess carcinogenic activity for the rat forestomach epithelium in the presence of NaNO2.

III. MECHANISTIC APPROACHES TO ANTIOXIDANT CARCINOGENICITY

Since the carcinogenicity of antioxidants was found to be mostly limited to the forestomach epithelium except in the catechol, 4-methylcatechol, and hydroquinone cases, approaches to elucidating mechanisms have been primarily directed toward this tissue. All of the carcinogenic phenolic antioxidants which target the forestomach were shown to induce cytotoxicity as well as hyperplasia. To examine whether the observed hyperplasia was due to excess regeneration associated with cytotoxicity or due to primary mitogenic effects, early forestomach lesions induced by BHA, caffeic acid, or 4-methoxyphenol in rats were investigated.


FIGURE 2 Sequential observation of labeling indices in rat forestomach epithelium treated with antioxidants. ○, 2% BHA; ●, 2% caffeic acid; ▲, 2% 4-methoxyphenol; --, basal diet.

DNA synthesis of the forestomach epithelium, expressed as the number of BrdU-labeled cells per 100 basal cells (labeling index), increased 12 hr after treatment with caffeic acid or 4-methoxyphenol. In the case of BHA, an increase in the labeling index was apparent 3 days after treatment. After 7 days of continuous antioxidant administration, the labeling index increased or continued to be high, especially in the groups treated with 4-methoxyphenol followed by caffeic acid and BHA (Fig. 2). Hyperplasia was observed 3 days after treatment with caffeic acid, but this change first became evident only later in the cases of BHA, sesamol, and 4-methoxyphenol. Evidence of toxicity, such as erosion or ulceration, developed in animals treated with caffeic acid or 4- methoxyphenol for 7 days, but were not found in those treated with BHA. A strongly elevated expression of cell proliferation-related c-fos oncogene expression in the forestomach epithelium was demonstrated 15 min after beginning treatment with BHA, but rapidly decreased thereafter. Another cell proliferation-related c-myc expression was similarly observed after 15 min of treatment, then decreased slowly. By the electron microscopical observation, the initial changes observed in the forestomach epithelial cells are the enlargement of nucleous and an increase in free ribosomes and polysomes in the basal layer. These changes are observed 24 hr after treatment with caffeic acid and 72 hr after treatment with BHA, without any cytotoxicity. These results strongly suggest that antioxidants primarily induce cell proliferation by direct stimulation and that cell proliferation is probably further enhanced by regeneration subsequent to cytotoxicity.      
Strong cell proliferation, however, does not always correlate with occurrence of carcinomas. It takes a long time (usually more than 1 year) for the development of forestomach carcinomas, and induced hyperplasia regresses after cessation of chemical treatment. We compared reversibility of rat forestomach lesions induced by an intragastric dose of 20 mg/kg body weight N-nitroso-N'-nitro-N-nitrosoguanidine (MMNG) once a week, 20 ppm N-methylnitrosourethane (MNUR) in the drinking water as genotoxic forestomach carcinogens, 2% BHA, 2% caffeic acid, or 2% 4- methoxyphenol in the diet as nongenotoxic carcinogens for 24 weeks.      
Forestomach lesions induced by genotoxic carcinogens did not regress 24 weeks after the removal of the carcinogen stimulus. In contrast, hyperplasia induced by nongenotoxic carcinogens clearly regressed after the cessation of insult. Preneoplastic atypical hyperplasia, observed at high incidences in rats treated with genotoxic carcinogens, was also evident in animals receiving nongenotoxic agents, even after their withdrawal, albeit at low incidences. These results indicate that even with nongenotoxic carcinogens, a heritable alteration at the DNA level could occur during strong cell proliferation and result in atypical hyperplasia development. This preneoplastic lesion might then progress to produce carcinomas. Indeed, weak forestomach genotoxic carcinogens show potent carcinogenicity under cell-proliferating conditions induced by BHA. Thus, when animals were treated with 2% BHA and given sc injections of 50 mg/kg body weight of the weak forestomach carcinogen 3,2'- dimethyl-4-aminobiphenyl (DMAB) once a week or an ip injection of 15 mg/kg body weight of Nmethylnitrosourea (MNU) once every 2 weeks for 22 weeks, the carcinogenic response was amplified. At week 24 the BHA treatment was associated with significant papilloma induction in the forestomach (40%), while no lesions were observed in the group given only DMAB. MNU alone did not induce forestomach carcinomas, but carcinomas were found in 75% of rats receiving BHA in combination with either of the genetoxic agents.       
Although antioxidant forestomach carcinogens are generally negative in the Ames test and in several in vitro and in vivo mutagenesis assays, they can show weak genotoxic activity under certain conditions. Three hours after a single intragastric administration of 40 mg BHA, no detectable DNA damage was present in the forestomach epithelium, but the oxidative metabolite tert-butylquinone (TBQ) did cause DNA damage at 1/1000 of the parent concentration level. Other oxidative BHA metabolites, 3-tert-butyl- 4,5-dihydroxyanisole and 3-tert-butylanisole-4,5- quinone, also showed DNA-damaging activity but were weaker than with TBQ. In vitro incubation of BHA with calf thymus DNA in the presence of an S9 mixture under acidic conditions results in DNA adducts as evaluated by a 32P-postlabeling assay.      
Quinone metabolites of BHA form DNA adducts without the presence of an S9 mixture. Recently, low levels of DNA adducts were demonstrated in the forestomach epithelium of animals treated with 2% BHA for 2 weeks. During prostaglandin H synthasemediated oxidative metabolism of phenolic compounds, active oxygen species could be produced. This finding was supported by the clear inhibition by aspirin of BHA-induced rat forestomach hyperplasia and ESR analyses which showed that prostaglandin H synthase administration resulted in a substantially accelerated metabolism of TBHQ into TBQ, which is accompanied by the formation of superxide anion, hydroxy radical, and hydrogen peroxide. Thus, it is conceivable that active oxygen species are responsible for antioxidant-induced cytotoxicity or carcinogenesis.      
However, the results of investigation of 8-hydroxyguanosine (8-OH-dG), which is a reliable marker of oxidative DNA damage by reactive oxygen species, in the forestomach epithelium in vivo have been equivocal. Caffeic acid also causes metaldependent DNA damage through H2O2 formation in vitro. In addition, food-derived mutagenic compounds could be formed in the stomach by interaction of amines and nitrite, or nitrite and phenolic compounds. Therefore, oxidative metabolites, active oxygen species, and food-derived mutagens might contribute weak genotoxicity which could act in concert with strong cell proliferation. The putative carcinogenic process driven by phenolic compounds in the forestomach epithelium is summarized schematically in Fig. 3.


FIGURE 3 Putative pathway of rat forestomach carcinogenesis induced by phenolic antioxidants. Possible contributing factors are indicated by brackets. Observed changes in the forestomach epithelium are shown to parentheses.

IV. MODIFICATION OF CARCINOGENESIS BY ANTIOXIDANTS IN TWO-STEP CARCINOGENESIS MODELS

Many antioxidants are capable of modifying chemical or ultraviolet carcinogenesis in a broad spectrum of organs. In addition to direct effects on the initiation and/or postinitiation neoplastic process, they can also exert an influence by blocking nitrosamine formation or reducing the activity of promoters such as 12-Otetradecanoylphorbol- 13-acetate (TPA) in mouse skin carcinogenesis. The mechanisms underlying modification appear to vary with the stage of carcinogenesis and with the carcinogen.

A. Modification of Carcinogenesis by Antioxidants in the Initiation Stage
In this stage, antioxidants could modify carcinogenesis by (1) altering the metabolic activation of procarcinogens, (2) altering detoxifying enzymes, (3) direct interaction with the proximate carcinogenic species, (4) trapping active oxygen species, or (5) influencing absorption of carcinogens from the gastrointestinal tract. BHA inhibits benzo[a]pyrene (BP)- or its proximate carcinogen (±)-trans-7,8-dihydrobenzo[a]pyreneinduced mouse forestomach and lung carcinogenesis in the initiation stage. In this case, inhibition of the cytochrome P450-dependent monooxygenase, which metabolizes BP via 7,8-dihydrodiol to the ultimate carcinogen 7,8-diol-9,10-epoxide, and induction of phase II enzymes, such as glutathione S-transferase, which detoxify the proximate carcinogen, eventually resulted in the decreased formation of diol epoxide-DNA adducts. The plant flavonoid ellagic acid inhibits BPinduced mouse lung carcinogenesis by intraperitoneal and/or oral administration. It also reduces the mutagenic activity of BP 7,8-diol-9,10-epoxide and 7,8-diol- 9,10-epoxide-induced mouse pulmonary tumor formation when administered prior to carcinogen. The observed inhibition may be due to ellagic acid decreasing hepatic and pulmonary cytochrome P450 levels, increasing hepatic glutathione S-transferase activity, and directly interacting with ultimate carcinogen. Many phenolic antioxidants, flavonoids, and seleno-compounds are thought to inhibit carcinogenesis by modifying metabolic pathways. On the other hand, BHA was found to enhance dibutylnitrosamine (DBN)-induced hepatocarcinogenesis, possibly due to enhancing cytochrome P450-mediated oxidation of DBN to proximate carcinogenic studies.      
In cases of active oxygen-mediated carcinogenesis, some antioxidants lower the tumor yield or cytotoxicity induced by carcinogens, but protective effects are not general. Continuous oral treatment with 500 ppm potassium bromate (KBrO3) induces renal cell tumors in 90% of rats when given for up to 2 years. Intraperitoneal or intragastric administration of KBrO3 induces cytotoxicity, lipid peroxidation, and increases in 8-hydroxydeoxyguanosine (8-OH-dG) formation at the target site of carcinogenesis, and therefore active oxygen might be involved in its carcinogenic action.      
Combined treatment with glutathione, cystein, or ascorbic acid, but not superoxide dismutase or vitamin E, protected against its associated oxidative DNA damage and nephrotoxicity. On the other hand, ferric nitrilotriacetate-induced nephrotoxicity and lipid peroxidation were protected against by vitamin E. It is known that this renal carcinogen generates hydroxy radicals in the presence of H2O2 in vitro, which cause DNA cleavage and base damage. It produces 8-OH-dG in the kidney DNA and causes lipid peroxidation and nephrotoxicity in the proximal tubules. Hepatocarcinogenesis induced by peroxisomal proliferators, in which excess production of H2O2 may be responsible, is also blocked by BHA and ethoxyquine, and the ascorbic acid derivative (CV 3611), N,N'-diphenyl-p-phenylenediamine, or BHT protect against liver tumor induction by a cholinedeficient diet in which lipid peroxidation and formation of 8-OH-dG are thought to be involved.      
Chlorophyllin inhibits mutagenesis by several carcinogens, such as the heterocyclic amine Trp-P-2, in the Ames test by absorbing carcinogens to form carcinogen-chlorophyllin complexes. In the rat, chlorophyllin accelerates the excretion of Trp-P-2 into feces. Although an inhibition action for chlorophillin in in vivo carcinogenesis has not been demonstrated to date, this chemical might be expected to exert beneficial effects.

B. Modification of Carcinogenesis by Antioxidants on the Promotion/Progression Stage
The modifying effects of antioxidants on carcinogenesis when administered after carcinogen treatment have been examined in various organs by a large number of investigators. Both promoting and inhibitory effects in various organs have been documented, dependent on the organ site and agent, as shown in Table II. Carcinogenic antioxidants usually strongly enhance carcinogenesis in their target organs. Thus BHA, caffeic acid, and 4-methylcatechol promote forestomach carcinogenesis, and catechol and 4- methylcatechol enhance both forestomach and glandular stomach carcinogenesis after pretreatment with MNNG. Both BHA and BHT promote rat urinary bladder carcinogenesis initiated with N-butyl-N-(4- hydroxybutyl)nitrosamine (BBN), and BHT enhances rat urinary bladder, esophagus, and thyroid carcinogenesis in animals pretreated with BBN, DBN, and DHPN, respectively. Without carcinogen pretreatment, they induce hyperplasia or increase the BrdUlabeling index in their target organs. Therefore, promotion effects appear closely related to their potency in inducing cell proliferation.      
On the positive side, BHA was found to inhibit DHPN-initiated lung carcinogenesis, 7,12-dimethylbenz[ a]anthracene (DMBA)-initiated mammary carcinogenesis, and diethylnitrosamine (DEN)-induced hepatocarcinogenesis. BHT also inhibits colon, kidney, and mammary carcinogenesis, and similar findings have been widely gained for synthetic as well as naturally occurring antioxidants. Modulation of ornithine decarboxylase (ODC) activity, generation of active oxygen species, interaction with calcium- and phospholipid-dependent protein kinase C, altered prostaglandin synthesis, and influence on intercellular communication are all factors that might play a role in antioxidant effects on cell proliferation and/or modulation of carcinogenesis.

TABLE II Modifying Effects of Antioxidants on Rat Carcinogenesis after Carcinogen Exposure

Target organ Antioxidant Synthetic BHA Synthetic BHT Synthetic TBHQ Synthetic PG Antioxidant Naturally occurring SA Naturally occurring α-TOC Naturally occurring PA Naturally occurring DDS
Esophagus
Forestomach
Glandular stomach
Colon
Liver
Lung
Kidney
Urinary bladder
Mammary gland NE NE
Thyroid

Note: TBHQ, t-butylhydroquinone; PG, propyl gallate; SA, sodium ascorbate; α-TOC, α-tocopherol; PA, phytic acid; DDS, diallyldisulfide. ↑, enhancement; →, no effect; ↓, inhibition, NE, not examined.

C. Antipromoting Activity of Antioxidants
Antipromotion effects have been primarily demonstrated in the two-stage mouse skin carcinogenesis model. In this system, mice are given a single topical application of DMBA, 3-methylcholanthrene, BP, or BP-7,8-diol-9,10-epoxide as an initiator, then receive a continuous topical treatment with TPA or teleocidin together with synthetic antioxidants such as BHA, BHT, α-tocophenol, or other naturally occurring antioxidants such as green tea polyphenols, curcumin, chlorogenic acid, caffeic acid, and ferulic acid. The observed inhibition of TPA-induced skin promotion may be partly due to a scavenging action of these antioxidants against TPA-induced active oxygens.      
TPA induces superoxide anion radicals and H2O2 release in human peripheral leukocytes in vitro and H2O2 and 8-OH-dG in mouse skin in vivo. These phenomena can be strongly inhibited by copper(II)- (3,5-disopropylsalicylate)2 (CuDIPS) and by nordihydroguaiaretic acid (NDGA), which are superoxide anion radical scavengers and detoxifiers. In an in vivo experiment, CuDIPS and NDGA significantly inhibited TPA-induced skin tumor promotion in mice initiated with DMBA. Induction of ODC in mouse epidermis also appears to be an important factor for TPA-induced skin tumor promotion. This TPAinduced ODC increase could be inhibited by lipoxygenase inhibitors NDGA, morin, fisetin, kaempferol, propyl gallate, esculetin, and BHA. In addition, morin or esculetin treatment was associated with significant inhibition of skin tumor promotion. Thus, the inhibitory effects of flavonoids on TPA-induced ODC induction and tumor promotion roughly paralleled their lipoxygenase inhibition. These results therefore suggest that antioxidants act by scavenging the superoxide anion radicals that are responsible for tumor promotion or by interfering with lipoxygenase in the epidermis induced by TPA.

D. Modification by Blocking Nitrosamine Formation
From the epidemiological viewpoint, a number of studies have suggested that an intake of nitrite and nitrate correlates with a high incidence of human gastric cancer and that this was due to formation of carcinogenic N-nitroso compounds in the stomach by the reaction of nitrite with amines present in foods and certain drugs. Some antioxidants have been shown to prevent nitrosation in vitro and tumor formation in animals by preventing this reaction between nitrite and amines to form N-nitroso compounds. Ascorbic acid and α-tocophenol are well-known inhibitors of nitrosation. For example, they significantly inhibit in vitro nitrosation of secondary amines such as morpholine, piperazine, diethylamine, and N-methylurea. The reaction of ascorbic acid with nitrite proceeds with the reduction of 2 mol of nitrite to nonnitrosating nitric oxide per mole of ascorbic acid, which is oxidized to dehydroascorbic acid. However, the nonnitrosating nitric oxide can, in the presence of oxygen, give rise to higher oxides of nitrogen, which are themselves powerful nitrosating species. Therefore, under certain conditions ascorbic acid can catalyze nitrosation. Nevertheless, sodium ascorbate at 11.5 or 23 g/kg in the diet gave 89-98% inhibition of lung adenoma induction when NaNO2 was applied with a piperazine, morpholine, or methylurea system. Inhibition of nitrosation by ascorbic acid can be observed by examination of human urine following sequential oral doses of nitrite and proline.      
However, the dose of antioxidant required to return the N-nitrosoproline excretion to basal levels was far in excess of the proline administered. Phenolic antioxidants and flavonoids are also capable of blocking nitrosamine formation. In one carcinogenesis study, gallic acid strongly inhibited adenoma induction in mouse lung by morpholine plus NaNO2. The mechanisms by which phenolic compounds inhibit nitrosation involve reduction of nitrite to nitric oxide, or formation of C-nitroso compounds and mutagenic diazoquinone. Therefore, the possibility remains that the C-nitroso compounds or diazoquinone formed could exert carcinogenic activity. Recently, continuous oral administration of ascorbic acid or some phenolic compounds, including catechol, hydroquinone, and gallic acid, and NaNO2 in combination induced strong cell proliferation or papillomas in rat forestomach epithelium. Therefore, much care and attention should be given to this type of cancer prevention.

V. CHEMOPREVENTION OF CARCINOGENESIS BY ANTIOXIDANTS

Antioxidants that have demonstrated an inhibitory effect in experimental chemical carcinogenesis have been proposed as possible chemopreventors in man. However, this has been suggested primarily on the basis of epidemiological findings, and the existence of adverse effects in different experimental models has indicated that care must be taken in application of antioxidants as chemopreventors.      
Necessary characteristics for an ideal chemopreventor include (1) ability to inhibit initiation activity, (2) ability to inhibit promotion or progression activity, (3) ability to block nitrosamine formation, (4) lack of genotoxicity, (5) lack of carcinogenicity, (6) not a carcinogen precursor, (7) lack of enhancing activity at any stage of carcinogenesis, (8) lack of toxicity, and (9) commercially available. Factors (5) to (8) might, however, be ignored if hazardous effects are only evident at doses much higher than the chemopreventive dose.      
Nevertheless, it may still be difficult to find chemopreventors which satisfy all these requirements. For example, sodium ascorbate can satisfy requirements 1, 2, 3, 5, 8, and 9, but not 4, 6, and 7, whereas α- tocopherol satisfies 1-6, 8, and 9 but not 7. The fact that antioxidants may show opposite effects in different organs, particularly in the promotion stage, means, furthermore, that a total body approach using different carcinogenic initiators is necessary for the reliable assessment of second-stage effects.      
Recently we established a multiorgan carcinogenesis model in which five different carcinogens were used as initiators. In this model, not only the enhancing but also the inhibitory effects of chemicals on carcinogenesis either in the initiating or in the promoting stage on major organs (liver, forestomach, small intestine, large intestine, lung, kidney, urinary bladder, and thyroid gland) can be examined in a single experiment, and its application for examining chemopreventive effects of several naturally occurring antioxidants suggested that green tea catechins (GTC) may be in fact possible chemopreventors. Significant decreases in the incidences of small intestinal tumors (adenomas and adenocarcinomas) were evident in the group treated with 1% GTC (content of catechins >91%, of these ( - )-epigallocatechin gallate >54%) during carcinogen exposure (13%) as compared with the carcinogen alone control value (57%). Multiplicities (average no. per rat) were also lower in groups treated with GTC both during (0.13 ± 0.35) and after (0.13 ± 0.48) carcinogen exposure than in the carcinogen alone case (1.07 ± 1.21). No significant differences in esophagus, forestomach, colon, liver, kidney, urinary bladder, lung, and thyroid gland lesion induction were observed.      
Subsequent treatment with 1.0% GTC also inhibited rat mammary tumor development and consequently increased the survival rate (93% vs. 33.3% in DMBA alone) after a single intragastric administration of 50 mg/kg body weight of DMBA. GTC potently lowers the hepatocarcinogenicity of glutamic acid pyrolisate 2-amino-6-methyldipyridol[1,2- a:3',2'-d] imidazole (Glu-P-1) as assessed in terms of number and areas of preneoplastic glutathione Stransferase placental form (GTC-P) positive foci in our medium term liver bioassay. Green tea extracts or green tea polyphenols also inhibit benzo[a]pyrene (BP)-, DMBA, 3-methylcholanthrene-, or ultraviolet light-induced tumor initiation and complete carcinogenesis in mouse skin, 12-O-tetradecanoylphobol-13- acetate caused tumor promotion in mouse skin, Nethyl- N'-nitro-N-nitrosoguanidine (ENNG)-induced mouse duodenal carcinogenesis, azoxymethaneinduced rat colon carcinogenesis, BP- and diethylnitrosamine- induced mouse lung and forestomach carcinogenesis, and 4-(methylnitrosamino)-1-(3-pyridyl)- 1-butanone-induced mouse lung carcinogenesis.      
These inhibitory effects were observed not only in the initiation stage but also in the promotion or progression stage in some cases. The lowest effective dose for protection to occur was 0.005% ( - )-epigallocatechin gallate in the ENNG case and this dose is almost comparable with the daily intake of GTC in green tea drinkers. The mechanism(s) underlying how GTC inhibits carcinogenesis is not fully understood, but inhibition of ornithine decarboxylase and lipoxygenese activities, enhancement of Phase II enzymes, inhibition of oxidative stress induced by promoters or carcinogens, direct interaction between GTC and ultimate carcinogens, inhibition of promoter-induced protein kinase C, reduction of activating enzymes, and stimulation of immunity in the target organs may all play roles. In addition, some epidemiological data indicate a reduced risk of colon tumors and gastric cancers among populations with high levels of green tea consumption.      
1-O-2,3,5-Trimethylhydroquinone (HTHQ) is a strong phenolic antioxidant. In the medium term liver bioassay for the detection of hepatocarcinogens or hepatopromoters in F344 male rats, treatment with Glu- P-1 alone was associated with a significant increase in the number (per cm2 liver) and area (mm2 per cm2 liver) of preneoplastic GST-P-positive foci (47.5 ± 8.9 and 11.1 ± 4.7, respectively). Combined treatment with 1.0% HTHQ significantly reduced the number and area of GST-P-positive foci (to 8.1 ± 2.1 and 0.6 ± 0.2), almost to control level values without chemicals (3.6 ± 1.6 and 0.3 ± 0.1). HTHQ is therefore expected to be a selective potent chemopreventor which could reduce the carcinogenicity of heterocyclic amines such as Glu-P-1 (Fig. 4).


FIGURE 4 Quantitative analyses of the effects of antioxidants on Glu-P-1-induced GST-P positive foci in the medium term liver bioassay. HTHQ, 1-O-2,3,5-trimethylhydroquinone; GTC, green tea catechin. Significantly different from the Glu-P-1 group at ***P < 0.001.

Since chemopreventors exert their actions in different organs, in different stages of carcinogenesis, and dependent on the carcinogen, intake of different chemopreventors in combination may prove to be important for the prevention of human cancer.

VI. EVALUATION OF ANTIOXIDANTS AS HUMAN HAZARDS

There are many synthetic and naturally occurring antioxidants in our environment. Humans may ingest considerable amounts of such compounds in foodstuffs, medicines such as vitamins C and E, and γ- oryzanol, or by absorption through the skin of antioxidant additives in cosmetics, antiseptics, disinfectants, and industrial chemicals. It is therefore possible that these antioxidants may indeed play a role in human carcinogenesis. Although there are some epidemiological and case control studies suggesting that high intake of antioxidants such as ascorbic acid, α-tocopherol, selenium, -carotene, and vegetables that contain vitamins A, C, and E may lower the mortality rate for certain cancer types in humans, no such studies have been performed for phenolic antioxidants.     
For human risk assessment of phenolic antioxidant exposure, and extrapolation from experimental data, it is of importance to take into account the target organ, dose level, and route of administration. BHA is carcinogenic for the rat, hamster, and possibly mouse forestomach epithelium, but this activity is strictly limited to this tissue, and no carcinogenic potential for other squamous epithelia, such as those lining the esophagus and oral cavity, or for glandular stomach has been found. Since humans do not have a forestomach, it appears most likely that such limited forestomach carcinogens would lack effects on human gastric epithelium. Moreover, the threshold carcinogenic dose of BHA in rats is 2% in the diet (only a small incidence of benign papillomas was induced at lower dose levels), a level that is exceedingly high as compared with the possible human exposure. The estimated daily dietary intake of BHA was reported to be less than 7 mg/person in a Canadian study and therefore the carcinogenic dose in animals is nearly 10,000 times higher than the likely human exposure level.      
On the other hand, catechol, which is present in certain foods (e.g., fruits, vegetables, coffee), in tobacco, in cosmetics such as hair dye, in film developers, and in wood smoke, promotes glandular stomach carcinogenesis and induces adenocarcinomas in the rat glandular stomach, which is anatomically and biologically similar to human gastric epithelium at a dose of 0.8% in diet. Catechol at a dose of 0.16% for 2 years also caused glandular stomach adenomas to develop at low incidence. A 0.16% dose level is equivalent to 5-7.5 g catechol per person per day. The amount of catechol and its conjugates actually excreted in urine in humans was reported to be 1.1-30 mg/day, but although the carcinogenic dose in rats is 250-6250 times higher than the estimated human exposure, this chemical might still be a factor for enhancing human gastric cancer.      
Thus, the promotion potential of antioxidants may be far more important for human environmental carcinogenesis than any complete carcinogen action. Experiments have shown that effective enhancement can be achieved at much lower levels than the carcinogenic dose, and since antioxidants can exert promoting potential in various organs that are not necessarily targets for carcinogenicity, as shown in Table II, this must be taken into account. Moreover, clear synergistic effects regarding promotion have been reported; that is, combined treatment with 0.5% caffeic acid, 0.16% catechol, 0.5% BHA, and 0.25% 2-tert-butyl-4-methylphenol in rats for 51 weeks in rats pretreated with MNNG induced an 80% incidence of forestomach squamous cell carcinomas, whereas the individual treatments resulted in only 13-27% incidences. In a long-term carcinogenicity study, rats were treated with 0.4% caffeic acid, 0.4% sesamol, 0.16% catechol, 0.4% BHA, and 0.4% 4- methoxyphenol either alone or in combination for 104 weeks. Although papillomas were found in 0-15.8% of the individual treatment groups, the incidence increased to 42.9% with the combined treatment.      
Such synergistic or additive effects in carcinogenicity or promotion of carcinogenesis have been observed not only in the forestomach. Moreover, the carcinogenic or hyperplasiagenic activity of BHA was enhanced by concurrent treatment with sodium ascorbate or vitamin A, but was inhibited by concomitant treatment with diethylmaleate, a glutathionedepleting agent, or aspirin, and some antioxidants potentiate carcinogenicity of genotoxic carcinogens possibly through metabolic activation. NaNO2 also is a factor that can greatly modify carcinogenicity of phenolic compounds.       
Therefore antioxidant effects may be considerably altered by changes in environmental or physiological conditions. Endogenous factors such as age, immunological condition, and other diseases in target organs will also influence the effective dose for promotion of carcinogenesis or carcinogenicity. Available data thus indicate that low concentrations of carcinogens or promoters, even if they do not show activity per se, may indeed be important for human environmental carcinogenesis.

Nobuyoki Ito
Masao Hirose
Katsumi Imaida
Nagoya City University Medical School, Nagoya, Japan

See Also
ANIMAL MODELS FOR COLON CANCER CHEMOPREVENTION ; CANCER RISK REDUCTION (DIET/SMOKING CESSATION/ LIFESTYLE CHANGES) ; CHEMOPREVENTION, PHARMACOLOGY OF ; MOLECULAR EPIDEMIOLOGY AND CANCER RISK ; NUTRITIONAL SUPPLEMENTS AND DIET AS CHEMOPREVENTIVE AGENTS

GLOSSARY
chemoprevention The primary prevention of carcinogenesis by chemicals. Any chemical agent that inhibits initiation, promotion, or progression, or more than one of these steps, is considered to be a chemopreventor.

forestomach The proximal part of the rodent stomach, situated between the esophagus and glandular part of the stomach and lined by squamous epithelium. Occupying about half of the stomach and separated from the glandular stomach by a limiting ridge, the forestomach is found in rats, mice, and hamsters.

initiation An irreversible alteration in the heritable material of target cells caused by carcinogens; this is considered the first step of the carcinogenesis process.

progression The process by which benign tumors or premalignant lesions progress to increasing malignant behavior.

promotion The process by which initiated cells grow to form benign tumors or precancerous lesions.

squamous cell carcinoma A malignant tumor originating from squamous epithelium and demonstrating squamous differentiation such as cornification and intercellular bridges. Characteristics include frequent invasion of adjacent tissues.

squamous cell papilloma A benign papillary tumor originating from squamous epithelium with fine or abundant connective tissue stroma. Structural and cellular atypia, or invasive growth, is absent.

Bibliography
Hirose, M., Imaida, K., Tamano, S., and Ito, N. (1994). Cancer chemoprevention by antioxidants. In "Food Phytochemicals for Cancer Prevention" (C.-T. Ho, M.-T. Huang, R. T. Rosen, and T. Osawa, eds.), Vol. 547, p. 122, ACS Books, Washington, DC.
Hirose, M., Tanaka, H., Takahashi, S., Futakuchi, M., Fukushima, S., and Ito, N. (1993). Effects of sodium nitrite and catechol, 3-methoxycatechol, or butylated hydroxyanisole in combination in a rat multiorgan carcinogenesis. Cancer Res. 53, 32.
Ito, N., Hirose, M., and Shirai, T. (1992). Carcinogenicity and modification of carcinogenic response by plant phenols. In "Phenolic Compounds in Food and Their Effects on Health II" (M.-T. Huang, C.-T. Ho, and C.-Y. Lee, eds.), Vol. 507, p. 270. ACS Books, Washington, DC.
Ito, N., Shirai, T., and Hasegawa, R. (1992). Medium-term bioassays for carcinogens. In "Mechanisms of Carcinogenesis in Risk Identification" (H. Vainio, P. N. Magee, D. B. McGregor, and A. J. McMichael, eds.), p. 353. International Agency for Research on Cancer, Lyon.
Ito, N., and Imaida, K. (1992). Strategy of research for cancer chemoprevention. Teratog. Carcinog. Mutag. 12, 79.
Ito, N., and Hirose, M. (1989). Antioxidants-carcinogenic and chemopreventive properties. Adv. Cancer Res. 53, 247.
Ito, N., Hirose, M., and Takahashi, S. (1993). Cell proliferation and forestomach carcinogenesis. Environ. Health Perspect. 101, Suppl. 5, 69.
Nera, E. A., Lok, E., Iverson, F., Ormsby, E., Karpinski, K. F., and Clayton, D. B. (1984). Short-term pathological and proliferative effects of butylated hydroxyanisole and other phenolic antioxidants in the forestomach of Fischer 344 rats. Toxicology 32, 197.
Rodrigues, C., Lok, E., Nera, E., Iverson, F., Page, D., Karpinski, K., and Clayton, D. B. (1984). Short-term effects of various phenols and acids on the Fischer 344 male rat forestomach epithelium. Toxicology 38, 103.
Schiderman, P. A. E. L., van Maanen, J. M. S., ten Vaarwerk, F. J., Lafleur, M. V. M., Westmijze, E. J., ten Hoor, F., and Kleinjans, J. C. S. (1993). The role of prostaglandin H synthase-mediated metabolism in the induction of oxidative DNA-damage by BHA metabolites. Carcinogenesis 14, 1297. 

