University of Southern California Dissertations and Theses

The role of genetic variation in insulin -like growth factors in prostate cancer risk: The multiethnic cohort

(USC Thesis Other)

The role of genetic variation in insulin -like growth factors in prostate cancer risk: The multiethnic cohort

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content THE ROLE OF GENETIC VARIATION IN INSULIN-LIKE GROWTH FACTORS IN PROSTATE CANCER RISK: THE MULTIETHNIC COHORT by Iona Cheng A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (EPIDEMIOLOGY) December 2005 Copyright 2005 Iona Cheng UMI Number: 3219809 3219809 2006 UMI Microform Copyright All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, MI 48106-1346 by ProQuest Information and Learning Company. ii DEDICATION This dissertation is dedicated to my family. iii ACKNOWLEDGEMENTS I am indebted to the participants of the Multiethnic Cohort, who have contributed to a better understanding of the genetic contributions to cancer susceptibility. I thank my committee members, Brian Henderson, Matthew Freedman, Malcolm Pike, Daniel Stram, Anna Wu, and Juergen Reichardt, for their mentorship and expertise. I thank my colleagues in the USC Department of Preventive Medicine and the Broad Institute at Harvard/MIT for their guidance and support. Finally, thank you to my family and friends for their tremendous patience and encouragement. This work was supported by National Cancer Institute grants CA 63464 and CA 54281. iv TABLE OF CONTENTS DEDICATION………………………………………………................... ii ACKNOWLEGEMENTS………………………………………............. iii LIST OF TABLES………………………………………………............. v LIST OF FIGURES………………………………………………….….. vi ABSTRACT..............................................................................................vii INTRODUCTION TO THE DISSERTATION..........................................1 PART I. LITERATURE REVIEW Section 1. IGF physiology.......................................................................11 Section 2. IGFs and IGFBPs in the circulation........................................22 Section 3. IGF system and prostate cancer..............................................32 a. Experimental studies………….....................................32 b. Epidemiologic studies………………...........................35 c. Molecular epidemiologic studies...................................46 Section 4. IGF1, IGFBP1, and IGFBP3 gene and protein organization…………………………………………...…...…51 Section 5. Haplotype-based approach......................................................53 PART II. DATA ANALYSES Section 1. Association study of common genetic variation in IGF1 and prostate cancer risk in the Multiethnic Cohort.........................58 Section 2. Haplotype analysis of IGFBP1 and IGFBP3 in relation to prostate cancer risk: the Multiethnic Cohort…………………85 Section 3. Association study of common genetic variation in IGF1, height, and circulating IGF-I levels: The Multiethnic Cohort……………………………………………………….105 v PART III. GRANT PROPOSAL Section 1. A Comprehensive genomic approach to characterize the role of IGF Receptor genes in relation to breast cancer risk: The Multiethnic Cohort................................................133 PART IV. CONCLUSION.....................................................................163 ALPHABETIZED BIBLIOGRAPHY....................................................169 vi LIST OF TABLES 1. Prospective studies of IGF-I and IGFBP3 levels and prostate cancer…………………………………………………………….36 2. IGF-I and IGFBP3 levels and prostate cancer risk………………37 3. IGF-I and IGFBP3 levels and advanced prostate cancer risk……39 4. IGF1 (CA) n and prostate cancer risk.............................................47 5. IGFBP3 A-202C and prostate cancer risk.....................................50 6. IGF1, IGFBP1, and IGFBP3 SNP characteristics.........................53 7. Sixty-four SNPs utilized for IGF1 genetic analysis.......................66 8. Common haplotypes in blocks 1-4 of IGF1 estimated by tSNPs in the multiethnic panel..................................................................72 9. Study characteristics of prostate cancer cases and controls in the Multiethnic Cohort Study...............................................................73 10. Associations between common haplotypes in block 1-4 of IGF1 and prostate cancer risk..................................................................74 11. Odds ratios and 95% CI of prostate cancer associated IGF1 haplotypes by ethnic group.............................................................76 12. Odds ratios and 95% CI of prostate cancer IGF1 snps……….…..78 13. Odds ratios and 95% CI of prostate cancer associated IGF1 snps by ethnic group…………..……………………………………......80 14. IGFBP1 and IGFBP3 SNPs utilized for genetic analysis….……..96 15. Common haplotypes in blocks 1-3 of IGFBP1 and IGFBP3 estimated by tSNPs in the multiethnic panel...................................98 16. Odds ratios and 95% CI of prostate cancer for IGFBP1 and IGFBP3 missense snps and IGFBP3 A-202C…….………...........99 vii 17. Ass ociation between common haplotypes in blocks 1 -3 of IGFBP1 and IGFBP3 and prostate cancer risk...........................100 18. Association between common haplotypes in blocks 1-3 of IGFBP1 and IGFBP3 and prostate cancer risk...........................101 19. Study characteristics for height analysis……………………….115 20. Association between IGF1 common haplotypes and height: short versus tall………………….…………...………………...118 21. Odds Ratios and 95% CIs for IGF1 common haplotypes associated with height by sex and ethnic group..........................119 22. Odds Ratios and 95% CIs for IGF1 SNPs associated with height ……………………………………………………...…...121 23. Odds Ratios and 95% CIs for IGF1 SNPs associated with height by sex and ethnic group……………………....…….…...123 24. Age, sex, and ethnicity adjusted geometric mean height by IGF1 SNPs among entire MEC study………………..…….......124 25. Association between prostate cancer and height…..…...….…...128 26. Grant proposal: timeline……………………………..…....……159 viii LIST OF FIGURES 1. IGF1 on chromosome 12q22-q23………………………………..51 2. IGFBP1 and IGFBP3 loci on chromosome 7p13-p12…………...53 3. IGF1 locus spanning 156 kb on chromosome 12 ………………..71 4. Haplotype structure of blocks 1-4 of IGF1 using tSNPs in all ethnic groups combined……………….……………………….…76 5. –Log P value for trend of the association between IGF1 SNPs and prostate cancer risk…………………………………………..77 6. IGFBP1 and IGFBP3 loci spanning 71 kb on chromosome 7.…..92 7. –Log P value for trend of the association between 29 IGF1 tSNPs and height ………..…….………………………………..120 8. Geometric mean IGF-I levels (ng/ml) and 95% CIs for IGF1 blocks 1-4 ………………………………………………………126 ix ABSTRACT Prostate cancer is the most common cancer diagnosed among men in the United States. The mortality rate of prostate cancer is second only to lung cancer as the most frequent cause of cancer deaths. The disease displays striking racial/ethnic variation in risk as it disproportionately affects African-American men. The only established risk factors for prostate cancer are age, ethnicity, and family history. Multiple lines of evidence from population and experimental studies provide strong support for the role of the Insulin-like Growth Factor (IGF) system in prostate cancer tumorigenesis. IGFs serve as regulatory peptides involved in cell proliferation, differentiation, and apoptosis. To search for genetic variants that impact prostate cancer susceptibility, we examined three candidate genes in the IGF system, IGF1, IGFBP1, and IGFBP3, for their association with prostate cancer among African-American, Native Hawaiian, Japanese, Latino, and White men from the Multiethnic Cohort Study. We utilized a comprehensive genetic strategy whereby we a) sequenced the exons of IGF1, IGFBP1, and IGFBP3 in 95 advanced prostate cancer cases to identify any new missense variants, b) thoroughly characterized the haplotype structure of each locus, and c) tested the association between IGF1, IGFBP1, and IGFBP3 haplotypes and genotypes and prostate cancer risk among 2,320 prostate cancer cases and 2,290 controls. To extend our investigation of IGF1, we examined the relationship between inherited differences in IGF1, circulating levels of IGF-I, and height, an indicator of IGF-I x bioactivity. We found no new missense SNPs in IGF1, IGFBP1, and IGFBP3. We identified four haplotype blocks for IGF1 and three haplotype blocks for the shared IGFBP1/IGFBP3 locus. We identified several IGF1 haplotypes associated with prostate cancer risk and two perfectly correlated SNPs rs7978742 (P trend = .002) and rs7965399 (P trend = .002) that were significantly associated with prostate cancer risk and could account for the haplotype findings. Common genetic variation in IGF1 was also significantly associated with height such that permutation testing revealed a gene-wise empirical P value < 0.01 indicating that a similar association between IGF1 and height would be observed by chance less than 1% of the time. We found no association with either IGFBP1/IGFBP3 haplotypes or genotypes and prostate cancer risk. Our studies indicate that inherited differences in IGF1 influence prostate cancer susceptibility and also variation in height. Future work will examine whether genetic variation in the IGF receptors, IGF1R and IGF2R, which mediate IGF bioactivity, influence the risk of breast cancer. 1 INTRODUCTION TO THE DISSERTATION This dissertation is written in accordance with the requirements of the Doctor of Philosophy (Ph.D.) degree in Epidemiology of the Department of Preventive Medicine at the University of Southern California Keck School of Medicine. This dissertation examines the relationship between common genetic variation in candidate genes of the insulin-like growth factor (IGF) family and prostate cancer risk. The dissertation is composed of three chapters: (I) a literature review of IGF physiology and epidemiologic studies addressing these topics; (II) a series of three manuscripts investigating these research topics; and (III) a grant proposal for future work. Prostate cancer is a common malignancy among men in the United States. It accounts for 33% of all newly diagnosed male malignancies and is the leading cancer site among men. In 2005, approximately 232,090 men will be diagnosed with prostate cancer and 30, 350 men will die of this disease (Society 2005). The annual mortality rate of prostate cancer is second only to lung cancer, accounting for 10% of all cancer deaths in males (Society 2005). It is estimated that 1 in 6 men in the United States will be diagnosed with prostate cancer during their lifetime (Society 2005). Prostate cancer is extremely rare before the age of 40, but its rate of increase with age is greater than with any other cancer (Cook, Doll et al. 1969). The incidence rate of prostate cancer increases approximately to the tenth or 2 eleventh power of age in men aged 50 years or older (Cook, Doll et al. 1969). The probability of developing prostate cancer among men less than 40 years of age is 0.01% (Society 2004). This increases to 2.3% in men ages 40 to 59 and 14.2% in men ages 60 to 79 (Society 2004). This steep slope of age-related incidence suggests that prostate cancer has a longer developmental period and a greater number of stages or mutational events than other common epithelial cancers (Cook, Doll et al. 1969; Stanford, Stephenson et al. 1999). The incidence rates of prostate cancer vary dramatically between racial/ethnic groups. African-American men have the highest incidence rates in the world. In the United States from 1998 to 2002, the age-adjusted incidence rate (standardized to 1970 U.S. Standard Population) was 62% higher among African- Americans (272.0 per 100,000) than Whites (167.7), while the incidence rates among Latinos (141.9) and Asians/Pacific Islanders (101.4) were lower, 15% and 40%, respectively, in comparison to Whites (Ries LAG 2005). Moreover, African-Americans have a higher mortality rate of prostate cancer (68.1 per 100,000) than Whites (27.9), Latinos (23.0), and Asian/Pacific Islanders (12.1) (Ries LAG 2005). The higher mortality rate among African-Americans has been attributed to their more frequent presentation with advanced-stage disease (Morton 1994). It has been suggested that lower utilization of screening services contributes to a delay in the detection of early-stage disease among African-American men. However, several studies have reported that factors related to health care use and 3 access cannot completely explain the higher proportion of advanced disease among African-Americans (Mettlin, Murphy et al. 1997; Delfino, Ferrini et al. 1998; Hoffman, Gilliland et al. 2001; Oakley-Girvan, Kolonel et al. 2003). In a large cohort study examining U.S. Census data for 23,334 African-American and White men, socio-economic factors failed to explain the higher rate of prostate caner deaths in African-Americans compared to Whites (Robbins, Whittemore