Animal Models for Colon Cancer Chemoprevention

Colorectal cancer is a tumor of colon and rectum, which occurs with high frequency in both men and women in Western countries. Most cases of colorectal cancers arise in a benign adenoma; some evidence suggests that some cancers arise directly from the mucosal cells. With regard to genetic mechanisms of colorectal cancer, the disease appears to result from an increase in the number of genetic mutations, mostly acquired, that accumulate in the genome of the evolving cancer cell. The histopathologic changes associated with the development of colorectal cancer are driven by the progressive accumulation of definable genetic changes, including the activation of one or more oncogenes plus the inactivation of several tumor suppressor genes and by endogenous and exogenous promoting agents. These genetic changes proceed from cellular hyperproliferation to small, benign adenomas to more dysplastic, larger adenomas and then can become cancer and ultimately metastasize. Identification and removal of these premalignant lesions are considered important in order to design a rational approach to reduce the incidence and mortality of colon cancer.

I. INTRODUCTION

In studies of various human diseases, it is critical that reliable animal models both chemically induced and transgenic are developed that demonstrate similarity to the human disease. Animal models are extremely valuable in our understanding of the human disease, but one has to be familiar with the limitations of the model system that is being used to study the human disease. Animal models should bear relevance to human colorectal cancer with similarities not only in terms of histopathology and molecular and genetic lesions during early and promotion/progression stages of carcinogenesis, but also adequacy of the model for prevention studies. The animal models should also reflect the efficacy of both effective and ineffective nutritional and chemopreventive agents that have been evaluated in humans. It should also be recognized that extrapolation of data obtained in animal model systems entails inherent sources of uncertainty that must be taken into account in predicting human responsiveness. This brief article discusses the carcinogeninduced colon cancer model to study the relationship between chemopreventive agents and colon carcinogenesis.

II. PRECLINICAL MODELS FOR COLON CANCER

Animal models have been developed to study the multiple environmental factors involved in the pathogenesis of cancer of the colon. These animal models are (a) induction of colon tumors in rats through aromatic amines such as 3,2 -dimethyl-4-aminobi-phenyl (DMBA); (b) derivatives and analogs of cycacin such as methyazoxymethanol (MAM), 1,2-dimethylhydrazine (DMH), and azoxymethane (AOM) in rats and mice of selected strains; (c) direct-acting carcinogens of the type of alkylureas, such as methylnitrosourea (MNU) or N-methyl-N`nitro-Nnitrosoguanidine (MNNG); and (d) heterocyclic amines such as 2-amino-3-methylimidazo[4,5-f]quinoline (IQ) and 2-amino-1-methyl-6-phenylimidazo[ 4,5-b]pyridine (PhIP). The spectrum of epithelial lesions induced in the colon by these carcinogens is similar to various types of neoplastic lesions observed in the colorectum of humans. Several studies have utilized these relevant animal models to investigate the modulation of colon carcinogenesis by nutritional and chemopreventive agents.

A. Alkylnitrosoureido Compounds
MNNG and MNU are direct alkylating agents, which do not require metabolic activation, and thus they are topical and potent carcinogens. Intrarectal instillation of NMU or MNNG induced colorectal tumors in rodent models. Because of the fact that biochemical activation is not required for these carcinogens, it is an ideal way of inducing colon tumors in animals and of studying modifying effects during the postinitiation stage of colon carcinogenesis without involving metabolism of the genotoxic, initiating carcinogen. Intrarectal administration of MNNG at a dose rate of 1-3 mg/rat/week for 20 weeks induced colon tumors in 100% of male F344 rats of which 43% tumors were adenocarcinomas and 57% were adenomas. The neoplasms were all located in the distal colon and rectum, as MNNG and MNU are locally acting carcinogens.       
No metastatic lesions were usually observed. Although MNNG and MNU given intrarectally provided the most reliable model for the topical and selective production of tumors in the distal colon and rectum, the major weakness of this model is that the technique of intrarectal injection requires highly skilled technicians and quantification of carcinogens instilled intrarectally is difficult.

B. Heterocyclic Amines
2-Amino-3,8-dimethylimidazo[4,5-f]quiniline (IQ), a heterocyclic aromatic amine produced from food pyrolysis, was first isolated from a variety of broiled or cooked fish and meat. Among a number of heterocyclic amines that have been demonstrated to be highly mutagenic and tumorigenic in rodent models, IQ and 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) have attracted a lot of attention because they demonstrate a multitarget organospecificity with specific cancer induction in Zymbal gland, skin, colon, oral cavity, and mammary gland of rodents. Colon tumors were induced in male F344 rats by administering PhIP daily in the diet at 100 and 400 ppm for 52 and 104 weeks. Although the colon tumor incidences were about 43 and 55% in animals given PhIP at 100 and 400 ppm, respectively, there was severe toxicity due to PhIP. We have also utilized heterocyclic amines such as IQ and PhIP to induce colon tumors. Results of these studies from our laboratory indicate that when these agents were administered daily in the diet for 52 weeks, the tumor incidence was very low, ranging between 5 and 28%.

C. 1,2-Dimethylhydrazine
1,2-Dimethylhydrazine is an effective carcinogen for the induction of tumors of the colon and rectum in rats and mice by systemic subcutaneous or intraperitoneal injections. The usefulness of this organospecific carcinogen, which induces selectively tumors in the colon, was confirmed by several laboratories. Despite the differences in doses, schedules, and animal strains used by different investigators, there is some consistency in their results in using DMH as a colon carcinogen. Subcutaneous injection of DMH at a dose rate of 20 mg/kg body weight, once weekly for 20 weeks induces colon adenomas and adenocarcinomas in about 60% of male F344 rats. However, the major weakness of this model is that multiple injections of DMH are required to induce colon tumors in the laboratory rodents.

D. Azoxymethane
Azoxymethane, a metabolite of DMH, has been used extensively by many investigators to induce colon tumors and to study the effects of nutritional factors and chemopreventive agents in colon carcinogenesis. AOM is a potent inducer of carcinomas of the large intestine in various strains of male and female rats. We have used a two-dose (15 mg/kg body weight, once weekly for 2 weeks sc) regimen in all our chemoprevention and nutritional studies since mid-1985.      
Our results and those of others indicate that the just mentioned dose regimen in male F344 rats induced colon tumors in about 80% of animals with a mean of three tumors/rat after 40-50 weeks following the second AOM treatment. Endoscopic examination of animals revealed that the first endoscopically visible colonic tumor can be detected 15 weeks after the AOM treatment and that the mean latency period of such tumors is about 20 weeks.      
Of all the model systems, use of Fischer (F344) rats and AOM seem to be appropriate because rat colons have light and electron microscopic morphology as well as histochemical properties that are quite similar to that of humans and biological behaviors of AOMinduced rat colon carcinomas have close similarity to those of human colon carcinomas. AOM-induced carcinomas metastasize to regional lymph nodes and liver, and these carcinomas are transplantable. Epithelial neoplasms of colon induced by AOM in F344 rats include both adenomas and adenocarcinomas.      
Based on our past experience with this model, about 70% of colon tumors are adenocarcinomas and the rest are adenomas. Histologically, adenomas of the colon are benign, with mild or moderate epithelial atypia. Malignant neoplasms of the colon (epithelial origin) are adenocarcinomas, the majority of which are well-differentiated, frank malignant tumors showing invasion across the line of the muscularis mucosa. Some of these are poorly differentiated, which are highly infiltrative, and often reach the intestinal wall and the serosa and may even invade the neighboring organs. In our past experience with this model, the extension of lesions into the adjacent peritoneal tissues can occur, and metastic lesions, if present, can be seen in the mesenteric lymph nodes, lung, or liver.       
Similar to the regional distribution of tumors in human colon, AOM treatment induces colonic tumors predominantly in the distal colon. AOM treatment also induces ras oncogene mutations at codon 12 of K- and H-ras and increases the expression of the ras family of protooncogenes that have been causally associated with colon tumor development. Enhanced ras oncogene expression has been observed in a variety of human colon tumors. AOM-induced colon tumors also demonstrate enhanced cyclooxygenase-2 and inducible nitric oxide synthase expression similar to human colonic tumors. Mutations in the tumor suppression gene, APC, are known to be early events in the colon cancer process in humans. APC gene mutations have been identified in patients with familial adenomatous polyposis, who have germline mutation in one of the APC alleles, and in sporadic colorectal cancer. Evidence in humans thus implicates the APC suppressor gene as causal in large bowel carcinogenesis. Studies indicating the presence of APC mutation in AOM-induced colon tumors in mice strengthens the concept that these models are appropriate for human colon cancer studies.

III. CHEMOPREVENTION STUDIES UTILIZING PRECLINICAL MODELS

The developmental strategies for chemoprevention have been markedly facilitated by the use of relevant animal models mimicking the neoplastic process that occurs in humans. The F344 rat model for AOM induced colon cancer has been used extensively to obtain critical information on the chemopreventive efficacy of several agents.

A. Phytochemicals
Several studies have demonstrated that generous consumption of vegetables reduces the risk of colon cancer. Although the nature of the constituents of these vegetables and other food items that are responsible for reduced risk has not been fully elucidated, it is clear that the plant foods contain chemopreventive agents, including several micronutrients, such as vitamins, and minerals and also contain nonnutrients, such as organosulfur compounds, polyphenols, and isoflavones, to cite a few. The diversity of these compounds is a positive feature, indicating that a variety of approaches to cancer prevention by these agents may be made so that the optimal selection will emerge.      
Mechanisms of chemopreventive activity of these agents range from inhibition of carcinogen activation to detoxification of the carcinogen, blockage of binding of critical carcinogen metabolites to DNA, scavenging reactive electrophiles, and inhibiting arachidonic acid (AA) metabolism. Phytochemicals and their substituted and synthetic analogs tested for their efficacy in the AOM colon cancer model assay included anethole trithione, oltipraz [5-(2-pyrazinyl)-4- methyl-1,2-dithiole-3-thione], diallyl sulfide, and curcumin, to cite a few.      
Administration of organosulfur compounds such as diallyl sulfide, oltipraz, or anethole trithione during the initiation and/or postinitiation stage significantly suppressed the incidence and multiplicity of AOMinduced colonic adenocarcinomas in male F344 rats (Table I). The inhibition of colon carcinogenesis by these agents was associated with an increase in the activities of detoxifying enzymes, such as glutathione S-transferase, quinone reductase, and UDPglutathione transferase in the colonic mucosa and tumors.

TABLE I Chemopreventive Efficacy of Naturally Occurring and Synthetic Agents against Azoxymethane-Induced Colon Carcinogenesis in F344 Rats

Chemopreventive agent Dosage μg/g diet (ppm) % Inhibition
Oltipraz 200 35
Curcumin 2000 42
Aspirin 200 32
Ibuprofen 400 45
Sulindac 320 55
Piroxicam 400 64
Celecoxib 1500 96
Piroxicam + DL-difluoromethylornithine 150 + 1000 80
1,4-Phenylenebis(methylene)selenocyanate 20 42

Curcumin [diferuloylmethane; 1,7-bis-(4-hydroxy- 3-methoxyphenyl)-1,6-heptadiene-3,5-dione)], which has been identified as the major pigment in turmeric, the powdered rhizome of Curcuma longa Lim, possesses both anti-inflammatory and antioxidant properties. Importantly, dietary administration of curcumin inhibited AOM-induced colon tumor incidence and multiplicity in a dose-dependent manner (Table I). Curcumin, given as a dietary supplement during the promotion/progression period, dramati- cally inhibited colon tumorigenesis, suggesting that curcumin may retard growth and/or the development of existing neoplastic lesions in the colon and that this agent may be an effective chemopreventive agent for individuals at high risk for colon cancer development, such as patients with polyps.

B. Nonsteroidal Anti-inflammatory Drugs
In recent years, attention has been drawn to the potential chemopreventive properties of nonsteroidal anti-inflammatory drugs (NSAIDs). Several casecontrol and cohort studies have provided unequivocal evidence for the inverse relationship between colon cancer and the use of NSAIDs, specifically aspirin. There has been ample and consistent experimental evidence from laboratory animal model studies to indicate that NSAIDs, including indomethacin, piroxicam, sulindac, aspirin, ibuprofen, and ketoprofen, inhibit chemically induced colon cancer (Table I). More importantly, piroxicam and sulindac administered during the promotion/progression stage significantly inhibit colon tumorigenesis. Results generated in this preclinical model assay provided baseline information for eventual clinical evaluation of the efficacy of NSAIDs in the late intervention/prevention protocols of colonic tumors in high-risk individuals, such as patients with sporadic colonic polyps of familial adenomatous polyposis (FAP).      
One of the mechanisms by which NSAIDs inhibit colon cancer is through the modulation of cyclooxygenase 1 and 2 (COX-1 and COX-2), which leads to a reduction of eicosanoid production, which in turn affects cell proliferation and tumor growth. However, these drugs can cause unwanted side effects, including gastrointestinal ulceration, bleeding, and renal toxicity, through the inhibition of constitutive COX-1 activity. Overexpression of COX-2 has been observed in colon tumors, and many commonly used NSAIDs have very little selectivity for COX-1 or COX-2; therefore, more specific yet minimally toxic inhibitors of COX-2 were developed and tested for chemopreventive efficacy. Celecoxib, a selective COX-2 inhibitor that induces very few toxic side effects, has been found to be significantly more effective than the commonly used NSAIDs in the chemoprevention of colon carcinogenesis in laboratory animal models; it may thus be an effective chemopreventive agent against colon cancer (Table I). A recent clinical trial has shown a reduction in adenomas in patients with FAP-administered celecoxib.

C. Combinations of Low Doses of Various Chemopreventive Agents
There is increasing interest in the use of combinations of low doses of chemopreventive agents that differ in mode of action, rather than administering single agents as a means of obtaining increased efficacy and minimized toxicity. This approach is extremely important when a promising chemopreventive agent demonstrates apparent efficacy but may produce toxic effects at higher doses. An example of combination of agents producing positive results in laboratory animal models has been a study in which the NSAID piroxicam and DFMO, a specific irreversible enzymeactivated or suicide inhibitor of ornithine decarboxylase (ODC), were evaluated for their chemopreventive efficacy. An important finding of the study was that the lowest dose levels of piroxicam (100 ppm) and DFMO (1000 ppm), when administered together, were more effective in inhibiting the incidence and multiplicity of colon adenocarcinomas than administration of these compounds as single agents, even at higher levels. These data strongly support the view that the use of combinations of chemopreventive agents having diverse actions should have beneficial applications in human cancer chemoprevention trials. This should be one of the approaches used in future research and human intervention trials.

D. Organoselenium
Epidemiological studies have also pointed to an inverse association between dietary selenium intake and colon cancer risk in humans. A randomized clinical trial demonstrated that supplementation of seleniumenriched brewer's yeast reduced the incidence and mortality from cancer of the colon. This finding was corroborated by studies with selenium supplementation of the diet in chemically induced colon carcinogenesis in laboratory animals. Humans ingest primarily organic forms of selenium, such as selenomethionine and selenocysteine, by eating grains, vegetables, and animal products. Chronic feeding of inorganic and certain organic forms of selenium at levels >5 ppm produced toxic effects. Therefore, substantial efforts were made to find and/or develop forms of organic selenium compounds that have the maximal chemopreventive efficacy and lowest possible toxicity. Studies in our laboratory have indicated that certain synthetic organoselenium compounds, such as 1,4-phenyulenebis(methylene)selenocyanate (p-XSC), hold great promise as chemopreventive agents because they have been found to be superior to historically used selenium compounds, such as sodium selenite and selenomethionine. More importantly, the chemopreventive efficacy of this agent is more pronounced when given along with a low-fat diet, thus making a strong case for the use of low-fat dietary regimens along with a chemopreventive agent as a desirable approach for primary prevention in the general population and for secondary prevention of colon cancer in high-risk individuals.

IV. CONCLUSION

An impressive body of observation supports the concept that chemoprevention has the potential to be a major component of colon cancer prevention and control. Accumulating evidence indicates that several NSAIDs, including aspirin, piroxicam, and sulindac, can reduce the incidence of colon cancer in laboratory animals and in humans. Celecoxib, a selective COX-2 inhibitor that induces very few toxic side effects as compared to traditional NSAIDs, has been found to be more effective than the commonly used NSAIDs against colon carcinogenesis in laboratory animal models. Studies have also indicated that the synthetic organoselenium compound p-XSC holds great promise as a chemopreventive agent because its chemopreventive index is higher than inorganic and naturally occurring organic forms of selenium. Growing knowledge of the mechanisms by which chemopreventive agents act offers opportunities to use combinations of specific chemopreventive agents, the aggregate action of which would be clinically beneficial while toxicity would be minimal. How to best use such knowledge toward prevention and control of colorectal cancer is a primary challenge for the future.

Bandaru S. Reddy
American Health Foundation, Valhalla, New York

See Also
CHEMOPREVENTION, PRINCIPLES OF ; CHEMOPREVENTION TRIALS ; COLORECTAL CANCER: EPIDEMIOLOGY AND TREATMENT ; HEREDITARY COLON CANCER AND DNA MISMATCH REPAIR ; NUTRITIONAL SUPPLEMENTS AND DIET AS CHEMOPREVENTIVE AGENTS ; TRANSGENIC MICE IN CANCER RESEARCH

GLOSSARY
cancer chemoprevention Intervention with chemical agents that may block the tumor initiation and promotion events that are the sequential stages of cancer development or delay the carcinogenic process.

preclinical efficacy Studies involving the evaluation of agents in a realistic laboratory animal model for cancer indicating the inhibition of carcinogenesis.

Bibliography
Clark, L. C., Combs, G. F., Turnbull, B. W., State, E. H., Chalker, D. K., Chow, J., Gover, R. A., Graham, G. F., Gross, E. G., Krongard, A., Lesher, J. L., Park, K., Sanders, B. S., Smith, C. L., and Taylor, J. R. (1996). Effect of selenium supplementation for cancer prevention in patients with carcinoma of the skin: A randomized controlled trial. J. Am. Med. Assoc. 276, 1957-1963.
Dubois, R. N., Radhika, A., Reddy, B. S., and Entingh, A. J. (1996). Increased cycloozygenase-2 levels in carcinogeninduced rat colonic tumors. Gastroenterology 110, 1259-1262.
Elwell, M. R., and McConnell, E. S. (1990). Small and large intestine. In "Pathology of Fischer Rat" (G. A. Boorman, S. L. Eustis, M. R. Wlwell, C. A. Montgomer, and Mackenzie, eds.), p. 43 Academic Press, San Diego.
Holt, P. R., Mokuolu, A. O., Distler, P., Liu, T., and Reddy, B. S. (1996). Regional distribution of carcinogen-induced colon neoplasia in the rat. Nutr. Cancer 25, 129-135.
Ito, N., Hasegawa, R., Sano, S., Tamano, S., Esumi, H., Takayama, S., and Sugimura, T. (1991). A new colon and mammary carcinogen in cooked food, 2-amino-1-methyl-6- phenylimidazo-[4,5-b]pyridine (PhIP). Carcinogenesis 12, 1503-1506.
Kawamori, T., Lubet, R., Steele, V. E., Kelloff, G. J., Kaskey, R. B., Rao, C. V., and Reddy, B. S. (1999). Chemopreventive effect of curcumin, a naturally-occurring antiinflammatory agent, during the promotion/progression stages of colon cancer. Cancer Res. 59, 597-601.
Kawamori, T., Rao, C. V., Siebert, K., and Reddy, B. S. (1998). Chemopreventive activity of celecoxib, a specific cyclooxygenase-2 inhibitor, against colon carcinogenesis. Cancer Res. 58, 409-412.
Kensler, T. W., Tsuda, H., and Wogan, G. N. (1999). United States-Japan workshop on new rodent models for the analysis and prevention of carcinogenesis. Cancer Epidemiol. Biomark. Prevent. 8, 1033-1037.
Kune, G. A., Kune, S., and Watson, L. F. (1988). Colorectal cancer risk, chronic illness, operations and mediations: Case-control results from the Melbourne Colorectal Cancer Study. Cancer Res. 48, 4399-4404.
Pozharisski, K. M. (1990). Tumors of the intestines. In "Pathology of Tumors in Laboratory Animals" (V. Turuso, and U. Mohr, Eds.), Vol. 1, pp. 159-198. IARC Scientific Publication No. 99, Lyon, France.
Rao, C. V., Rivenson, A., Simi, B., and Reddy, B. S. (1995). Chemoprevention of colon carcinogenesis by dietary curcumin, a naturally occurring plant phenolic compound. Cancer Res. 55, 2219-2225.
Rao, C. V., Rivenson, A., Simi, B., Zang, E., Kelloff, G., Steele, V., and Reddy, B. S. (1995). Chemoprevention of colon carcinogenesis by sulindac, a non-steroidal antiinflammatory agent. Cancer Res. 55, 1464-1472.
Rao, C. V., Tokumo, K., Kelloff, G., and Reddy, B. S. (1991). Inhibition by dietary oltipraz of experimental intestinal carcinogenesis induced by azoxymethane in male F344 rats. Carcinogen. (Lond.) 12, 1051-1056.
Reddy, B. S., Hirose, Y., Lubet, R. A., Steele, V. E., Kelloff, G. J., Paulson, S., Siebert, K., and Rao, C. V. (2000). Chemoprevention of colon cancer by specific cyclooxygenase-2 inhibitor, celecoxib, administered during different stages of carcinogenesis. Cancer Res. 60, 293-297.
Reddy, B. S., Maruyama, H., and Kelloff, G. J. (1987). Doserelated inhibition of colon carcinogenesis by dietary piroxicam, a nonsteroidal anti-inflammatory drug, during different stages of rat colon tumor development. Cancer Res. 47, 5340-5346.
Reddy, B. S., Narisawa, T., Maronpot, R., and Weisburger, J. (1975). Animal models for the study of dietary factors and cancer of the large bowel. Cancer Res. 35, 3421-3426.
Reddy, B. S., Nayini, J., Tokumo, K., Rigotty, J., Zange, E., and Kelloff, G. (1990). Chemoprevention of colon carcinogenesis by concurrent administration of piroxicam, a nonsteroidal antiinflammatory drug, with D,L- -difluoromethylornithine, an ornithine decarboxylase inhibitor in diet. Cancer Res. 50, 2562-2568.
Reddy, B. S., Rao, C. V., Rivenson, A., and Kelloff, G. J. (1993). Inhibitory effect of aspirin on azoxymethaneinduced colon carcinogenesis in F344 rats. Carcinogen. (Lond.) 14, 1493-1497.
Reddy, B. S., Rao, C. V., Rivenson, A., and Kelloff, G. (1997). Chemoprevention of colon carcinogenesis by organosulfur compounds. Cancer Res. 53, 3494-3498.
Reddy, B. S., and Rivenson, A. (1993). Inhibitory effect of Bifidobacterium longum on colon, mammary, and liver carcinogenesis induced by 2-amino-3-methylimidazo[4,5-f] quinoline, a food mutagen. Cancer Res. 53, 3914-3918.
Reddy, B. S., Rivenson, A., El-Bayoumy, K., Upadhyaya, P., Pittman, B., and Rao, C. V. (1997). Chemoprevention of colon cancer by synthetic organoselenium compounds, 1,4-phenylenebis(methylene)selenocyanate and p-methoxy benzyl selenocyanate, in low and/or high fat fed F344 rats. J. Natl. Cancer Inst. 89, 506-516.
Reddy, B. S., Tokumo, K., Kulkarni, N., Aliga, C., and Kelloff, G. (1992). Inhibition of colon carcinogenesis by prostaglandin synthesis inhibitors and related compounds. Carcinogenesis 13, 1019-1023.
Reddy, B. S., Wantanabe, K., and Weisburger, J. H. (1997). Effect of high-fat diet on colon carcinogenesis in F344 rats treated with 1,2-dimethylhydrazine, methylazoxymethanol acetate, or methylnitrosourea. Cancer Res. 37, 4156-4158.
Singh, J., Hamid, R., and Reddy, B. S. (1997). Dietary fat and colon cancer: Modulation of cyclooxygenase-2 by types and amount of dietary fat during the postinitiation stage of colon carcinogenesis. Cancer Res. 57, 3465-3470.
Singh, J., Kulkarni, N., Kelloff, G., and Reddy, B. S. (1994). Modulation of azoxymethane-induced mutational activation of ras protooncogens of chemopreventive agents in colon carcinogenesis. Carcinogenesis 15, 1317-1323.
Thun, M. H., Namboodiri, M. M., and Heath, C. W. (1991). Aspirin use and reduced risk of fatal colon cancer. N. Engl. J. Med. 325, 1593-1596.

Tobacco Carcinogenesis

Tobacco carcinogenesis is the process by which tobacco products and their constituents interact with cells to cause cancer. Cigarettes are the main tobacco product worldwide. Manufactured cigarettes are available in all countries, but in some areas of the world, roll-your-own cigarettes are still popular. Other smoked products include kreteks, clove-flavored cigarettes popular in Indonesia, and "sticks" which are smoked in Papua, New Guinea. Bidis, a small amount of tobacco wrapped in temburni leaf and tied with a string, are very popular in India and neighboring areas and have recently taken hold in the United States.       
Cigars are presently increasing in popularity, and pipes are still used. A substantial amount of tobacco is consumed worldwide in the form of smokeless tobacco products. These include chewing tobacco, dry snuff used for nasal inhalation, moist snuff, which is placed between the cheek and gum, a popular practice in Scandinavia and North America, and pan or betel quid, a product used extensively in India. All products are complex mixtures of defined compounds, many of which are capable of inducing tumors. The compounds that induce tumors are called carcinogens. The study of cancer induction by tobacco products and their constituents results in the elucidation of mechanisms by which many of the common human cancers develop.

I. INTRODUCTION

Due to the addictive power of nicotine, worldwide tobacco use is staggering. According to estimates by the World Health Organization, there are about 1100 million smokers in the world, representing approximately one-third of the global population aged 15 years or higher. China alone has approximately 300 million smokers, about the same number as in all developed countries combined. Globally, about 47% of men and 12% of women smoke. Smoking prevalence varies widely by country. For example, in Korea, 68% of men smoke daily, while the corresponding figure for Sweden is 22%. Among women, the highest smoking prevalence is in Denmark, where 37% of women smoke, while in many Asian and developing countries, prevalence is reported to be less than 10%. Although smoking prevalence is lower in the less developed countries in general, it is expected that this will increase markedly as smoking takes hold and larger numbers of young smokers grow older.      
Tobacco in all forms was consumed to the extent of 6.5 billion kg annually in the period 1990-1992, and there were six trillion cigarettes sold. It does not appear that tobacco use will disappear in the near future. Worldwide, smoking is estimated to have caused about 1.05 million cancer deaths in 1990. About 30% of all cancer death in developed countries is caused by smoking. The corresponding figure for developing countries is 13%. Lung cancer is the dominant malignancy caused by smoking, with 514,000 lung cancer deaths attributed to smoking in developed countries in 1995.      
Smoking is also an important cause of bladder cancer, cancer of the renal pelvis, oral cancer, oropharyngeal cancer, hypopharyngeal cancer, laryngeal cancer, esophageal cancer, and pancreatic cancer. Other cancers that may be caused by smoking include renal adenocarcinoma, cancer of the cervix, myeloid leukemia, colon cancer, and stomach cancer. In the United States, about 30% of all cancer death is caused by smoking, similar to worldwide estimates for developed countries. Lung cancer was rare at the beginning of the 20th century. However, the incidence and death rates increased as smoking became more popular. In the United States, the lung cancer death rate in 1930 for men was 4.9 per 100,000. By 1990, this had increased to 75.6 per 100,000. The lung cancer death rate can be shown to parallel the curves for cigarette smoking prevalence, with an approximate 20-year lag time. In 1964, the first Surgeon General’s report on the health consequences of cigarette smoking was published. Following this landmark report, smoking prevalence began to decrease in the United States, but there has been no change since 1990. There are still 47 million adult smokers in the United States.     
Unburned tobacco is a cause of oral cavity cancer. The annual mortality from tobacco chewing in south Asia, where it is used primarily in the form of betel quid, is estimated to be of the order of 50,000 deaths per year. Oral cavity cancer is the leading cancer killer in India. Snuff dipping, as practiced in North America, is an accepted cause of oral cavity cancer as well. The prevalence of snuff dipping has increased markedly in recent years in the United States, especially among young males.

II. TUMOR INDUCTION IN LABORATORY ANIMALS

One goal of scientists studying tobacco carcinogenesis  has been to replicate in laboratory animals the effects  of tobacco products that are observed in humans.  This has been challenging for a variety of  reasons, which can be summarized simply: laboratory  animals will not voluntarily use tobacco products the  way humans do.            
The International Agency for Research on Cancer  has reviewed and summarized this work. According  to their conclusions, experimental studies evaluating  the ability of cigarette smoke and its condensate to  cause cancer in laboratory animals have collectively  demonstrated that there is sufficient evidence that  inhalation of tobacco smoke, as well as topical application  of tobacco smoke condensate, causes cancer  in experimental animals. The Syrian golden hamster  has been the model of choice for inhalation  studies of cigarette smoke because it has a low background  incidence of spontaneous pulmonary tumors  and little interfering respiratory infection. Inhalation  of cigarette smoke has repeatedly caused carcinomas  in the larynx of hamsters and this model system has  been widely applied. It is the most reliable model  for induction of tumors by inhalation of cigarette  smoke. Studies in mice, rats, and dogs have been less  frequent.            
According to the International Agency for Research  on Cancer, there are a number of operational  problems inherent in inhalation studies of cigarette  smoke. The smoke must be delivered in a standardized  fashion and this has been accomplished in different  ways. Both whole body exposure and noseonly  exposure designs have been used. Generally, a  2-s puff from a burning cigarette is diluted with air  and forced into the chamber. Animals will undergo  avoidance reactions and will not inhale the smoke.  Thus, the dose to the lung is less than in humans,  which partially explains the occurrence of larynx tumors  rather than lung tumors in hamsters. Unlike  humans, rodents are obligatory nose breathers. Their  nasal passages are more complex than those of humans,  thereby affecting particle deposition in the respiratory  tract. Tobacco smoke is irritating and toxic,  creating further problems in inhalation studies with  rodents.            
Inhalation studies have reproducibly demonstrated  that cigarette smoke, especially its particulate phase,  causes laryngeal carcinomas in hamsters. Some experiments  with mice resulted in low incidences of  lung tumors, in tests of both mainstream smoke and  environmental tobacco smoke. Evidence shows that  gas phase components of cigarette smoke may be tumorigenic  in mice. Respiratory tract tumors were  produced in one long-term exposure of rats to cigarette  smoke. Studies in rabbits and dogs were equivocal.  Treatment-related tumors other than those of  the respiratory tract have not been consistently  observed.            
Cigarette smoke condensate (CSC) has been tested  extensively for tumor induction. CSC is produced by  passing smoke through cold traps. The material in the  traps is recovered by washing with a volatile solvent,  which is then evaporated. Some volatile and semivolatile constituents may be lost during this process.  CSC is roughly equivalent to cigarette total particulate  matter (TPM), the material collected on a glass  fiber filter that has had smoke drawn through it. The  term "tar," which is often used in official reports on  cigarette brands, is equivalent to TPM but without  nicotine and water.              
CSC generation and collection techniques have  been standardized. CSC has been widely tested for  carcinogenicity in mouse skin. Consistently, CSC induces  benign and malignant skin tumors in mice. This  bioassay has been employed to evaluate the carcinogenic  activities of cigarettes of different designs and  to investigate subfractions of CSC. Mouse skin studies  led to the identification of carcinogenic polycyclic  aromatic hydrocarbons in cigarette smoke. The overall  carcinogenic effect of CSC on mouse skin appears  to depend on the composite interaction of tumor initiators  and enhancing factors, such as tumor promoters  and cocarcinogens.            
Many studies have evaluated tumor induction in  rodents by extracts of unburned tobacco. Although  some positive results have been obtained, there is  presently no widely accepted and reproducible model  for the induction of oral cavity cancer in rodents by  tobacco extracts, despite strong human data. There  are probably cofactors that contribute to human oral  cancer upon tobacco use, which are not reproduced  in animal studies.

III. CARCINOGENS IN TOBACCO AND TOBACCO SMOKE

A second important goal of scientists in the field of tobacco carcinogenesis has been the identification of carcinogens in tobacco products. When these compounds are identified, studies can be designed to investigate the mechanisms by which they cause cancer, which in turn can provide important insights on the ways in which tobacco products cause cancer in humans.      
When cigarette tobacco is burned, mainstream smoke and sidestream smoke are produced. Mainstream smoke is the material drawn from the mouth end of a cigarette during puffing. Sidestream smoke is the material released into the air from the burning tip of the cigarette plus that which diffuses through the paper. The mainstream smoke emerging from the cigarette is an aerosol containing about 1 × 1010 particles/ ml, ranging in diameter from 0.1 to 1.0 μm (mean diameter 0.2 μm). About 95% of the smoke is made up of gases-- predominantly nitrogen, oxygen, and carbon dioxide. For chemical analysis, the smoke is arbitrarily separated into a vapor phase and a particulate phase, based on passage through a glass-fiber filter pad called a Cambridge filter.
In addition to nitrogen, oxygen, and carbon dioxide, the gas phase contains substantial amounts of carbon monoxide, water, argon, hydrogen, ammonia, nitrogen oxides, hydrogen cyanide, hydrogen sulfide, methane, isoprene, butadiene, formaldehyde, acrolein, pyridine, and other compounds. The particulate phase contains more than 3500 compounds, and most of the carcinogens. Some major constituents of the particulate phase include nicotine and related alkaloids, hydrocarbons, phenol, catechol, solanesol, neophytadienes, fatty acids, and others. Many of the components are present in higher concentration in sidestream smoke than in mainstream smoke; this is especially true of nitrogen-containing compounds. However, a person’s exposure to sidestream smoke is generally far less than to mainstream smoke because of dilution with room air.


FIGURE 1 Structures of organic pulmonary carcinogens in tobacco smoke.

TABLE I Summary of Carcinogens in Cigarette Smoke

Type Number of compounds
Polycyclic aromatic hydrocarbons 10
Aza-arenes 3
N-Nitrosamines 10
Aromatic amines 4
Heterocyclic aromatic amines 8
Aldehydes 2
Miscellaneous organic compounds 18
Inorganic compounds 7
Total 62

There are over sixty carcinogens in cigarette smoke that have been evaluated by the International Agency for Research on Cancer and for which there is "sufficient evidence for carcinogenicity" in either laboratory animals or humans. The types of carcinogens, based on their chemical classes, are listed in Table I. Carcinogens specifically associated with lung cancer are listed in Table II. The 20 compounds included in this list have been found convincingly to induce lung tumors in at least one animal species and have been positively identified in cigarette smoke. Structures of the organic compounds are shown in Fig. 1. These compounds are most likely involved in lung cancer induction in people who smoke.

TABLE II Pulmonary Carcinogens in Cigarette Smoke

Carcinogen class Compound Amount in mainstream cigarette smoke (ng/cigarette) Sidestream/mainstream smoke ratio Representative lung tumorigenicity in species
Polycyclic aromatic hydrocarbons Benzo[a]pyrene 20-40 2.5-3.5 Mouse, rat, hamster
  Benzo[b]fluoranthane 4-22   Rat
  Benzo[j]fluoranthane 6-21   Rat
  Benzo[k]fluoranthane 6-12   Rat
  Dibenzo[a,i]pyrene 1.7-3.2   Hamster
  Indeno[1,2,3-cd]pyrene 4-20   Rat
  Dibenz[a,h]anthracene 4   Mouse
  5-Methylchrysene 0.6   Mouse
Aza-arenes Dibenz[a,h]acridine 0.1   Rat
  7H-Dibenzo[c,g ]carbazole 0.7   Hamster
N-Nitrosamines N-Nitrosodiethylamine ND-2.8 40 Hamster
  4-(Methylnitrosamino)-1-(3-pyridyl)-1-butanone 80-770 1-4 Mouse, rat, hamster
Miscellaneous organic compounds 1,3-Butadiene 20-70 × 103   Mouse
  Ethyl carbamate 20-38   Mouse
Inorganic compounds Nickel 0-510 13-30 Rat
  Chromium 0.2-500   Rat
  Cadmium 0-6670 7.2 Rat
  Polonium-210 0.03-1.0 pCi 1.0-4.0 Hamster
  Arsenic 0-1400   None
  Hydrazine 24-43   Mouse

Polycyclic aromatic hydrocarbons are condensed ring aromatic compounds that are formed during all incomplete combustion reactions, such as those that occur in the burning cigarette. Among these, benzo[a]pyrene (BaP) is the most extensively studied compound. Its presence in cigarette smoke and ability to induce lung tumors upon local administration or inhalation are firmly established. It causes lung tumors in mice, but not in rats, when administered systemically. In studies of lung tumor induction by implantation in rats, BaP is more carcinogenic than several other polycyclic aromatic hydrocarbons in tobacco smoke.      
Aza-arenes are nitrogen-containing analogs of polycyclic aromatic hydrocarbons. Two aza-arenes, dibenz[a,h]acridine and 7H-dibenzo[c,g]carbazole, are pulmonary tumorigens when tested by implantation in the rat lung and instillation in the hamster trachea, respectively. The carcinogenic activity of dibenz[a,h]acridine is less than that of BaP, whereas that of 7H-dibenzo[c,g]carbazole is greater than BaP. The levels of both compounds in cigarette smoke are relatively low.      
N-Nitrosamines comprise a large group of potent carcinogens. Among these, N-nitrosodiethylamine is an effective pulmonary carcinogen in the hamster, but not the rat. Its levels in cigarette smoke are low compared to those of other carcinogens. 4-(Methylnitrosamino)- 1-(3-pyridyl)-1-butanone (NNK) is a potent lung carcinogen in rats, mice, and hamsters. NNK is called a tobacco-specific N-nitrosamine because it is a chemical derivative of nicotine, and thus occurs only in tobacco products. It is the only compound in Table II that induces lung tumors systemically in all three commonly used rodent models. The specificity of NNK for tumor induction in the lung is remarkable; it induces lung tumors independent of the route of administration and in both susceptible and resistant strains of mice. The systemic administration of NNK to rats is a reproducible and robust method for the induction of lung tumors. Cigarette smoke contains substantial amounts of NNK, and the total dose experienced by a smoker in a lifetime of smoking is remarkably close to the lowest total dose shown to induce lung tumors in rats. Levels of NNK and total polycyclic aromatic hydrocarbons in cigarette smoke are similar.      
Lung is one of the multiple sites of tumorigenesis by 1,3-butadiene in mice, but is not a target in the rat. Ethyl carbamate is a well-established pulmonary carcinogen in mice but not in other species. Nickel, chromium, cadmium, and arsenic are all present in tobacco and a percentage of each is transferred to mainstream smoke; arsenic levels are substantially lower since discontinuation of its use as a pesticide in 1952. Metal carcinogenicity depends on the valence state and anion; these are poorly defined in many analytical studies of tobacco smoke. Thus, although some metals are effective pulmonary carcinogens, the role of metals in tobacco-induced lung cancer is unclear. Levels of polonium-210 in tobacco smoke are not believed to be great enough to appreciably impact lung cancer in smokers. Hydrazine is an effective lung carcinogen in mice and has been detected in cigarette smoke in limited studies.      
Considerable data indicate that polycyclic aromatic hydrocarbons and NNK play very important roles as causes of lung cancer in people who smoke. The other compounds discussed earlier may also contribute, but probably to a lesser extent.      
Polycyclic aromatic hydrocarbons and Nnitrosamines such as NNK and N'-nitrosonornicotine (NNN) are probably involved as causes of oral cavity cancer in smokers. N-Nitrosamines such as NNN and NDEA are likely causes of esophageal cancer in smokers. The risk of oral cavity cancer and esophageal cancer in smokers is markedly enhanced by the consumption of alcoholic beverages. NNK is also believed to play a prominent role in the induction of pancreatic cancer in smokers, whereas aromatic amines such as 4-aminobiphenyl and 2-naphthylamine are the most likely causes of bladder cancer.       
Cigarette smoke and CSC are tumor promoters, e.g., they enhance the carcinogenicity of tumor initiators when administered subsequent to the initiators. The majority of the tumor-promoting activity seems to be due to uncharacterized weakly acidic compounds. Substantial levels of cocarcinogens, which enhance the carcinogenicity of tumor initiators when applied together with the initiators, are present in cigarette smoke. Catechol is prominent among these. In addition, cigarette smoke contains high levels of acrolein, which is toxic to the pulmonary cilia, and other agents such as nitrogen oxides, acetaldehyde, and formaldehyde, which could contribute indirectly to pulmonary carcinogenicity through their toxic effects.       
While cigarette smoke is extraordinarily complex, unburned tobacco is simpler. With respect to carcinogens, the tobacco-specific nitrosamines NNK and NNN are the most prevalent strong cancer-causing agents in unburned tobacco products. A mixture of NNK and NNN induces oral tumors in rats, and consequently these compounds are considered to play an important role as causes of oral cavity cancer in people who use smokeless tobacco products.


FIGURE 2 Scheme linking nicotine addiction and lung cancer via tobacco smoke carcinogens and their induction of multiple mutations in critical genes. PAH, polycyclic aromatic hydrocarbons; NNK, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butane.

IV. MECHANISMS OF TUMOR INDUCTION

The mechanisms by which tobacco causes cancer can best be illustrated by considering the relationship between cigarette smoking and lung cancer because it is here that the most information is available. The overall framework for discussing this information is illustrated in Fig. 2. Carcinogens form the link between nicotine addiction and cancer. Nicotine addiction is the reason that people continue to smoke. While nicotine itself is not considered to be carcinogenic, each cigarette contains a mixture of carcinogens. Thus, cigarettes are disastrous nicotine delivery devices. Most of the carcinogens in cigarette smoke require metabolic activation, i.e., they must be enzymatically transformed by the host into reactive intermediates in order to exert their carcinogenic effects. There are competing detoxification pathways, which result in harmless excretion of the carcinogens. The balance between metabolic activation and detoxification differs among individuals and will affect cancer risk.      
A great deal is known about mechanisms of carcinogen metabolic activation and detoxification. The metabolic activation process leads to the formation of DNA adducts, which are carcinogen metabolites bound covalently to DNA, usually at guanine or adenine. There have been major advances in our understanding of DNA adduct structure and its consequences in the past two decades and there is now a large amount of mechanistic information available. If DNA adducts escape cellular repair mechanisms and persist, they can cause miscoding, resulting in a permanent mutation in DNA. This occurs when DNA polymerase enzymes read an adducted DNA base incorrectly, resulting in the insertion of the wrong base.       
Other errors can also occur due to the presence of DNA adducts. Cells that contain damaged DNA may be removed by apoptosis, or programmed cell death. If a permanent mutation occurs in a critical region of an oncogene or tumor suppressor gene, it can lead to activation of the oncogene or deactivation of the tumor suppressor gene. Oncogenes and tumor suppressor genes play critical roles in the normal regulation of cellular growth. Changes in multiple oncogenes or tumor suppressor genes result in the production of aberrant cells with loss of normal growth control. Ultimately, this leads to lung cancer. While the sequence of events has not been well defined, there can be little doubt that these molecular changes are important. There is now a large amount of data on mutations in the human K-ras oncogene and p53 tumor suppressor gene in lung tumors from smokers.      
Blocking any of the horizontal steps in Fig. 2 may lead to decreased lung cancer, even in people who continue to smoke. The following discussion considers some of these steps in more detail.      
Upon inhalation, cigarette smoke carcinogens are enzymatically transformed to a series of metabolites as the exposed organism attempts to convert them to forms that are more readily excreted. The initial steps are usually carried out by cytochrome P450 (P450) enzymes, which add oxygen to the substrate. These enzymes typically are responsible for the metabolism of drugs, other foreign compounds, and some endogenous substrates. Other enzymes, such as lipoxygenases, cyclooxygenases, myeloperoxidase, and monoamine oxidases, may also be involved, but less frequently.       
The oxygenated intermediates formed in these initial reactions may undergo further transformations by glutathione-S-transferases, uridine-5'-diphosphateglucuronosyltransferases, sulfatases, and other enzymes, which are typically involved in detoxification. Some of the metabolites produced by P450s react with DNA or other macromolecules to form adducts. Metabolic pathways of BaP and NNK, representative pulmonary carcinogens in cigarette smoke, have been extensively defined through studies in rodent and human tissues.      
The major metabolic activation pathway of BaP is conversion to a reactive diol epoxide metabolite called BPDE; one of the four isomers produced is highly carcinogenic and reacts with DNA to form adducts with N2- of deoxyguanosine. The major metabolic activation pathways of NNK and its main metabolite, 4- (methylnitrosamino)-1-(3-pyridyl)-1-butanol (NNAL), occur by hydroxylation of the carbons adjacent to the N-nitroso group ( -hydroxylation), which leads to the formation of two types of DNA adducts: methyl adducts, such as 7-methyguanine or O6-methylguanine, and pyridyloxobutyl adducts.  
Considerable information is available on pulmonary carcinogen metabolism in vitro, both in animal and in human tissues, but fewer studies have been carried out on uptake, metabolism, and adduct formation of cigarette smoke lung carcinogens in smokers. Various measures of cigarette smoke uptake in humans have been used, including exhaled carbon monoxide, carboxyhemoglobin, thiocyanate, and urinary mutagenicity.     
BaP has been detected in human lung; no differences between smokers and nonsmokers were noted. 1-Hydroxypyrene and its glucuronide, urinary metabolites of the noncarcinogen pyrene, have been widely used as indicators of polycyclic aromatic hydrocarbon uptake. 1-Hydroxypyrene levels in smokers are generally higher than in nonsmokers. Overall, there is considerable evidence that pulmonary carcinogens in cigarette smoke are taken up and metabolized by smokers as well as by nonsmokers exposed to environmental tobacco smoke.       
Less than 20% of smokers will get lung cancer. Susceptibility will depend in part on the balance between carcinogen metabolic activation and detoxification in smokers. This is an important area requiring further study. Most investigations have focused on the metabolic activation pathways by quantifying DNA or protein adducts. There is considerable data demonstrating the activation of BaP to DNA adducts in the lungs of smokers. Earlier investigations demonstrated that cigarette smoke induces aryl hydrocarbon hydroxylase (AHH) activity and proposed a relationship between AHH activity and lung cancer. AHH metabolizes BaP and is equivalent to P450 1A1. Cigarette smoking induces expression of this enzyme. Lung tissue from recent smokers with elevated AHH activity metabolically activated BaP to a greater extent than lung tissue from nonsmokers or ex-smokers. DNA adduct levels correlated with AHH activity in the same samples. Collectively, these results support the existence of a cigarette smoke-inducible pathway leading to BaP-DNA adducts in smokers’ lungs (Fig. 2).      
Several studies have detected 7-methylguanine in human lung. Levels were higher in smokers than in nonsmokers in two studies, suggesting that NNK may be one source of these adducts. While 7-methylguanine is not generally considered as an adduct that would lead to miscoding in DNA and the introduction of a permanent mutation, other methyl adducts that do have miscoding properties, such as O6-methylguanine, are formed at the same time, but at lower levels. Pyridyloxobutylated DNA also has been detected in lung tissue from smokers, reflecting metabolic activation of NNK or NNN. The detection of methyl and pyridyloxobutyl adducts in DNA from smokers’ lungs is consistent with the ability of human lung tissue to metabolically activate NNK, but the quantitative aspects of the relationship of metabolism to DNA adduct levels are unclear.      
DNA repair processes are important in determining whether DNA adducts persist. Because smoking is a chronic habit, one would expect a steady-state DNA adduct level to be achieved by the opposing effects of damage and repair. Mechanisms of DNA repair include direct repair, base excision repair, and nucleotide excision repair. With respect to smoking and lung cancer, direct repair of O6-methylguanine by O6- methylguanine-DNA alkyltransferase and nucleotide excision repair of polycyclic aromatic hydrocarbon- DNA adducts would appear to be the most relevant processes.      
As indicated in Fig. 2, the direct interaction of metabolically activated carcinogens with critical genes such as the p53 tumor suppressor gene and the K-ras oncogene is central to the hypothesis that specific carcinogens form the link between nicotine addiction and lung cancer. The p53 gene plays a critical role in the delicate balance of cellular proliferation and death. It is mutated in about half of all cancer types, including over 50% of lung cancers, leading to loss of its activity for cellular regulation. Point mutations at guanine (G) are common. In a sample of 550 p53 mutations in lung tumors, 33% were G → T transversions, whereas 26% were G → A transitions. (A purine → pyrimidine or pyrimidine → purine mutation is referred to as a transversion, whereas a purine → purine or pyrimidine → pyrimidine mutation is called a transition.) A positive relationship between lifetime cigarette consumption and the frequency of p53 mutations and of G → T transversions on the nontranscribed DNA strand also has been noted.       
These observations are generally consistent with the fact that most activated carcinogens react predominantly at G and that repair of the resulting adducts would be slower on the nontranscribed strand, thus supporting the hypothesis outlined in Fig. 2. Mutations in codon 12 of the K-ras oncogene are found in 24-50% of human primary adenocarcinomas but are rarely seen in other lung tumor types. When K-ras is mutated, a complex series of cellular growth signals is initiated. Mutations in K-ras are more common in smokers and ex-smokers than in nonsmokers, which suggests that they may be induced by direct reaction of the gene with an activated tobacco smoke carcinogen. The most commonly observed mutation is GGT → TGT, which typically accounts for about 60% of the codon 12 mutations, followed by GGT → GAT (20%) and GGT → GTT (15%).       
The p16INK4a tumor suppressor gene is inactivated in more than 70% of human nonsmall cell lung cancers via homozygous deletion or in association with aberrant hypermethylation of the promoter region. In the rat, 94% of adenocarcinomas induced by NNK were hypermethylated at the p16 gene promoter. This change was frequently detected in hyperplastic lesions and adenomas, which are precursors to the adenocarcinomas induced by NNK. Similar results were found in human squamous cell carcinomas of the lung. The p16 gene was coordinately methylated in 75% of carcinoma in situ lesions adjacent to squamous cell carcinomas that had this change. Methylation of p16 was associated with loss of expression in tumors and precursor lesions, indicating functional inactivation of both alleles. Aberrant methylation of p16 has been suggested as an early marker for lung cancer. The expression of cell cycle proteins is related to the p16 and retinoblastoma tumor suppressor genes; NNK induced mouse lung tumors appear to resemble human nonsmall cell lung cancer in the expression of cell cycle proteins. The estrogen receptor gene is also inactivated through promoter methylation. There was concordance between the incidence of promoter methylation in this gene in lung tumors from smokers and from NNK-treated rodents.      
Loss of heterozygosity and exon deletions within the fragile histidine triad (FHIT) gene are associated with smoking habits in lung cancer patients and have been proposed as a target for tobacco smoke carcinogens. However, point mutations within the coding region of the FHIT gene were not found in primary lung tumors.      
Collectively, evidence favoring the sequence of steps illustrated in Fig. 2 as an overall mechanism of tobacco-induced cancer is extremely strong, although there are important aspects of each step that require further study. These include carcinogen metabolism and DNA binding in human lung, the effects of cigarette smoke on DNA repair and adduct persistence, the relationship between specific carcinogens and mutations in critical genes, and the sequence of genetic changes leading to lung cancer.      
Using a weight of the evidence approach, specific polycyclic aromatic hydrocarbons such as BaP and the tobacco-specific nitrosamine NNK can be identified as probable causes of lung cancer in smokers, but the contribution of other agents, such as those listed in Table II, cannot be excluded. The chronic exposure of smokers to the DNA-damaging intermediates formed from these carcinogens is consistent with our present understanding of cancer induction as a process that requires multiple genetic changes. Thus, it is completely plausible that the continual barrage of DNA damage produced by tobacco smoke carcinogens causes the multiple genetic changes that are associated with lung cancer. While each dose of carcinogen from a cigarette is extremely small, the cumulative damage produced in years of smoking is substantial.

V. SUMMARY

Although substantial progress has been accomplished in reducing the tobacco habit, worldwide use of tobacco products is still immense, due mainly to the addictive power of nicotine, arguably the single compound responsible indirectly for more cancer death than any other chemical. Tobacco products cause about 30% of all cancer death. Since the mid-1950s, studies in tobacco carcinogenesis have identified the major carcinogens in tobacco smoke and have elucidated the overall framework by which these carcinogens cause cancer in people. This series of steps, as illustrated in Fig. 2, is well established, although many details remain unclear. Blocking any of the horizontal steps in Fig. 2, even in people who continue to smoke, would result in decreased cancer mortality. Identification of individuals particularly susceptible to the carcinogenic properties of tobacco products would also be important. Rational approaches are now possible in this regard and, even if only partially successful, could have a major impact on cancer mortality because of the sheer magnitude of the epidemic of cancer death caused by tobacco products.

Acknowledgments
Our studies in tobacco carcinogenesis are supported by Grants CA-44377, CA-81301, and CA-85702 from the National Cancer Institute. The author is an American Cancer Society Research Professor, supported by Grant RP-00-138.

Stephen S. Hecht
University of Minnesota Cancer Center

See Also
BLADDER CANCER: EPIDEMIOLOGY ; CANCER RISK REDUCTION (DIET/SMOKINGCESSATION/LIFESTYLECHANGES) ; CARCINOGEN-DNA ADDUCTS ; CARCINOGENESIS: ROLE OF ACTIVE OXYGEN AND NITROGEN SPECIES ; CHEMICAL MUTAGENESIS AND CARCINOGENESIS ; ESOPHAGEAL CANCER: RISK FACTORS AND SOMATIC GENETICS ; HORMONAL CARCINOGENESIS ; LUNG, LARYNX, ORAL CAVITY, AND PHARYNX ; MOLECULAR EPIDEMIOLOGY AND CANCER RISK

GLOSSARY
carcinogen Any compound that is capable of inducing tumors in laboratory animals or humans.

DNA adduct A covalent binding product formed between a chemical and DNA.

metabolic activation Process by which a carcinogen is converted to a more reactive form that can bind to DNA.

metabolic detoxification Process by which a carcinogen is converted to a form that is excreted without reacting with DNA.

nitrosamines Compounds having a nitroso group bound to the nitrogen of a secondary amine.

polycyclic aromatic hydrocarbons A group of compounds consisting of more than two condensed benzene rings, generally formed in the incomplete combustion of organic matter.

Bibliography
Blot, W. J., and Fraumeni, J. F., Jr. (1996). Cancers of the lung and pleura. In "Cancer Epidemiology and Prevention" (D. Schottenfeld and J. Fraumeni, eds.), pp. 637-665. Oxford Univ. Press, New York.
Doll, R. (1996). Cancers weakly related to smoking. Brit. Med. J. 52, 35-49.
Guengerich, F. P., and Shimada, T. (1998). Activation of procarcinogens by human cytochrome P450 enzymes. Mutat. Res. 400, 201-213.
Guerin, M. R., Jenkins, R. A., and Tomkins, B. A. (1992). "The Chemistry of Environmental Tobacco Smoke: Composition and Management." Lewis Publishers, Chelsea, MI.
Hecht, S. S. (1998). Biochemistry, biology, and carcinogenicity of tobacco-specific N-nitrosamines. Chem. Res. Toxicol. 11, 559-603.
Hecht, S. S. (1999). Tobacco smoke carcinogens and lung cancer. J. Natl. Cancer Inst. 91, 1194-1210.
Hoffmann, D., and Hecht, S. S. (1990). Advances in tobacco carcinogenesis. In "Handbook of Experimental Pharmacology" (C. S. Cooper and P. L. Grover, eds.), pp. 63-102. Springer-Verlag, Heidelberg.
Hoffmann, D., and Hoffmann, I. (1997). The changing cigarette, 1950-1995. J. Toxicol. Environ. Health 50, 307-364.
International Agency for Research on Cancer (1985). Tobacco habits other than smoking: Betel quid and areca nut chewing and some related nitrosamines. In "Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans," Vol. 37. IARC, Lyon.
International Agency for Research on Cancer (1986). Tobacco smoking. In "Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans," Vol. 38. IARC, Lyon.
Sekido, Y., Fong, K. W., and Minna, J. D. (1998). Progress in understanding the molecular pathogenesis of human lung cancer. Biochim. Biophys. Acta 1378, F21-F59.
Smith, T. J., Livingston, S. D., and Doolittle, O. J. (1997). An international literature survey of "IARC group I carcinogens" reported in mainstream cigarette smoke. Food Chem. Toxicol. 35, 1107-1130.
World Health Organization (1997). "Tobacco or Health: A Global Status Report." World Health Organization, Geneva.

Multistage Carcinogenesis

Ahallmark of carcinogenesis, the process of tumor development in an organism, is a long latent period with no clinical evidence of disease. The agedependent incidence of diagnosed cancers in humans suggests that carcinogenesis commonly proceeds via four to seven independent rate-limiting steps. Both animal carcinogenesis models and the analysis of human clinical samples support this stepwise progression of tumorigenesis. The genetic and biochemical defects that occur during this period to transform gradually a normal cell that maintains strict control over both intracellular and intercellular events to a cell mass with abnormal growth potential and an ability to invade adjacent tissues remain incompletely understood. Determining the temporal sequence of specific etiologically relevant events in tumorigenesis has been greatly aided by clinical and histopathological identification of a range of distinct stages in the progression of malignancies, i.e., from precursor lesion to metastasis. This recognizable biological progression must reflect a molecular progression within the genetic complement of the cell, which normally maintains multiple independent barriers to each stage of the malignant conversion process. During the past two decades, it has become clear that breaching these barriers depends on the gradual accumulation of irreversible alterations in an unknown number of genes.       
The molecular functions of these genes are broadly categorized as either inhibiting or promoting tumor development, i.e., tumor suppressor genes, whose functional activity is switched off or downregulated; or dominant-acting oncogenes, whose functional activity is switched on, upregulated, or otherwise deranged. Genes from both functional groups are normally involved in the homeostatic regulation of various cell processes and in coordinating communication and compatibility with neighboring cells. Current models of malignant transformation in specific tumor types are focusing on the identification of the precise genetic perturbations at each stage and elucidating the impact these molecular defects have on proliferation, differentiation, and intercellular relationships of the tumor population. Exploitation of this knowledge should result in more effective strategies for the diagnosis and treatment of the cancer patient and in the development of more specific preventive measures for the individual at high risk for developing a neoplasm.

I. EARLY OBSERVATIONS

The search for a relevant animal model of human carcinogenesis led to a series of classical experiments in the 1940s that defined how the process would be viewed for many years to come. It was observed that benign papillomas and malignant carcinomas could be induced in the skin of mice by application of a single subcarcinogenic dose of polycyclic hydrocarbons followed by a secondary treatment consisting of repeated wounding or application of an irritant such as croton oil. Neither treatment alone resulted in carcinogenesis, and reversing the order of application eliminated the effect. Surprisingly, tumor formation occurred when the secondary treatment was applied up to 1 year following the initial exposure to a carcinogen. Rous and Kidd pioneered the understanding of this latency phenomenon by conceptually dividing the process of carcinogenesis in the mouse skin model into two distinct stages: initiation and promotion.       
They defined initiation, a rapid process producing no apparent morphological change, as a priming event involving DNA damage resulting in irreversible genetic alterations that confer upon cells the ability to form tumors when subsequently exposed to a promoting agent. Promotion, characterized by clonal expansion of the initiated cells and dramatic morphological and biochemical changes, was considered to be an epigenetic phenomenon due to the finding that it could be reversed in the absence of continued treatment. Besides the application of chemical promoters such as phorbol esters, many diverse stimuli were found to have tumor-promoting effects, including UV irradiation and repeated physical abrasion. A common theme of promoting events appeared to be skin irritation.      
In 1964 Foulds described initiation and promotion as part of a larger continuous carcinogenesis process of "progression." Later investigators redefined progression as the third stage of carcinogenesis, following promotion and characterized by a higher degree of malignancy as evidenced by an increased ability to proliferate and invade local tissues and a propensity to metastasize to distant sites. The progression stage also correlates with severe genetic damage, including visible karyotypic alterations in the majority of cells. Although it was unclear whether this phenomenon was a cause or an effect of neoplastic transformation, it had clinical importance; i.e., in many tumor types, DNA aneuploidy is a poor prognostic factor. By the mid-1970s the observation that all cells in many primary tumors exhibited the same abnormal karyotype or similarities in the karyotype of marker chromosomes prompted the idea that most neoplasms arise from a single cell of origin and that genetic instability acquired during the process of neoplasia results in genetic variability within the original clone, allowing for subsequent selection of more aggressive sublines.      
As described earlier, the multiple barriers a cell must overcome to become fully malignant may explain why cancer is a relatively rare event. However, it was quickly realized that based on known mutation rates for nongermline cells (~10-7 per gene/cell division), the combination of four to six genetic events necessary for the neoplastic transformation of a cell would be mathematically so rare as to virtually preclude any spontaneous tumor development during an average human lifetime. Cancer development must therefore be a self-accelerating process in which the first mutational event or events caused genetic instability, leading to an increased mutation rate. The identification of critical genomic lesions in human carcinogenesis is complicated by the background of diverse genetic defects, including not only point mutations but also deletions, amplifications, and rearrangements of genes and chromosomes present in most biopsied tumors. Although the three-stage mouse skin model of carcinogenesis was useful, it was recognized that a more precise understanding of the molecular events in carcinogenesis was needed.