et al. 2000). Furthermore, the racial/ethnic variation in prostate cancer rates was evident prior to the dramatic increase in incidence during the late 1980’s and early 1990’s due to prostate-specific antigen (PSA) testing. In 1984, the age-adjusted incidence rate of prostate cancer among African-Americans (178.9 per 100,000) was 1.6 times greater than among U.S. Whites (109.9 per 100,000) (Ries, Eisner et al. 2003). Recent studies of African countries using improved cancer registry data report prostate cancer as a common disease despite the absence of screening programs (Ogunbiyi and Shittu 1999; Echimane, Ahnoux et al. 2000; Wabinga, Parkin et al. 2000). Of all male cancers, prostate cancer is the principal cancer site among men in Nigeria (11%) and the Ivory Coast (15.8%), and is second only to Kaposi’s sarcoma in Uganda (10.8%) (Ogunbiyi and Shittu 1999; Echimane, Ahnoux et al. 2000; Wabinga, Parkin et al. 2000). Among Afro-Caribbean men in Tobago, screen-detected prostate cancer was observed to be three-times higher than similarly screened Whites (Bunker, Patrick et al. 2002). In Kingston, Jamaica 4 (1984-1989), prostate cancer rates were approximately 20% higher than the rates of U.S. African-American men (Glover, Coffey et al. 1998). The lowest incidence rates of prostate cancer are seen in Asian countries such as China (3.0 per 100,000), Singapore (14.4 per 100,000), and Japan (12.7 per 100,000) (Parkin, Whelan et al. 2002). Studies have estimated a substantially smaller difference in prostate cancer rates between Japanese-American and native Japanese men in comparison to their published incidence rates if similar detection practices were followed (Shimizu, Ross et al. 1991; Shibata, Whittemore et al. 1997). With adjustment for detection differences, the incidence rate in Japan is estimated to be three to four times that of the reported rates, which is closer to the rate observed in Japanese-Americans (64.7 per 100,000) (Shibata, Whittemore et al. 1997). Presently, PSA is considered the most useful clinical marker for prostate cancer diagnosis (Polascik, Oesterling et al. 1999). However, PSA is limited by its lack of specificity in distinguishing between organ-specific and cancer-specific disease. The widespread adoption of PSA screening has lead to dramatic increases in the incidence rates of prostate cancer in the U.S., which has made international comparison of incidence rates difficult. A study of PSA levels among nonscreened low-risk Singapore-Chinese and high-risk U.S. men reve aled comparable PSA levels between the two groups (Cheng to be published Jul 2005). This suggests either the presence of a considerable amount of undiagnosed prostate cancer among the Singapore-Chinese, which are likely to be of the 5 nonaggressive type given their low mortality, or that PSA may a poor marker of disease in this population. Two large-scale randomized screening trials, the Prostate, Lung, Colorectal and Ovary (PLCO) cancer trial in the USA and the European Randomized Screening for Prostate Cancer (ERSPC) trial in Europe are currently evaluating the efficacy of PSA screening in reducing prostate cancer mortality. Results from these trials will be vital in determining the benefit of PSA population screening for prostate cancer. Family history is an established risk factor for prostate cancer. Men who have an affected brother or father have a two-fold increase in risk of developing the prostate cancer, and larger risk estimates are observed among men with multiple affected family members (Steinberg, Carter et al. 1990). An earlier age of onset has been found to be associated with a family history of prostate cancer among first-degree relatives (Bratt 2002). This hereditary influence of prostate cancer is estimated to account for 5% to 10% of all prostate cancer cases and ~40% of prostate cancer diagnoses in men less than 55 years of age (Carter, Beaty et al. 1992; Bratt 2002). Familial clustering and early age of onset of prostate cancer highlight a genetic susceptibility to the disease. Linkage studies have mapped seven prostate cancer susceptibility loci on chromosomes 1, 8, 17, 20, and X (Simard, Dumont et al. 2002). However, confirmatory studies are still needed and the contribution of highly penetrant susceptibility genes is estimated to account for only a small proportion of all prostate cancer cases (Carter, Beaty et al. 1992). 6 Migration studies suggest an environmental contribution to prostate cancer risk. Prostate cancer incidence rates in immigrants have been reported to approach the rates of the host countries (Kolonel, Hankin et al. 1986; Grulich, McCredie et al. 1995; Kolonel, Altshuler et al. 2004). These studies have suggested a Western lifestyle of high fat, meat, and dairy intake may be possible contributors to prostate cancer risk. However, in spite of the numerous studies of dietary factors and prostate cancer risk, only dietary fat has been consistently associated with prostate cancer risk (Fleshner, Bagnell et al. 2004). Variation in saturated fat has been reported to explain approximately 10% to 15% of the differences in prostate cancer rates between African-Americans, Whites, and Asian-Americans (Whittemore, Kolonel et al. 1995). No other environmental contributors to prostate cancer risk have been well-established. Weight and height have been investigated extensively in relation to risk of prostate cancer. Recently, two large cohort studies conducted in Norway and the U.S. (American Cancer Society) of 950,000 and 900,000 men respectively, reported that obesity as measured by body mass index (BMI) was positively associated with increased prostate cancer risk (Calle, Rodriguez et al. 2003; Engeland, Tretli et al. 2003). Height was also found to be associated with prostate cancer risk in these two large studies (Rodriguez, Patel et al. 2001; Engeland, Tretli et al. 2003). Collectively, these findings imply that weight may also be associated with prostate cancer risk given the relationship between weight, height, and BMI. 7 Androgens play an important role in prostate development and tumorigenesis. The growth and function of the prostate is regulated by the conversion of testosterone (T) to dihydrotestosterone (DHT) by the 5-alpha reductase enzyme. Both T and DHT have been shown to induce prostate cancer growth in animal models (Noble 1977). In addition, anti-androgen therapy is a well-accepted treatment method for advanced prostate cancer. Epidemiological studies, however, have not demonstrated a convincing association between circulating androgens and prostate cancer risk. This lack of an association has been attributed to methodological limitations related to accuracy of assessment of circulating androgens (Ntais, Polycarpou et al. 2003) and to the fact that circulating androgens may not accurately reflect androgen levels in the prostate. Insulin-like growth factor-I (IGF-I) is a potent mitogen that regulates cell proliferation, differentiation, and apoptosis. Prospective studies have demonstrated that men in the highest quartile of circulating IGF-I have an increased risk of prostate cancer compared to men in the lowest quartile of circulating IGF-I (OR meta = 1.22; 95% CI: 0.91-1.62) (Schaefer 1998; Harman, Metter et al. 2000; Lacey, Hsing et al. 2001; Chan, Stampfer et al. 2002; Woodson, Tangrea et al. 2003; Stattin, Rinaldi et al. 2004; Chen, Lewis et al. 2005). Such findings have prompted the hypothesis that higher circulating levels of IGF-I may relate to elevated IGF activity in the prostate and correspond to increasing epithelial turnover and opportunities for genetic alterations leading to cellular transformation. 8 The insulin-like growth factor (IGF) system is comprised of several molecules that form a highly regulated network: two ligands (IGF-I and IGF-II), two types of cell membrane receptors (IGF1R and IGF2R), six binding proteins (IGFBP-1 through IGFBP-6), and several binding proteases and IGFBP-related proteins. IGF-I and IGF-II mediate their biological actions through activation of the IGF1R (LeRoith and Roberts 1991). The IGF2R does not appear to have signaling activity but rather limits mitogenic effects by sequestering IGF-II (Oates, Schumaker et al. 1998). IGFBPs serve to regulate the bioavailability of IGFs in the circulation and tissues, while a number of proteases reduce the carrying capacity of these binding proteins through cleavage activity, which results in the release of free IGFs. For common complex disorders such as prostate cancer, it is believed that common genetic variants may be responsible for a majority of disease (common variant/common disease hypothesis) (Smith and Lusis 2002). However, to date, no single gene variant has been consistently associated with prostate cancer risk. While linkage and genome-wide association studies have been proposed for identifying potential variants associated with risk, the haplotype-based approach offers one powerful method to investigate the genetic contribution to disease. This approach utilizes a comprehensive strategy to survey common genetic variation in candidate genes that may contain disease-associated variants (Rioux, Daly et al. 2001; Haiman, Stram et al. 2003). An important advantage of this method is that the causal variant does not have to be identified and tested directly. Instead, this 9 method involves localizing physical regions of the chromosome that may harbor disease-related variants (Haiman, Stram et al. 2003). As accumulating evidence supports the involvement of the IGF system in prostate cancer, the genetic contribution of the IGF system to prostate cancer risk is of particular interest. This dissertation utilizes a haplotype-based approach to investigate how genetic variation specifically in IGF1, IGFBP1, and IGFBP3 relate to prostate cancer susceptibility. This dissertation also examines the association between circulating levels of IGF-I and also the association between height and IGF1 variants. Chapter I of this dissertation reviews the literature on the role of IGF1, IGFBP1, and IGFBP3 as prostate cancer susceptibility genes. Relevant topics include: (a) IGF physiology; (b) cross-sectional studies of the determinants of circulating IGFs and IGFBPs; (c) epidemiologic studies of the levels of IGFs and IGFBPs and risk of prostate cancer; and (d) association studies of genetic variation in the IGF family genes and prostate cancer risk. Chapter II consists of a series of three papers that examines the biological predisposition to prostate cancer in the Hawaii-Los Angeles Multiethnic Cohort Study (MEC). The MEC is a prospective study of over 215,000 participants with the purpose of evaluating environmental and genetic contributions to racial/ethnic differences in cancer risk. It is a collaborative study between the University of Hawaii and the University of Southern California (Principal Investigators: Drs. Laurence Kolonel and Brian Henderson). Recruitment to the study began in 1993 10 and was completed in 1996. Incident cancer cases are ascertained through linkage with the Surveillance, Epidemiology, and End Results (SEER) cancer registries in Hawaii and Los Angeles, as well as with the California State Cancer Registry. Identified men with prostate cancer are contacted by telephone and asked to provide a blood specimen. Similarly, a random sample of control men in the MEC, are selected to provide blood specimens. This study population considered here is comprised of the five major racial/ethnic groups in the MEC: African- Americans, Hawaiians, Japanese-Americans, Latinos, and Whites. This study sample will allow for evaluation of the genetic and phenotypic contribution of the IGF-family of genes to racial/ethnic differences in prostate cancer risk. The first and second papers of Chapter II utilize a haplotype-based approach to evaluate the association between prostate cancer risk and genetic variation in IGF1, IGFBP1, and IGFBP3. The third paper examines whether genetic variation in IGF1 influences differences in height and circulating levels of IGF-I. Chapter III is a grant proposal to study the contribution of common genetic variation in the IGF1R and IGF2R to breast cancer risk among African- Americans, Native Hawaiians, Japanese-Americans, Latinas, and Whites. In addition, this grant proposes to evaluate breast cancer risk respective to interactions between the common haplotypes of IGF1R and IGF2R and previously characterized haplotypes of IGF1, IGFBP1, and IGFBP3. 