II. IDENTIFYING THE MOLECULAR EVENTS IN MULTISTAGE CARCINOGENESIS

A. Oncogenes
The study of oncogenic viruses such as the Rous sarcoma virus led to the discovery of specific viral genes that were responsible for cell transformation. At the same time, investigators found that DNA isolated from human carcinomas and other tumors was able to induce neoplastic transformation at high efficiencies when transfected into transformation-sensitive "normal" mouse NIHT3 cells. In the 1970s, a group of cellular transforming genes, termed "oncogenes," was identified by homology to the transforming genes of retroviruses and by the biological activity of tumor cell DNA in transfection assays. Transfection of NIHT3 cells with mos (the normal cellular homologue of the transforming gene of the Moloney sarcoma virus) or with H-ras (the normal cellular homologue of the transforming gene of the Harvey sarcoma virus) under the control of viral transcriptional regulatory sequences resulted in cellular transformation.      
These findings suggested that oncogenesis was the result of dominant genetic alterations in which the functional activity of these genes was upregulated or expressed in an abnormal form. Protooncogenes (normal cellular homologues of transforming genes) were found to be (1) highly conserved in vertebrate evolution, (2) expressed in a highly regulated manner, and (3) key players in the growth control mechanisms of normal cells. It seemed logical that the derangement or overexpression of one or a combination of cell growth-related genes could transform cells in a "growth gone wrong" scenario. Revisiting the chemical carcinogenesis in mouse skin model, it was found that introducing the ras oncogene into keratinocytes via transducing retroviruses was tantamount to chemical initiation: subsequent application of promoting agents to infected cells resulted in papilloma formation.      
In addition, point mutations in the H-ras oncogene were invariably found in methylnitrosourea induced breast tumors in rats. It appeared as if a critical lesion in carcinogenesis had at last been identified. An important caveat in the oncogene theory of cancer was that cell transformation by transfection of a single oncogene was only observed under certain limited experimental conditions. The established rodent cell lines (such as NIHT3) used in the original transfection experiments were already phenotypically immortal and therefore partially transformed. Additionally, the results could not be duplicated in human cell lines, cautioning against oversimplification of the carcinogenic process in humans. As the number of oncogenes associated with human cancers increased, researchers were frustrated by the inability to associate a specific genetic lesion with a particular tumor type to a degree that indicated causality. Analysis of human tumors confirmed that there was no one particular oncogene that was necessary, let alone sufficient, for any given type of cancer. With the advent of transgenic mice bearing oncogenes, it was conclusively demonstrated that simply harboring oncogene mutations in a particular cell lineage is insufficient for tumor development.

B. Lessons from the Rb Gene
The examination of hereditary cancers by Knudson revealed a new paradigm important to the pathogenesis of cancer. If the stages of carcinogenesis were defined by genetic events, could a person with a hereditary disposition to cancer be already "initiated" by virtue of an inherited genetic lesion? In the case of retinoblastoma, the defective gene was localized to a band on the long arm of chromosome 13. Knudson's analysis suggested that, in contrast to oncogenes, this "antioncogene" acts in a recessive manner, with one normal allele being adequate to protect against tumorigenesis.      
Thus, two separate mutational events were needed for retinoblastoma formation, one to inactivate each copy of an antioncogene. In 1989 Weinberg refined Knudson's hypothesis by providing a more sophisticated molecular model for the process. Weinberg based his model on the insights following the molecular cloning and analysis of the retinoblastoma gene Rb. At the time, little was known about the precise function of protooncogenes in normal cells or about their regulation. When primary cell cultures (as opposed to partially transformed immortal lines such as NIHT3) are transfected with ras, only small numbers of cells acquire the oncogene and they do not proliferate to form visible foci. If, however, the transfection includes acquisition of neomycin resistance, subsequent selection results in a pure population of transformed cells whose growth proceeds in an uncontrolled manner. It had also been observed that while implantation of cells transformed by an oncogenic virus (e.g., Harvey sarcoma virus) onto the back of a mouse resulted in rapidly growing squamous carcinomas, cell proliferation resulting in tumor formation could be nearly completely prevented by implanting the transformed cells together with a fourfold excess of normal fibroblasts. Weinberg proposed that these observations could be explained if neighboring normal cells exert a constant inhibitory effect on the growth of transformed cells. Therefore, a critical event in carcinogenesis is when cells gain the ability to overcome the limiting effect of their normal tissue environment by ignoring or neutralizing the effect of inhibitory growth signals. He suggested that a number of key genes in growth regulatory pathways could contribute to carcinogenesis by suffering mutations resulting in their inactivation or downregulation. He termed this class of antioncogenes "tumor suppressor" genes.      
Weinberg and colleagues identified the Rb antioncogene, central to the development of retinoblastomas, as a putative tumor suppressor. Knudson's analysis of normal tissues and retinoblastomas from the same patients had suggested that inactivation of both alleles was the critical event in tumorigenesis. It was known that the viral E1A oncoprotein from human adenovirus type 5 formed complexes with the Rbencoded protein and that the region of E1A responsible for its ability to bind to the Rb gene product was crucial to the tumorigenic properties of E1A. The idea that loss of the Rb gene product enabled deregulation of cell growth was supported by experiments showing that introducing a cloned copy of the Rb gene into retinoblastoma cells restored normal growth control.
Subsequent work has validated the Rb gene product as an important regulator of cell growth; it is part of a cellular pathway responding to extracellular antigrowth factors such as transforming growth factor β (TGF-β) and it controls the activity of the EF2 transcription factors responsible for activation of the genes essential for progression from G1 into S phase. The pRb pathway has proved to be central to the cellular antigrowth signaling circuit and is disrupted in some manner in the majority of human cancers. The appealing notion that loss of regulation of a gene that in some manner controls cellular growth through genetic or epigenetic mechanisms was essential for cell transformation allowed carcinogenesis to be described as the net result of the combination of at least two molecular events: activation of an oncogene and inactivation of a tumor suppressor gene. This was consistent with the multistage nature of animal models of carcinogenesis. Disruption of one cellular pathway triggered proliferation, and a complimentary disruption conferred upon transformed cells the ability to overcome inhibitory effects of their normal neighbors. Tumorigenesis would be the result of sustained, uncontrolled growth that rendered the cell population susceptible to other mutagenic events.

C. The Colorectal Carcinogenesis Paradigm
Colorectal cancer was the first human tumor type in which the oncogene activation/tumor suppressor gene inactivation model was conclusively validated. In 1990, Fearon and Vogelstein published an elegant model for the development of colorectal cancer that could be broadly applied to the entire field of carcinogenesis research (Fig. 1). This tumor type was uniquely suited for the study of multistep carcinogenesis because of the availability of tissue samples representing all clinical stages of the disease (i.e., from very small adenomas to large metastatic carcinomas). During the past decade, this model has not only proven its relevance but has stimulated a wide range of important advances in the study of tumor progression.


FIGURE 1 Adaptation of Fearon and Vogelstein's pivotal model of colorectal carcinogenesis.

In order to gain understanding of the different clinical stages of the disease at a molecular level, Fearon and Vogelstein analyzed data from a wide range of molecular pathological studies of colorectal cancers. By correlating the clinical stages with observed genetic derangements, they identified four key sites: ras gene mutations and deletions of chromosomes 5q, 17p, and 18q. Mutations in the ras gene were found in about half of all colorectal carcinomas and adenomas greater than 1 cm in size. Familial adenomatous polyposis, an inherited disease that predisposes patients to colorectal tumor formation, was linked to a site (now known to be the locus of the APC gene) on chromosome 5q. Allelic losses of chromosome 5q were evident in 20-50% of colorectal carcinomas. The functional inactivation of the p53 protein is seen in more than half of all human cancers studied. The p53 tumor suppressor gene maps to the common region of loss on chromosome 17p found in colorectal tumors.      
Consistent with findings in other adult tumors, Fearon and Vogelstein found that more than 75% of colorectal carcinomas exhibit the loss of a large portion of chromosome 17p. Finally, they noted that chromosome 18q was lost in more than 70% of these carcinomas and almost half of late adenomas. The DCC gene maps to the common region of loss, and DCC was recognized as a cell adhesion molecule. The nature of the prevalent genetic defects in colorectal carcinogenesis reiterated the requirement for mutational activation of an oncogene and mutational inactivation of a tumor suppressor in carcinogenesis.      
Indeed, this study supported the concept that genetic losses appear to be more important genetic gains during carcinogenesis. Consistent with the estimated four- to six-step mechanism of human carcinogenesis, a correlation between the number of genetic aberrations and the stage of the tumor was observed. While disruptions of at least four to five genes (all of the aforementioned key sites plus one additional allelic loss) appeared in colorectal carcinomas, most early adenomas contained an average of only two. They also confirmed that different, specific sets of genes were likely to be involved (but not necessarily involved) at different stages of colorectal tumorigenesis; however, they found exceptions from each stage that suggested that the process was more complex than could be explained by a particular set of genetic lesions. This led them to conclude that the total accumulation of changes is more important than their temporal order.      
Fearon and Vogelstein noted that in some cases a mutation in the p53 gene appeared to dominate the wild-type allele through oligomerization of the mutant protein with the wild-type protein, resulting in inactivation of normal p53 function. This demonstrates that a mutation in one allele of a tumor suppressor gene sometimes exerted its effect in a dominant manner. We now know that the protein encoded by the p53 gene is a central component in the biochemical circuits controlling cell proliferation and programmed cell death (apoptosis). Removal of p53 might confer a selective growth advantage via growth deregulation and insensitivity to apoptotic signals, with a concomitant increase in the mutation rate and/or chromosomal instability leading to the eventual loss of the corresponding wild-type allele through localized mutation, mitotic recombination, or chromosomal loss. This would statistically account for the formation of sporadic tumors, which would be difficult to explain using the recessive model for tumor suppressors in which two unrelated mutational events are required, one to inactivate each allele.

D. The Cutaneous Malignant Melanoma Paradigm
The colorectal model is useful for gaining a deeper understanding of neoplastic development in virtually all types of tissues. For example, the same type of analysis can be used to examine the development of cutaneous malignant melanoma in terms of successive genetic changes. Molecular analysis of the clinically evident biological phases in melanoma development shown in Fig. 2, i.e., the development of nevi displaying architectural disorder and cytologic atypia (i.e., atypical or dysplastic nevi), the unregulated proliferation of melanocytes within the epidermis in melanoma in situ, the acquired competence to invade and proliferate within the dermis in primary invasive melanoma, and the development of metastatic capacity, has proven invaluable in our present understanding of this disease.      
It appears that one of the earliest events in the malignant transformation of the melanocyte is the disruption of genetic integrity and the triggering of dynamic genetic instability. During this stage of melanoma development, DNA aneuploidy can be used to distinguish melanomas in situ from nonmalignant atypical nevi (which do not normally exhibit aneuploidy). Early melanoma cell populations with near-diploid chromosome complements are not uncommon; however, upon analysis, they can be shown to bear subtle genetic abnormalities, most probably in genes involved in maintaining genetic stability (e.g., genes critical to DNA repair, replication, cell cycle, chromosome maintenance, and mitosis).


FIGURE 2 Biological stages of development for cutaneous malignant melanoma.

Melanocytes acquire genetic disruptions via two major routes: (i) spontaneous endogenous damage due to deamination of pyrimidines, the generation of oxidative free radicals, infidelity in DNA replication, defects in DNA repair, and metabolism of toxic or mutagenic substances and (ii) exogenous damage by ultraviolet radiation (UVR). A number of efficient repair enzymes continually monitor DNA before, during, and after replication for a range of accumulated defects. Derangement of genes associated with repair of DNA damage is typically found in many types of cancer. DNA repair genes map to chromosomes that are often perturbed in melanomas, e.g., 3p and 7, possibly implicating these genes in the observed genetic instability of these lesions.      
Deregulated cell proliferation is a phase critical to the propagation of genomic disturbances. Because epidermal melanocytes rarely divide in adult skin, damage to DNA probably contributes less to the process of melanoma development than to tumorigenesis in other tissues. However, exposure of melanocytes to UVR results in a transient and limited number of cell divisions that accelerate the development of a protective skin tanning by increasing the mean density of epidermal melanocytes. Concomitantly, UVR inflicts DNA damage by provoking an increase in lipid peroxidation and free radical formation and by inducing single strand breaks and pyrimidine dimers in DNA. Thus, following sun exposure, the melanocyte is faced with two conflicting signals: (i) cease replication of DNA and repair of UVRinduced damage or (ii) proceed with UVR-initiated cell division. Normally, melanocytes will respond by arresting the cell at G1, activating DNA repair systems, and allow the cell to divide only when repairs are completed. If the damage is severe enough, the melanocyte may be driven into apoptosis and eliminated.      
However, in some instances, the intrinsic differences in repair efficiencies for a particular DNA defect or premature resumption of DNA synthesis on a damaged template results in the melanocyte repairing most, but not all, of the UVR-induced damage. One possible outcome of this incomplete repair is the formation of a premalignant melanocyte harboring a critical but biologically inert genetic lesion, which may become the first step in carcinogenesis if it is followed by a complementary lesion produced via a subsequent error in normal cell division.      
The development of deregulated proliferation of melanocytes within the epidermis is a key clinical feature that differentiates melanoma in situ from normal and atypical nevi. As in situ melanomas continue to proliferate, they can accumulate additional genetic defects. The connection between abnormal proliferation and malignant progression is underscored by the observation that any melanocytic lesion (e.g., atypical nevus, primary, or metastatic melanoma) with a disproportionately high number of cells traversing the S phase has a worse prognosis than lesions in which the S-phase fraction is similar to normal tissue. Linking genetic instability with a loss of G1/S transition control is evident from data showing that euploid melanomas have a lower percentage of cycling cells in S phase than aneuploid tumors, and this feature correlates with longer survival. How transition from the normal G0 state of the melanocyte to G1 and S phase is accomplished in a deregulated manner appears to involve the evolution of a subpopulation of cells that have lost control of the G1/S phase transition due in part to gene defects in cell cycle regulatory proteins (e.g., p16INK4A, p15INK4B, and PITSLRE proteins) and to loss of genes regulating cell senescence (several of which have been mapped to chromosomes 1,6,7,9, and 11).      
Sometime during progression, in situ melanoma cells that are restricted to growth in the epidermis spontaneously acquire an invasive phenotype and penetrate the underlying dermal layer. The clinical relevance of this new propensity is that it shows a strong positive correlation with the development of widespread metastases and increasing mortality rates. How a melanoma in situ progresses to an invasive melanoma is unknown, but it is clear from model systems that unrestrained growth of the cells alone is insufficient. Current evidence suggests that the development of melanoma cell invasion is driven by the evolution of specific biological traits, e.g., (i) melanoma-directed dysregulation of the surrounding normal tissue interactions and architecture allowing physical invasion, (ii) the ability of melanoma cells to abrogate or attenuate inhibitory growth and motility signals from the normal tissue promoting invasion, and (iii) the production by melanoma cells of paracrine and autocrine growth factors and cytokines (and their receptors) allowing altered growth and motility.      
In order for physical invasion to occur, the melanoma cell must disrupt the extracellular matrix of the dermis prior to metastatic spread. More than a simple static barrier, the extracellular matrix plays a complex role in maintaining normal homeostasis of the skin by providing structural integrity and by generating biochemical signals that control cell adhesion, growth, differentiation, and migration. Invasive tumor cells must neutralize both of these barriers to affect tissue invasion. Several mechanisms have been identified in melanoma. Derangements in the expression and/or activation of proteolytic enzymes have been found that can disrupt the physical integrity of the extracellular matrix. Melanoma cells promote changes in the expression and assembly of major matrix components. Additionally, the expression and/or function of cell surface integrins and other molecules important to cell-cell communication has been found to be altered or disrupted. Contact with the extracellular matrix during the invasive stage of tumor progression alters the expression of a wide range of genes in many cell types, including melanomas. This interaction is complex, and a fuller understanding of it is necessary for the elucidation of the molecular basis of invasiveness.      
Melanoma cells must also breach the biochemical barriers to invasion. The ability of melanoma cells to acquire invasive potential correlates with the acquisition of resistance to inhibitory factors produced by dermal fibroblasts and possibly infiltrating inflammatory cells. The mechanisms by which invasive melanoma cells become resistant to the inhibitory effect of cytokines and interleukins such as interleukin-6 (IL-6), tumor necrosis factor α (TNF-α), and TGF-β are unknown. It may involve alterations in the receptors for these molecules and/or in other genes that comprise a signal-transducing pathway that engages these molecules. Many of the growth factors, integrins, and inhibitory proteins that are perturbed in melanoma transduce signals to various cellular compartments via protein phosphorylation. Emerging data suggest a connection between defects in members of the protein tyrosine kinase family of genes and altered signal transduction in the pathogenesis of melanoma. Melanoma cells have been shown to express and secrete growth factors, providing autocrine stimulation of melanoma proliferation. In addition, the expression of a number of such factors is inducible in epidermal cells by UVR, including IL-1, IL-6, IL-8, and TNF-α. In this manner, "normal" cells within the immediate tissue environment can play a major role in the growth deregulation of transformed cells. We are beginning to appreciate that tumors are not merely collections of transformed cells. They are complex tissues in which heterogeneous mixtures of normal cells such as fibroblasts, endothelial cells, and immune cells interact continuously with cancer cells.      
The use of heterotypic organ culture systems may ultimately prove more fruitful than a traditional cell culture for elucidating the mechanisms of neoplasia and for screening potential therapeutic agents. The final stage of neoplasia, i.e., the development of metastatic competence and migration of tumor cells from the primary site to specific distant organs, is the most poorly understood in terms of relevant molecular defects. Many critical aspects are required for the cell to acquire metastatic competence, including the expression and/or dysfunction of numerous biologic pathways involved in the ability of the cell to invade tissue, survive, and proliferate within the invaded tissue, migrate through the tissue, inhibit immune defenses, orchestrate the development of new blood vessels, escape into the bloodstream, and reemerge in a distant organ. Studies are beginning to establish dominant roles for specific families of genes such as the cadherins in the metastatic process. In melanoma cells, switching cadherin expression from E-cadherin to N-cadherin results in the loss of keratinocyte control over these cells and establishment of communication with fibroblasts and endothelial cells. Disruption of growth factor signaling pathways, notably endothelin-3 and stem cell factor and their receptors, is also associated with the development of metastatic potential in melanomas.      
In addition to a series of intrinsic molecular defects within the tumor cell, the study of melanoma has also provided evidence for another critical factor that may either suppress or aid tumor development in an as yet unpredictable manner: i.e., host immune competence and responsiveness. The potentially important role of this phenomenon in tumor growth is evidenced by spontaneous tumor regression in melanoma patients, and the observation that UVR in animal model systems can have a profound stimulatory effect on the outgrowth of tumor cells by transiently suppressing host immunological defense mechanisms. At present, relatively little is known about the interplay between immune suppression and specific genetic defects in tumor cells and how this interchange impacts upon the growth and spread of tumor cells. Presumably, through a separate series of genetic defects, the tumor cell may also develop the ability to resist or abrogate host immunity.

III. REFINING THE MODEL

A. Simplifying Complexity
The two tumor models discussed earlier clearly show that the evolution of genetic instability, deregulated proliferation, and invasive and metastatic competence are complicated events developing, not as the result of a single genetic defect, but driven by an accumulation of molecular alterations in an overlapping succession of cell subpopulations. This disruption of the cellular genome results in defects in numerous regulatory mechanisms that control normal cell homeostasis, and because these mechanisms are complex, it is not surprising that an impressive diversity of cancer genotypes and phenotypes exists. Since the 1980s, more than 100 oncogenes and tumor suppressor genes have been identified, including representatives from virtually all of the major regulatory circuits of the cell. Components of the pathways regulating the cell cycle (e.g., Rb, E2Fs, p21INK4A, p15INK4B), apoptosis (e.g., p53, Bax, Bcl-2), inflammatory response (cyclooxygenase-2), cell communication (e.g., integrins, cadherins), and genetic stability (e.g., mismatch repair-associated proteins, telomerase) are disrupted, damaged, or abnormally regulated in neoplasia.      
Nonnuclear genetic defects may also play a role in carcinogenesis. The occurrence of somatic mutations of the mitochondrial genome in human colorectal cancer has been examined. Mutations of this type have the potential to interfere with normal oxidative phosphorylation, a disruption possibly accompanied by an increase in the level of cellular reactive oxygen species (which are known to affect DNA damage leading to mutations). Due to the nature of mitochondrial replication, it is conceivable that the entire mitochondrial population within a clonal cell population may become homogeneous if the end result of the mutated mitochondrial genome is to confer a selective growth advantage.      
The carcinogenic effects of epigenetic events further complicate the picture, such as the binding of chemical promoters of carcinogenesis to cellular receptors to regulate certain gene products. Cancer cells display a variety of epigenetic mechanisms by which they circumvent normal barriers to neoplasia, e.g., the disruption of the FAS death signal pathway by upregulation of a nonsignaling decoy receptor that titrates signals away from the apoptosis pathway. Hypermethylation of the promoter regions of cancerassociated genes is another important epigenetic mechanism of carcinogenesis. Methylation of CpG islands in the promoter region of a gene can block its expression; aberrant hypermethylation of cancerrelated gene promoters has been observed in several types of cancers.     
There have been frequent attempts to simplify and codify this apparent diversity of mechanisms through categorization. The most recent and intellectually satisfying systems group cancer-associated genes into categories representing their functional activities. Classical tumor suppressor genes such as p53, Rb, and APC, which prevent cancer through direct control of growth have been termed "gatekeepers." Genes that suppress neoplasia in an indirect manner by maintaining the fidelity of the genome, including DNA mismatch repair genes, spindle checkpoint genes such as BUB1 and MAD1, or DNA damage checkpoint genes that include ATM, Brca1, and Brca2, have been called "caretakers."      
A third pathway to cancer (carcinogenesis via "landscaper" defects) has been proposed based on the observation that, under certain conditions, normal cells proximal to a rapidly proliferating defective cell population are at an increased risk of neoplastic transformation due to the biochemical factors present in their abnormal microenvironment.      
Although these classification systems have proven useful, they fail to address the cancer problem in a manner that connects them in a meaningful way to cell phenotype and biological behavior and that would lead to a unified understanding of the malignant process. Such an understanding is crucial to the eventual development of comprehensive therapeutic or preventive strategies. We have been fortunate in having available for study the colorectal cancer and malignant melanoma paradigms, which are amenable to analysis at multiple clinical stages, in the pursuit of a molecular mechanism of multistage carcinogenesis.      
Both these models focus on the genetic and biological characteristics a cell must acquire to overcome cellular and tissue barriers to oncogenesis. This theme is reflected in current models of carcinogenesis. During the next few years, aided in part by the sequencing of the human genome and recent technical advances in sophisticated and sensitive analysis of global expression of genes and proteins, such models will be greatly refined as we continue to link the phenotype and biological behavior of a tumor cell with specific genes and signaling circuits.

B. Current Models
Recent models of tumorigenesis have attempted to synthesize our understanding of the molecular events underlying the stages of neoplasia with our increasing knowledge of the central molecular circuitry of the cell. Examination of tissues from various types of cancers has shown that when cells progress from a preneoplastic state through advanced malignancy, they acquire a set of characteristics that are the hallmarks of cancer. In a current model proposed by Hanahan and Weinberg, carcinogenesis can be viewed as a process in which disruption of each key cellular circuit results in the acquisition by the cell of a new capability, enabling the cell to successfully breach one of the anticancer defense mechanisms of the organism. Malignancy is thus achieved through genetic and epigenetic disruptions, resulting in the acquisition of a set of six acquired capabilities essential for fully malignant neoplastic transformation (Table 1). They argue that virtually all genetic and epigenetic disruptions that contribute to carcinogenesis result in the acquisition of one or more of these capabilities and they reiterate Fearon and Vogelstein's observation that the precise order and nature of genetic and epigenetic derangements may not always be important for carcinogenesis.

TABLE I Hanahan and Weinberg's Model of Multistage Carcinogenesis

Acquired capabilitya Common mechanisms Specific examples
Self-sufficiency in growth signals Deregulation of growth signal pathways via disruption of extracellular signals, signal transducers, and corresponding intracellular circuits Production of PDGF, bFGF, or TGFα (autocrine stimulation), upregulation or truncation of EGF-R/erbB, upregulation of HER2/neu, SOSRas-Raf-MAP kinase mitogenic cascade mutants, defects in herterotypic signaling
Insensitivity to antigrowth Derangement of the pathways that recognize and signals respond to the antigrowth signals that normally force proliferating cells into the G0 stage or direct them to terminally differentiate and lose their proliferative, capacity Rb gene mutation or pRb sequestration. Other disruptions of the pRb signaling circuit, including downregulation and/or mutation of TGFβ receptors, p16INK4A, p15INK4B and CDK4. Overexpression of c-myc, erbA
Evading apoptosis Defects in sensors directing apoptosis or effectors carrying it out p53 tumor suppressor gene abnormalities, upregulation of bcl-2, disruption of FAS deathsignaling, IGF-2 gene expression
Limitless replicative potential Circumvention of telomere shortening Upregulation of telomerase expression, induction of alternate mechanisms of telomere maintenance
Sustained angiogenesis Disregulation of angiogenic factors, altered expression, or proteolytic modification of pro- and anti-angiogenic signaling factors Activation of ras, upregulation of VEGF, FGF1/2, downregulation of thrombospondin-1, β-interferon
Tissue invasion and metastasis Changes in expression of cell-cell adhesion molecules, integrins, growth factors, and extracellular proteases Switch in expression from E-cadherin to Nmetastasis cadherin, preferential expression of α3β1 and αVβ3 integrins, downregulation of endothelin-3

aThe process is viewed as the gradual acquisition of six key capabilities.

The expansion of our understanding of the molecular circuitry of the cell has facilitated an in-depth exploration of how different mutational and epigenetic routes can be followed that lead to the same end and how phenotypically identical cancers can have very different genotypes. We now recognize that knowledge of the unique molecular signature of a given tumor, rather than its histological identity, may prove to be the pivotal factor in designing the most effective therapeutic regimen. It has also become clear why some mutations are more oncogenic than others. The study of multifunctional proteins such as p53 and pRb indicates that it is because the gene product involved is central to more than one circuit, hence deregulation at one key point allows the cell to breach multiple barriers to neoplasia simultaneously.

IV. IMPLICATIONS FOR CANCER PREVENTION AND THERAPY

The recent refinements of the carcinogenesis model should result in superior prognostic and diagnostic modeling as well as improved clinical management of the patient with cancer. By focusing on biological characteristics common to most forms of neoplasms, new therapies can be developed that target the acquisition of these cancer-associated biological phenomena and the underlying genetic abnormalities.      
Antiangiogenic therapies, for example, hold the promise of being applicable to a wide variety of cancers. Through definition and elucidation of the specific molecular pathways that can lead to cancer and understanding how a cell interconnects and organizes signaling pathways in a hierarchical system to control the behavior of a cell, we may be able to identify pharmacological targets that mitigate or attenuate those circuits that often become permuted and/or rerouted during cancer development. The great strides made in large-scale gene expression analyses of cells and tissues should materially and quickly provide targets for novel therapies. The ongoing genetic mapping studies of polymorphisms in putative cancer susceptibility genes will further allow exciting opportunities to not only tailor cancer treatment, but also to identify those who may be at increased risk for a primary or secondary cancer. These individuals are potential candidates for cancer chemoprevention clinical trials.

Anthony P. Albino
Ellen D. Jorgensen
The American Health Foundation, Valhalla, New York

See Also
ANIMAL MODELS FOR COLON CANCER CHEMOPREVENTION ; CHEMOPREVENTION TRIALS ; COLORECTAL CANCER: MOLECULAR AND CELLULAR ABNORMALITIES ; MELANOMA: BIOLOGY ; MOLECULAR EPIDEMIOLOGY AND CANCER RISK ; RETINOBLASTOMA TUMOR SUPPRESSOR GENE

GLOSSARY
apoptosis Programmed cell death via biochemical circuits responding to aberrations or defects in the cell. A normal defense against the propagation of mutant or damaged cells.

carcinogenesis The process of tumor development in an organism.

clonal expansion The selective replication of a mutated cell within a population resulting in the eventual genetic homogeneity of the cell population.

initiation The first step in the three-step model of multistage carcinogenesis in which an irreversible genetic alteration occurs, sensitizing the cell to promoting agents.

kryotypic instability The inability to maintain genomic integrity, as evidenced by chromosome rearrangement, truncation, and loss.

oncogene Implicated in cancer development when its expression is upregulated, activated, or deranged.

progression Third step in the three-step model of multistage carcinogenesis in which accumulated genetic alterations result in the ability of affected cells to invade local tissues or to metastasize to distant sites.

promotion Second step in the three-step model of multistage carcinogenesis in which initiated cells are stimulated to proliferate by promoting agents.

protooncogene Normal cellular homologue of an oncogene.

transformation Process by which normal cells become neoplastic.

tumor suppressor Gene implicated in cancer development when its expression is turned off, downregulated, or disrupted.

Bibliography
Albino, A. P., Reed, J. A., and Fountain, J. W. (1998). Melanoma: Molecular biology. In "Cutaneous Oncology" (S. J. Miller and M. E. Mahoney, eds.). Blackwell Science, Oxford.
Cooper, G. M. (1982). Cellular transforming genes. Science 218, 801-806.
Fearon, E. R., and Vogelstein, B. (1990). A genetic model for colorectal carcinogenesis. Cell 61, 759-767.
Foulds, L. (1964). Tumour progression and neoplastic development. In "Cellular Control Mechanisms and Cancer" (P. Emmelot and O. Mühlbock, eds.), pp. 242-258. Elsevier, Amsterdam.
Hanahan, D., and Weinberg, R. (2000). The hallmarks of cancer. Cell 100, 57-70.
Herlyn, M., Berking, C., Li, G., and Satyamoorthy, K. (2000). Lessons from melanocyte development for understanding the biological events in naevus and melanoma formation. Melanoma Res. 10, 303-312.
Kinsler, K. W., and Vogelstein, B. (1996). Lessons from hereditary colorectal cancer. Cell 87, 159-170.
Kinsler, K. W., and Vogelstein, B. (1997). Gatekeepers and caretakers. Nature 386, 761-763.
Kinsler, K. W., and Vogelstein, B. (1998). Landscaping the cancer terrain. Science 280, 1036-1037.
Knudson, A. G., Jr. (1985). Hereditary cancer, oncogenes, and antioncogenes. Cancer Res. 45, 1437-1443.
Loeb, L. (1991). Mutator phenotype may be required for multistage carcinogenesis. Cancer Res. 51, 3075-3079.
Nowell, P. (1976). The clonal evolution of tumor cell populations. Science 194, 23-28.
Renan, M. J. (1993). How many mutations are required for tumorigenesis? Implications from human data. Mol. Carcinogen. 7, 139-146.
Rous, P., and Kidd, J.G. (1941). Conditional neoplasms and subthreshold neoplastic states. J. Exp. Med. 73, 365-390.
Weinberg, R. (1989). Oncogenes, antioncogenes, and the molecular bases of multistep carcinogenesis. Cancer Res. 49, 3713-3721.