11 PART I. LITERATURE REVIEW Over half a century ago, in Salmon and Daughaday’s seminal study, it was observed that a circulating factor mediated the growth promoting effects of growth hormone (GH) (Salmon and Daughaday 1957). From this first work, the somatomedin hypothesis was developed to understand how factors secreted from the pituitary gland regulated somatic growth. Two decades later, insulin-like growth factor-I (IGF-I) and insulin-like growth factor-II (IGF-II) were purified (Posner, Guyda et al. 1978; Zapf, Schoenle et al. 1978). Since these early studies, it has become evident that the GH/IGF system plays a fundamental role in regulating somatic growth and recent work has investigated the possible influence of insulin-like growth factors on neoplastic states. This overview highlights the physiology of the IGF system and biologic and epidemiologic research investigating the role of insulin-like growth factors in relation to prostate cancer development. Section 1. IGF Physiology Insulin-like Growth Factors IGF-I and IGF-II are growth factor peptides that share 62% homology in their amino acid sequences, and are approximately 50% homologous to insulin (Stewart and Rotwein 1996; Yu and Rohan 2000). The majority of circulating IGFs are produced by the liver, while numerous tissues are able to synthesize locally IGF-I and IGF-II (Khandwala, McCutcheon et al. 2000). Hepatic 12 production of IGF-I is primarily regulated by pituitary stimulated-growth hormone, while synthesis of IGF-II is independent of GH regulation (Jones and Clemmons 1995). IGFs are unique peptides in their ability to serve as endocrine factors as well as tissue growth factors with paracrine and autocrine effects (Jones and Clemmons 1995). Numerous agents are involved with the expression of IGFs. Several hormones in addition to growth hormone are able to influence IGF levels such as adrenocorticotropic hormone (ACTH), estrogen, follicle-stimulating hormone (FSH), lutenizing hormone (LH), parathyroid hormone (PTH), progesterone, and testosterone (Rotwein 1999). Other growth factors also regulate expression such as epidermal growth factor (EGF), fibroblast growth factor (FGF), platelet- derived growth factor (PDGF), and transforming growth factor- (TGF) (Rotwein 1999). Dietary factors may influence circulating IGF-I levels, although their role in IGF-I production at the tissue level is unclear (Thissen, Ketelslegers et al. 1994; Naranjo, Yakar et al. 2002). Insulin-like Growth Factor Receptors The biological actions of IGF-I and IGF-II are mediated by their binding to the insulin -like growth factor 1 receptor (IGF1R) (Jones and Clemmons 1995). IGF-I has a two- to fifteen-fold higher affinity for IGF1R than IGF-II (D'Ercole 1996). IGF1R is composed of two chains that comprise the extracellular ligand binding domain and two chains that comprise the intracellular tyrosine kinase 13 domain and extracellular region (Massague and Czech 1982). Activation of the receptor by IGF-I and IGF-II directly phosphorylates various substrates such as insulin receptor substrate-1 (IRS-1) and src-homology 2/collagen-proteins (Shc), which triggers downstream signaling mechanisms such as the ras/MAPK pathway (through Grb-2/Sos) and the PI3-K/Akt pathway (through the p85 regulatory subunit) (Jones and Clemmons 1995). Activation of the Ras/MAPK leads to cellular growth, while PI3-K/Akt signaling is implicated in governing anti- apoptotic effects (Ahmad, Singh et al. 1999; Gooch, Van Den Berg et al. 1999). IGF1R signaling also influences cell motility and invasion through phosphorylation of FAK (a tyrosine kinase that transmits integrin-mediated- signals) and paxillin and CAS (proteins involved with cytoskeleton rearrangement) (Gray, Stenfeldt Mathiasen et al. 2003). A hybrid receptor resulting from dimerization of the IGF1R and the insulin receptor (IR) has been identified. These hybrids are comprised of two extracellular - and two transmembrane -subunits linked by disulfide bonds. These receptors retain high affinity for IGF-I but reduced affinity for insulin. The presence of two isoforms of the IR, IRA and IRB, further complicate the effects of these hybrid receptors given their differential binding affinities for IGF-II (LeRoith and Roberts 2003). The IRA isoform has a high affinity for IGF-II and mediates proliferative effects, while the IRB isoform mediates metabolic effects in response to insulin (Frasca, Pandini et al. 1999). Recently, IGF1R/IRA hybrids have been shown to bind insulin, IGF-I, and IGF-II, whereas the IGF1R/ I RB 14 hybrids do not bind insulin, and binds IGF-I with high affinity and IGF-II with low-affinity; this results in greater cell proliferation and migration in IGF1R/IRA hybrids than IGF1R/IRB (Pandini, Frasca et al. 2002). Insulin-like growth factor 2 receptor (IGF2R), also known as the Mannose-6-Phosphate (M6P) receptor, is a single-chain transmembrane receptor that binds IGF-II with a five hundred-fold higher affinity than IGF-I (Khandwala, McCutcheon et al. 2000). IGF2R also binds ligands containing mannose-6- phosphate molecules such as renin, proliferin, thyroglobluin, transforming growth factor 1 (TGF1), retinoic acid, and urokinase-type plasminogen activator receptor (uPAR) (Jones and Clemmons 1995; Kang, Li et al. 1997; Godar, Horejsi et al. 1999). Binding of IGF-II to IGF2R results in degradation of IGF-II by internalization and transport to lysosomes. By removing IGF-II from the extracellular environment and precluding its activation of the IGF1R, IGF2R is believed to reduce the biological effects of IGF-II (Oates, Schumaker et al. 1998; Braulke 1999; Hassan 2003). Because of this unique feature of the IGF2R, it is considered a potential tumor suppressor. Several studies report IGF2R signaling activity (Kornfeld 1992; Minniti, Kohn et al. 1992; Nishimoto 1993; Ikezu, Okamoto et al. 1995; Groskopf, Syu et al. 1997; Zhang, Tally et al. 1997; Ikushima, Munakata et al. 2000; Tsuruta, Eddy et al. 2000; McKinnon, Chakraborty et al. 2001). However, this remains controversial as the cytoplasmic domain of the IGF2R does not demonstrate any known enzymatic activity and studies have reported contrary evidence (Sakano, 15 Enjoh et al. 1991; Korner, Nurnberg et al. 1995; Ghosh, Dahms et al. 2003). Additionally, the mitogenic effects of IGF-II are mediated through interaction with the IGF1R and IGF1R/IRA (Frasca, Pandini et al. 1999). It is generally accepted that the IGF2R does not have transmembrane signaling and is not involved in mitogenesis; previous reports may be explained by activity of IGF1R or IGF1R/IRA hybrid activation (DaCosta, Schumaker et al. 2000). Regulation of IGF1R expression involves an important balance of stimulatory and inhibitory factors. Stimulating factors include growth factors (PDGF, FGF), hormones such as estrogens, oncogenes, and transcription factors (Sp1, Stat5b) (Rotwein 1999; Woelfle, Billiard et al. 2003). Elevated levels of IGFs and insulin promote reduced IGF1R expression that is believed to act through negative feedback signals (Yang, Hoeflich et al. 1996; Hernandez- Sanchez, Werner et al. 1997). Additional inhibitory factors include tumor suppressors proteins such as Wilms’ tumor protein (WT1) and p53 (Rotwein 1999). Regulation of IGF2R expression has not been studied extensively. Evidence suggests expression of IGF2R is developmentally regulated, as larger amounts of IGF2R are present in fetal than postnatal organs and soluble receptor levels are high during childhood, fall during adolescence, and remain constant throughout adulthood (Funk, Kessler et al. 1992; Costello, Baxter et al. 1999). Both IGFs and insulin have shown to increase IGF2R expression (Jones and Clemmons 1995). Estradiol has been reported to decrease IGF2R levels, and 16 recently platelet-derived growth factor (PDGF) has shown to increase IGF2R expression (Mathieu, Vignon et al. 1991; Weiner, Chen et al. 2000). The biological effects of IGFs include both acute anabolic actions on protein and carbohydrate metabolism and long-term effects on cell proliferation, differentiation, and apoptosis (Jones and Clemmons 1995). IGF-I is able to stimulate amino acid and glucose uptake as well as protein and glycogen synthesis (Jones and Clemmons 1995). A wide variety of cells demonstrate proliferation in response to IGF-I. This mitogenic effect is a result of increase DNA synthesis and cyclin D1 expression, which advances the cell cycle from Gap 1 (G1) to Synthesis (S) phase (Dufourny, Alblas et al. 1997; Minshall, Arkins et al. 1997). In addition to promoting cell proliferation, IGF-I is also able to inhibit cellular apoptosis. Through stimulating anti-apoptotic Bcl expression and suppressing pro-apoptotic Bax expression, IGF-I increases the relative amount of Bcl/Bax, which suppresses apoptosis initiation (Minshall, Arkins et al. 1997; Wang, Ma et al. 1998). IGFs also play an important role in fetal development and differentiation. Studies of IGF1, IGF2, IGF1R, and IGF2R knockout mice demonstrate the importance of the IGF system in prenatal and postnatal growth (Liu, Baker et al. 1993; Powell-Braxton, Hollingshead et al. 1993; Wang, Fung et al. 1994). IGF-I null mice display severe growth retardation and most die following birth (Powell- Braxton, Hollingshead et al. 1993). In a transgeneic mouse model that exclusively deleted the IGF1 gene in the liver, serum IGF-I levels decreased 75% while femur length and weight declined 6%, yet overall growth velocity was maintained 17 (Yakar, Liu et al. 1999). This suggests that endocrine production of IGF-I is essential for the production of circulating levels of IGF-I, and autorcine/paracrine production of IGF-I is important for postnatal growth. IGF-II activity appears to regulate embryonic and fetal stages of development (Schultz, Hahnel et al. 1993; Stewart and Rotwein 1996; Lighten, Hardy et al. 1997). Genomic imprinting of IGF-II gene has been observed in mice and humans (DeChiara, Robertson et al. 1991; Ohlsson, Nystrom et al. 1993), whereby paternal allele is commonly expressed and has been associated with somatic overgrowth (Weksberg, Smith et al. 2003). Insulin-like Growth Factor Binding Proteins Six IGFBPs (IGFBP-1 to IGFBP-6) have been identified. The majority of IGFBPs have a higher affinity for IGF-I than IGF-II with the exception of IGFBP- 6, which has a 100-fold higher affinity for IGF-II rather than IGF-I (Baxter and Saunders 1992; Baxter 1994). Various IGFBP post-translational modifications such as phosphorylation, glycosylation, and proteolysis may influence their effect on IGF bioactivity (Jones and Clemmons 1995). IGFBPs are produced ubiquitously in various tissues and bind most of the IGFs to form a binary complex, while limiting the amount of free IGFs (Wetterau, Moore et al. 1999). In the circulation, IGFBP-3 is the principal binding protein that maintains IGF-I levels in conjunction with the acid-labile subunit (ALS) (Baxter 1994). IGFBP-1, IGFBP-2, and IGFBP-4 are also present in the circulation and have the ability to 18 cross the endothelial barrier; these proteins are believed to be responsible for IGF transport from the circulation to peripheral tissues (Baxter, Hizuka et al. 1993; Jones and Clemmons 1995; Rajaram, Baylink et al. 1997). The regulatory mechanisms for IGFBPs involve a large array of factors some of which are tissue specific (Yu and Rohan 2000). Factors which have been found to be involved with regulation of IGFBPs include glucocorticoids, retinoic acid, cytokines (e.g. TNF, IL-1, IL-6), growth factors (e.g. FGF, EGF, IGFs, TGF-), and hormones (e.g. GH, PTH, FSH, estradiol) (Martin, Coverley et al. 1995; Oh, Muller et al. 1995; Olney, Wilson et al. 1995; Collett-Solberg and Cohen 1996; Kelley, Oh et al. 1996; Gabbitas and Canalis 1997; Chambery, de Galle et al. 1998; Demori, Balocco et al. 2004). Insulin is the principal regulator of decreasing IGFBP-1 expression, while GH is primarily involved with increasing IGFBP-3 levels (Hasegawa, Cohen et al. 1995; Gucev, Oh et al. 1997). IGFBPs have complex functions involved with either suppressing or enhancing the effects of IGF-I and IGF-II. The principal roles of IGFBPs are: (1) transportation of IGFs; (2) protection of IGFs from degradation; and (3) regulation of the interaction between IGFs and their receptors (IGFRs) (Yu and Rohan 2000). All IGFBPs are able to suppress IGF-mediated mitogenic effects by competitively binding to IGF-I and IGF-II and preventing their interactions with the IGF1R. IGFBPs are able to enhance IGF effects through steady and slow release of free IGFs, which in turn activate the receptors. This also protects IGF1Rs from downregulation in response to high IGF levels (Conover and Powell 19 1991). This activity has been observed in IGFBP-1, IGFBP-3, and IGFBP-5 (Grimberg and Cohen 2000). In addition to these IGF-dependent actions, IGFBPs have various IGF- independent actions that include inhibition and promotion of cell growth as well as induction of apoptosis (Ferry, Cerri et al. 1999). These IGF-independent effects are mediated through the interaction of IGFBPs with specific cell-surface receptors and molecules. IGFBP-1 binds to the 51-integrin receptor that has been shown to stimulate cell migration in Chinese hamster ovary cells (Jones, Gockerman et al. 1993). IGFBP-3 has been found to inhibit cell growth independent of IGF-I (Valentinis, Bhala et al. 1995; Zadeh and Binoux 1997). These IGF-independent effects of IGFBP-3 are mediated through cell surface proteins or receptors