Hormonal Carcinogenesis

The endocrine system and its numerous hormones function to integrate genetically encoded developmental programs and environmental signals and regulate the myriad biochemical processes, such as cell proliferation, differentiation, and death, required to maintain homeostasis within boundaries compatible with life. It is becoming increasingly clear that aberrations in the control of these hormone-regulated processes contribute to the genesis of many forms of cancer. In specific instances, hormones also appear to act through nonreceptor-mediated mechanisms to contribute to carcinogenesis. Steroid hormones, in particular, are implicated in the etiology of cancers of the breast, female reproductive tract, and male reproductive tract, as well as benign tumors of the anterior pituitary gland. Both endogenous and exogenous steroid hormones appear to contribute to carcinogenesis. In addition, much effort is being focused on defining the potential of environmental chemicals, both naturally occurring and man made, to enhance or inhibit carcinogenesis via hormonal mimicry. This article summarizes our current knowledge of hormonal carcinogenesis, focusing on cancers of the breast and other tissues where hormones play primary, if not causative, roles in cancer etiology.

I. BREAST CANCER

A. Endogenous Estrogens, Progestins, and Breast Cancer Etiology
It is well documented that estrogens and progestins act through direct and indirect pathways to stimulate development of the breast at puberty and during pregnancy. In addition, the breast epithelium exhibits a cyclical pattern of cell proliferation and death throughout each menstrual cycle; cell proliferation is highest during the luteal phase of the menstrual cycle, when circulating estradiol is submaximally elevated and circulating progesterone is at its peak. Data from the study of mouse models in which the genes encoding either estrogen receptor (ER) α or progesterone receptor (PR) are disrupted through insertional mutagenesis indicate that ERα and PR are required for the growth-promoting actions of estrogens and progestins within the mammary epithelium.      
The report of Cooper in 1836 that breast cancers regress in size at the beginning of each menstrual cycle and at menopause and the report of Beatson in 1896 that oophorectomy leads to regression of breast cancers are now recognized as providing the first demonstrated links between ovarian hormones and breast cancer. Subsequent epidemiologic and laboratory studies provide overwhelming evidence inextricably linking ovarian steroids to the genesis of breast cancer. Both early menarche and late onset of menopause are associated with an increased risk of breast cancer. Moreover, oophorectomy prior to menopause significantly reduces breast cancer risk, with the greatest reduction in risk being observed in women 35 years of age or younger. Several, but not all, studies have demonstrated a positive correlation between the level of circulating estrogens in postmenopausal women and breast cancer risk. Clinical trials have demonstrated that tamoxifen, a selective estrogen receptor modulator (SERM), reduces the risk of breast cancer in asymptomatic women at high risk of developing the disease. In addition, inhibitors of aromatase, the enzyme that catalyzes the production of estrogens from androgen precursors, are gaining use in the treatment of breast cancer and may provide additional effective agents for the prevention of these cancers. Together, these data indicate that lifetime exposure to endogenous ovarian hormones is directly correlated with breast cancer risk and demonstrate that these hormones, particularly the estrogens, play an important, if not causative, role in breast cancer development.      
Several epidemiologic studies have attempted to identify associations between specific allelic variants of genes that encode enzymes involved in estrogen metabolism and breast cancer risk. The hypothesis being tested in these studies is that genes with variant alleles that impact expression and/or function of the encoded protein may be low penetrance, but high frequency, modifiers of breast cancer risk. Cytochrome P450c17 (CYP17) catalyzes two steps in the pathway leading to sex steroid biosynthesis. A variant allele encoding CYP17 has been identified that is associated with increased levels of circulating estradiol and progesterone. This variant CYP17 allele has been associated with increased breast cancer risk in some, but not other, studies.       
Cytochrome P4501A1 (CYP1A1) catalyzes the hydroxylation of a variety of substrates, including polycyclic aromatic hydrocarbons and the estrogens, 17β- estradiol (E2) and estrone (E1). Variant alleles of CYP1A1 have been identified that may differ in activity and/or inducibility. The m1 variant, which exhibits a single nucleotide polymorphism in the 3'-untranslated region, has been associated with an increased breast cancer risk in African-American and Chinese women. However, no risk modification was associated with the m1 allele in other studies. The m2 variant, a single nucleotide polymorphism that results in a single amino acid change, Ile462Val, in the heme-binding domain, did not significantly modify breast cancer risk in the general population, but was associated with an increased risk in specific subsets of smokers. Estrogens are also substrates for cytochrome P4501B1, which catalyzes the hydroxylation of E2 and E1 preferentially on C-4 of the aromatic A ring.      
Two single nucleotide polymorphisms have been identified that impact amino acid sequence in the heme-binding domain of CYP1B1: m1, Val432Leu; and m2, Asn453Ser. Neither the m1 nor the m2 variant was observed to impact significantly breast cancer risk in Caucasian or African-American populations. However, homozygosity for the m1 allelic variant of CYP1B1 was associated with an increased breast cancer risk in a Chinese population. Catechol-Omethyl-transferase (COMT) catalyzes the methylation of various catechols, including the 2-hydroxylated and 4-hydroxylated metabolites of E2 and E1. A single nucleotide polymorphism in exon 4 of COMT results in an amino acid change, Val158Met, that reduces COMT activity. Two studies have associated homozygosity for the low-activity COMT allele with an increased breast cancer risk in postmenopausal, but not premenopausal, women; another study demonstrated an association between the low-activity COMT allele and increased breast cancer risk in premenopausal, but not postmenopausal, women; and yet another study did not associate the COMT genotype to breast cancer risk. The failure of these genetic epidemiology studies to produce consensus regarding the associations of variant alleles for genes encoding these different enzymes in the etiology of breast cancer most probably results from several factors, including (1) het-erogeneity within or between the populations studied; (2) gene-gene and gene-environment interactions that occlude modifier effects; and (3) modest population sizes in several of the studies.      
Additional studies will be required to establish firmly whether these genes act as low penetrance modifiers of breast cancer risk. In addition, it must ultimately be demonstrated that any associations between genotype at a particular genetic locus and cancer risk result from loss or gain of function of the gene residing at that locus or the protein encoded by that gene.

B. Exogenous Estrogens, Progestins, and Breast Cancer Etiology
Exogenous hormones also modify breast cancer risk. Several studies indicate that current or recent use of estrogen replacement therapies (estrogen alone) or hormone replacement therapies (estrogen and progestin) by menopausal women is associated with an increased breast cancer risk. The increase in breast cancer is greatest in women using regimens containing both estrogen and progestin. Breast cancer risk also appears to be slightly increased by the current use of oral contraceptives. Finally, use of the synthetic estrogen, diethylstilbestrol (DES), during pregnancy is associated with an increased breast cancer risk.

C. Estrogen Receptor and Breast Cancer Etiology
The ER, in particular ERα, is central to estrogen action in the breast. Much effort has been invested toward the identification of mutations and polymorphisms in the gene encoding ERα and attempts to correlate these genetic differences with breast cancer risk, pathology, ER status, or responsiveness to hormonal therapy. Although mutant forms of the ER have been identified that differ from the wild-type receptor in ligand-binding and/or transcriptional activation activity, little evidence exists to suggest that somatic mutations in the ER play a major role in breast cancer etiology. However, loss of ER expression does often occur during progression from a hormonedependent to a hormone-independent stage. Specific ERα polymorphisms have been associated with an increased breast cancer risk in two of four published studies, suggesting that variant forms of ERα may act as low penetrance modifiers of breast cancer risk.

D. Potential Mechanisms of Hormonal Carcinogenesis in the Breast
It is well established that estrogens and progestins interact in the regulation of cell proliferation and cell survival in the breast. By stimulating cell proliferation and/or enhancing cell survival within the breast epithelium, ovarian steroids may increase the probability of somatic mutations, as well as enhance the growth and/or survival of transformed cells. These mutations could arise through the actions of environmental or endogenous mutagens, exposure to ionizing radiation, or replication errors during DNA synthesis. Mammographic density correlates with the proportion of the breast volume composed of the epithelial and stromal parenchyma as well as with breast cancer risk. Use of estrogens with progestins in menopausal hormone replacement regimens is associated with increased mammographic density, an increased rate of cell proliferation within the breast epithelium, and an increased risk of breast cancer. Moreover, the observation that the SERM tamoxifen significantly reduces breast cancer incidence in women at increased risk indicates that ER-mediated pathways play important roles in breast cancer etiology. Together, these studies strongly link the actions of ovarian steroids in the regulation of cell proliferation and/or survival in the breast epithelium to the genesis of breast cancer. At this time, the mechanisms through which ovarian hormones regulate proliferation and survival within the breast epithelium are not completely understood.      
Both estrogens and progestins induce expression of cyclin D1 in breast epithelial cells and may thereby directly stimulate progression of those cells through the G1 phase of the cell cycle. In addition, estrogens regulate the production of several growth factors, which in turn act through autocrine and/or paracrine mechanisms to stimulate cell cycle progression and/or enhance cell survival within the breast epithelium. It is becoming increasing clear that interactions between stromal and epithelial cells are required for normal development and function of the breast and that the estrogens and progestins act through the stroma to exert some of their stimulatory effects on the breast epithelium.      
It is hypothesized that specific estrogen metabolites may be genotoxic carcinogens that contribute to the development of breast cancers and other cancers by inducing mutations in oncogenes and/or tumor suppressor genes. An often proposed mechanism of estrogen carcinogenesis is that E2 and/or E1 is hydroxylated at either the C-2 or the C-4 position to the corresponding catechol estrogen, which is subsequently oxidized through a semiquinone intermediate to the estrogen quinone, which could bind covalently to DNA. Alternatively, the catechol estrogen and/or estrogen semiquinone may undergo redox cycling and generate superoxide radicals, which could damage DNA. An enzyme(s) capable of generating the catechol estrogens 4-hydroxyestradiol and 4-hydroxyestrone is expressed in human breast cancers, and CYP1B1 has been identified as an estrogen-4-hydroxylase expressed in the MCF-7 breast cancer cell line.      
However, the carcinogenicity of catechol estrogens in the rodent mammary gland has yet to be established, and estrone-3,4-quinone did not induce mammary cancers when injected directly into the mammary gland of female rats. In addition, several studies have demonstrated that estrogens and catechol estrogens lack mutagenic activity in standardized bacterial and mammalian cell assays. Therefore, the contribution of potentially genotoxic estrogen metabolites to the etiology of breast cancer remains to be defined.      
Treatment of a mouse mammary epithelial cell line with 16α-hydroxyestrone induces anchorage independent growth, suggesting that this estrogen metabolite has the potential to induce neoplastic transformation in this mammary cell model. The ability of estrogens and estrogen metabolites to induce neoplastic transformation in vitro has been extensively studied by T. Tsutsui and J.C. Barrett using Syrian hamster embryo cells. Transformation of Syrian hamster embryo cells is associated with the induction of aneuploidy, which may result from binding of estrogens or specific estrogen metabolites to β-tubulin and disruption of microtubule assembly during mitosis. More recently, the transforming potential of different metabolites of E2, including the 2-hydroxy and 4- hydroxy catechols of E2 and E1, as well as 16α- hydroxyestrone, in Syrian hamster embryo cells has been associated with the induction of mutations, chromosomal aberrations, and/or aneuploidy.      
In summary, estrogens and progestins most probably act through multiple mechanisms to induce and/or promote the development of breast cancers (Fig. 1). Pathways dependent on the ER are clearly implicated by the ability of tamoxifen to reduce breast cancer incidence in women at increased risk. It is also clear that the actions of these hormones in regulating cell proliferation and survival within the breast epithelium contribute to the genesis of breast cancers. Data are emerging to suggest that genes encoding enzymes involved in steroid hormone metabolism may be low penetrance modifiers of breast cancer risk. In this regard, breast cancer risk could be impacted by genetic variation in (1) the amounts of active hormones produced; (2) the specific types of hormones or metabolites produced; (3) the timing of hormone production; and/or (4) the location of hormone production. Estrogens may be metabolized to genotoxic forms that can bind to DNA and induce mutations. Alternatively, estrogens may induce aneuploidy or other forms of genomic instability. Additional studies in validated animal models are required to define precisely the mechanisms through which hormones contribute to breast cancer etiology.


FIGURE 1 Possible pathways of hormonal carcinogenesis in the breast. Genetic variation in genes encoding enzymes involved in estrogen biosynthesis (1) or estrogen excretion (2) may impact the type or amount of steroidal estrogens acting on target cells. Endogenous and exogenous estrogens may be metabolized to reactive forms that can bind macromolecules, including DNA, and proteins such as tubulin (3). Genetic variation in genes encoding specific detoxifying enzymes may impact the clearance of potentially genotoxic estrogen metabolites (4). Genetic variation in the genes encoding ERα or ERβ may impact the binding of estrogens to its receptor, receptor levels, or receptor signaling. (5) Although not as intensely studied, genetic variability in progesterone metabolism or signaling may also contribute to breast cancer etiology.

E. Animal Models of Estrogen-Induced Mammary Cancer
The first demonstration that estrogens induce mammary cancers in rodents was by Lacassagne in the 1930s. However, rodent models of estrogen-induced mammary carcinogenesis did not achieve the same degree of prominence in the research community as animal models in which mammary cancers are induced by carcinogens, such as dimethylbenz[a]anthracene (DMBA) or N-methyl-N-nitrosourea (MNU). Studies suggest that rat models of estrogen-induced mammary carcinogenesis can provide novel and physiologically relevant insights into the role of estrogens and other hormones in the etiology of breast cancer.

1. ACI Rat as a Model of Estrogen-Induced Mammary Cancer
The ACI rat was developed by Dunning and colleagues in 1926 from a cross between the August and the Copenhagen (COP) rat strains. Although these investigators first demonstrated the ability of DES to induce mammary cancers in the ACI rat in 1947, this model was used only sporadically over the next 50 years.      
Data from our laboratory indicate that the female ACI rat exhibits a unique propensity to develop mammary cancers when treated continuously with physiologic levels of E2, released from subcutaneous Silastic tubing implants. Palpable mammary cancers begin to appear as early as 70 days following the initiation of E2 treatment, median latency is approximately 140 days, and virtually 100% of the treated animals exhibit palpable cancers within 210 days of treatment. Ovariectomy significantly inhibits the development of E2-induced mammary cancers, suggesting a requirement for another ovarian factor(s), possibly progesterone, in E2-induced mammary carcinogenesis. Focal regions of atypical epithelial hyperplasia are common in the mammary glands of E2- treated ACI rats, and these are probable precursors to carcinoma (Fig. 2). Relative to the surrounding mammary epithelium, cells within the focal regions of atypical hyperplasia and the mammary cancers exhibit an increased expression of progesterone receptor (PR). Carcinomas induced in ACI rats by E2 are estrogen dependent and commonly exhibit aneuploidy, two features that are common to breast cancers in humans.

Continuous E2 Treatment from Subcutaneous Implant
-------------------------------------------------------------------→


FIGURE 2 Histologic progression of estrogen-induced mammary carcinogenesis in the ACI rat. (A) The mammary gland of the ovary-intact, virgin female ACI rat is composed of ductal structures extending throughout the mammary fat pad. (B) Continuous treatment with E2 rapidly induces marked lobular development and hyperplasia. (C) Focal regions of atypical hyperplasia begin to appear within the hyperplastic epithelium following approximately 8 weeks of E2 treatment. (D) Palpable mammary carcinomas appear as early as 10 weeks following the initiation of E2 treatment, but the median latency to the appearance of the first palpable mammary carcinoma is 20 weeks. (E) Invasive features are observed in a fraction of the E2-induced mammary carcinomas.

Relative to the ACI rat strain, females of the COP strain, which is closely related genetically to ACI, and the Brown Norway (BN) strain, which is unrelated, are much less susceptible to E2-induced mammary cancers (Table I). These strain differences provide an avenue toward elucidation of the mechanisms through which estrogens induce mammary cancer development. For example, we have demonstrated that the proliferative response of the ACI mammary epithelium to E2 significantly exceeds that of the COP mammary epithelium, suggesting one possible mechanism for the differing susceptibilities of these rat strains to E2-induced mammary cancers. Genetic studies from our laboratory indicate that the highly susceptible ACI phenotype behaves as an incompletely dominant trait in crosses to either of the resistant COP or BN strains. Although most of the E2-treated ACI/COP F1 and ACI/BN F1 progeny develop mammary cancers when treated with E2, latency is significantly delayed relative to that observed in the parental ACI strain. We have mapped genetic modifiers of susceptibility to E2-induced mammary cancers to rat chromosome 5 in reciprocal crosses between ACI and COP rat strains and to chromosomes 5, 18, and 2 in a cross between ACI and BN strains. The gene encoding Cdkn2a is a candidate for the modifier residing on chromosome 5. Cdkn2a is an inhibitor of the cyclin D-dependent kinase Cdk4 and Cdk6 and would therefore be expected to inhibit the mitogenic actions of estrogens exerted through the cyclin D/Cdk/retinoblastoma pathway. We have demonstrated that the expression of Cdkn2a is markedly downregulated in both the focal regions of atypical hyperplasia and the mammary cancers induced in ACI rats by E2, relative to the surrounding epithelium. We anticipate that the ultimate identification of the genetic modifiers of susceptibility to E2-induced mammary cancers in this rat model will reveal significant information regarding the mechanisms through which estrogens contribute to breast cancer in humans. These data strongly suggest that the ACI rat provides a unique and physiologically relevant animal model for investigating the mechanisms of estrogen-induced mammary carcinogenesis, as well as for studying interactions among hormonal, genetic, and environmental determinants of breast cancer susceptibility.

TABLE I Unique Susceptibility of the Female ACI Rat to Estrogen-Induced Mammary Carcinogenesisa

Rat strain Mean latency (days) Incidence (%) Palpable tumors/rat
ACI 141 100 3.9
COP 208 44 0.4
BN 369 20 0.2

aSilastic tubing implants containing 17β-estradiol were inserted subcutaneously when ovary intact rats of the indicated inbred rat strains were 9 weeks of age.

2. Noble Rat as Model of Estrogen/Androgen-Induced Mammary Cancer
The Noble (NBL) rat is another unique model for hormone-induced mammary carcinogenesis. Noble first demonstrated the ability of E1 to induce mammary cancers in males and females of this rat strain in 1940. In a subsequent study, Noble demonstrated that the susceptibility of the NBL rat to E1-induced mammary cancers declined markedly as a function of the age at which treatment was initiated. Whereas 100% of the E1- treated NBL rats developed mammary carcinoma when treatment was initiated at 2 weeks of age, less than 10% of animals treated at 9 weeks of age were tumor positive following 13 months of E1 treatment. More recently, the combination of E2 and testosterone propionate has been demonstrated by S.A. Li and colleagues to induce a high incidence of mammary cancers in male NBL rats, whereas mammary cancers did not develop in the same time period in rats treated with E2 or testosterone propionate alone. Similarly, studies by Y. C. Wong and colleagues indicate that E2 benzoate plus testosterone propionate induces a high incidence of mammary cancers in female NBL rats. These studies, in conjunction with those on the ACI rat summarized earlier, suggest that the NBL is less sensitive to mammary cancer induction by E2 or E2 benzoate relative to ACI rats treated with E2. Moreover, the requisite for coadministration of androgen and estrogen for rapid development of mammary cancers in the NBL rat may illustrate a fundamental mechanistic difference between the NBL and ACI models of E2-induced mammary carcinogenesis. Therefore, each of these inbred rat strains may ultimately reveal different insights into the mechanisms of hormonal carcinogenesis in the mammary gland.

II. CANCERS OF THE FEMALE REPRODUCTIVE TRACT

A. Endometrial Cancer
Estrogens stimulate cell proliferation within the uterine endometrium, and most cell proliferation in this tissue occurs during the follicular phase of the menstrual cycle, when circulating estrogens are increasing. An increased risk of endometrial cancer is associated with late menopause and obesity, both of which are generally believed to result in prolonged exposure of the uterus to estrogens. Use of combination oral contraceptives decreases the risk of endometrial cancer, whereas use of sequential oral contraceptive formulations is associated with an increased risk. Estrogen replacement therapy to alleviate symptoms of menopause significantly increases risk. However, hormone replacement regimens containing both estrogens and a progestin do not increase the risk of endometrial cancer. Moreover, tamoxifen exhibits agonistic properties in the uterus, and use of tamoxifen has been strongly associated with an increased risk of endometrial cancers. A recent genetic epidemiology study provides suggestive evidence that allelic variants in the gene encoding ERα may modify the risk of endometrial cancer. Together, these data strongly implicate both endogenous and exogenous estrogenic hormones in the etiology of cancers of the uterine endometrium.      
In the mouse, disruption of the gene encoding ERα, but not ERβ, drastically alters normal development of the uterus and eliminates responsiveness of the uterine tissues to estrogens. Treatment of newborn CD-1 mice with DES has been shown to induce a 90% incidence of uterine adenocarcinoma by 18 months of age. Development of uterine cancers in this model requires a functional ovary. More recently, the estrogens E2 and 17α-ethinylestradiol, as well as the catechol estrogens 2-hydroxyestradiol and 4-hydroxyestradiol, have been demonstrated to induce uterine adenocarcinoma in this CD-1 mouse model. Similarly, tamoxifen has been demonstrated to induce uterine adenocarcinoma when administered to neonatal Wistar rats.      
Data from these rodent models strongly suggest that estrogens play a direct role in inducing endometrial cancers and are consistent with the hypothesis that estrogens act through a genotoxic mechanism in the neonatal uterus to initiate carcinogenesis. Alternatively, the administered estrogens may act through an ER-mediated mechanism to trigger developmental changes in the neonatal uterine epithelium and/or stroma that predispose to adenocarcinoma in the adult animals.

B. Cervical and Vaginal Cancers
The synthetic estrogen DES has been inextricably linked to clear cell carcinoma of the cervix/vagina of young women who were exposed in utero when their mothers were treated with DES during the first trimester of pregnancy. This carcinogenic action appears to be quite specific because cancers at other sites are not significantly increased in women exposed to DES in utero. An association between exogenous estrogens and cervical cancers in mice has been demonstrated. For example, DES, administered prenatally or neonatally, induces vaginal adenocarcinoma in CD-1 mice. Endogenous ovarian steroid hormones, oral contraceptives, and hormone replacement regimens are not generally believed to be strong determinants of cervical cancer risk.

C. Myometrial Tumors and Cancers
Uterine leiomyoma is the most commonly observed uterine tumor in premenopausal women. Ovarian steroids have been linked to the genesis of these benign uterine tumors as well as to the development of malignant uterine leiomyosarcomas. Leiomyoma and, less frequently, leiomyosarcoma develop spontaneously in the Eker rat. Development of these tumors is markedly inhibited by ovariectomy or treatment with SERMs. Continuous treatment with E2 induces leiomyoma in guinea pigs. Ovariectomy increases the incidence of leiomyoma in this animal model and decreases the latency to tumor appearance. Tumor regression occurs in response to cessation of E2 treatment or treatment with the SERM raloxifene. Finally, treatment of Syrian hamsters with the combination of DES and testosterone propionate has been reported to induce uterine leiomyosarcoma. Together, these epidemiologic and laboratory studies indicate a role of estrogens in the development of tumors of the uterine myometrium.

III. PROSTATE CANCER

Abundant data link androgenic hormones to the etiology of prostate cancer. Orchiectomy leads to involution of the prostate and is recognized as an effective therapeutic regimen for the treatment of prostate cancers. Prospective studies have not demonstrated strong correlations between prostate cancer risk and circulating levels of testosterone, dihydrotestosterone, androstenedione, and several other androgenic hormones. However, several studies have demonstrated that variant alleles of the gene encoding the androgen receptor (AR) may modify prostate cancer risk. The AR gene contains within exon 1 a polymorphic CAG repeat that encodes a variable length glutamine tract that appears to impact the ability of the AR to activate gene transcription. Prostate cancer risk appears to be inversely correlated with the number of CAG repeats in the AR. A polymorphic GGC repeat in exon 1 of AR, encoding a variable length glycine tract, may also modify prostate cancer risk.      
Evidence is emerging to suggest that genes encoding enzymes involved in androgen biosynthesis or catabolism may also modify prostate cancer risk. The enzyme 5α-reductase type 2 (SRD5A2) catalyzes the conversion of testosterone to the more active dihydrotestosterone, and the gene encoding this enzyme, SRD5A2, is expressed exclusively in the prostate. Finasteride, an inhibitor of SRD5A2, is currently being evaluated in a large cooperative trial for its ability to prevent prostate cancer. A variant SRD5A2 allele, which encodes threonine in place of alanine at codon 49 (Ala49Thr), is associated with a significantly increased risk of prostate cancer in African-American and Hispanic men. In addition, the Ala49Thr variant is associated with a more aggressive disease phenotype and poorer prognosis relative to the wild-type allele. Interestingly, the Ala49Thr variant of SRD5A2 encodes an enzyme with an increased capacity to produce dihydrotestosterone relative to the wild-type allele.      
Two other SRD5A2 allelic variants, Val89Leu and a polymorphic dinucleotide repeat in the 3'- untranslated region, do not appear to significantly impact androgen metabolism or prostate cancer risk. As discussed in Section IA, CYP17 catalyzes two reactions in the biosynthetic pathway leading to androgens and estrogens. Data on CYP17 as a modifier of prostate cancer risk are inconsistent, with two studies associating the A2 allelic variant of CYP17 with an increased risk of prostate cancer and two studies associating the A1 allele with increased risk.      
Epidemiologic studies associate vitamin D deficiency with prostate cancer, and numerous laboratory studies indicate a role of the hormonally active form of vitamin D, 1,25-dihydroxyvitamin D3, in the regulation of proliferation and survival of prostate cancer cell lines. 1,25-Dihydroxyvitamin D3 functions through a receptor that is a member of the nuclear receptor superfamily. Several allelic variants of the vitamin D receptor have been identified, and studies suggest that these variants may be associated with an altered prostate cancer risk. Although not all of these studies concur with regard to the associations between specific polymorphisms and risk, together they suggest that the gene encoding the vitamin D receptor may act as a low penetrance modifier of prostate cancer development.      
Noble was the first to report that continuous treatment with either testosterone or testosterone in combination with estrogen induces the development of prostate cancers in the NBL rat. A series of studies by S.-M. Ho and colleagues indicates that treatment of intact NBL rats with testosterone and E2 for 16 weeks induces a high incidence of epithelial dysplasia in the dorsolateral lobe of the prostate. Relative to the surrounding epithelium, dysplastic lesions exhibit several changes in gene expression that may contribute to or result from their development, including increased expression of Ha-ras and progesterone receptor mRNAs. Induction of the dysplastic lesions is inhibited by treatment with bromocryptine, suggesting that prolactin contributes to their development. Interestingly, the location of the hormone-induced lesions within the prostate appears to be dependent on the amount and/or type of estrogen administered. For example, testosterone in combination with increased E2 induces atypical hyperplasia in the ventral lobe of the prostate. Similarly, treatment of NBL rats with testosterone plus DES induces dysplasia in the ventral prostate.      
Studies by Bosland and colleagues indicate that long-term treatment of NBL rats with testosterone and E2 induces a 100% incidence of adenocarcinoma in the dorsolateral, but not ventral, prostate of the NBL rat. A DNA adduct of unknown structure was detected in the dorsolateral, but not ventral, prostate of NBL rats treated with testosterone and E2 for 16 weeks or longer. More recently, Y. C. Wong and colleagues have published a series of studies on the induction of carcinoma in the ventral prostate of the NBL rat by continuous treatment with testosterone and high-dose E2.      
The mechanisms of prostate cancer induction by testosterone and estrogens are not currently known. It is clear that androgens play an important role in the regulation of cell proliferation and survival in the prostate gland of the human as well as rodent species, and it is probable that these actions contribute to the etiology of prostate cancers. For example, it is hypothesized that by stimulating proliferation within the prostate epithelium, androgens promote the development of cancers that are induced by endogenous or environmental carcinogens, possibly including genotoxic estrogen metabolites. The identification of specific mutations induced in the rat prostate as a consequence of combined testosterone and E2 treatment would provide strong support for this hypothesis. It is also hypothesized that the hormonal environment during embryonic development may contribute to the genesis of prostate cancers later in life. This hypothesis is based on data from numerous studies that indicate that prenatal exposure to naturally occurring estrogens, synthetic estrogens, or xenoestrogens impacts development of the male reproductive tract, including the prostate, in rats and mice. Each of these actions of androgenic and estrogenic hormones could contribute to prostate cancer development.

IV. PITUITARY TUMORS

Benign tumors of the anterior pituitary gland are common in humans, occurring in approximately 20-25% of individuals randomly evaluated at autopsy. Prolactin (PRL)-producing tumors, prolactinomas, represent the most common type of pituitary tumor. Estrogens have been implicated in the etiology of prolactinoma. The incidence of these tumors in females is approximately twice that of males, and several anecdotal observations link the development of prolactinoma to clinical or environmental exposures to estrogens. Supporting a role of estrogens in the etiology of prolactinoma are numerous reports that administered estrogens induce development of PRLproducing pituitary tumors in mice and rats. The literature on the role of estrogens in pituitary tumorigenesis in the human and rat has been reviewed by Spady and co-workers.      
Different inbred rat strains exhibit marked differences in sensitivity to the pituitary tumor-inducing actions of estrogens. The Fischer 344 (F344) strain is the most sensitive of the inbred rat strains. Continuous treatment of male or female F344 rats with DES or E2 for 6 weeks or longer results in 10- to 20-fold increases in pituitary mass. This increase in pituitary mass results from increased proliferation and enhanced survival within the PRL-producing lactotroph population and correlates with both absolute lactotroph number and the level of PRL in the circulation. Estrogeninduced pituitary tumors are usually defined histologically as diffuse lactotroph hyperplasias, although adenomatous foci have been described in some studies.      
In our laboratory, we have only rarely observed a pituitary carcinoma in a rat treated continuously with estrogen for long periods of time. Although less sensitive than F344, the Wistar-Furth, ACI, NBL, and COP rat strains also develop PRL-producing pituitary tumors when treated with estrogens. In contrast, the BN rat strain is highly insensitive to the pituitary growth-promoting actions of estrogens.      
The differing sensitivities of different inbred rat strains to the pituitary tumor-inducing actions of estrogens allow the underlying mechanisms to be studied using genetic techniques. The first studies of this type, by Wiklund and Gorski, indicated that the high sensitivity of the F344 strain to DES resulted from the actions of multiple, independently segregating genes. Wendell and Gorski have subsequently mapped within the rat genome the locations of several modifiers of DES-induced pituitary growth in F344 x BN intercrosses. Our laboratory has mapped additional modifiers of DES-induced pituitary growth in ACI x COP and ACI x BN intercrosses. Together, these genetic studies indicate that the actions of estrogens in the regulation of lactotroph proliferation and the survival and development of PRL-producing pituitary tumors are controlled by many genes. The ultimate goals of these genetic studies are to identify these genes and establish their functions. It is possible that at least some of these genes will impact carcinogenesis in other estrogen-responsive tissues.