that have yet to be fully characterized. In breast cancer cells, IGFBP-3 has demonstrated specific, dose-dependent binding to cell surface proteins of 20, 26 and 50 kDa that inhibit cell growth independently of IGF-I stimulation (Oh, Muller et al. 1993). A putative receptor has been cloned that is associated with increased cell surface binding of IGFBP-3 and subsequent apoptosis (Ingermann, Kim et al. 2000). A recent study in breast cancer cells has identified IGFBP-3 independent activation of caspases, a family of proteins that transmit apoptotic signals (Kim, Ingermann et al. 2004). In prostate cancer cells, binding of IGFBP-3 to the nuclear retinoid X receptor- (RXR-) has been observed to influence transcriptional activity and mediated IGFBP-3 effects on apoptosis (Liu, Lee et al. 2000). IGFBP-5 binds to the extracellular matrix (ECM) 20 in fibroblasts with a decreased affinity for IGF -I and potentiates the effect of IGF- I on cell growth (Jones, Gockerman et al. 1993). IGFBP proteases may modulate activities of IGFBPs. Proteolysis of IGFBP by IGFBP proteases results in IGFBP fragments that typically show reduced affinity for IGFs. This leads to increased levels of unbound IGFs that can then activate the receptors (Grimberg and Cohen 2000). IGFBP proteases are categorized into three types: kallikrein-like serine proteases, cathepsins, and matrix metalloproteinases (MMPs). Prostate specific antigen (PSA) is a kallikrein-like serine protease that is able to cleave IGFBP-3 and IGFBP-5 (Cohen, Graves et al. 1992; Collett-Solberg and Cohen 1996). Cathepsin D is an acidic activated protease that can proteloyze all six IGFBPs (Braulke, Claussen et al. 1995; Claussen, Kubler et al. 1997; Marinaro, Hendrich et al. 1999). MMPs are peptide hydrolases involved in degradation of components of the extracellular matrix such as collagen and proteoglycans (Grimberg and Cohen 2000). Numerous metalloproteinases such as collagenase, gelatinase A, stromelysin 1, gelatinase B, and disintegrin are involved in the fragmentation of IGFBP2, IGFBP3, IGFBP4, and IGFBP5 (Fowlkes, Enghild et al. 1994; Collett-Solberg and Cohen 1996; Kubler, Cowell et al. 1998; Manes, Llorente et al. 1999). IGFBP-1, IGFBP-3, and IGFBP-5 are secreted as phosphorylated proteins. Phosphorylation does not appear to affect IGF binding of IGFBP-3 and IGFBP-5, however, phosphorylated IGFBP-1 has a six-fold higher affinity for IGF-I than the nonphosphorylated form (Jones, D'Ercole et al. 1991; Hoeck and Mukku 21 1994; Coverley and Baxter 1995). Phosphorylated IGFBP-1 may contribute to suppressing IGF-I effects, while nonphosphorylated IGFBP-1 may enhance IGF-I activity. Phosphorylation of IGFBP-1 has been to found to have no effect of IGF- II (Jones, D'Ercole et al. 1991). The IGFBP family has expanded to include a group of IGFBP-related proteins (IGFBP-rPs) that bind IGF-I and IGF-II with low affinity (Baxter, Binoux et al. 1998; Hwa, Oh et al. 1999). IGFBP-rPs are cystein-rich proteins that share structural and functional similarities to the IGFBPs (Bork 1993; Oh, Nagalla et al. 1996; Hwa, Oh et al. 1999). These proteins share the N-terminal domain of IGFBPs, but differ from the common midregion and C-terminal structure of the IGFBPs (Hwa, Oh et al. 1999). They bind to IGFs with a 100-fold lower affinity than the IGFBPs (Hwa, Oh et al. 1999). IGFBP-rP1 also known as Mac25, prostacyclin-simulating factor (PSF), and tumor-adhesion factor (TAF) was the first protein IGFBP-rP to be identified (Oh, Nagalla et al. 1996). IGFBP-rP1 appears to have multiple functional roles involved with growth inhibition and proliferation (Damon, Haugk et al. 1997; Hwa, Tomasini-Sprenger et al. 1998; Hwa, Oh et al. 1999). IGFBP-rP2, formerly known as connective tissue growth factor (CTGF), is implicated in fibrosis and muscle repair (Hwa, Oh et al. 1999). IGFBP-rP2 may also be involved in epithelial growth of mammary cells as it appears to be an important downstream effector of TGF, an inhibitor of cell proliferation (Yang, Kim et al. 1998). IGFBP-rP3 to IGFBP-rP9 have been 22 proposed, yet their structures and functions have not been confirmed (Wetterau, Moore et al. 1999). Section 2. IGFs and IGFBPs in the Circulation In adults, IGF-I levels range from 100 and 200 ng/ml, while levels of IGF- II range from 400 to 600 ng/ml (Yu and Rohan 2000). More than 90% of circulating IGFs are bound to IGFBP-3 and an acid-labile subunit (ALS), a liver- derived glycoprotein, to form a 150 kd ternary complex. GH stimulates hepatic production of all three components of this ternary complex (Rosenfeld, Pham et al. 1994). This IGF-I/IGFBP-3/ALS complex reduces the ability of IGF-I to cross the endothelial barrier, and prolongs the half-life of IGF-I from approximately ten minutes to fifteen hours (Guler, Zapf et al. 1989). The remaining IGFBPs bind most of other IGFs and less than 1% of IGF-I and IGF-II circulate in the free from (Baxter 1994). The high affinity and specificity of IGFBPs for IGFs serve to regulate the bioavailability of free IGFs in the circulation and tissues. Circulating IGF-I levels in normal individuals display considerable heterogeneity. However, levels within each individual are fairly constant and exhibit no apparent diurnal variation (Daughaday, Kapadia et al. 1987). IGFBP-3 is the predominant binding protein in the circulation and levels are relatively constant (Yu and Rohan 2000). In contrast, circulating levels of IGFBP-1 and IGFBP-2 vary considerably in response to metabolic state (Underwood, Thissen et al. 1994; Smith, Underwood et al. 1995). The regulators of IGFBP-4, IGFBP-5, 23 and IGFBP-6 in the circulation have not yet been determined (Wetterau, Moore et al. 1999; Schneider, Wolf et al. 2002; Zhou, Diehl et al. 2003). Twin studies have estimated the genetic contribution of inter-individual variation in circulating levels of IGFs and IGFBPs. In a study of adult twins of the Finnish Twin Cohort (32 monozygotic and 47 dizygotic twins), 38% of the variance of IGF-I was attributable to genetic effects, 66% for IGF-II, and 60% for IGFBP-3 (Harrela, Koistinen et al. 1996). There was no significant genetic influence on IGFBP-1 levels (Harrela, Koistinen et al. 1996). However, in a larger study of 248 adult twin pairs from the Swedish Adoption/Twin Study, the heritability estimate for IGFBP-1 was 36% and the estimate for IGF-I was 63% (Hong, Pedersen et al. 1996). Age and Gender At birth serum IGF-I levels are approximately 50% of adult levels (Daughaday and Rotwein 1989). IGF-I slowly increases during childhood to a peak pubertal level that is two to three times the adult range (Cara, Rosenfield et al. 1987). This rise during puberty is correlated with the increase in production of sex steroids (Rosenfield, Furlanetto et al. 1983). After puberty there is a steady decline to average adult levels that are reached during the third decade of life (Daughaday and Rotwein 1989). A slow decline in IGF-I levels is observed during adulthood. In an 8-year longitudinal study of young adult men, ages 20 to 34 years, IGF-I levels decreased approximately 6 ng/ml per year (Gapstur, Kopp 24 et al. 2004). In two studies of men ages 25 to 64 and 21 to 80 years, IGF-I levels decreased an average of 2 ng/ml per year (Yamamoto, Sohmiya et al. 1991; Landin-Wilhelmsen, Wilhelmsen et al. 1994), while a smaller decline , 0.11 ng/ml decrease per year, was reported in a study of men 50 to 97 years of age (Goodman-Gruen and Barrett-Connor 1997). IGF-II levels also similarly increase from birth to a peak level at puberty and thereafter levels remain fairly stable (Yu, Mistry et al. 1999). IGFBP-3 and ALS levels display similar age-related changes as IGF-I (Juul, Dalgaard et al. 1995; Juul, Moller et al. 1998; Yu, Mistry et al. 1999). IGFBP-1 and IGFBP-2 levels are high at birth and decline from childhood to puberty, after which they remain relatively constant or slightly increase (Nystrom, Ohman et al. 1997; Yu, Mistry et al. 1999). IGFBP-4 and IGFBP-6 levels appear to increase with age, while IGFBP-5 is suggested to decrease with age (Mohan, Farley et al. 1995). IGF-I levels vary across genders in adult populations. Numerous studies have reported higher IGF-I levels among men in comparison to women (Juul, Bang et al. 1994; Landin-Wilhelmsen, Wilhelmsen et al. 1994; Kaklamani, Linos et al. 1999; Lukanova, Soderberg et al. 2002; Probst-Hensch, Wang et al. 2003; Sandhu, Gibson et al. 2004), Studies have found IGFBP - 3 levels to be lower in men than women (Kaklamani, Linos et al. 1999; Lukanova, Soderberg et al. 2002; Probst-Hensch, Wang et al. 2003); and one report also observed lower levels of IGFBP-1 among men than women (Sandhu, Gibson et al. 2004) 25 Race and Ethnicity In the largest study of 622 African-American and 792 White young men of the CARDIA male hormone study (ages 20-34 years), longitudinal changes in IGF-I and IGFBP-3 levels were evaluated over an 8-year period (Gapstur, Kopp et al. 2004). At all three measurement periods (year 2, year 7, and year 10) the age-adjusted mean IGF-I and IGFBP-3 levels were significantly lower among African-Americans in comparison to Whites (P < 0.0001). Yet, there were no differences between both groups in the 8-year change in IGF-I or IGFBP-3 levels. For the first measurement at year 2, African-Americans had a lower IGF-I level (mean + se = 231 + 3 ng/ml) than Whites (250 + 3 ng/ml). IGFBP-3 levels at this time point were also lower for African-American (3401 + 31 ng/ml) in comparison to Whites (3762 + 27 ng/ml). In two small studies of adult men, no significant racial/ethnic differences in IGF-I levels were observed although IGFBP-3 levels were consistently found to be lower in African-American men in comparison to White men (Platz, Pollak et al. 1999; Winter, Hanlon et al. 2001). In the Health Professionals Follow-up Study of 42 African-American, 52 Asian, and 55 White men of the ages between 47 and 78 years at blood collection, Whites were found to have the highest IGF-I level (mean + sd = 239 + 74 ng/ml) followed by Asians (213 + 70 ng/ml) then African-Americans (205 + 74 ng/ml) (Platz, Pollak et al. 1999). In a study of men ages 35 to 69 years that included 85 African-Americans and 84 Whites, who have at least one affected first-degree 26 relative with prostate cancer, found no significant differences in IGF-I levels, 162 + 49 ng/ml and 172 + 53 ng/ml, respectively (Winter, Hanlon et al. 2001). Diet Nutritional factors are important regulators of IGF-I levels in the circulation (Rudman, Kutner et al. 1985). It is well-established that protein-calorie malnutrition lowers IGF-I levels and energy intake increases levels of IGF-I in a calorie restricted state (Thissen, Ketelslegers et al. 1994). Recent studies have examined the role of dietary influences on the IGF system among well-nourished populations. Protein intake has been associated with increased IGF-I levels in several studies (Holmes, Pollak et al. 2002; Giovannucci, Pollak et al. 2003; Larsson, Wolk et al. 2005), but not other smaller studies (Baibas, Bamia et al. 2003; Gunnell, Oliver et al. 2003; Vrieling, Voskuil et al. 2004; Maskarinec, Takata et al. 2005). In the larger studies, the Health Professionals Follow-up (n=753) and Nurses Health studies (n=1037), IGF-I levels increased 15% and 11%, respectively, among individuals at the top quintile of protein intake compared to those at the bottom quintile (Holmes, Pollak et al. 2002; Giovannucci, Pollak et al. 2003). A study of 696 men (226 meat-eaters, 237 vegetarians, and 233 vegans) reported a 9% decrease in IGF-I levels among vegans in comparison to vegetarians and meat-eaters (Allen, Appleby et al. 2000). No consistent relationship between carbohydrate intake and IGF-I levels have been observed(Kaklamani, Linos et al. 1999; Pollak 2000; Holmes, Pollak et al. 2002; Baibas, Bamia et al. 2003; Giovannucci, Pollak et al. 2003; Heald, Cade et 27 al. 2003; Larsson, Wolk et al. 2005; Maskarinec, Takata et al. 2005). The relationship between IGF-I and fat intake remains unclear, as several studies suggested increasing fat intake to be associated with higher IGF-I levels (Kaklamani, Linos et al. 1999; Baibas, Bamia et al. 2003; Gunnell, Oliver et al. 2003; Heald, Cade et al. 2003), while some report an inverse association (Holmes, Pollak et al. 2002; Giovannucci, Pollak et al. 2003), and others find no association (Larsson, Wolk et al. 2005; Maskarinec, Takata et al. 2005). A positive relationship between dairy intake and IGF-I levels have been reported by several studies (Heaney, McCarron et al. 1999; Ma, Giovannucci et al. 2001; Holmes, Pollak et al. 2002; Giovannucci, Pollak et al. 2003; Gunnell, Oliver et al. 2003). In an intervention study of 254 men and women, a 10% increase in IGF-I levels was associated with increased milk intake in comparison to usual diets (Heaney, McCarron et al. 1999). Similarly, higher calcium intake has been associated with increased IGF-I levels (Ma, Giovannucci et al. 2001; Holmes, Pollak et al. 