V. OTHER HORMONEINDUCED CANCERS

Numerous reports illustrate the ability of administered estrogens to induce cancers at a wide variety of tissue sites in rats, mice, hamsters, and other animal species. The incidence and type of cancers vary markedly as a function of species, genetic strain, type of estrogen used, route of estrogen administration, and timing of estrogen administration. For example, Noble described a wide variety of cancers or tumors that develop in NBL rats in response to continuous treatment with E1. The affected tissues include the adrenal cortex, mammary gland, pituitary gland, ovary, uterus, cervix, vagina, testis, prostate, thymus, adipose, pancreas, and salivary gland, as well as lymphoid tissues. Noble recognized that dependence on estrogens for growth is a common feature of most estrogen-induced tumors.      
Some studies suggest that exogenous estrogens may increase the risk of liver cancers in humans. Estrogens are potent tumor promoters in the rat liver, and the SERM tamoxifen acts as a complete carcinogen in the rat liver.      
Several, but not all, estrogens induce renal cancers in orchiectomized Syrian hamsters. These cancers appear to be of epithelial origin, and their induction by estrogens is markedly suppressed by testosterone, progesterone, and SERMs. Continuous estrogen treatment is associated with histologically evident cytotoxicity and compensatory regeneration in the hamster kidney. It has been suggested that these actions contribute significantly to estrogen-induced renal carcinogenesis. A role of potentially genotoxic estrogen metabolites has also been suggested to initiate the development of kidney cancers in this animal model.

VI. CONCLUDING REMARKS

Based on a wealth of data from epidemiologic and laboratory studies, steroid hormones can be considered to be the cause of at least a subset of specific cancers in humans. The evidence that endogenous hormones cause cancer is most compelling for cancers of the breast and prostate. Consequently, strategies based on blocking the actions of estrogens and androgens are currently being used or are being evaluated for the prevention of cancers of the breast and prostate, respectively. Exogenous hormones contribute significantly to the etiology of cancers of the breast, uterine endometrium, and uterine cervix. Despite many years of intense study, the mechanisms through which these hormones exert their carcinogenic actions are not well defined. Possible contributory mechanisms include, but are not limited to the (1) stimulation of cell proliferation; (2) inhibition of apoptosis; (3) alteration of hormonally regulated developmental programs; (4) genotoxic actions of specific hormone metabolites; and (5) abrogation of genome stability. It is probable that multiple and/or different mechanisms contribute to the genesis of different cancer types and that genetic background impacts on these mechanisms.      
Currently existing animal models of hormoneinduced cancers should continue to serve as valuable tools for elucidating the mechanisms of hormonal carcinogenesis, and additional, physiologically relevant, animal models need to be developed. Geneticsbased approaches can be utilized to compare different inbred strains of rats and mice that exhibit marked differences in susceptibility to specific hormoneinduced cancers. Novel transgenic and knockout mouse models can be developed to evaluate the roles of specific genes as modifiers of susceptibility to hormone-induced cancers. It is critical to identify causative mutations for the different hormoneinduced cancers to establish firmly whether hormones or their metabolites act through genotoxic mechanisms to initiate carcinogenesis. The ongoing human, mouse, and rat genome projects are providing vast information as well as new tools to support these efforts. It is hoped that significant improvements in the prevention and treatment of hormone-associated cancers are on the horizon.

James D. Shull
University of Nebraska Medical Center

See Also
BREAST CANCER ; ENDOCRINE TUMORS ; ENDOMETRIAL CANCER ; ESTROGENS AND ANTIESTROGENS ; CHEMICAL MUTAGENESIS AND CARCINOGENESIS ; HORMONE RESISTANCE IN PROSTATE CANCER ; PROSTATE CANCER ; STEROID HORMONES AND HORMONE RECEPTORS

GLOSSARY
allele One form of a gene for which different forms exist in a population.

allelic variant One of two or more forms of a specific gene existing within a population.

cytochrome P450 A large family of enzymes that catalyze oxidation/reduction reactions on a wide variety of endogenous and exogenous substrates.

genetic modifier A gene that through the actions of its protein product alters cancer risk in a quantifiable manner.

genotoxic DNA damage or mutagenesis.

inbred strain A mouse or rat strain in which genetic variation is eliminated through multiple generations of stringent inbreeding.

low penetrance A gene for which a specific allelic variant does not uniformly confer the specific phenotype.

polymorphism Genetic variability within a population.

xenoestrogens Exogenous chemicals that can bind to the estrogen receptor and mimic the actions of estrogens.

Bibliography
Bosland, M. C., Ford, H., and Horton, L. (1995). Induction at high incidence of ductal prostate adenocarcinomas in NBL/Cr and Sprague-Dawley Hsd: SD rats treated with a combination of testosterone and estradiol-17β or diethylstilbestrol. Carcinogenesis 16, 1311-1318.
Carthew, P., Edwards, R. E., Nolan, B. M., Martin, E. A., Heydon, R. T., White, I. N. H., and Tucker, M. J. (2000). Tamoxifen induces endometrial and vaginal cancer in rats treated in the absence of endometrial hyperplasia. Carcinogenesis 21, 793-797.
El-Bayoumy, K., Ji, B.-Y., Upadhyaya, P., Chae, Y.-H., Kurtzke, C., Rivenson, A., Reddy, B. S., Amin, S., and Hecht, S. S. (1996). Lack of tumorigenicity of cholesterol epoxides and estrone-3,4-quinone in the rat mammary gland. Cancer Res. 56, 1970-1973.
Harvell, D. M. E., Strecker, T. E., Tochacek, M., Xie, B., Pennington, K. L., McComb, R. D., Roy, S. K., and Shull, J. D. (2000). Rat strain specific actions of 17β-estradiol in the mammary gland: Correlation between estrogen-induced lobuloalveolar hyperplasia and susceptibility to estrogeninduced mammary cancers. Proc. Natl. Acad. Sci. USA 97, 2779-2784.
Herbst, A. L. (1999). Diethylstilbestrol and adenocarcinoma of the vagina. Am. J. Obstet. Gynecol. 181, 1576-1578.
Jaffe, J. M., Malkowicz, S. B., Walker, A. H., MacBride, S., Peschel, R., Tomaszewski, J., Van Arsdalen, K., Wein, A. J., and Rebbeck, T. R. (2000). Association of SRD5A2 genotype and pathological characteristics of prostate tumors. Cancer Res. 60, 1626-1630.
Li, J., and Li, S. A. (1996). Estrogen carcinogenesis in the hamster kidney: A hormone-driven multistep process. In "Cellular and Molecular Mechanisms of Hormonal Carcinogenesis" (J. Huff, J. Boyd, and J. C. Barrett, eds.), pp. 255-267.
Wiley-Liss, New York. Liehr, J. G. (2000). Is estradiol a genotoxic mutagenic carcinogen? Endocr. Rev. 21, 40-54.
Makridakis, N. M., Ross, R. K., Pike, M. C., Crocitto, L. E., Kolonel, L. N., Pearce, C. L., Henderson, B. E., and Reichardt, J. K. (1999). Association of missense substitution in SRD5A2 gene with prostate cancer in African- American and Hispanic men in Los Angeles, USA. Lancet 354, 975-978.
Newbold, R. R., Bullock, B. C., and McLachlan, J. A. (1990). Uterine adenocarcinoma in mice following developmental treatment with estrogens: A model for hormonal carcinogenesis. Cancer Res. 50, 7677-7681.
Newbold, R. R., and Liehr, J. G. (2000). Induction of uterine adenocarcinoma in CD-1 mice by catechol estrogens. Cancer Res. 60, 235-237.
Noble, R. L. (1975). The development of prostatic adenocarcinoma in Nb rats following prolonged sex hormone administration. Cancer Res. 37, 1929-1933.
Shull, J. D., Spady, T. J., Snyder, M. C., Johansson, S. L., and Pennington, K. L. (1997). Ovary intact, but not ovariectomized, female ACI rats treated with 17β-estradiol rapidly develop mammary carcinoma. Carcinogenesis 18, 1595- 1601.
Spady, T. J., McComb, R. D., and Shull, J. D. (1999). Estrogen action in the regulation of cell proliferation, cell survival, and tumorigenesis in the rat anterior pituitary gland. Endocrine 11, 217-233.
Telang, N. T., Suto, A., Wong, G. Y., Osborne, M. P., and Bradlow, H. L. (1992). Induction by estrogen metabolite 16 alpha-hydroxyestrone of genotoxic damage and aberrant proliferation in mouse mammary epithelial cells. J. Natl. Cancer Inst. 84, 634-638.
Tsutsui, T., Tamura, Y., Yagi, E., and Barrett, J C. (2000). Involvement of genotoxic effects in the initiation of estrogen- induced cellular transformation: Studies using Syrian hamster embryo cells treated with 17β-estradiol and eight of its metabolites. Int. J. Cancer 86, 8-14.
Weiderpass, E., Persson, I., Melhus, H., Wedrén, S., Kindmark, A., and Baron, J. A. (2000). Estrogen receptor α gene polymorphism and endometrial cancer risk. Carcinogenesis 21, 623-627.
Wong, Y. C., Wang, Y. Z., and Tam, N. N. (1998) The prostate gland and prostate carcinogenesis. Ital. J. Anat. Embryol. 103, 237-252.

Chemical Mutagenesis and Carcinogenesis

Experimental carcinogenesis and human epidemiology studies have clearly identified specific chemicals that can act as human carcinogens. Certain chemicals have been associated with increased human cancer incidence in both occupational and environmental exposure settings.

I. INTRODUCTION

It is now widely recognized that exposure to chemicals in the workplace and the environment can contribute to human cancer risk. This was first postulated in 1775 by Dr. Percival Pott, who recognized an association between occupational exposure to chemicals in soot and an increased incidence of scrotal cancer in London chimney sweeps. Other sporadic clinical observations of this type throughout the 19th century also suggested an association between certain occupational exposures to specific chemical agents and increased human cancer. However, it was not until the 20th century that science and medicine actively investigated this aspect of cancer etiology. Boveri first hypothesized that cancer was a genetic disease in 1921, prior to the discovery of the genetic material. In 1915, Yamagawa and co-workers demonstrated that application of coal tar could induce tumors in animals.      
In the 1930s Kenneway and co-workers demonstrated that pure chemicals isolated from coal tar could also induce animal tumors. Parallel discoveries in the 1950s of the structure of the DNA double helix and its establishment as the hereditary material on the one hand and mutagenic potential of ionizing radiation and certain chemical carcinogens in humans and experimental systems on the other set the stage for extensive investigations into the relationship between chemically induced mutations and human cancer. Public perceptions, and the regulatory policies resulting from them, have had a strong influence on the direction of chemical carcinogenesis studies and their interpretation. The infamous Delaney Clause of the 1950s led to an intensive focus on the potential carcinogenicity of food additives and other contaminants, and also established the paradigm of the animal tumor model as a test system for predicting carcinogenicity of chemicals in humans. Similarly, the environmental movement of the 1960s and 1970s, which began with the publication of Rachel Carson's Silent Spring and the subsequent discovery of Love Canal, led to the establishment of the Federal Environmental Protection Agency and "Superfund," clean air, and clean water legislation. This movement was accompanied by a growing public concern that the widespread use of pesticides and exposure to chemicals from toxic waste sites could cause cancer, and more generally to the belief that chemicals in the environment, particularly man-made chemicals, were responsible for a major fraction, perhaps the majority, of human cancers. The 1980s saw the elucidation of the first oncogenes, genes that appeared to be responsible for the initiation of cancer as first predicted by Boveri. This era also saw the development of the Ames Salmonella bacterial mutagenesis assay and hundreds of similar genetic toxicology assays. These developments firmly established the basic paradigm for the field of chemical carcinogenesis; that chemicals that can cause mutations are presumed to be carcinogens. By extension, it was predicted that any chemical or physical agent that could covalently damage DNA could also cause mutations through its DNAdamaging mechanism, and hence could also be a carcinogen.      
The data that followed over the next decade appeared to strongly support this central tenet, as the vast majority of chemicals that were initially tested for DNA damage or mutations were also shown experimentally to be animal carcinogens. However, most of the chemicals that were initially tested were either known or suspected carcinogens, were agents of concern, or were closely related structurally to these known and suspected carcinogens.      
Over the most recent decade, our understanding of the molecular basis of cancer has substantially improved. In addition, investigations into the molecular basis of chemical carcinogenesis, as well as more extensive human cancer epidemiology studies using modern molecular tools, have greatly expanded our knowledge in this area. This has led to a reevaluation of the overall role of chemicals in human cancer, as well as a modification of the basic paradigms that form the initial foundations of the field. In contrast to earlier predictions, estimates suggest that perhaps only 4-8% of human cancers can be directly attributed to specific chemical exposures. Moreover, these exposures are primarily occupationally related or involve specific environmental contaminants, and it is therefore unlikely that environmental exposures to most chemicals contribute significantly to the overall burden of human cancer. The reasons for this are several fold.      
First, while many chemicals have the potential to be genotoxic (DNA damaging), mutagenic, or carcinogenic in various test systems, the levels of most environmental human exposures are usually too low, or are too transient, to have an appreciable impact on cancer relative to other risks. Second, it is also now clear that humans have many protective mechanisms that can act as practical threshold barriers to most minor chemical exposures. These include physiological barriers to chemicals, xenobiotic metabolism and other pharmacokinetic mechanisms of detoxification, a highly efficient DNA repair system, various active apoptotic mechanisms, and various immune and other active surveillance mechanisms. These thresholds are often not present, or are exceeded or bypassed, in many test systems, including rodent bioassays. For example, some agents are only carcinogenic in animals when administered by large bolus intraperitoneal or intramuscular injections that bypass the normal dose, time course, and route of exposure that would otherwise have resulted in detoxification. Chemicals are also usually given to test animals at or near their maximum tolerated dose (MTD), whereas people are usually exposed to these same chemicals at levels that are thousands or even millions of times lower. About half of all chemicals that have been tested in rodents, whether natural or man-made, have been shown to increase tumor formation. However, the list of chemicals that have been tested to date is highly biased.       
Third, based on a growing database, it is now also clear that while there are considerable overlaps among groups of chemicals that are genotoxic, mutagenic, or carcinogenic, these properties are not equivalent. For example, not all chemicals that are genotoxic are mutagenic. This is likely to be due to both the ability of the cell to recognize certain kinds of damage with extremely high efficiency at low levels and the observation that not all types of damage are equally mutagenic to the cell during DNA replication. Similarly, there are many chemicals that are mutagenic but not genotoxic. For example, certain metal ions that do not cause DNA damage can interact with and directly alter the fidelity of DNA polymerase, at least in vitro, and thereby lead to mutations due to replication errors. Moreover, DNA-intercalating agents such as ethidium bromide, which can transiently stack between the base pairs of the DNA, can also cause mutations in the form of frameshifts (additions or losses of DNA base pairs), without any covalent modification of DNA.      
Most importantly, there is a large list of chemicals of concern that are carcinogenic, based on evidence in experimental animals or as demonstrated directly in human epidemiology studies, but that have not been found as overtly genotoxic or mutagenic in experimental test systems. Approximately one-third of chemical mutagens are negative as chemical carcinogens in animals, and vice versa. Examples of carcinogenic agents that are not genotoxic or mutagenic include the so-called solid-state carcinogens (e.g., smooth plastic implants) and the carcinogenic metals, arsenic and cadmium. In fact, in the case of cadmium and arsenic, it has been difficult to demonstrate that they are carcinogenic as single agents in animal models, despite strong epidemiological evidence that they are carcinogenic in humans.
Moreover, many chemicals that were initially thought to act by genotoxic and mutagenic mechanisms may, in fact, act primarily through other as yet unclear mechanisms that do not involve DNA damage or mutations per se. For example, although the human lung toxin and carcinogen chromium(VI) is moderately positive for DNA damage and mutations in certain experimental test systems, there is growing evidence that it may act as a human carcinogen primarily through cell signaling and tumor promoter-associated mechanisms rather than or in addition to its genotoxic or mutagenic effects. It is therefore important, in assessing the overall carcinogenic potential of a chemical, to consider whether it is a genotoxin, mutagen, or animal or human carcinogen separately and to also classify the type of mechanism involved. Thus, chemically induced DNA damage, chemical mutagenesis, and chemical carcinogenesis are considered separately.

II. CHEMICALLY INDUCED DNA DAMAGE

A. Types of Chemical Interactions with DNA
Two basic types of chemical interaction with DNA are noncovalent and covalent binding. Examples of the primary types of noncovalent interactions include (a) ionic interactions, such as when Mg 2 and other cations interact with the negative phosphate groups on the outside of the DNA helix; (b) minor groovebinding chemicals, such as the dye Hoechst 33258; and (c) intercalating agents, such as ethidium bromide and bleomycin, which are planar aromatic compounds capable of stacking between the parallel base pairs inside the DNA helix. These various noncovalent interactions can each have toxicological consequences. For example, such interactions can affect the structure of the DNA helix or disrupt DNA- protein interactions within chromatin, leading to changes in DNA replication and RNA expression.      
They can also alter the fidelity of DNA and RNA polymerases, in the former case increasing the probability of mutations. Noncovalent chemical-DNA interactions are not considered damage perse; however, the noncovalent interactions of some chemicals can subsequently lead to the generation of covalent DNA damage. For example, intercalation of bleomycin into DNA and subsequent binding of iron to the drug can lead to the generation of reactive oxygen species that can damage the DNA at sites adjacent to the drug intercalation site. At high levels, this damage will kill the cell, which is the basis for the use of bleomycin as an anticancer drug, but at lower levels this damage can also result in mutations. Covalent interactions occur when the chemical, or a portion of the chemical, is covalently adducted to the DNA helix or when it causes other types of covalent modifications, such as oxidative damage. Chemicals that can directly adduct to DNA are often referred to generically as "DNA-alkylating agents." The various types of DNA damage are discussed in further detail.

B. Types of DNA Damage
1. Spontaneous DNA Damage
In the context of this topic there are two basic classes of DNA damage to consider, i.e., so-called "spontaneous" or background damage and chemically induced damage. Spontaneous DNA damage includes deamination, loss of bases to form abasic sites, and oxidative damage. Deamination, i.e., chemical loss of amino groups on bases, occurs frequently and spontaneously. It has been estimated that approximately 10,000 deamination events occur per cell per day, on average, in humans and other mammals. Deamination usually leads to the formation of unusual bases, e.g., cytosine deamination forms uracil (which is the thymine analog in RNA but not a normal DNA base) and adenine deamination forms hypoxanthine.
These deaminated bases are potentially mutagenic. If left unrepaired, some abnormal bases can form an alternative base pair with the wrong partner during replication, leading to a mutation at that site. For example, uracil can base pair with adenine because it is a structural analog of thymine so that deamination of cytosine to uracil can result in a transition from a C-G base pair to a T-A base pair if undetected. However, the common deamination products are detected by specific enzymes called DNA glycosylases, each of which recognizes a specific inappropriate base in DNA (e.g., uracil glycosylase). Some deamination events do not have a mutagenic consequence, e.g., deamination of guanine leads to the formation of xanthine, which, if unrepaired, normally will still be base paired preferentially with cytosine during replication. From the standpoint of mutations, the most important deamination is of 5-methylcytosine, as this leads to formation of thymine, which is a normal DNA base. If this mismatch is left unresolved, it can lead to a transition mutation to a T-A base pair during replication.      
In fact, it has been observed that sites of 5- methylcytosine are typically "hot spots" for mutation, with a much higher mutation frequency than other sites. Repair of these mismatches requires a complex repair process that queries each strand to identify the parental and daughter strands of the helix. This allows it to determine whether the site should be methylated to determine which base is incorrect.      
Abasic sites occur when a glycosidic bond in DNA weakens and the helix releases the base at that site. This also occurs at a high spontaneous level, with approximately 20,000 purines and 1000 pyrimidines lost per cell per day on average. These abasic sites are also recognized and repaired by the cellular repair machinery. Background oxidative DNA damage and radiation-induced DNA damage are often classified as "spontaneous" damage because they are part of the normal background. A low level of reactive oxygen species (primarily hydroxyl radical and superoxide) is generated chemically or enzymatically in cells on a continuous basis, and some of these reactive byproducts will attack DNA to cause oxidative base damages such as 8-oxodeoxyguanosine. Similarly, background levels of ultraviolet (UV) and ionizing radiation are encountered by virtually all organisms. Low levels of background DNA damage (pyrimidine dimers and 6-4 photoproducts from UV; oxidized bases, and single and double strand breaks from ionizing radiation) can occur from this radiation. However, these lesions are also normally recognized and repaired by the cellular machinery.       
It is important to consider spontaneous or background levels of DNA damage in considering chemically induced DNA damage for several reasons. First, it is clear that all the cells in our bodies are continuously challenged by these background levels of DNA damage, which, if left unrepaired, are potentially mutagenic and therefore possibly carcinogenic. Thus, normal individuals will have low background levels of DNA damage and mutations even in the absence of chemical exposure, which must be taken into account when assessing the added impact of chemical exposures that might contribute to these levels. One obvious way that this might occur is that chemicals can directly damage DNA and thereby induce mutations.      
However, additional mechanisms might include altering important cellular processes, such as DNA damage recognition and repair, replication fidelity, cell cycle checkpoints, apoptotic mechanisms, or overall cell proliferation rates, each of which might strongly influence the mutation rate from these background forms of DNA damage. Alternatively, one must consider very low levels of chemically induced DNA damage in the context of the background damage and repair that is always present. This additional damage may be inconsequential with respect to increasing the overall rate of mutation until a specific threshold of damage is attained such that the probability of mutation is increased to a measurable level. Understanding the relationship between background and chemically induced DNA damage and mutations has important implications for risk assessment, particularly for determining the appropriate model for extrapolating from measurable effects at high-dose human or animal exposures to possible risks from very low-dose exposures.

2. Chemically Induced DNA Damage
Chemically induced DNA damage includes oxidative damage, simple alkylation, bulky monoadducts, and DNA cross-links. Chemically induced oxidative DNA damage can result from the generation of hydrogen peroxide in the cell or from redox reactions generated by chemicals such as menadione, bleomycin, and certain metals inside the cell. Simple alkylation typically occurs when a reactive parent compound has a reactive methyl or ethyl group that can covalently interact a DNA base, leaving the alkyl group behind.
Examples of alkylation agents include direct-acting compounds, such as methylor ethyl-methanesulfonate, and indirect-acting agents, such as methyl- or ethyl-nitrosourea, which require activation to a reactive intermediate metabolite by the phase I cytochrome P450 system. Simple alkylation normally involves the addition of a methyl or ethyl group to the N2, N7, and O6 of guanine but also occurs at other nucleophilic sites of DNA. The O6-methylguanine adduct in particular is highly mutagenic if unrepaired. Many organic mutagen-carcinogens form bulky monoadducts with DNA, typically following metabolism of the parent compound to a reactive intermediate by the phase I cytochrome P450 system. Examples of chemical carcinogens that do this are mycotoxin, aflatoxin B1 (AFB1), and the polycyclic aromatic hydrocarbon, benzo[a]pyrene (BaP), each of which has been shown to induce bulky monoadducts that cause specific mutations at the site of adduction.      
With the advent of modern molecular biology techniques, it has been possible to determine the precise mutations that result from each type of chemical adduct and to compare this mutational pattern with the mutations found in oncogenes and tumor suppressor genes of tumors produced by that chemical in experimental animals. For example, BaP is metabolized to the reactive intermediate benzo[a]prene-7,8,- diol-9,10-epoxide (BPDE), which forms an adduct at the N2 of guanine and results in a G-C to T-A mutation. Use of site-specific adducts of BPDE in shuttle vectors has shown that this lesion almost exclusively results in G-C to T-A mutations at the site of the lesion.       
Cells in culture or tumors from animals treated with BaP or BPDE have certain activated oncogenes, and the mutations that occur in these oncogenes are G-C to T-A mutations. Oncogenes that have been synthesized to contain specific G-C to T-A mutations have been shown to be sufficient to transform cells to tumorigenic cells in culture, indicating that this type of mutation in and of itself can contribute to the carcinogenic process. Thus, these empirical observations, at least in the example of BaP, match the predictions of the basic paradigm linking chemical genotoxicity to mutagenesis and carcinogenesis, i.e., that certain chemicals can cause cancer by attacking and damaging DNA, generating DNA adducts that cause specific mutations in critical cancer genes that, in turn, can contribute to a cancer cell genotype and phenotype.      
DNA cross-links are another category of DNA damage induced by chemicals and some forms of radiation. The three basic types are interstrand (covalent linkage of two bases on opposite DNA strands), intrastrand (linkage of bases on the same strand), and DNA-protein cross-links. Different bifunctional agents can induce one, two, or all three types of crosslinks, in addition to other types of DNA lesions, with varying efficiencies. For example, the human lung carcinogen chromium(VI) induces both DNA interstrand and DNA-protein cross-links, in addition to the formation of Cr-DNA monoadducts and the generation of reactive oxygen species that can also damage DNA. Similarly, the cancer chemotherapy agents mitomycin C and cisplatin can each induce a spectrum of monoadducts, DNA interstrand cross-links, and DNA intrastrand cross-links. DNA interstrand cross-links, in particular, are very difficult for the cell to repair because both strands of DNA are involved, and if left unrepaired these lesions can be lethal to the cell. This is believed to be the principal basis for the ability of mitomycin, cisplatin, and similar crosslinking agents to kill cancer cells. However, at lower doses it is not clear whether DNA interstrand crosslinks are mutagenic. Similarly, DNA-protein crosslinks do not appear to be strongly associated with an increase in mutations. These lesions are generally well recognized and repaired by the cellular machinery.      
Most chemically induced DNA intrastrand cross-links are also easily recognized and repaired, as they usually cause severe distortions of the DNA helix. UV light can cause thymine dimers and 6-4 photoproducts, two specific forms of interstrand cross-links. Thymine dimers are also recognized and repaired in cells, whereas the 6-4 photoproduct appears to be poorly recognized and is very mutagenic if left unrepaired. Formation of this adduct is probably the primary basis for the increased risk of skin cancer in heavily sunexposed people and in xeroderma pigmentosum (XP) patients who are deficient in specific repair enzymes (discussed in further detail later). Ionizing radiation and some chemicals can cause single and double strand breaks in the DNA helix. Single strand breaks are usually repaired by the cellular machinery and are a normal intermediate in many DNA replication and maintenance processes. Double strand breaks can be repaired, although poorly and with slow kinetics, as this involves a more complex repair process, and so these are considered lethal to a dividing cell. This is one of the primary cellular lesions induced by therapeutic ionizing radiation that leads to tumor cell cytotoxicity.      
Assays for DNA damage are generally of two types: they are either very specific for a particular type of damage or they detect overall levels of DNA damage without information about the type of damage. For example, carcinogen-binding assays with radiolabeled compounds provide information about covalent adduction but not other types of DNA damage (e.g., oxidative damage or strand breaks). Oxidative damage can be assessed by HPLC analysis of altered bases or use of lesion-specific antibodies. DNA alkaline elution is a technique that can be used to measure DNA interstrand and DNA-protein cross-links, as well as frank strand breaks and so-called "alkali labile sites," i.e., regions of weakness in the DNA helix, presumably as a result of damage, that are elaborated in the presence of strongly alkaline conditions. Assays using 32P postlabeling for altered bases can be used to assess a spectrum of adducts from a given agent, but the technique must be customized to each agent and its DNA damage products, and many agents are not amenable to this analysis for technical reasons. Similarly, HPLC and antibody-based detection of adducts can be developed, but these methods are agent specific.      
Sister chromatid exchanges occur at a low level in all dividing cells but are increased by many different chemical and physical agents that damage DNA. These events can be detected by differential staining or antibody techniques. However, it is not clear whether chemically induced exchanges between identical chromatids represent DNA damage, successful or unsuccessful repair of damage, or whether there are mutations associated with these events (e.g., by unequal exchange of sequences). Light microscopy can be used to measure increases in chromosomal aberrations, and formation of micronuclei and "unscheduled DNA synthesis" (i.e., DNA synthesis not associated with replication, which is presumed to be DNA repair in response to induced DNA damage) in treated cells can also be used to assess the effects of genotoxic agents, but these techniques also do not distinguish individual lesions. It is therefore often useful to use several complementary assays when assessing DNA damage from a particular chemical.

C. Repair of DNA Damage
Eukaryotic DNA repair is a highly efficient and coordinated process that normally protects cells and organisms from chemical or genetic alterations to the genetic material. This is a continuous process in the cell that counteracts the background levels of DNA damage that occur in each cell every day. That DNA repair is critical both for suppressing the cancer process and for maintaining life itself can be illustrated by two related observations. First, as mentioned earlier, it has been estimated that in the absence of any other chemical or physical agent, each cell of the body loses or experiences damage to several thousand DNA bases per day through spontaneous chemical degeneration of the DNA helix, which would be potentially mutagenic or lethal were it not repaired on an ongoing basis.      
These tens of thousands of potentially mutagenic lesions are occurring in each of the trillions of cells in our body on a daily basis. However, the average life span of U.S. citizens is now expected to extend well into the eighth decade, and cancer is not a significant risk for most of us until the sixth or seventh decade of life. Second, among the hereditary mutations described to date that are known to predispose humans to cancer, the majority involve defects in genes whose proteins mediate specific steps of DNA repair. For example, individual mutations in eight different genes all result in the clinical disease xeroderma pigmentosum. XP patients have a high rate of skin cancer from even low levels of sun exposure, and all of the XP mutations occur in genes involved in DNA excision repair.      
Studies of the different "complementation groups" of these XP mutations have given us major insights into the specific genes and proteins involved in DNA excision repair in mammals. Mutations in various other DNA repair genes are also associated with an increased risk of cancer and other diseases, including a gene involved in mismatch repair that predisposes to colorectal cancer (hereditary nonpolyposis colon cancer, HNPCC) and genes associated with Bloom's syndrome, Fanconi's anemia, and trichithiodystrophy.      
There are four basic types of repair, i.e., direct repair, base excision repair (BER), (poly)nucleotide excision repair (NER), and postreplication repair (recombination and mismatch repair). Sexual reproduction has been described as a fifth form of DNA repair, and the ultimate form of genomic DNA repair, as the pairing of alleles from two different individuals provides an opportunity for the genome of the offspring to contain at least one intact copy of each critical gene. Direct repair involves chemically restoring a damaged base to its original structure, e.g., by reversal of a UV light-induced lesion by the photoreactivation enzyme, photolyasem, or removal of a methyl or ethyl group from guanine by O6-methylguanine methyltransferase. BER involves removal of a damaged base from the DNA helix and subsequent replacement by the repair machinery. Different lesions are recognized by specific enzymes, e.g., uracil glycosylase. Nucleotide excision repair (more properly polynucleotide excision repair) involves recognition of a wide variety of DNA damages, excision of a patch of DNA surrounding the lesion by an enzyme complex, and repair of the gap by the replication machinery. Mismatch repair is a process that is linked to DNA replication, whereby mismatches created by DNA polymerase infidelity are corrected as part of the replication process. HNPCC genes are involved in mismatch repair. Why HNPCC mutations would specifically predispose people to colon cancer rather than a general increase in overall cancer risk is not clear. These examples of genetic predispositions to cancer illustrate how important DNA repair is for suppressing carcinogenesis from background and chemically induced DNA damage.