2002; Gunnell, Oliver et al. 2003). Cross-sectional studies of alcohol and IGF-I levels have been inconsistent (Kaklamani, Linos et al. 1999; Holmes, Pollak et al. 2002; Probst-Hensch, Wang et al. 2003; Vrieling, Voskuil et al. 2004; Larsson, Wolk et al. 2005), while in the one intervention study of 53 postmenopausal women showed lower IGF-I levels among those consuming 2 drinks a day in comparison to one or no drinks (Lavigne, Baer et al. 2005). Overall, based on mostly cross-sectional studies, a consistent relationship between protein intake and IGF-I levels has been observed. Also, dairy and 28 calcium intake have been associated with IGF-I levels. No consistent relationship between IGF-I levels and fat and carbohydrate intake has been demonstrated. Anthropometric Characteristics IGF-I levels are positively correlated with childhood height (Blum, Albertsson-Wikland et al. 1993; Juul, Bang et al. 1994). However, the relationship between IGF-I levels and height among adults remains unclear. IGF-I levels have been shown to increase with adult height in some studies (Signorello, Kuper et al. 2000; Chang, Wu et al. 2002; Teramukai, Rohan et al. 2002; Gapstur, Kopp et al. 2004) but not others (Kaklamani, Linos et al. 1999; Voskuil, Bueno de Mesquita et al. 2001; Holmes, Pollak et al. 2002; Gunnell, Oliver et al. 2003; Gunnell, Oliver et al. 2004; Sandhu, Gibson et al. 2004; Wolk, Larsson et al. 2004). Similarly, the relationship between BMI and IGF-I levels among adults is not well-understood. Although, many studies have reported no association between BMI and IGF-I levels (Benbassat, Maki et al. 1997; O'Connor, Tobin et al. 1998; Kaklamani, Linos et al. 1999; Lukanova, Toniolo et al. 2001; Voskuil, Bueno de Mesquita et al. 2001; Holmes, Pollak et al. 2002; Gunnell, Oliver et al. 2003), a non-linear relationship has been reported by some studies. In a study of 445 men and 391 women, the highest levels of IGF-I was seen among subjects with a BMI from 24 to 26 and levels declined to the lowest IGF-I levels among subjects having a BMI < 20 and BM >30 (Lukanova, Soderberg et al. 2002). In well- nourished Western populations, IGF-I levels have been reported to decline with 29 an increase in BMI (O'Connor, Tobin et al. 1998; Chang, Wu et al. 2002; Gapstur, Kopp et al. 2004; Gomez, Maravall et al. 2004), while in non-Western countries such as Japan and China, IGF-I levels tend to increase with higher levels of BMI (Teramukai, Rohan et al. 2002; Probst-Hensch, Wang et al. 2003). In the only intervention study of 56 Japanese men with a BMI>25, who were assigned to a one-year exercise program and an increase in daily walking by >1000 steps, an increase in IGF-I levels was observed that was correlated with a significant reduction in visceral fat area and a non-significant decline in subcutaneous fat area and BMI (Kunitomi, Wada et al. 2002). At baseline these overweight men had significantly lower IGF-I levels in comparison age-matched normal weight men (18<BMI<25) (Kunitomi, Wada et al. 2002). Lifestyle Factors The relationship between physical activity and IGF-I levels is not well understood. Intervention studies have shown an increase in IGF-I levels associated with resistance and intense training in young and middle-aged adults (Borst, De Hoyos et al. 2001; Kraemer, Durand et al. 2004), while no association has been observed among older adults (Borst, Vincent et al. 2002); strength training has seen to lower IGF-I levels among middle aged women (Schmitz, Ahmed et al. 2002). Cross-sectional studies of physical activity and IGF -I levels have been inconsistent and may be due to age-related differences and types of physical activity assessed (leisure-time, general physical activity, training) 30 (Lukanova, Toniolo et al. 2001; Voskuil, Bueno de Mesquita et al. 2001; Chang, Wu et al. 2002; Holmes, Pollak et al. 2002; Teramukai, Rohan et al. 2002; Manetta, Brun et al. 2003; Probst-Hensch, Wang et al. 2003; Gapstur, Kopp et al. 2004). Smoking has been inconsistently associated with IGF-I levels (Goodman- Gruen and Barrett-Connor 1997; Lukanova, Toniolo et al. 2001; Chang, Wu et al. 2002; Holmes, Pollak et al. 2002; Teramukai, Rohan et al. 2002; Baibas, Bamia et al. 2003; Probst-Hensch, Wang et al. 2003; Gapstur, Kopp et al. 2004). Hormonal factors Several hormonal factors are suggested to be associated with levels of IGFs and IGFBPs. The most consistent findings are an inverse relationship between sex hormone-binding globulin (SHBG) and IGF-I and IGFBP-3, and a positive correlation with IGFBP-1 (Erfurth, Hagmar et al. 1996; Pfeilschifter, Scheidt-Nave et al. 1996; Janssen, Stolk et al. 1998; Leifke, Gorenoi et al. 2000; Chokkalingam, Pollak et al. 2001; Kaaks, Lukanova et al. 2003; Lukanova, Lundin et al. 2004). Testosterone levels have been seen to increase with higher levels in IGF-I in most studies (Erfurth, Hagmar et al. 1996; Pfeilschifter, Scheidt-Nave et al. 1996; Leifke, Gorenoi et al. 2000; Lukanova, Lundin et al. 2004) but not all (Chan, Stampfer et al. 1998; Kaaks, Lukanova et al. 2003). In a cross-sectional study of prostate cancer cases and controls of the Northern Sweden Disease Cohort, IGF-I and IGFBP-3 was estimated to explain approximately 10% of the variation in SHBG and 5% of the variation in 31 testosterone (Kaaks, Lukanova et al. 2003). Two studies, respectively, have reported a positive association between IGF-I levels and -androstanediol glucuoronide (3-diol G) (Pfeilschifter, Scheidt-Nave et al. 1996; Chokkalingam, Pollak et al. 2001) and dihydroepiandrosterone sulfate (DHEAS) (Janssen, Stolk et al. 1998; Lukanova, Lundin et al. 2004). IGF-I and IGFBP-3 levels in the MEC The lifestyle and dietary determinants of circulating IGF-I and IGFBP-3 have been recently examined in the MEC (DeLellis, Rinaldi et al. 2004). Plasma levels of IGF-I were significantly higher in 490 men (177.6 ng/ml) in comparison to 465 postmenopausal women (142.3 ng/ml) (P < 0.0001). Among both genders, IGF-I and IGFBP-3 levels were found to significantly decrease with age (men and women P < 0.0001). The age-adjusted mean IGF - I levels for females were: African American (154.4 ng/ml), Japanese (148.5 ng/ml), Hawaiian (144.8 ng/ml), White (143.8 ng/ml) and Latino (122.8 ng/ml). The age-adjusted mean IGF-I levels among males were: Hawaiian (190.1 ng/ml), White (180.4 ng/ml), Japanese (180.0 ng/ml), African American (176.5 ng/ml), Latino (169.3 ng/ml). For men, height was positively associated with IGF-I and IGFBP-3 levels. The age and race-adjusted IGF-I least squares (LS) me an level for men 65 inches and taller was 195.8 ng/ml and for men shorter than 61 inches was 159.3 ng/ml. For women, high BMI was inversely associated with IGF-I and IGFBP-3. Women having a BMI >30 had an age and race-adjusted LS mean IGF-I level of 126.8 32 ng/ml in comparison to women with a BMI < 23 whose mean level was 149.6 ng/ml. Physical activity displayed no clear relationship with IGF-I and IGFBP-3 levels. Total fat was inversely associated with IGFBP-3 in men but not women, and total fat was not associated with IGF-I in either sex. Fat from meat intake was inversely associated with IGF-I and IGFBP-3 levels among men. Alcohol consumption was inversely associated with IGFBP - 3 levels in women, but no relationship was observed with IGF-I. Among men current smokers had lower IGF-I levels compared to never and past smokers. Section 3. IGF system and prostate cancer Carcinogenesis is characterized by uncontrolled cellular growth and the spread of aberrant cells. This course of pathogenesis is facilitated by the disruption of regulatory systems that mediate normal cell proliferation and death. As the IGF systems plays a key role in regulating cellular growth, it has been hypothesized that elevated IGF-I bioactivity may lead to increased cell turnover and opportunities for genetic alterations eventually resulting in transformation. Multiple lines of evidence from experimental and epidemiological studies provide support for the role of the IGF system in prostate cancer tumorigenesis. A. Experimental Studies The prostate stroma predominately secretes IGF-I, and both stromal and epithelial cells are responsive to IGF-I growth-promoting effects through the 33 interaction of IGF-I with the IGF1R (Cohen, Peehl et al. 1991; Cohen, Peehl et al. 1994). IGF-I, IGF-II, and IGF1R mRNA and protein levels have been found to significantly increase from normal to prostatic intraepithelial neoplasia (PIN) to prostate cancer tissue (Cardillo, Monti et al. 2003). Interestingly, in normal prostate tissue levels of IGF-I and IGF-II were observed to be higher in the stroma than the epithelium, while in prostate cancer tissue both IGF-I and IGF-II levels were increased in malignant epithelial cells (Cardillo, Monti et al. 2003). IGF1R expression has been observed in the progression of prostate cancer to advanced andrgoen-independent metastatic disease (Hellawell, Turner et al. 2002). In vitro studies have demonstrated IGF-I to be a potent mitogen for both androgen-dependent and androgen-independent prostate cancer cell lines (Pietrzkowski, Mulholland et al. 1993). In prostate cancer cells, IGF-I has been found to activate the androgen receptor and stimulate PSA secretion in the absence of androgens (Culig, Hobisch et al. 1994). Peptide analogues of IGF-I are able to down-regulate IGF1R and inhibit proliferation of prostate cancer cells (Pietrzkowski, Mulholland et al. 1993). Additionally, inhibition of IGF1R gene expression in prostate cancer cells demonstrate increased apoptosis; and PTEN, a tumor suppressor, appears to regulate apoptosis and proliferation through suppressing IGF1R synthesis (Grzmil, Hemmerlein et al. 2004; Zhao, Dupont et al. 2004). Recently, androgens have been shown to upregulate IGF1R expression and sensitive prostate cancer cells to IGF-I activity (Pandini, Mineo et al. 2005). IGF2R appears to have opposing roles in prostate cancer cell growth, whereby 34 decreased proliferation is associated with the binding activity of IGF-II, while increased growth is attributable to binding of M6P-ligands (Schaffer, Lin et al. 2003). Increase autocrine production of IGF-I has been observed in prostate cancer cells that metastasize to the bone, the primary metastasis site for prostate cancer (Rubin, Chung et al. 2004). In vivo studies have demonstrated that IGF-I is involved in normal prostate development and tumorigenesis. IGF-I deficient mice (IGF - / - null) exhibit smaller and less developed prostate glands than their wild-type littermates (Ruan, Powell-Braxton et al. 1999). Systemic administration of IGF-I in rats (400µg/rat/day) has been found to significantly increase ventral prostate weight in comparison to controls (Torring, Vinter-Jensen et al. 1997). Reduction of ventral prostate weight in rats by administration of finasteride, a competitive inhibitor of 5-reductase, is associated with an inhibition of IGF-I and IGF1R gene expression and an up-regulation of IGFBP-3 expression (Huynh, Seyam et al. 1998). Animal models of prostate cancer progression have repo rted a large increase in IGF-I and IGF1R expression (Kaplan, Mohan et al. 1999; Nickerson, Chang et al. 2001). Transgenic mice with targeted IGF-I expression in basal epithelial cells demonstrate IGF1R activation and tumor progression similar to human prostate cancer with 50% of these mice developing prostatic carcinomas (DiGiovanni, Kiguchi et al. 2000). In a mouse model of prostate cancer progression, a mutation that inactivates the growth hormone releasing hormone receptor and decreases circulating growth hormone and IGF-I levels, has been 35 shown to reduce the prostate cancer progression and improve survival (Majeed, Blouin et al. 2005). B. Epidemiologic Studies Several prospective studies have examined the relationship between circulating levels of IGF-I and prostate cancer risk (Table 1) (Chan, Stampfer et al. 1998; Schaefer 1998; Harman, Metter et al. 2000; Stattin, Bylund et al. 2000; Lacey, Hsing et al. 2001; Stattin, Stenman et al. 2001; Chan, Stampfer et al. 2002; Woodson, Tangrea et al. 2003; Stattin, Rinaldi et al. 2004; Chen, Lewis et al. 2005). Table 2 presents a summary of the seven prospective studies. In the first report of the Physicians’ Health Study a significant 4-fold increased risk of prostate cancer was observed among men in the highest quartile of IGF-I compared to those in lowest quartile (Chan, Stampfer et al. 1998). A subsequent investigation of the Baltimore Longitudinal Study on Aging reported similar findings when comparing the highest to lowest tertiles of IGF-I (OR=3.11; 95% CI: 1.11-8.74) (Harman, Metter et al. 2000). The Northern Sweden Health and Disease Cohort also reported higher IGF-I levels to be positively associated with an increase risk of prostate cancer (OR=1.47; 95% CI: 0.81-2.64) (Stattin, Rinaldi et al. 2004). These three studies were consistent with the finding that elevated IGF-I levels are associated with increased prostate cancer risk. However, additional