III. CHEMICALLY INDUCED MUTATIONS

A. Types of Mutations
The three basic classes of genetic mutations are point mutations, clastogenic mutations, and aneuploidy events. Point mutations are arbitrarily defined as changes in the DNA sequence of 10 bp or less. These can include single base pair changes (purine-to-purine and pyrimidine-to-pyrimidine transitions and purine-pyrimidine transversions) and small insertions or deletions of 1-10 bp. Transitions and transversions can result in changes in the coding of an individual amino acid if it occurs in the first or second codon position, but can be silent mutations if in the "wobble" position of many amino acid codons. These mutations can also result in creation of stop codons that result in mRNA and protein truncation. These can also cause changes in other types of genetic information, such as alterations in binding sites for transcription factors and other regulatory proteins within promoter regions, methylation sites within promoter regions, and mRNA splice sites that define the final mRNA structure and sequence. Insertions or deletions that are not divisible by three and that occur within coding regions of base pairs will result in frameshifting, such that the remainder of the coding sequence will have a dramatically altered amino acid information. This often results in the creation of premature termination signals and truncated gene products.      
Clastogenic mutations include insertions or deletions of greater than 10 bp, inversion of a sequence of DNA within the same chromosome, duplication of DNA sequence (which can include entire gene segments or gene clusters), and gene amplification. Amplification events can result in homogeneously staining regions (HSRs) that are visible in chromosome karyotypic analysis, or extrachromosomal "minichromosomes" that contain high copy numbers of an amplified gene or genetic cluster. Amplification events are inducible even in normal cells, at least in cell culture, and also reversible, suggesting that they represent a survival strategy for upregulating certain important genes during times of severe stress.      
Translocation events can occur, where regions of different chromosomes swap locations. This is a common event in cancer and appears to be important in the etiology of certain cancers. The classic example of this is the "Philadelphia chromosome," which is a translocation of the short "p" arms of chromosomes 9 and 22 +(9;22) and a hallmark of chronic myelogenous leukemia. Many translocations in cancer result in the juxtapositioning of an oncogene with a strong promoter region, resulting in substantial upregulation of oncogene expression (e.g., BCL-2/IG). Other translocations result in either truncation of a gene product, which can be an activation event for certain oncogenes (e.g., MYB), or creation of a chimeric gene product (e.g., PML-RARα in acute promyelocytic leukemia or BRC-ABL in chronic myelogenous leukemia).      
Aneuploidy involves gain or loss of one or more entire chromosomes. Only two human aneuploidy events are compatible with life if they are germline or occur somatically in the early embryo, namely, trisomy 21 (Down's syndrome) and XO (Turner's syndrome). However, most malignant cancers arising from somatic mutations exhibit extensive genetic rearrangement and aneuploidy events, resulting in what is called "loss of heterozygosity" (LOH) or, more properly, loss of dizygosity as the genes that remain are haploid (single copy). This genetic instability and increasing haploidy is considered to provide a selective advantage to the tumor, but whether these events are primarily causal in development of a fully malignant phenotype, and/or are a result of processes that occur in a more advanced cancer phenotype, is still not fully understood.

B. Chemically Induced Mutations
As described earlier, certain chemicals are genotoxic, i.e., they can directly or indirectly cause covalent DNA damage. Certain forms of damage potentially result in mutations if unrepaired. The biochemical mechanism for this, at least for the handful of chemicals that have been investigated at this level of mechanistic detail, is primarily by causing DNA polymerase to misread a template base (usually containing, or immediately adjacent to a site of covalent DNA damage) and causing polymerase to insert the incorrect base into the daughter strand when forming a base pair. Lesions that are most mutagenic appear to not only cause DNA polymerase to make this mistake, but also form an alternative base pair that is poorly recognized as being incorrect by the DNA replication machinery, thereby eluding postreplication mismatch repair as well. For example, the BPDE lesion bound to the N2 of guanine can cause the damaged base to rotate about its glycosidic bond, flipping the base over.      
This results in polymerase adding an adenine at the site to form a base pair with the "Hoogsteen" face of the flipped guanine, ultimately resulting in the G-C to T-A transversion that is characteristic of BPDE mutations. Similarly, the O6-methylguanine lesion, which is caused by many simple alkylating agents, can continue to base pair with a protonated cytosine, such that the site appears normal to the machinery, but O6-methylguanine can also be inappropriately base paired with thymine, resulting in a G-C to A-T transition. Moreover, the cellular machinery appears to poorly recognize the alternative thymine-containing base pair as well. Thus, the O6-methylguanine adduct can persist without detection, and this lesion can cause mutations during replication that are poorly de- tected. The prevalence of this adduct in background levels of oxidative DNA damage and its high mutagenic potential are probably why evolution has developed a specific repair pathway (O6-methylguanine methyltransferase) that recognizes and removes this lesion from DNA.     
As mentioned previously, chemicals can influence mutations in the absence of inducing overt DNA damage. One class of agents that can do this are chemicals that noncovalently interact with DNA. DNAintercalating agents such as ethidium bromide, acridine orange, bleomycin, actinomycin, and the various anthracyclines (e.g., daunorubicin, doxorubicin) can insert between the base pairs of DNA, forming stable interactions by virtue of their planar multiring structure and the formation of favorable electronic interactions with the ring systems of the purine and pyrimidine DNA bases. However, this intercalation has the effect of "stretching" the DNA helix along its long axis, and also altering DNA-protein interactions along the face of the helix. Intercalating agents can cause frameshift mutations, especially in sequential runs of the same base, e.g., a run of four or more adenines. Adjacent bases in these runs can share electronic pairing of the opposite bases, such that if DNA polymerase adds or omits a base on the opposite strand, the error can be difficult to detect. Intercalating agents appear to be able to increase the probability of these frameshift events occurring, as one typically sees insertion or deletion of single bases in these contiguous runs. Minor groove-binding drugs can also influence the rate of mutation, although the mechanism for this is not clear. Other agents, especially certain metal ions, can influence the fidelity of DNA polymerase itself, at least in in vitro and cell culture systems. Agents that block or suppress specific steps in DNA repair can increase mutations, presumably resulting from background and spontaneous DNA damage events.      
Another class of nongenotoxic chemicals that can increase mutagenesis in the presence of other agents are called "comutagens." An example of this is 2,3,7,8- dibenzo-p-dioxin (TCDD, dioxin) and similar polyhalogenated hydrocarbons, which are strong inducers of specific isozymes of cytochrome P450. This can lead to increased activation of other agents such as BaP, resulting in much greater mutagenesis following a low-level exposure than would otherwise occur. Another example is agents that can influence DNA damage recognition and/or repair of DNA lesions. Arsenic, which is not genotoxic or mutagenic per se, is a strong comutagen when present with other genotoxic agents such as BaP or AFB1. The precise mechanism for this is not clear, but arsenic has been shown to influence the expression of DNA repair genes and to affect the fidelity of DNA replication and repair. Thus, one might predict that arsenic is most carcinogenic in combination with other agents, such as cigarette smoke or UV irradiation, and animal data and human epidemiology studies support this prediction.       
As mentioned earlier, although the vast majority of mutations seen in human tumors are large insertions, deletions, rearrangements, and aneuploidy events, the vast majority of tests for genotoxicity and mutagenesis, and therefore the vast majority of our information about chemical carcinogens, have focused on their ability to cause point mutations, especially single base pair transitions and transversions. The development of the Salmonella reversion assay by Bruce Ames and colleagues in the early 1980s (the "Ames test") led to an ability to rapidly screen chemicals for their potential to cause simple point mutations in this bacterial system. While this provided a high throughput and rapid and inexpensive screening system, there are two apparent drawbacks of this assay. First, it is a prokaryotic system, which differs in important respects from human cells in its uptake, metabolism, and DNA machinery.      
Second, it focuses on reversion mutations (i.e., reverting from a mutant to a wild-type sequence) under strong selective pressure for specific phenotypes. Thus, there is a strong bias to only observing certain mutations in this system. Similar assays have also been developed in yeast and other lower eukaryotes, utilizing both reverse and forward (wild-type to mutant) mutation screening, and in mammalian cells with screens for both forward and reverse mutations. Use of mammalian forward mutation systems addresses some concerns about the bacterial and yeast systems, but these systems still have limitations with respect to metabolism of most promutagens. This latter issue has been partially addressed by the development of cell lines that have been transfected with specific P450 isozymes to provide key metabolic activation steps.      
Most of these assay systems also use immortalized cell lines, either derived from tumors or transformed to an immortalized phenotype by a viral or other genetic change. These cellular systems, being simple, uniform monolayers, also lack the multicellular, three dimensional aspects of whole animal tissues, and also lack other pharmacokinetic and pharmacodynamic properties of intact animals. Nonetheless, these various assays have provided a great deal of information about the mutagenic potential of hundreds of chemicals to establish a growing database of information.      
Recent development of transgenic mice and rats (e.g., the "Big Blue" mice and rats) has allowed the screening of mutations in shuttle vectors following in vivo exposure, which largely alleviates the problems of cell culture systems. However, these systems still focus predominantly on single point mutations and selectable markers. In vivo systems that can look at other mutational events in vivo are limited but include dominant lethal, heritable translocation, and mouse spot test and specific locus assays. However, these systems are time-consuming, expensive, and have low throughput.      
The use of diverse genetic toxicology assays over the past two decades has provided a database of several hundred chemicals that are positive for DNA damage and/or point mutations. However, we still have little mechanistic insight into how chemicals might cause or influence the larger mutations that are the hallmark of human clinical tumors, and this remains an important goal. Vogelstein and co-workers have described a genetic model for colon cancer that is likely to be a relevant genetic paradigm for most cancers. According to this model, it requires four to seven distinct genetic changes in separate genes to progress from a normal colonic epithelium to a malignant and metastatic colon cancer. Although there appears to be a favored order to this process, these events do not strictly require a specific order of occurrence, but rather it is an accumulation of these genetic changes that results in cancer. Interestingly, the majority of these genetic events, and some of the key steps in the process, involve the inactivation of tumor suppressor genes rather than the activation of oncogenes.      
The precise mechanism for this inactivation is not well understood but is clearly critical to elucidate. It will be important in future studies to examine the genetic mechanisms underlying this process, e.g., focusing on how genotoxic and nongenotoxic chemicals can lead to genetic and phenotypic changes in cells that promote the carcinogenic process in addition to or in the absence of point mutations. In particular, a critical need remains for the development of moderate to high throughput molecular in vivo assays that can provide information on large mutations and other genetic events important in human cancers.

IV. CHEMICAL CARCINOGENESIS

A. Chemical Carcinogenesis in Animals
Kenneway first demonstrated in the 1930s that a pure chemical could cause cancer in animals. Experimental evidence of tumorigenicity in animals has remained the gold standard for determining whether something is considered a potential human chemical carcinogen, particularly in the absence of strong human epidemiology data. The mouse or rat 2-year tumorigenicity study is currently the primary assay for experimentally assessing chemical carcinogenesis in animals for the purposes of human risk assessment. Generally, chemicals are administered at the maximum tolerated dose (MTD), and one-half the MTD for the lifetime of the animals, typically by gavage or in the diet, and tumor incidence is assessed in comparison to control animals. It is important to note that the background incidence of tumors in the control animals of these studies can be substantial. For example, in the two most widely used strains, B6C3F1 mice and F344 rats, the overall incidence of tumors can be as high as 50-60% and individual tissues can have a tumor incidence of 10-50%. Thus, a chemical must increase the tumor burden above this spontaneous background. Agents are considered positive that increase overall incidence, shift the overall tissue distribution or specific sites of tumors, or decrease the time to tumors.      
It is important to note that these long-term rodent tumor assays do not usually provide information about involved mechanisms. However, other short- and long-term experimental animal tumor assays can provide such information. The mouse two-stage skin cancer assay and similar assays developed in mouse lung and rat liver have provided a paradigm for assessing the basic mode of action of different chemicals. In the mouse two-stage skin cancer assay, four basic phases of carcinogenesis have been defined, initiation, early promotion (promotion I), late promotion (promotion II), and progression. Different chemicals can act during one or more of these phases and can be operationally defined by their mode of action in this model. For example, most genotoxic mutagens act as classic "initiating agents" in this assay, as a single application of one of these chemicals will usually result in increased skin tumors in the absence of any other treatment.      
A classic "tumor promoter," such as the phorbol ester, tetradecanoyl phorbol acetate (TPA), will usually not cause skin tumors by itself, but will substantially increase the tumor incidence of another agent if applied daily for several weeks or months following application of an initiating agent. Tumor promotion can experimentally be divided into an early and late phase in this system, the first of which appears to be reversible (tumor incidence decreases if chemical promotion is terminated) and the second of which is usually irreversible. The final phase of tumor progression is a period of time in which existing tumors progress from a nonmalignant to a fully malignant and metastatic phenotype. "Progressors" are chemicals that enhance this last phase of the process.      
"Complete carcinogens" are initiating agents that cause a substantial increase in fully malignant tumors in virtually all treatment animals, and therefore appear to be able to push cells through the entire process, without any additional help during the promotion or progression phases. A "cocarcinogen" in this assay is one that is not an initiating agent by itself, but enhances tumorigenesis when applied simultaneously with or prior to an initiating agent. Conversely, an "anticarcinogen" is one that can suppress the potency of another initiating chemical when given in combination. Likewise, other chemicals can be identified in this system that act as "chemopreventive agents" by being able to suppress the promotion and/or progression phases following initiation by another chemical. Much of our experimental evidence for the mechanism of action of chemicals as carcinogens is derived from these multistage tumor models.      
Another mechanism of action that has been described for some chemicals that induce tumors in long-term rodent assays appears to involve the induction of cell proliferation. Many agents that are only tumorigenic in animals at their MTD may be directly inducing cell proliferation or causing tissue damage leading to compensatory cell proliferation. In the case of chemicals that only act by this mode of action, it seems unlikely that this will also occur at the lower doses that humans are typically exposed to (often at thousands or millions of times lower than the MTD).       
Thus, there is some concern about determining the carcinogenic potential of some chemicals only at or near their MTD in these rodent assays. It is well known that cell proliferation is a strongly cocarcinogenic process. For example, in the rat liver, tumor induction by an initiator such as AFB1 can be substantially increased by giving the AFB1 in conjunction with a partial hepatectomy in which a large portion of the liver is surgically removed, leading to rapid proliferation of the liver to its fully restored size over the subsequent 72 h. Similarly, hepatitis B is strongly synergistic with AFB1 as a risk factor in human liver cancer, which is likely due to the continual proliferative response induced by this chronic viral infection.       
Conversely, it has been shown that mutagenesis from chemical and physical genotoxins can be strongly suppressed in mammalian cell culture by inhibiting cell division, presumably by providing an opportunity for cells to repair DNA damage before DNA replication, whereas the same treatment in cells that are rapidly dividing induces high levels of mutagenesis. Thus proliferation alone can be a strongly cocarcinogenic process, and agents or conditions that involve continual cell proliferation, such as chronic viral infections, other inflammatory responses, or chronic tissue injury, may strongly influence both background and chemically induced carcinogenesis. An additional mechanism of carcinogenesis that can be influenced by chemicals may involve disruption of cell-cell communication, which may be most important during the later phases of carcinogenesis when cells are acquiring a malignant and metastatic phenotype.      
Because of the large expense of long-term rodent tumor bioassays, many of the chemicals that have been screened to date were already strongly suspected of being carcinogenic in humans, based on human epidemiology studies or short-term assays such as those described in the preceding sections. The majority of these chemicals are genotoxic mutagen-carcinogens, as this class of agents is of considerable concern. However, more than a third of the chemicals that have been shown to be positive in long-term tumor assays are not overtly genotoxic or mutagenic. These nongenotoxic chemicals likely are acting through other epigenetic mechanisms as described earlier. There is concern about both false-positive and false-negative results in these assays. For example, there are agents that are positive in these long-term rodent assays but do not appear to be carcinogens in humans who are exposed to lower doses and different exposure conditions.      
Conversely, many chemicals that are known to be human carcinogens based on strong epidemiology evidence have been negative or equivocal in longterm rodent tumor assays (e.g., the carcinogenic metals arsenic and cadmium) or require nonphysiological exposures (e.g., intraperitoneal, intratracheal or implantation) to be positive (e.g., nickel and chromium). Thus, while animal tumorigenicity data are useful in assessing the carcinogenic potential and mechanism of action of many chemicals, there are limitations to these systems both for screening and for human risk assessment that are important to consider when evaluating individual agents.      
One of the more controversial uses of animal tumorigenicity data is in quantitative human risk assessment. This involves a mathematical extrapolation from the tumor incidence that is observed at high doses in rodents (often at only one or two doses at or near the MTD) to the potential cancer risk that might occur at the very low doses encountered by humans. This risk extrapolation from high to low doses may be required to extend over a dose range of three to six orders of magnitude. The conservative assumptions used by the U.S. Environmental Protection Agency and most other state and federal regulatory agencies that perform these risk assessments is that the dose- response will be linear over this range. This linear, low-dose extrapolation model was initially based on the so-called "one-hit" hypothesis, i.e., that one molecule of a chemical mutagen-carcinogen can theoretically interact with one critical target (i.e., DNA base) within a cell, thereby causing a change (mutation) that produces a cancer cell that can give rise to a tumor. We have little or no empirical data to support or reject this or other alternative models regarding the shape of the dose-response curve. However, as discussed earlier, it is likely that there are practical thresholds to the carcinogenic response to chemicals in vivo, and our current model of the multihit, multistage nature of the cancer process in humans suggests that the one-hit model is not a valid risk assessment model. Thus, it is likely that these risk assessments substantially overestimate risk at low doses. There has been a move toward the use of mechanistic information, when available, to do risk assessments of specific chemicals. For example, it appears that virtually all of the biological effects of TCDD can be attributed to its interaction with a cellular receptor, the Ah receptor, leading to activation of the receptor as a transcription factor. The Kd of TCDD for the Ah receptor in humans is known, and we would predict, based on basic receptor pharmacology principles, that there would be little or no biological effect of TCDD in humans at doses well below those that would be required to bind half or more of the available receptors. As additional mechanistic information such as this becomes available for individual chemicals, it will be increasingly possible to do meaningful, mechanistically based risk assessments that more accurately predict actual cancer risk to the human population in occupational and environmental exposure settings.

B. Evidence for Chemically Induced Cancers in Humans
There is substantial evidence from human epidemiology studies that chemicals can increase cancer in humans. Cigarette smoking alone is estimated to be responsible for approximately 30% of all cancer deaths, and this is clearly related to the hundreds of toxic chemicals in cigarette smoke, which collectively and individually have been shown to cause cancer in animals.      
Many of these chemicals are genotoxic mutagens, such as BaP, and others act as classic tumor promoters, comutagens, and cocarcinogens in the assays described earlier. Other major causes of cancer deaths include diet (35%), infection (10%), reproductive and sexual behavior (7%), and alcohol (3%) and, therefore, like tobacco use, are presumed to be largely preventable by identifying and instituting appropriate lifestyle changes. Human exposure to chemicals has been associated with cancer risk in environmental and occupational settings, but these exposures overall are estimated to contribute to only a small fraction of total cancer deaths. Approximately 4% of all cancer deaths have been attributed to occupational exposures, presumably involving exposures to individual chemicals and chemical mixtures in most cases. Similarly, exposure to pollution, industrial products, medicines or medical procedures, and food additives has been estimated to only account for an additional 2-4% of total cancer deaths. This is evident by examining data for death rates from major individual cancers over the past century. If occupational and environmental chemical exposures were responsible for a large and growing number of cancers, one might predict that cancer death rates should have increased from the 1950s onward, when there was a concomitant increase in manufacturing, use, and environmental release of industrial chemicals. However, with the exception of increases in male and female lung cancer from cigarette smoking, and a few other exceptions, most individual cancer death rates have been relevantly constant or have declined over the past century. Thus, the vast majority of cancers appear to be attributable to other causes. In particular, it is unlikely that exposure to trace levels of chemicals in the environment--in air, water, soil, as pesticide residues on food, and similar background exposures--contribute significantly to individual cancer risk, despite a public perception that this is the case.      
However, certain chemicals have been linked to human cancer in specific occupational or environmental settings. Hill first described an association between tobacco snuff and nasal polyps in 1761. Pott was the first to describe an occupational exposure of chimney sweeps to soot and their increased incidence of scrotal cancer. In the 1920s, the Germans described a strong association between occupational exposure to chromium dusts and an increase in lung and other respiratory tract cancers. This led to the development of industrial hygiene practices to reduce exposure as well as alterations in the manufacturing processes.      
Workers who began their employment in the chromium ore industries after the institution of these practices have demonstrated cancer risks similar to the general population. Occupational exposure to several other metals, particularly nickel, arsenic, and cadmium, has also been strongly associated with increased cancer risk, especially respiratory tract cancers. However, with the exception of arsenic (discussed later), there does not appear to be an increased cancer risk from exposure to the much lower environmental levels of these metals, particularly via noninhalation routes of exposure.      
Epidemiological studies of past occupational exposures have provided the best evidence for a link between individual chemicals and human cancer. Groups of workers in the late 19th and early 20th century who were exposed to 2-naphthylamine during its purification demonstrated a very high incidence of bladder cancer, some up to 100%. Similarly, occupational benzene exposure has been strongly associated with an increased incidence of acute myelogenous leukemia. Occupational exposure to vinyl chloride, an important chemical used in the manufacture of plastics, has been associated with an increased risk of liver angiosarcoma. This increased risk is fairly low even in exposed workers; however, the relative rarity of this particular cancer allowed its increased incidence to be detected. The relationship between increased human cancer risk and occupational exposure to other chemicals has been more controversial. This includes agents such as phenoxy herbicides and their contamination products (including TCDD), formaldehyde, and several organic solvents. The various conflicting studies published to date for some of these agents suggest that if there is an increased risk, it may be relatively low even in occupational settings.      
Occupational exposure to asbestos has been clearly associated with an increased risk of lung carcinoma and mesothelioma. There appears to be a strong synergy between asbestos and cigarette smoking for the former neoplasm. In fact, asbestos alone appears to be a weak or nonsignificant risk factor for this cancer in the absence of smoking, whereas asbestos exposure alone is clearly associated with mesothelioma risk and smoking appears to play a smaller or inconsequential role in this disease. This illustrates a common issue in both occupational and environmental exposures, namely, that humans are exposed to many agents simultaneously.      
Moreover, these agents can act additively, synergistically, or even antagonistically in contributing to overall cancer risk. Wood dust exposure, especially in the furniture manufacturing industries, is associated with an increased risk of lung cancer and other lung diseases and exhibits an interaction with smoking. The mechanistic basis for increased cancer risk from wood dust is not known, but may involve both a physical component (small, respirable fibers) and one or more specific chemical components.      
Use of certain drugs in medical therapies has also been associated with an increased risk of cancers. The most well-known example of this is the use of various DNA-damaging agents in cancer chemotherapy. By the nature of their mechanism of action, it is predicted that the majority of these drugs will have carcinogenic potential. The therapeutic goal when using these agents is to saturate the more rapidly dividing target tumor cells with sufficient DNA damage to arrest cell division and induce cell death. However, other normal dividing tissues are also receiving considerable DNA damage during systemic treatment with these agents, and surviving nontarget cells would be expected to have increased mutations that may contribute to carcinogenesis in these tissues, giving rise to so-called "second site" neoplasms. Ironically, because of the latency period of these tumors, this is usually only observed in long-term, disease-free survivors of the initial treatment. For example, in children with cancer, approximately 3-12 percent of survivors will develop second site cancers by 25 years after treatment. In adults, the risk of second site cancers is highly dependent on initial tumor type and mode of treatment. The greatest risk is for survivors of Hodgkin's disease, in which there is a 10-15% risk of second site tumors at 15 years after therapy. Not surprisingly, the risk of second site tumors is highest for those cancer patients receiving both radiation therapy and chemotherapy. A few other drugs in noncancer therapeutic treatments have also been shown to increase cancer risk. Use of diethylstilbestrol (DES) is a well-known but unusual example of hormonally induced cancer, where the adult female offspring of DES-treated mothers have an increased risk of clear cell vaginal adenocarcinoma. Other hormone therapies have also been associated with increased cancer risk, especially liver cancer risk associated with the use of estrogens in premenopausal women. Phenacatin use has been associated with an increased risk of renal carcinoma, and use of the immunosuppressive drug azothioprine has been associated with an increased risk of lymphoma, skin cancer, and Kaposi's sarcoma.      
Environmental exposure to certain chemicals is associated with an increased cancer risk. Exposure of certain populations to AFB1 through contaminated diet is clearly associated with an increased risk of liver cancer, although as mentioned earlier, the risk is synergistically greater with simultaneous infection by hepatitis B virus. AFB1 exposure is endemic in tropical areas of Africa, China, and South and Central America. Regulation of AFB1 levels in the food supply in the United States and the Western world has largely eliminated this risk to those populations. Environmental exposure to arsenic in drinking water in various regions throughout the world is also an important factor in the increased incidence of several cancers, as well as an increased risk of vascular disease and type 2 diabetes. Exposure to arsenic in drinking water occurs primarily as a result of leaching of arsenic from natural geological sources into well water. Arsenic is odorless, tasteless, and colorless and is therefore usually undetectable without chemical analysis. Certain areas of South America, southeast Asia, Europe, Asia, and North America contain appreciable arsenic levels in groundwater. Long-term exposure to arsenic through drinking water has been associated with elevated incidences of skin, lung, bladder, liver, kidney, and other cancers and can approach 10-fold above control levels in some areas.      
Chemical carcinogens that are associated with lifestyle include those in tobacco products, Betel nut chewing, and consumption of alcoholic beverages. As mentioned previously, dietary factors have been associated with up to 35% of all cancers. Many of these "dietary factors" may involve specific chemicals in the diet, apart from contaminants such as AFB1 or arsenic. These may include certain components of fats or fatty acids; various specific chemical inducers or inhibitors of cytochrome P450s, such as some of the natural plant chemicals found in dark green vegetables; heterocyclic amines and other food mutagens, especially those produced by cooking; polycyclic aromatic hydrocarbons produced by charring of meat and other foods; and nitrosamines found in certain foods.      
Conversely, certain plant chemicals may act as chemopreventive agents, such as some of those identified in certain fruits, vegetables, and teas. However, while it may be possible to optimize our diets to achieve the best balance between factors that enhance and sup- press carcinogenesis, we may be limited in the extent to which this is possible because virtually all foods appear to contain both types of agents. However, specific chemicals are being identified from the diet that may be useful pharmacologically in pure form as chemopreventive agents.

V. SUMMARY

Experimental carcinogenesis and human epidemiology studies have clearly identified specific chemicals that can act as human carcinogens. Certain chemicals have been associated with increased human cancer incidence in both occupational and environmental exposure settings. Experimental cell and animal systems have been useful in helping identify potential human carcinogens, as well as in determining their basic mechanisms of action. Many of the chemicals that are known carcinogens act by causing covalent DNA damage that can lead to mutations in critical oncogenes and tumor suppressor genes, which in turn can contribute to the carcinogenic process. However, many other chemicals initiate or promote carcinogenesis through non-genotoxic mechanisms that may also be important, including cell proliferation and disruption of cell-cell communication. Moreover, chemicals that are identified in experimental systems as being positive for a particular end point such as DNA damage, mutations, cell transformation, and animal tumor formation, while having the potential to be carcinogenic, may not represent a major cancer risk in humans due to differences between these end points and the actual cancer process or limitations in the assay systems that do not allow direct comparisons with human exposures. In this regard, it is difficult to do meaningful human risk assessment when one must extrapolate from high-dose animal exposures to the more typical low human exposures using theoretical models that are based on unproven hypotheses and assumptions. There is a clear need to understand better the mechanisms of action of nonovertly genotoxic carcinogens. It will be important to determine the mechanistic basis for the induction of large clastogenic mutations and aneuploidy events, which are a hallmark of human cancers, and to develop effective screening assays for assessing the potential of chemicals to induce these types of genetic events.

Joshua W. Hamilton
Dartmouth Medical School

See Also
ANIMAL MODELS FOR COLON CANCER CHEMOPREVENTION ; CARCINOGEN-DNA ADDUCTS ; CARCINOGENESIS: ROLE OF ACTIVE OXYGEN SPECIES ; CELLULAR RESPONSES TO DNA DAMAGE ; HORMONAL CARCINOGENESIS ; MOLECULAR EPIDEMIOLOGY AND CANCER RISK ; MISMATCH REPAIR: BIOCHEMISTRY AND GENETICS ; MULTISTAGE CARCINOGENESIS ; TOBACCO CARCINOGENESIS

GLOSSARY
carcinogen Any chemical or physical agent that increases cancer burden by increasing the incidence, altering the tissue distribution, increasing the malignant or metastatic potential, or decreasing the latency period of cancers in an individual or a population.

genotoxin Any chemical or physical agent that directly or indirectly causes DNA damage, i.e., a covalent chemical modification to a DNA molecule.

mutagen Any chemical or physical agent that directly or indirectly leads to a heritable alteration in the genetic sequence of bases in DNA.

Bibliography
Ames, B. N., and Gold, L. S. (1998). The prevention of cancer. Drug Metab. Rev. 30, 201-223.
EPA. (1996). "Proposed Guidelines for Carcinogen Risk Assessment." Office of Research and Development, U.S. Environmental Protection Agency, Washington, DC.
Garner, R. C. (1998). The role of DNA adducts in chemical carcinogenesis. Mutat. Res. 402, 67-75.
Hoffmann, G. R. (1996). Genetic toxicology. In "Casarett and Doull's Toxicology, the Basic Science of Poisons" (C. D. Klaassen, ed.), 5th Ed., pp. 269-300. McGraw-Hill, New York.
Hussain, S. P., and Harris, C. C. (2000). Molecular epidemiology and carcinogenesis: Endogenous and exogenous carcinogens. Mutat. Res. 462, 311-322.
Kinzler, K. W., and Vogelstein, B. (1996). Lessons from hereditary colorectal cancer. Cell 87, 159-170.
Kirsch-Volders, M., Aardema, M., and Elhajouji, A. (2000). Concepts of threshold in mutagenesis and carcinogenesis. Mutat. Res. 464, 3-11.
Klaunig, J. E., Kamendulis, L. M., and Xu, Y. (2000). Epigenetic mechanisms of chemical carcinogenesis. Hum. Expt. Toxicol. 19, 543-555.
Nickoloff, J. A., and Hoekstra, M. F. (1998). "DNA Damage and Repair," Vol. II. Humana Press, Totowa, NJ.
Pitot, H. C., III, and Dragan, Y. P. (1996). Chemical carcinogenesis. In "Casarett and Doull's Toxicology, the Basic Science of Poisons" (C. D. Klaassen, ed.), 5th Ed., pp. 201-268. McGraw-Hill, New York.