studies have not confirmed this positive relationship between IGF-I levels and. 36 Table 1. Prospective studies of IGF-I and IGFBP3 levels and prostate cancer Author Year Study Design Source #Cases/#Controls Age (years) Time blood draw to dx Assay Chan 1998 Physicians' Health Study U.S. 152/152 40-84 in 1982 6 mos-9.5 yrs ELISA Q1=63.9,Q2=58.9, Q3=59.0, Q4=59.3 mean=7 yrs Scahefer 1998 Kaiser U.S. 45/719 60-91 years @ study start 1-21 yrs RIA Harman 2000 Baltimore Longitudnial Study on Aging U.S. 72/127 cases mean=64.8 @ sample 3-17 yrs RIA controls mean=65.7 @ sample mean=9 yrs Stattin 2001 Northern Sweden Health and Disease Cohort Study Sweden 149/298 cases mean=59.7 @ recuitment <1 mo-10 yrs IRMA controls mean=59.6 @ recruitment median=3.85 yrs Lacey 2001 WA County Serum Blood Bank U.S. 30/60 N/A 6-18 yrs ELISA Chan* includes 1998 2002 Physicians' Health Study U.S. 530/534 40-84 in 1982 6 mos-13 yrs ELISA mean=9 yrs Woodson 2003 ATBC Study Finland 100/400 50-69 @ entry 5-12 yrs ELISA median=9 yrs Stattin* includes 2001 2004 Northern Sweden Health and Disease Cohort Study Sweden 281/560 80% >58 yrs @ blood draw cases <59 yrs mean=6.1+2.9 IRMA controls >59 yrs mean=4.6+2.8 Chen 2004 Cardiovascular Health Study U.S. 174/174 > 65 years 1-9 yrs IRMA mean=3.4 yrs 37 Table 2. IGF-I and IGFBP3 levels and prostate cancer risk Author IGF-I (ng/ml) OR (95%CI) IGFBP-3 (ng/ml) OR (95%CI) IGF-I: IGFBP-3 (ng/ml) OR (95%CI) Adjusted/Matched Variables Chan Q1: 99.4-184.8 1.00 Q1: median=2234 1.00 -- -- age, smoking, duration of follow-up, IGFBP-3, IGF1 Q4: 293.8-499.6 4.32 (1.76-10.6) Q4: median=3473 0.41 (0.17-1.03) -- -- Scahefer Q1 1.00 -- -- -- -- age Q4 0.81 (0.36-1.80) -- -- -- -- Harman Q1: 10-125 1.00 Q1: 1.3-2.5 mg/ml 1.00 -- -- age, date of blood draw, IGF-I, IGF-II, IGFBP-3, PSA Q3: 159-450 3.11 (1.11-8.74) Q3: 3-5.5 mg/ml 0.76 (0.3-1.94) -- -- Stattin Q1: mean=160.2 1.00 Q1: mean=2021 1.00 -- -- age, date of survey, residency, fasting time, IGFBP-3 Q4: mean=283 1.32 (0.73-2.39) Q4: mean=3133 1.19 (0.66-2.15) -- -- Lacey Q1 1.00 Q1 1.00 -- -- age, race, date of blood draw Q4 0.7 (0.2-2.3) Q4 1.1 (0.3-3.8) -- -- Chan* includes 1998 Stage C & D: Q1 vs. Q4 1.00 Stage C & D: Q1 vs. Q4 1.00 Stage C & D: Q1 vs. Q4 1.00 age, smoking, IGFBP-3, IGF-I 5.1 (2.0-3.2) 0.2 (0.1-0.6) 2.5 (1.19-5.23) Stage A & B: Q1 vs. Q4 1.00 Stage A & B: Q1 vs. Q4 1.00 Stage A & B: Q1 vs. Q4 1.00 1.2 (0.7-2.2) 1.0 (0.6-1.8) 1.35 (0.86-2.13) Woodson Q1: med=65.4 (3.9-87.3) 1.00 Q1: med=5805 (4875-9850) 1.00 Q1: med=0.045 (.021-0.05) 1.00 age, BMI, intervention group, IGFBP-3, IGF-I Q4: med=210.2 (167.5-395.7) 0.52 (0.23-1.16) Q4: med=2156 (236-2620) 1.93 (0.83-4.49) Q4: med=0.017 (.070-0.13) 0.54 (0.29-1.01) Stattin* includes 2000Q1<166 1.00 Q1<2020 1.00 -- -- age, date at recruitment, IGFBP-3, IGF-I Q4>254 1.47 (0.81-2.64) Q4>2696 1.04 (0.63-1.74) -- -- Chen Q1: med=87 (38-112) 1.00 Q1: med=2169 (1098-2601) 1.00 Q1: med=0.126 (0.057-0.145) 1.00 age, race, year of entry, year of blood draw, insulin, BMI, IGFBP-3, IGF-I Q4: med=259 (193-917) 0.77 (0.03-1.84) Q4: med=4118 (2796-6081) 0.63 (0.29-1.37) Q4: med=0.253 (0.215-0.647) 0.75 (0.41-1.36) 38 prostate cancer risk (Schaefer 1998; Lacey, Hsing et al. 2001; Woodson, Tangrea et al. 2003; Chen, Lewis et al. 2005). Two small prospective studies of 45 and 30 cases, respectively, found no association between IGF-I levels and prostate cancer (Schaefer 1998; Lacey, Hsing et al. 2001). The Finnish ATBC study of long-term smokers (n=100 cases) found no evidence to support an association between IGF- I levels and prostate cancer (Woodson, Tangrea et al. 2003). Also, a U.S. study of 174 cases from the Cardiovascular Health Study observed no association between IGF-I levels and prostate cancer risk (Chen, Lewis et al. 2005). A crude meta- analysis of the seven studies suggests a modest positive relationship between IGF- I levels and risk of prostate cancer (OR meta = 1.22; 95% CI: 0.91-1.62) In the follow-up report of the Physicians’ Health Study, the largest study to date of 530 cases (including subjects from their original report (Chan, Stampfer et al. 1998)), the association between higher IGF-I levels and prostate cancer risk was observed only among advanced disease (Chan, Stampfer et al. 2002). Based on 112 advanced prostate cancer cases in this study, men at the top 25% of IGF-I levels had 5-fold significant risk of advanced disease compared to men the bottom 25% of IGF-I levels. Table 3 presents the findings of three studies of IGF-I and advanced prostate cancer. The Northern Sweden Health and Disease Cohort observed an non-significant increased risk of advanced disease comparing the top and bottom quartiles of IGF-I (OR=1.53; 95% CI: 0.68-3.44) (Stattin, Rinaldi et al. 2004), while a report of 52 advanced cases of the Cardiovascular Health Study 39 Table 3. IGF-I and IGFBP-3 levels and advanced prostate cancer risk Author Year Study Design # Advanced Cases Advanced Disease IGF-I Q1 vs. Q4 IGFBP-3 Q1 vs. Q4 Chan 1998 Physicians' Health Study 49 Stage C & D 10.0 (1.4-74.5) 0.2 (0.02-1.2) Chan 2002 Physicians' Health Study 112*include 1998 Stage C & D 5.1 (2.0-13.2) 0.2 (0.1-0.6) Stattin* includes 2000 2004 Northern Sweden Health and Disease Cohort Study N/A locally advanced, lymph node or 1.53 (0.68-3.44) 1.87 (0.66-5.30) bone metastasis, or PSA > 50 ng/ml Chen 2004 Cardiovascular Health Study 52 Stace C & D or Grade 3 0.17 (0.05-0.62) 0.24 (0.07-0.84) OR 95% CI 40 found higher levels of IGF-I associated with a reduced risk of advance disease (OR=0.17; 95%CI: 0.05-0.62) (Chen, Lewis et al. 2005). Three prospective studies have also examined the relationship circulating levels of IGFBP-3 and the molar IGF-I/IGFBP-3 ratio and prostate cancer risk (Table 2). Results from the Physicians’ Health Study suggest a decline in risk of prostate cancer with increasing levels of IGFBP-3 (Chan, Stampfer et al. 2002). The reports of the Baltimore Aging Study and Cardiovascular Health study support this inverse relationship, while findings of the ATBC study report a positive association between IGFBP-3 levels and prostate cancer risk, and no association was observed in the Sweden Health and Disease Cohort (Harman, Metter et al. 2000; Woodson, Tangrea et al. 2003; Stattin, Rinaldi et al. 2004; Chen, Lewis et al. 2005). The molar IGF-I/IGFBP-3 ratio has been investigated as a possible indicator of bioavailable IGF-I; the Physicians’ Heath Study reported a higher IGF-I/IGFBP3 ratio to be associated with increased risk, however subsequent studies have not confirmed this observation (Chan, Stampfer et al. 2002; Woodson, Tangrea et al. 2003; Chen, Lewis et al. 2005). Interestingly, in the Physicans’ Health Study, men in the highest and lowest tertiles of IGF-I and IGFBP-3, respectively, had a large significant increased risk of advanced disease in comparison to men in the lowest tertiles of both proteins (OR=9.5; 95% CI: 1.9-48.4) (Chan, Stampfer et al. 2002). 40 The cumulative evidence of the prospective studies suggest that higher circulating levels of IGF-I are associated with at most a modest increase in overall prostate cancer risk with a larger effect on advanced disease. These studies are unclear as to the role of IGFBP-3 and risk of disease, although the largest study does suggest that higher IGFBP-3 levels are associated with a lower prostate cancer risk. Possible explanations for the inconsistent findings include issues related to study design, study population, and IGF measurements. These issues are further discussed below. Epidemiological Design and Study Populations Prospective studies whereby blood specimens are collected from men prior to their diagnosis are most informative as they minimize the influence of disease on IGF-I levels. Yet, this does not preclude the possible effect of preclinical disease on IGF-I levels. This is especially relevant to prostate cancer given the prevalence of screen-detected disease. In the Northern Sweden Health and Disease Cohort Study, 80% of cases and 20% of controls had PSA levels above 4 ng/ml at the time of blood collection, suggesting the presence of preclinical prostate cancer among these men. The Physicians’ Health Study and Northern Sweden Health and Disease Cohort analyzed a subset of their cases with at least a five year lag-time between blood collection and diagnosis, both studies found a positive relationship between IGF-I levels and prostate cancer risk as observed in their analysis of all prostate cancer cases (Chan, Stampfer et al. 1998; 41 Stattin, Rinaldi et al. 2004). In contrast, the WA county and ATBC studies, which included only men with a minimum of five year lag-time, found no association between IGF-I levels and prostate cancer (Lacey, Hsing et al. 2001; Woodson, Tangrea et al. 2003). Other issues related to differences in study populations may explain these discrepant findings. Prostate cancer is a slow-growing disease, and it remains unclear what time interval is sufficient to mitigate the influence of preclinical disease on circulating levels of IGF-I. The possibility of undetected prostate cancer leading to the misclassification of disease status among control subjects may also bias results. The elevated PSA levels (>4 ng/ml) among 20% of the controls in the Northern Sweden Health and Disease Cohort (Stattin, Bylund et al. 2000) suggests the presence of latent disease and the possibility of misclassification of disease status. If misclassified control subjects had higher IGF-I levels, then an attenuation of the risk estimate of the positive relationship between IGF-I and prostate cancer would occur. Conversely, if controls with undetected disease had lower IGF-I levels, an overestimation of positive IGF-I-prostate cancer association would be observed. Heterogeneous study populations and small study size may also account for some of the differences in study results (Table 1). Both the Kaiser and WA County studies examined a small number of cases, 45 and 30, respectively, which may have limited their ability to detect an association (Schaefer 1998; Lacey, Hsing et al. 2001). The ATBC study was comprised of long-term smokers (average 20 cigarettes daily for 36 years) and given the unclear relationship 42 between smoking and IGF-I, smoking practices may explain some of the inconsistencies in results (Woodson, Tangrea et al. 2003). The age of the study population is another complexity in evaluating the association between IGF-I and prostate cancer: IGF-I levels slowly decrease with age, yet prostate cancer is associated with a late age of onset. As the Kaiser and Cardiovascular Health studies were comprised of older study populations (> 60 years at the start for both studies), the different age distributions between these studies and others may limit comparability of results (Schaefer 1998; Chen, Lewis et al. 2005). Most of the prospective studies were of predominantly Caucasian subjects, with the exception of the Cardiovascular Healthy Study in which 24% of the study population was African-American. As African-American men are at increased risk of prostate cancer, it is of particular interest, how circulating levels among this population relate to prostate cancer susceptibility. Confounding Although a clear relationship between potential confounders for IGF-I and prostate cancer has yet to be firmly established, all potential covariates such as age, ethnicity, height, and BMI as well as dietary factors should be addressed (Table 2). The principal confounder that may influence the association between IGF-I levels and prostate cancer risk is the level of IGFBP-3. IGFBP-3 is the primary binding protein associated with circulating IGF-I, and an inverse relationship between IGFBP-3 levels and prostate cancer has been observed in 43 several studies (Harman, Metter et al. 2000; Chan, Stampfer et al. 2002; Chen, Lewis et al. 2005). The two smaller studies, Kaiser and WA County, did not adjust for IGFBP-3 levels which may also account for the lack of association observed in these studies (Schaefer 1998; Lacey, Hsing et al. 2001). A larger increase in IGF-I-associated prostate cancer risk was observed after adjustment of IGFBP-3 in the Physicians’ Health Study (adjusted OR=4.3; unadjusted OR=2.4) and the Baltimore Study (adjusted OR=3.11; unadjusted OR=1.65), however this effect has not been observed in other studies (Chan, Stampfer et al. 1998; Harman, Metter et al. 2000; Stattin, Rinaldi et al. 2004; Chen, Lewis et al. 2005). IGF-I measurement Circulating IGF-levels are measured by several different laboratory assays that are available from commercial and academic sources. Commonly used assays in epidemiologic studies are enzyme-linked immunosorbent assay (ELISA), immunoradiometric assay (IRMA), and radioimmunoassay (RIA). These assays are based on specific immunological recognition of the IGF-I molecule, and utilize enzymatic activity and radioactivity for quantification. These assays were originally developed to distinguish between aberrant and normal somatic developmental conditions and are considered suitable for epidemiological investigations (Pollak 2003). However, variability in assay specificity and systematic differences between assays may contribute to discrepancies in results. 44 The current epidemiological studies and results are heterogeneous in regards to assay methodologies (Table 1). Accurate assessment of circulating IGF-I is also complicated by its association with IGFBPs. IGFBPs may be present in the circulation in different glycosylation and phosphorylation states, as well as proteolytic fragments, all of which may interfere with IGF-I measurement. Several extraction strategies have been developed to separate IGF-I from its associated binding proteins prior to assay measurement. However, not all extraction methods have been found to efficiently remove all IGFBPs and associated fragments (Chestnut and Quarmby 2002). In addition, there is no current ‘gold standard’ reference method for IGF measurement. Because of these methodological complexities, IGF-I levels may be imprecisely measured and result in misleadingly elevated or lowered levels. Different distributions in IGF-I levels are observed across studies and within similar methodologies of IGF-I measurement. This also poses a limitation in comparability of results across studies. Two studies using ELISA assays display a different range in the lower quartile of IGF-I, yet a similar range is observed in the upper quartile: Physicians’ Health Study, Q1=99.4-184.8 ng/ml; Q4=293.8-499.6 ng/ml and ATBC Study: Q1=3.9-87.3 ng/ml; Q4=167.5-395.7 ng/ml (Chan, Stampfer et al. 1998; Woodson, Tangrea et al. 2003). Studies using IRMA also report different average IGF-I levels in the lower quartiles of IGF-I yet similar averages are observed in the upper quartile: Sweden Health and Disease Cohort, Q1 mean=160 and Q4 mean=283, Cardiovascular Health Study, 45 Q1 median=87 and Q4 median=259 (Stattin, Byl und et al. 2000; Chen, Lewis et al. 2005). Additionally, different study groups were used for classification of quartile categories across studies. The Physicians’ Health and Cardiovascular Health studies used control subjects to determine their quartile distribution, while the Baltimore Study and Sweden Heath and Disease Cohort used both controls and cases combined; the remaining studies did not provide detailed information. As epidemiological studies use a single measurement of IGF-I to define the level of exposure, it is important to consider how well a single measurement reflects the level of exposure at the time of interest. Several studies have conducted serial IGF-I measurements in a subset of subjects and observed that a single measurement is predictive of habitual levels up to one year (Goodman- Gruen and Barrett-Connor 1997; Chan, Stampfer et al. 1998; Kaaks, Toniolo et al. 2000). In contrast, a serial serum analysis over a three year period in the ATBC study, observed a significant 18% increase in IGF-I among cases in comparison to a 4% decrease among controls (Woodson, Tangrea et al. 2003). Gapstur et al. examined IGF-I levels over an eight-year period among young men ages 20-34 and estimated the correlation between IGF-I levels measured 5 or 8 years apart to range from 0.39 to 0.47 (Gapstur, Kopp et al. 2004). Platz et al. examined IGF-I levels among older men, 47 to 78 years, collected on average 3 years apart and reported a correlation of 0.70 (Platz, Pollak et al. 1999). To what extent a single measurement captures the appropriate level of IGF-I exposure remains unclear 46 and further evaluation of multiple IGF-I measurements over extended periods of time is needed. C. Molecular Epidemiologic Studies A dinucleotide cytosine-adenosine (CA) n repeat polymorphism, approximately 1 kb upstream from the transcription start site of IGF1 has been identified (Rotwein 1991). A range from 10 to 24 repeats has been observed, with the 19 (CA) n repeat sequence being most common. The functional significance of this polymorphism is not yet known; although it is located in a region of the promoter that has been found to contain regulatory elements in rats that may alter IGF1 transcriptional activity (West, Arnett et al. 1996). This polymorphism has been examined for its relationship with circulating levels of IGF-I in two population-based studies (Allen, Davey et al. 2002; Rietveld, Janssen et al. 2003). In the EPIC study of 696 men from U.K., there was no association between (CA) 19 and circulating levels (Allen, Davey et al. 2002); while in the Rotterdam study of 189 men and women, homozygous carriers of the (CA) 19 repeat had the highest IGF-I levels in comparison to subjects with either shorter or longer (CA) n repeats (Rietveld, Janssen et al. 2003). Five studies have investigated the (CA) 19 polymorphism in relation to prostate cancer risk (Table 4) (Nam, Zhang et al. 2003; Li, Cicek et al. 2004; Neuhausen, Slattery et al. 2005; Schildkraut, Demark- Wahnefried et al. 2005; Tsuchiya, Wang et al. 2005). The results are inconsistent and the lack of uniform analysis of the (CA) n repeat limits the comparability of 47 Table 4. IGF1 (CA) n repeat and prostate cancer risk Author Year Study Design Source #Cases/#Controls Genotyping Method IGF1 (CA)n classification # Cases # Controls OR (95% CI) Adjusted/Matched variables & Notes Nam 2003 hospital based case-control study U.S. 483/548 DNA sizing (CA) 19 noncarrier 64 156 1.00 age, PSA, free:total PSA, DRE, ethnicity, family history, and obstructive voiding systems heterozygous (CA) 19 230 373 1.30 (0.9-2.2) study subjects PSA> 4ng/ml or abnormal DRE homoygous (CA) 19 189 275 1.46 (0.9-1.9) Li 2004 family based case-control study of brothers U.S. 440/480 DNA sequencing 18 repeats 65 67 0.97 (0.67-1.41) age 19 repeats 548 586 1.00 20 repeats 170 180 0.99 (0.78-1.27) Neuhausen 2005 hospital based case-control study U.S. 199/267 DNA sizing (CA) 19 noncarrier/het (CA) 19 115 156 1.0 (1.0-1.1) age homoygous (CA) 19 78 107 1.00 (CA) 19 noncarrier 35 20 1.00 age, BMI, height, diabetes, and race Schildkraut 2005 hosptial based case-control study U.S. 100/93 DNA sizing heterozygous (CA) 19 39 33 0.6 (0.3-1.4) homoygous (CA) 19 20 28 0.3 (0.1-0.7) Tsuchiya 2005 hospital based case-control study Japan 303/262 DNA sizing <17 repeats 68 75 1.00 age 18 repeats 56 70 0.90 (0.56-1.46) 19 repeats 136 82 1.85 (1.20-2.83) >20 repeats 43 35 1.38 (0.79-2.41) (CA) 19 noncarrier 155 174 1.00 heterozygous (CA) 19 130 82 1.78 (1.25-2.53) homoygous (CA) 19 18 6 3.36 (1.30-8.67) 48 results across studies. Comparing subjects homozygous for the (CA) 19 repeat to noncarriers of the (CA) 19 repeat, two studies reported a positive association associated with homozygous state, while one study observed an inverse association with the homozygous (CA) 19 repeat (Nam, Zhang et al. 2003; Schildkraut, Demark-Wahnefried et al. 2005; Tsuchiya, Wang et al. 2005). The effect of (CA) n repeat length on prostate cancer risk remains uncertain, and inconsistencies in study design and inadequate sample sizes may explain the variability in results. Additionally, the lack of Recently, Tran et al. compared two different laboratory methods for genotyping the (CA) n repeat polymorphism: 1) DNA sizing, which compares molecular size of PCR products to a DNA standard by gel electrophoresis; 2) direct DNA sequencing, where the number of repeats are counted in sequences of DNA (Tran, Bharaj et al. 2004). Tran et al. demonstrated that DNA sizing is less sensitive than DNA sequencing in measuring differences in the number of alleles. This misclassification of repeat length may explain the inconsistent reports of IGF1 (CA) 19 and prostate cancer risk as well as the findings relating to circulating levels. Only one study of prostate cancer risk used the more sensitive DNA sequencing method; and this study found no association between (CA) n repeat length and risk (Li, Cicek et al. 2004). Two studies used DNA sequencing to investigate circulating levels of IGF- I and the (CA) 19 polymorphism, with one report finding an association between IGF-I levels and homozygote carriers of the (CA) 19 , and the other study observing no association (Kato, Eastham et al. 2003; Rietveld, Janssen et al. 2003). 49 A single nucleotide polymorphism located in the promoter region of IGFBP3 (nucleotide -202 A/C) has been reported to correlate with circulating IGFBP-3 levels, with higher IGFBP-3 levels observed in the presence of the A allele (Deal, Ma et al. 2001; Schernhammer, Hankinson et al. 2003; Ren, Cai et al. 2004; Le Marchand, Kolonel et al. 2005). An in vitro study suggests elevated promoter activity of the A allele in comparison to the C allele (Deal, Ma et al. 2001). Four studies (Table 5) have investigated the association between IGFBP3 (A-202C) and prostate cancer risk and found no association (Nam, Zhang et al. 2003; Wang, Habuchi et al. 2003; Li, Cicek et al. 2004; Schildkraut, Demark-Wahnefried et al. 2005). Although, in the study by Wang et al. the C allele was reported to be associated with advanced prostate cancer in comparison to localized disease (Wang, Habuchi et al. 2003). Summary In summary, evidence from in vivo studies demonstrating IGF-I associated prostate tumorigenesis and the larger epidemiologic studies reporting increased prostate cancer risk associated with higher circulating levels of IGF-I, provide support for the hypothesis that IGF-I plays a role in prostate cancer development. Studies that have investigated the role of genetic variation in the IGF1 and IGFBP-3 and prostate cancer risk have focused only on a few polymorphisms and have yet to systematically examine how genetic diversity in these genes may influence prostate cancer susceptibility. Examining whether inherited differences 50 Table 5. IGFBP3 A-202C and prostate cancer risk Author Year Study Design Source #Cases/#Controls Genotyping Method IGFBP3 A-202C # Cases # Controls OR (95% CI) Adjusted/Matched variables & Notes Nam 2003 hospital based case-control study U.S. 483/548 RFLP AA 135 219 1.00 age, PSA, free:total PSA, DRE, ethnicity, family history, and obstructive voiding systems AC 233 399 0.70 (0.5-0.9) study subjects PSA> 4ng/ml or abnormal DRE CC 115 186 0.90 (0.6-1.2) Wang 2003 hospital based case-control study Japan 307/227 RFLP AA 189 152 1.00 age AC 100 105 0.77 (0.55-1.09) CC 18 15 0.96 (0.66-1.40) Li 2004 famiy based case-control study of brothers U.S. 440/480 RFLP AA 97 139 0.76 (0.40-1.44) age AC 217 225 0.81 (0.51-1.30) CC 126 115 1.00 Schildkraut 2005 hospital based case-control study U.S. 100/93 RFLP AA 18 23 1.00 age, BMI, height, diabetes, and race AC/CC 55 41 1.9 (0.9-4.1) 51 relate to prostate cancer risk is essential in understanding disease etiology and may be useful in prevention and development of targeted treatment regimens. Section 4. IGF1, IGFBP1, and IGFBP3 Gene and Protein Organization IGF1 The IGF1 gene is located on chromosome 12q22-q23 and spans approximately 84.6 kb and is comprised of 4 exons (Figure 1). IGF1 shares a high degree of conservation at the molecular level with approximately 80% homologous amino acid sequence shared with other mammalian species (Kelley, Schmidt et al. 2002). A large number of evolutionary conserved regions (ECR) is observed in the mouse and rat, 38 and 74, respectively (defined as at least 200 bp with 80% identify to the human genome) (http://www.dcode.org/). IGF-I molecules have been found in species whose origins extend over 550 million years (LeRoith, Kavsan et al. 1993). The IGF-I protein is comprised of a 70 amino-acid singe chain: B amino-terminal domain of 29 amino-acids, central C domain of 12 amino-acids, A domain of 21 amino acids, and a carboxy-terminal D domain of 8 amino acids (Rotwein 1999). Figure 1. IGF1 on chromosome 12q22-q23 (human genome assembly 17) 52 IGFBP1 and IGFBP3 IGFBP1 and IGFBP3 genes are located in close proximity on chromosome 7p13-p12, separated by a 19 kb distance and transcription oriented in a tail-to-tail configuration (Figure 2) (Ekstrand, Ehrenborg et al. 1990; Allander, Bajalica et al. 1993). Both genes are associated with the homeobox A (HOXA) cluster, which are genes that encode transcription factors essential for early development (Kelley, Schmidt et al. 2002). The close genomic relationship between IGFBP1 and IGFBP3 suggest that these genes were derived by duplication of an ancestral IGFBP (Kelley, Schmidt et al. 2002). IGFBP1 and IGFBP3 spans 5.2 kb and 9.0 kb, respectively. All IGFBPs have four exons. IGFBPs share overall approximately 50% homologous protein sequence, and up to 80% homology is conserved between corresponding IGFBPs of different mammalian species (Lamson, Giudice et al. 1991; Shimasaki and Ling 1991). The amino acid sequence of IGFBP1 and IGFBP3 can be divided into 3 regions of cysteine clusters; these cysteine residues and their spatial configuration are highly conserved among the six IGFBPs (Lee, Giudice et al. 1997). It is hypothesized that regions 1 and 3 are necessary for optimal binding of IGFs, and region 2 serves as a hinge that allows the protein to make a hairpin fold such that regions 1 and 3 can interact with a single ligand (Lee, Giudice et al. 1997). 