Carcinogenesis: Role of Reactive Oxygen and Nitrogen Species

Carcinogenesis is a complex multistage process often taking decades until malignancy appears. Conventionally, the carcinogenic process has been divided into three main stages: initiation, promotion, and progression. Initiation requires an irreversible genetic damage causing mutations in transcribed genes. Promotion consists of a potentially reversible oxidant-mediated conversion step followed by a clonal expansion of the initiated cells into benign tumors, which can progress to malignancy when they acquire many additional genetic changes. Those genetic changes include modification of DNA bases, insertions and deletions, genetic instability consisting of loss of heterozygosity, chromosomal translocations, and sister chromatid exchanges, activation of oncogenes, and suppression of tumor suppressor genes. Although cell initiation is a frequent occurrence, tumor promotion and progression usually require a long time because of all of the genetic changes that have to accumulate within the same few cells.      
It has been known for many years that antioxidants inhibit formation of tumors, even though they might not decrease DNA adducts, thought to be the initiating lesions. Hence, antioxidants are likely to interfere with the oxidant formation during promotion and/or progression stages of tumor development. Since then, numerous publications have shown that various types of reactive oxygen species (ROS) are generated during all stages of carcinogenesis, but especially during that long time required to take an initiated cell to a fully disseminated cancer. This long period between the initiation stage and cancer development provides a wide-open window of opportunity to interfere with and suppress the carcinogenic process. This can be accomplished more readily when processes of tumor promotion/progression and factors responsible for them are known. Some of those factors are discussed in this entry.

I. INTRODUCTION

The importance of oxidative stress to human cancer is underscored by the existence of cancer-prone syndromes characterized by the formation of high levels of oxidants and/or impairment(s) in their degradation or repair of the DNA damage they evoke. Those human congenital syndromes include Fanconi's anemia, xeroderma pigmentosum, Bloom's syndrome, ataxia telangectasia, Wilson's disease, and hemochromatosis, among others. The last two conditions also point to the contribution of an excess of bioavailable transition metal ions, such as copper and iron, to an overload of oxygen radicals in the liver and to the progression and outcome of those diseases. Hussain and colleagues clearly showed that livers of patients with Wilson's disease and hemochromatosis contain increased levels of inducible nitric oxide synthase (iNOS), as well as mutations in the p53 tumor suppressor gene, especially G:C to T:A transversions at codon 249. This finding is particularly important because hydrogen peroxide (H2O2)/iron treatment, as well as lipid peroxide-derived mutagenic aldehydes, such as 4-hydroxynonenal (HNE), cause the same type of mutations at codon 249. High levels of etheno compounds are present in liver DNA from patients having these two diseases. Interestingly, HNE causes formation of etheno derivatives in DNA and induces mutations at the 249 codon of the p53 gene in HNEtreated lymphoblastoid cells. Notably, not only congenital syndromes exhibit mutations in the p53 gene, but also noncancerous colon tissues from ulcerative colitis patients, a colorectal cancer-prone chronic inflammatory disease. Frequencies of p53 mutations at codons 247 and 248 were strongly correlated with the progression of the disease being appreciably higher ininflamed than in noninflamed tissues. These studies strengthen the idea that chronic inflammation contributes to cancer, and the results are consistent with the hypothesis that p53 mutations at 247-249 codons are due to chronic inflammation-associated oxidative stress.      
Oxidants are continuously formed and degraded during various normal cellular processes. ROS are required for the normal functioning of an organism and for the protection from invading bacteria. They are produced or utilized during mitochondrial respiration, metabolism of fats and xenobiotics, melanogenesis, and other peroxidatic reactions. Many types of cellular defenses keep these oxidants under control, but when they fail, extensive repair systems remove the damage from DNA and try to restore its integrity. ROS produced in small amounts can serve as second messengers in signal transduction. However, when ROS are formed at a time when they are not needed and/or in amounts exceeding antioxidant defenses and DNA repair capacity, then ROS can contribute to various diseases, with cancer being prominent among them. Tumor promoters and complete carcinogens induce chronic inflammation, ROS production, and oxidative DNA damage. Some of the oxidized DNA base derivatives are mutagenic, cytotoxic, and cross-linking agents. They can also cause hypomethylation, known to induce accelerated expression of some genes. This process might perhaps account for the ROS-mediated enhanced levels of various growth and transcription factors, and genes involved in antioxidant defenses.

II. OXIDANTS: REACTIVE OXYGEN SPECIES (ROS) AND REACTIVE NITROGEN SPECIES (RNS)

Extensive amounts of reactive oxygen and nitrogen species (ROS and RNS) are produced during inflammation, a normal bactericidal and tumoricidal process. However, chronic inflammation contributes to the long-lasting pathologic effects of the carcinogenic course of action, regardless of the type of cancer. ROS and RNS are generated by the activated "professional" phagocytic cells polymorphonuclear leukocytes (PMNs, neutrophils, granulocytes) and both circulating and resident macrophages. Target cells (i.e., epidermal keratinocytes or hepatocytes) can also form them in response to treatment with appropriate stimuli, such as tumor promoters or allergens. Although absolute amounts of ROS and RNS are much lower than those produced by stimulated phagocytes, these oxidants can set off a release and a synthesis of a cascade of various cytokines and chemotactic factors, which are instrumental to the initiation of inflammatory responses by phagocytic cells. ROS and RNS also mediate increased formation of growth and transcription factors needed for the accelerated growth of initiated cells.    
Major ROS include superoxide anion radicals (O2.-), their dismutation product hydrogen peroxide (H2O2), hypochlorous acid/hypochlorite (HOCl/ OCl-), singlet oxygen (1O2), and hydroxyl radicals (∙OH) (Table I). RNS are also produced during inflammation and contribute to its pathologic effects, with nitric oxide (∙NO), a product of iNOS, being a major player. The distinction between ROS and RNS is further blurred because of an avid interaction between O2∙- and ∙NO, which produces peroxynitrite (ONOO-, a potent oxidant) at a near diffusion rate. One of the most potent oxidants is ∙OH. This can be generated by ionizing radiation and by iron autooxidation, as well as through the reduction of H2O2 by transition metal ions, such as iron or copper, in the well-known Fenton or Haber-Weiss reactions. Although to a lesser extent, ∙OH can also be formed in the absence of transition metal ions, when there is an opportunity for a homolytic scission. Because ∙OH are the most potent oxidizing species, they cause damage only within a short distance from their generation.      
For this reason, H2O2, a neutral and not very reactive oxidant, is most likely responsible for reaching nuclear DNA and forming ∙OH-like species at sites harboring transition metal ions. Even H2O2 generated outside of the target cells (i.e., by PMNs or macrophages) can pass through membranes of the target cells and cause damage in the neighboring or distant cells. Other ROS cannot cross cell membranes because they are charged (i.e., O2∙-), lipophilic (1O2) reacting with membrane lipids, or react too rapidly (HOCl/OCl-) with amino groups within membranes forming much longer acting chloramines. Of the RNS, ∙NO can cross membranes. Although very reactive, ONOO- can diffuse within cells and might cross membranes of certain cells via anion transporters.

TABLE I Major ROS and RNS Produced during Inflammation

ROS or RNS Secondary ROS, RNS, and other reactive species
O2∙-   H2O2, ∙ OH, ONOO-
H2O2   ∙OH, HOCl/OCl-
HOCl/OCl-   1O2, RNHCl, RNCl2, Cl2, NO2Cl
∙OH   Lipid peroxides
Lipid peroxides   Lipid hydroperoxides, aldehydes
∙NO   ONOO-, N2O3, NOx
ONOO-   1O2, NO2-, NO3-, NOx

III. DIRECT OXIDATION OF DNA BASES
Both ROS and RNS can induce a plethora of genetic changes. Much is already known about the formation of oxidatively modified bases in cellular DNA. Figure 1 shows structures of oxidized DNA bases selected from well over 30 that are produced. The most extensively studied DNA base derivatives are thymidine glycol (dTG), 5-hydroxymethyl-2'-deoxyuridine (HMdU), and 5-hydroxy-2'-deoxyuridine (5-HdU), 5-hydroxy- 2'-deoxycytidine (5-HdC), and 5-hydroxymethyl-2'- deoxycytidine (HMdC), 8-hydroxy-2'-deoxyguanosine (8-OHdG), 8-hydroxy-2'-deoxyadenosine (8-OHdA), and their open-ring pyrimidine derivatives FaPy G and FaPy A. Many of these oxidized base derivatives are mutagenic (i.e., HMdU, 5-HdC, and 8-OHdG), whereas others block DNA replication (i.e., dTG). These oxidized bases can be formed directly by ROS, such as a site-specific attack by ∙OH (or ∙OH-like species), 1O2 (i.e., 8-OHdG), or peroxyl radicals (HMdU). HOCl/OCl- can oxidize DNA bases through the evolution of 1O2 in the presence of excess H2O2 and chlorinate cytosine and adenine residues in DNA and tyrosines in proteins. ONOO- oxidizes the guanine moiety in DNA to 8-OHdG but it reacts even more readily with 8-OHdG, which results in oxazolone derivative, cyanuric acid, and oxaluric acid as the end products. Hence, 8-OHdG might not be the best marker of oxidative stress and oxidative DNA damage if there is a possibility for ONOO- formation.


FIGURE 1 Selection of oxidized DNA nucleosides formed in DNA of humans, animals, and/or mammalian cells grown in culture. FAPYdG, formamido-aminopyrimidine product of opening of the imidazol ring of dG.

IV. INDIRECT ROS-MEDIATED MODIFICATION OF DNA BASES

In addition to the direct attack of oxidants on DNA bases, they can also be indirectly modified through interactions with electrophilic species produced by a ROS attack on other molecules or during ROS formation by such molecules.

A. Etheno-Base Derivatives Formed in DNA due to Lipid Peroxidation
ROS attacking unsaturated lipids cause the formation of lipid peroxides, which upon decomposition release aldehydes, such as malondialdehyde (MDA) and HNE. Both MDA and HNE are mutagenic and capable of binding to the amino group-containing dG, dA, and/or dC to form etheno derivatives, which contribute an additional ring to purines and pyrimidines (Fig. 2). Etheno compounds are present in DNA isolated from various types of tumors or from nontumorous sections of tumor-bearing animals and humans attesting to the existence of ROS and lipid peroxides in those tumors or tissues.


FIGURE 2 Selected products of interactions between aldehydes derived from lipid hydroperoxides (i.e., malondialdehyde or 4-hydroxynonenal) and amino group-containing 2'-deoxyribonucleosides in DNA of humans and animals.

B. Quinone-DNA Adducts Formed Due to ROS
An important example of processes leading to reactions of quinones with DNA is the redox cycling of the quinone-semiquinone couple derived from catecholestrogens. These evolve during estrogen metabolism by cytochrome P450 1A1 (CYP 1A1). Interestingly, lipid hydroperoxides acting as cofactors for peroxidatic oxidation also contribute to the formation of catecholestrogens (Fig. 3). Oxidation of semiquinone to quinone by molecular oxygen produces O2 ∙- and H2O2 and, in the presence of iron or copper, forms ∙OH, which can oxidize unsaturated lipids to lipid peroxides or, if formed in situ, cause oxidation of DNA bases. Thus, during metabolic estrogen transformation, several types of DNA modification should be anticipated: oxidized bases, etheno products derived from aldehydes released by lipid hydroperoxides, and quinone adducts. It is no wonder that estrogens have been implicated as contributors to hormonal cancers in humans.      
Another example of the importance of quinones in carcinogenesis and ROS generated during metabolism is benzo[a]pyrene (BP), a polycyclic aromatic hydrocarbon (PAH) and a major pollutant. Long and co-workers showed that up to 50% of female knockout mice lacking NAD(P)H:quinone oxidoreductase (NQO1, enzyme reducing quinones by two electrons to the nonredox-cycling hydroquinones) develop tumors when initiated with BP and promoted with phorbol ester 12-O-tetradecanoylphorbol-13-acetate (TPA) by the end of the experiment at 31 weeks, whereas the wild-type mice had none. Some (25%) of these tumors in NQO1(-/-) mice progressed to malignancy. BP is metabolically activated by CYP 1A1 and related enzymes to a variety of oxygenated derivatives, including three BP quinines (Fig. 4), which in the presence of P450 reductase form semiquinones. Hence, again quinone-semiquinone couples can redox cycle and generate ROS. These results also underscore the important role that antioxidant enzymes, such as NQO1, play in carcinogenesis.


FIGURE 3 Redox cycling of quinone-semiquinone couples leading to the formation of three general types of modified bases in DNA: quinone-DNA adducts, oxidized bases, and etheno compounds, which are products of lipid hydroperoxide (LPH)-derived aldehydes' reactions with amino group-containing DNA bases. This redox cycling is illustrated by the metabolism of estradiol, a physiological hormone, which is suspected as a contributor to human cancers, such as those of breast, endometrium, and prostate. It also applies to other carcinogens, which are metabolized to quinones and semiquinones that can redox cycle. Adapted from JNCI Monograph No. 27, p. 86, 2000, published by the Oxford Univ. Press, with permission from the publisher.

C. Effects of Inflammation-Derived ROS
on Metabolism of Exogenous Carcinogens Stansbury and colleagues showed that HOCl/OCl- (a powerful oxidant generated during inflammation by H2O2-mediated oxidation of chloride ions catalyzed by myeloperoxidase, an enzyme released by PMNs during activation) oxidizes BP 7,8-dihydrodiol to pyrene dialdehyde (Fig. 4), a novel BP-derived product that binds to DNA. These results point to ROS generated during inflammation as having a potential to change PAH metabolism and form DNAbinding species different from diol epoxides or quinones (Fig. 4). Oxidation and activation of polyaromatic amines by HOCl/OCl- have been known for many years. These could be some of the reasons why chronic inflammation contributes to carcinogenesis.


FIGURE 4 Metabolism of benzo[a]pyrene (BP, a carcinogenic polycyclic aromatic hydrocarbon) to metabolites that can form adducts with DNA bases and depurinating adducts, as well as contribute to the oxidation of bases in DNA through the redox cycling of quinone-semiquinone couples. The newly described pyrene dialdehyde potentially can form mono-adducts with DNA bases, as well as form intra- and interstrand DNA cross-links and DNA-protein cross-links. Pyrene dialdehyde is produced by oxidation of BP 7,8-dihydrodiol with HOCl/OCl- generated by activated PMNs during inflammation.

D. Oxidative Deamination and Nitration of DNA Bases
Although ∙NO has many physiological functions, when it is generated during chronic inflammation, it can oxidatively deaminate 5-methylcytosine, cytosine, guanine, and adenine, leading to the formation of thymine, uracil, xanthine, and hypoxanthine moieties in DNA. If not repaired, deaminated bases can mispair during DNA replication and cause mutations. Moreover, the nitrosating potential of ∙NO contributes to the formation of carcinogenic nitrosamines.     
In the presence of O2∙-, ∙NO produces a powerful oxidant, ONOO-, which oxidizes dG and 8-OHdG and also has nitrating properties evident from the formation of 8-nitroguanine in DNA and 3- nitrotyrosine in proteins. Upregulated iNOS and 3-nitrotyrosine have been found in tumor tissues and, in some cases, such as in metastatic melanoma, were correlated with poor survival of the patients, according to Ekmekcioglu and co-workers.

V. CARCINOGEN-MEDIATED INFLAMMATORY AND OXIDATIVE RESPONSES

As already mentioned, carcinogens, such as PAHs, can be metabolized by ROS-mediated pathways and they also contribute to ROS formation. Virtually all types of complete carcinogens and tumor promoters tested cause oxidative DNA base damage in vivo and, when analyzed, H2O2 is also evident in the target tissues. Carcinogens have been known to evoke inflammation thought to be necessary for tumorigenesis. An important question arises as to the mechanism(s) by which inflammatory responses are induced. Tumor promoters, such as TPA, induce inflammation very rapidly through protein kinase C, which provides a signal for a plethora of responses. A prominent response among them is a rapid release of interleukin (IL)-1α from suprabasal keratinocytes, which initiates a cascade of other cytokines and chemotactic factors synthesis, including IL-1α, tumor necrosis factor (TNF)-α, IL-8, granulocyte/macrophage-colony stimulating factor, and others. Frenkel and colleagues and Li and co-workers showed that similar to TPA, 7,12- dimethylbenz[a]anthracene (DMBA, a carcinogenic PAH) also induces IL-1α and TNF-α release and causes mRNA upregulation, which leads to further cytokine production in mouse skin. DMBA-induced IL-1α was responsible for PMN infiltration into mouse skin, as >65% of that infiltration was inhibited by preinjection of mice with anti-IL-1α antibody (Ab).      
Interestingly, anti-IL-1α Ab did not have appreciable effects on the incidence of DMBA-induced papillomas or carcinomas, but very potently inhibited the carcinoma volume, whereas anti-TNF-α Ab suppressed the incidence and volume of benign tumors but not carcinomas. DMBA mediated a substantial increase in HMdU and 8-OHdG in mouse skin, levels of which declined over time but remained elevated after tumors appeared. It will be important to establish whether DMBA-induced oxidative DNA base damage is modulated by anti-IL-1α and/or anti- TNF-α Ab and whether these two Abs have comparable or different effects on that DNA base damage.      
Upregulation of proinflammatory cytokines causes infiltration of phagocytic cells into the affected area. Circulating phagocytes are primed by IL-1α, whereas infiltrating cells may be further primed by TNF-α, which increases ROS production when those cells are activated by an appropriate stimulus. It is not yet known which of the DMBA metabolites is responsible for the start of this inflammatory process and whether the same or a different metabolite is needed for PMN activation. However, H2O2, HMdU, and 8- OHdG formation in DMBA-treated mouse skin attests to that activation. It appears then that immune responses to carcinogen treatment are very important in the determination of which pathway predominates and culminates in a preferential formation of benign and/or malignant tumors. Elevated levels of inflammatory cytokines have been found in many types of human cancer, including ovarian, lung, skin, and liver, as well as in many chronic inflammatory conditions, which are known cancer risk factors.

VI. FACTORS CONTRIBUTING TO CYTOKINE UPREGULATION

Upregulation of inflammatory cytokines' mRNAs requires binding of transcription factors, such as activator protein (AP)-1 and/or nuclear factor (NF)-κB, to the specific sites in the promoter or enhancer regions of the cytokine genes. Formation and activation of transcription factors required for cytokine upregulation depend on the redox status of the cell and its effect on those factors. Increased H2O2 production causes rapid phosphorylation of the inhibitor (IκB) and its dissociation from the NF-κB complex, which allows for NF-κB translocation from cytosol into the nucleus. Before binding to the appropriate consensus sequence giving a signal for the initiation of the transcription process, NF-κB must be reduced by the thioredoxin system, which itself depends on glutathione (GSH), a major cellular reductant in the nucleus. Often, more than one binding site is needed for the transcription to occur or more than one type of a transcription factor is required. AP-1, another transcription factor, is formed from c-fos and c-jun, two immediate early genes produced under conditions of oxidative stress. Interestingly, AP-1 binding also requires reduced sulfhydryl groups in the cysteine residues, which is accomplished under conditions of increased reductive power of GSH, NAD(P)H, and antioxidant enzymes. For this reason, the AP-1 pathway is often referred to as an antioxidant-regulated pathway. Interestingly, recent results show that the presence of oxidized bases at the recognition sequence can modulate binding of the transcription factors and, thus, interfere with gene expression, including those of cytokines. This is a very exciting area of investigation, which might show how oxidation of bases at specific sites of DNA could interfere directly with or enhance the carcinogenic process.

VII. OXIDIZED DNA BASES IN HUMAN CANCER

Elevated levels of oxidized purines were found in DNA isolated from human breast tumors, whereas increased levels of oxidized pyrimidines (HMdU) were present in white blood cell (WBC) DNA of women at high risk for breast cancer, as well as those diagnosed with breast cancer. Changes in the diet to lower the consumption of fat and increase that of fruits and vegetables caused a statistically significant decline in HMdU present in WBC DNA of the dieting high risk subjects. The formation of oxidized bases in the DNA of breast cancer patients is due to extensive oxidative stress that is evident in breast cancer patients, whose phagocytes produce more O2∙- and H2O2 than those of healthy controls (Ray et al.). At the same time, antioxidant enzymes superoxide dismutase (SOD) and GSH peroxidase are appreciably increased, whereas catalase is decreased. The high SOD content could rapidly dismutate O2∙- to H2O2, which cannot be completely degraded to water because of the decreased catalase activity. Thus, the accumulated H2O2 can migrate into other cells and their nuclei, where it generates ∙OH-like species in a site-specific manner that oxidize DNA bases. Results obtained by Gowen and colleagues further underscore the importance of oxidative DNA base damage to breast cancer. The authors discovered that the BRCA1 gene product is required for the transcription-coupled repair of oxidative DNA base damage. BRCA1 mutations confer susceptibility to human breast cancer, which provides strong evidence that oxidative DNA base damage contributes to breast cancer, at least in women carrying the mutated BRCA1 gene.      
In comparison to healthy controls, autoantibodies (aAb) that recognize HMdU were increased in sera of women diagnosed with breast cancer, thus providing another proof of oxidative stress and a biologic response to the consequences of that stress in cancer. More importantly, sera of apparently healthy women, who 1 to 6 years after that blood donation, were diagnosed with breast, colon, or rectal cancers, and sera of those at high risk for cancer contained elevated anti-HMdU aAbs. Thus, the enhanced presence of oxidized bases (i.e., HMdU) in WBC DNA, anti- HMdU aAb in serum, and possibly other antioxidized DNA base aAb can potentially serve as biomarkers of susceptibility to cancer. These biomarkers may allow preventive measures to be undertaken before an overt malignancy develops and, therefore, may also serve as efficacy markers of cancer-preventive agents.

VIII. CONCLUSIONS

Studies of factors contributing to the carcinogenic process, especially those modulated by oxidative stress, are by necessity correlative in nature. However, a discovery of more specific inhibitors and an increased use of molecular biology techniques have started to yield more direct proof of the involvement of oxidative stress and oxidative DNA base damage in carcinogenesis. Oxidative stress is characterized by increased ROS production, decreased antioxidant defenses, and impaired DNA repair capacity. All are evident during chronic inflammation evoked by a variety of agents or health conditions, many of which are known risk factors for different types of cancer. In summary, there is ever increasing scientific evidence that changes occurring during chronic inflammation, which include increased expression of inflammatory cytokines and downregulation of protective cytokines leading to further exacerbation of oxidative stress, play a direct role in the process of carcinogenesis. Oxidative modification of DNA bases present at specific sites of tumor suppressor genes, such as p53, which plays a pivotal role in cell cycling and apoptosis, causes a decline in their expression. Oxidized bases are also capable of changing binding patterns of transcription factors, and thus affect gene expression of many genes, which can lead to the disregulation of genetic machinery responsible for normal cell functioning.

Acknowledgments
Support by NCI (CA37858), NIA (AG14587), and NIEHS (ES00260) is gratefully acknowledged.

Krystyna Frenkel
New York University School of Medicine

See Also
CHEMICAL MUTAGENESIS AND CARCINOGENESIS ; HORMONAL CARCINOGENESIS ; HYPOXIA AND DRUG RESISTANCE ; TOBACCO CARCINOGENESIS

GLOSSARY
autoantibodies Antibodies recognizing antigen present in the same organism.

bioavailable Form that is not sequestered and is available for biochemical or chemical reactions within cells.

complete carcinogen A carcinogen that causes cell initiation and promotes cell transformation and growth to benign tumors, as well as mediates progression to malignancy.

chemokines Agents that even in minute amounts cause directed migration of cells.

cytokines Proteins produced by cells in response to inflammatory stimuli that affect the same or other cells in their growth, activation, priming, and death. Cytokines are an important part of the intercellular communication network.

growth factors Low molecular weight substances produced by cells that induce growth of the same (autocrine response) or other (paracrine response) cells.

inflammation A very complex process that is initiated by a release of cytokines and chemokines leading to the infiltration of phagocytic cells. Upon stimulation, phagocytes generate copious amounts of reactive oxygen and nitrogen species, proteases, as well as other enzymes and proteins. It can be manifested by fever, edema, hyperplasia, phagocytic infiltration, and oxidative stress.

oxidative stress Excessive production of and/or impaired removal of oxidants with a concomitant decrease in reducing capacity of cells.

oxidative modification of DNA bases Direct oxidation is characterized by a decrease in electron density. Such a modification can occur by oxidation of a double bond in a normal DNA base, of a methyl group, or by the addition of oxygen to a free pair of electrons. Indirect oxidative modification occurs when another molecule or macromolecule is oxidized and the product or by-product of that oxidation binds to DNA.

reactive oxygen species (ROS) and reactive nitrogen species (RNS) These are important players in normal physiology and signal transduction. They also participate in a plethora of damaging reactions, which lead to the disregulation of normal cell controls, thus contributing to various pathologies, including cancer.

transcription factors Factors needed for the transcription of genes into mRNA by binding to the specific recognition sequences in the promoter or enhancer regions of DNA.

tumor promoters Agents that support the development of initiated cells into transformed cells and tumors.

tumor suppressor genes Genes that prevent cell transformation and growth of tumors. Mutation of such genes usually abrogates their normal functioning and facilitates tumor development.

Bibliography
Ardestani, S. K., Inserra, P., Solkoff, D., and Watson, R. R. (1999). The role of cytokines and chemokines on tumor progression: A review. Cancer Detect. Prev. 23, 215-225.
Arrigo, A. P. (1999). Gene expression and the thiol redox state. Free Radic. Biol. Med. 27, 936-944.
Bartsch, H. (2000). Studies on biomarkers in cancer etiology and prevention: A summary and challenge of 20 years of interdisciplinary research. Mutat. Res. 462, 255-279.
Bolton, J. L., Trush, M. A., Penning, T. M., Dryhurst, G., and Monks, T. J. (2000). Role of quinones in toxicology. Chem. Res. Toxicol. 13, 135-160.
Cavalieri, E., Frenkel, K., Rogan, E., Roy, D., and Liehr, J. G. (2000). Estrogens as endogenous genotoxic agents: DNA adducts and mutations. J. Natl. Cancer Inst. Monogr. 27, 75-93.
Ekmekcioglu, S., Ellerhorst, J., Smid, C. M., Prieto, V. G., Munsell, M., Buzaid, A. C., Grimm, E. A. (2000). Inducible nitric oxide synthase and nitrotyrosine in human metastatic melanoma tumors correlate with poor survival. Clin. Cancer Res. 6, 4768-4775.
Frenkel, K. (1997). Carcinogenesis: Role of active oxygen species. In "Encyclopedia of Cancer" (J. R. Bertino, ed.), pp. 233-245.
Academic Press, San Diego. Frenkel, K. (1992). Carcinogen-mediated oxidant formation and oxidative DNA damage. Pharmacol. Ther. 53, 127-166.
Frenkel, K., Karkoszka, J., Glassman, T., Dubin, N., Toniolo, P., Taioli, E., Mooney, L., and Kato, I. (1998). Serum autoantibodies recognizing 5-hydroxymethyl-2-deoxyuridine, an oxidized DNA base, as biomarkers of cancer risk in women. Cancer Epidemiol. Biomark. Prevent. 7, 49-57.
Frenkel, K., Wei, L., and Wei, H. (1995). 7,12-Dimethylbenz[ a]anthracene induces oxidative DNA modification in vivo. Free Radic. Biol. Med. 19, 373-380.
Ghosh, R., and Mitchell, D. L. (1999). Effect of oxidative DNA damage in promoter elements on transcription factor binding. Nucleic Acids Res. 27, 3213-3218.
Gowen, L. C., Avrutskaya, A. V., Latour, A. M., Koller, B. H., and Leadon, S. A. (1998). BRCA1 required for transcription- coupled repair of oxidative DNA damage. Science 281, 1009-1012.
Hussain, S. P., Raja, K., Amstad, P. A., Sawyer, M., Trudel, L. J., Wogan, G. N., Hofseth, L. J., Shields, P. G., Billiar, T. R., Trautwein, C., Hohler, T., Galle, P. R., Phillips, D. H., Markin, R. Marrogi, A. J. and Harris, C. C. (2000). Increased p53 mutation load in nontumorous human liver of wilson disease and hemochromatosis: oxyradical overload diseases. Proc. Natl. Acad. Sci. USA 97, 12770-12775.
Klein, C. B., Snow, E. T., and Frenkel, K. (1998). Molecular mechanisms in metal carcinogenesis: Role of oxidative stress. In "Molecular Biology of Free Radicals in Human Diseases" (O. I. Aruoma and B. Halliwell, eds.), pp. 79-137. OICA International, London.
Li, X., Li, C. and Frenkel, K. (2001). Differences in the effects of inflammatory cytokines interleukin (IL)-1α and tumor necrosis factor (TNF)-α on 7,12-dimethylbenz(a)anthracene (DMBA)-induced carcinogenesis. Proc. Am. Assoc Cancer Res. 42, 806.
Li, X., Eckard, J., Shah, R., Malluck, C. and Frenkel, K. (2002). Interleukin-1α up-regulation in vivo by a potent carcinogen 7,12-dimethylbenz(a)anthracene (DMBA) and control of DMBA-induced inflammatory responses. Cancer Res. 62, 417-423.
Long, D. J., Waikel, R. L., Wang, X. J., Perlaky, L., Roop, D. R., and Jaiswal, A. K. (2000). NAD(P)H:quinone oxidoreductase 1 deficiency increases susceptibility to benzo(a)-pyrene-induced mouse skin carcinogenesis. Cancer Res. 60, 5913-5915.
Lyons, C. R. (1995). The role of nitric oxide in inflammation. Adv. Immunol. 60, 323-371.
Marnett, L. J. (2000). Oxyradicals and DNA damage. Carcinogenesis 21, 361-370.
Neurath, M. F., Becker, C., and Barbulescu, K. (1998). Role of NF-kappaB in immune and inflammatory responses in the gut. Gut 43, 856-860.
Qian, S., and Buettner, G. (1999). Iron and dioxygen chemistry is an important route to initiation of biological free radical oxidations: An electron paramagnetic resonance spin trapping study. Free Radic. Biol. Med. 26, 1447-1456.
Ray, G., Batra, S., Shukla, N. K., Deo, S., Raina, V., Ashok, S. and Husain, S. A. (2000). Lipid peroxidation, free radical production and antioxidant status in breast cancer. Breast Cancer Res. Treat. 59, 163-170.
Stansbury, K. H., Noll, D. M., Groopman, J. D., and Trush, M. A. (2000). Enzyme-mediated dialdehyde formation: An alternative pathway for benzo[a]pyrene 7,8-dihydrodiol bioactivation. Chem. Res. Toxicol. 13, 1174-1180.