53 Figure 2. IGFBP1 and IGFBP3 loci on chromosome 7p13-p12 (human genome assembly 17) Genetic Variation in IGF1, IGFBP1, and IGFBP3 With recent efforts to characterize the genetic variation in the human genome, over ten million single nucleotide variants known as single nucleotide polymorphisms (SNPs) have been catalogued. The large genetic variation in these genes has yet to be comprehensively examined; and their genetic contribution to prostate cancer susceptibility is unknown. Table 6. IGF1, IGFBP1, and IGFBP3 SNP characteristics IGF1 IGFBP1 IGFBP3 Exons 251 UTR 64 9 11 Intrageneic 226 34 26 Total 292 48 38 Chromosome length (kb) 84.6 5.2 9.0 SNP density (SNP/bp) 1/289 1/108 1/237 Section 5. Haplotype-based Approach SNPs and Linkage Disequilibrium Individuals are 99.9% identical at the nucleotide level (Sachidanandam, Weissman et al. 2001). The majority of the 0.1% difference consists of single 54 nucleotide polymorphisms (SNPs). SNPs occur every 100 to 300 bases in the 3- billion base human genome and over 10 million SNPs have been catalogued to date (http://www.ncbi.nlm.nih.gov/SNP/). A series of SNP alleles along a stretch of a single chromosome is described as a haplotype. Over successive generations, recombination and mutational events will result in the rearrangement of an ancestral haplotype with a new SNP allele (Olivier 2003). This results in the new allele being located near the ancestral haplotype, and as both are inherited together, they are considered to be tightly linked; a property referred as linkage disequilibrium (LD) (Ardlie, Kruglyak et al. 2002). LD is a function of time and the recombination distance between SNPs, where the extent of LD decreases with the increasing distance between SNPs and the number of generational events (Ardlie, Kruglyak et al. 2002). Other factors that may also impact LD include: genetic drift, population growth, admixture/migration, population structure, natural selection, variable recombination rates, variation mutation rates, and gene conversion (Ardlie, Kruglyak et al. 2002). The D’ and r 2 statistics, both ranging from 0 to 1, estimate the extent of linkage between SNP pairs and are the most frequently used measures of LD. A D’ = 1 indicates complete LD, where a SNP pair have not been separated by recombination, and a D’ < 1 signifies a disruption in the ancestral haplotype (Ardlie, Kruglyak et al. 2002). The D’ statistic estimates the differences between the observed two-locus haplotype frequency and the expected frequency if the alleles segregated at random (Ardlie, Kruglyak et al. 2002). Estimates of D’ may be inflated due to small sample sizes and low 55 allele frequencies (Weiss and Clark 2002). The r 2 statistic quantifies the correlation between two SNP alleles, where an r 2 = 1 indicates perfect LD and no recombination; the information of one allele provides complete or redundant information of the other allele (Ardlie, Kruglyak et al. 2002). The r 2 measure in contrast to D’ is not influenced by differences in allele frequencies, however, low values of r 2 does not necessarily signify strong evidence of recombination (Ardlie, Kruglyak et al. 2002). Previous studies have compared the LD patterns among non-African and African populations (Tishkoff, Dietzsch et al. 1996; Frisse, Hudson et al. 2001; Wall 2001; Weiss and Clark 2002). These studies have shown that European populations tend to have longer range LD in comparison to African populations. Reich et al. (2001) examined the LD patterns of 19 randomly selected genomic regions in non-African populations (U.S. and European) and an African-Yoruban population (Reich, Cargill et al. 2001). LD was estimated to extend 60 kb around common alleles in the U.S. population and a much shorter range in LD was observed among Yorubans. The authors hypothesized that the LD in northern Europeans was shaped by either an extreme founder effect or population bottleneck approximately 27,000-53,000 years ago. Haplotype Blocks and haplotype-tagging SNPs Recent studies of LD have reported that the human genome can be parsed into discrete haplotype blocks, regions of strong LD flanked by sites of historical 56 recombination (Daly, Rioux et al. 2001; Patil, Berno et al. 2001; Dawson, Abecasis et al. 2002; Gabriel, Schaffner et al. 2002). These regions exhibit limited haplotype diversity, where a few number of haplotypes account for the majority of the observed chromosomes. In the largest survey of the human genome, Gabriel et al. (2002) examined 51 genomic regions for a total distance of ~13 MB (1 SNP every 7.8 kb) (Gabriel, Schaffner et al. 2002). The authors estimated that half of the genome is comprised of blocks 22 kb or larger for African populations and 44 kb or larger for European and Asian populations. These haplotype blocks could be detected by 6-8 SNPs (minor allele frequency > 10% ; 1 SNP every 5-10 kb), and within each block three to five haplotypes could account for more than 90% of the common (>5%) haplotype diversity. This research has spurred an international effort of 9 groups in U.S., Canada, U.K., Japan, and China, the International Hap Map project, to generate a detailed map of the common haplotype structure of the human genome with the goal of identifying a subset of SNPs that captures the complete haplotype diversity. Many SNPs are informatively equivalent within regions of strong LD, thus only a subset of SNPs, known as haplotype-tagging SNPs (htSNPs), is necessary to describe the full complement of haplotypes. This allows for predicting the allele of a neighboring SNP based on knowing the SNP allele of its linked partner. Stram et al provide an efficient method to select htSNPs based on estimating the haplotype frequencies using the Excoffier-Slatakin E-M algorithm and calculating the uncertainty of the number of common haplotypes by a formal calculation of 57 R 2 h , the coefficient of determination (Stram, Haiman et al. 2003). A set of haplotype-tagging SNPs are then selected that maximize the predictability of the common haplotypes with a minimum R 2 h. Once these htSNPs are identified they may be tested in association studies to examine the relationship between haplotypes and disease risk (Stram, Haiman et al. 2003). Association Studies Since SNPs comprise the vast majority of human genetic variation, these single nucleotide variations are thought to regulate the majority of phenotypic variation in the population, including disease risk, the Common Disease Common Variant hypothesis. This hypothesis is the basis of association studies whereby disease-associated variants are identified by linkage disequilibrium to determine the ancestral haplotype in which the causal variant first arose. This approach has been shown to be a successful and powerful strategy for identifying genomic regions that harbor disease-associated variants in studies of diabetes, Chron’s disease, and cancer (Rioux, Daly et al. 2001; Haiman, Stram et al. 2003; Florez, Burtt et al. 2004; Pearce, Hirschhorn et al. 2005). 58 PART II. DATA ANALYSIS Section 1. Association study of common genetic variation in IGF1 and prostate cancer risk in the Multiethnic Cohort Iona Cheng 1 , Daniel O Stram 1 , Kathryn L Penney 2,3,4,6 , Malcolm Pike 1 , Loic Le Marchand 9 , Laurence N. Kolonel 9 , Joel Hirschhorn 2,3,5,10 , David Altshuler 2,3,4,6,7 , Brian E Henderson 1 , Matthew L Freedman 2,3,4,6,8 1 Department of Preventive Medicine, Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, Los Angeles, CA 90033, USA 2 Program in Medical and Population Genetics, Broad Institute of Harvard and Massachusetts Institute of Technology, Cambridge, MA 02139, USA 3 Department of Genetics, 4 Medicine, and 5 Pediatrics, Harvard Medical School, Boston, MA 02115, USA 6 Department of Molecular Biology, 7 Diabetes Unit, and 8 Hematology-Oncology, Massachusetts General Hospital, MA 02214, USA 9 Cancer Etiology Program, Cancer Research Center of Hawaii, University of Hawaii, Honolulu, HI 96813, USA 10 Division of Genetics and Endocrinology, Children’s Hospital and Department of Pediatrics, Boston, MA 02115, USA Address for correspondence: Matthew Freedman, Program in Medical and Population Genetics, Broad Institute of Harvard and MIT, Cambridge, MA 02139, USA (e-mail: freedman@broad.mit.edu) 59 Abstract Background: Insulin-like growth factor-I (IGF-I) is a mitogen involved in regulating cellular growth in the prostate. Higher circulating levels of IGF-I have been associated with increased risk of prostate cancer. It has yet to be determined, however, whether inherited variation at the IGF1 locus is associated with prostate cancer risk. Methods: To thoroughly evaluate the role of genetic variation in IGF1 in relation to prostate cancer risk, we: 1) sequenced IGF1 exons in 95 advanced prostate cancer cases to identify any amino acid altering variants; 2) characterized linkage disequilibrium patterns and common haplotypes at the IGF1 locus by genotyping 64 SNPs spanning 156 kb in 349 controls from five ethnic groups; and 3) conducted an association study of IGF1 haplotypes and genotypes among 2,320 prostate cancer cases and 2,290 controls from the Multiethnic Cohort. Results: No missense SNPs were found after sequencing 95 prostate cancer cases. We identified four blocks of strong linkage disequilibrium and selected 29 tagging SNPs to predict both the common IGF1 haplotypes and the remaining SNPs. Haplotype analysis revealed nominally significant effects in each of the four haplotype blocks: haplotype 1B (OR = 1.21; 95% CI: 1.04-1.40), haplotype 2C (OR = 1.24; 95% CI: 1.06-1.44), haplotype 3C (OR = 1.25; 95% CI: 1.03- 1.50), and haplotype 4D (OR = 1.19; 95% CI: 1.02-1.39). SNP analysis revealed two perfectly correlated SNPs located in block 1 that were associated with prostate cancer risk and also could explain the haplotype findings: rs7978742 (P trend = 0.002) and rs7965399 (P trend = 0.002). Permutation testing of the 29 60 tSNPs provided a gene-wise empirical P value of 0.056. Conclusion: Our findings suggest that inherited variation in IGF1 may predispose men to prostate cancer risk. 61 Introduction Insulin-like growth factor-I stimulates cellular proliferation and inhibits apoptosis (Jones and Clemmons 1995). Prostate epithelial cells exhibit mitogenic and anti-apoptotic effects in response to IGF-I in vitro (Cohen, Peehl et al. 1991; Pietrzkowski, Mulholland et al. 1993; Ngo, Barnard et al. 2003), and in vivo studies have demonstrated the importance of IGF-I in normal prostate development and tumorigenesis (Kaplan, Mohan et al. 1999; Ruan, Powell- Braxton et al. 1999; DiGiovanni, Kiguchi et al. 2000). Prospective studies report that men at the upper 25% of circulating levels of IGF-I have an increased risk of prostate cancer compared to men at the lowest 25% of IGF-I levels (OR meta = 1.22; 95% CI: 0.91-1.62) (Schaefer 1998; Harman, Metter et al. 2000; Lacey, Hsing et al. 2001; Chan, Stampfer et al. 2002; Woodson, Tangrea et al. 2003; Stattin, Rinaldi et al. 2004; Chen, Lewis et al. 2005). In the largest study of 530 prostate cancer cases, an extension of the initial report of the Physicians’ Health Study, men in the highest quartile of IGF-I in comparison to those in the lowest quartile had a five-fold significant increased risk of prostate cancer of an advanced stage (Chan, Stampfer et al. 2002). Previous studies that have investigated the role of genetic variation in IGF1 in relation to prostate cancer risk have focused solely on a (CA) n repeat located approximately 1 kb upstream from the IGF1 transcription start site (Nam, Zhang et al. 2003; Li, Cicek et al. 2004; Neuhausen, Slattery et al. 2005; Schildkraut, Demark-Wahnefried et al. 2005; Tsuchiya, Wang et al. 2005). These 62 studies have been inconsistent with two studies reporting a positive association between men homozygous for the (CA) 19 repeat and prostate cancer risk, while one study finding an inverse association, and two other reports observing no association (Nam, Zhang et al. 2003; Li, Cicek et al. 2004; Neuhausen, Slattery et al. 2005; Schildkraut, Demark-Wahnefried et al. 2005; Tsuchiya, Wang et al. 2005). To date, no studies have systematically examined the genetic diversity of IGF1 in relation to prostate cancer risk, or done so in sample sizes adequate to evaluate even modest genetic effects. To more comprehensively evaluate this relationship, we conducted a large nested case-control study of 2,320 cases and 2,290 controls from the Multiethnic Cohort Study (MEC). Methods The Multiethnic Cohort (MEC) is a large population-based cohort study of over 215,000 men and women from Hawaii and Los Angeles. The cohort is comprised of predominantly five racial/ethnic groups: African-Americans, Native Hawaiians, Japanese, Latinos, and Whites. Participants between the ages of 45 and 75 years were recruited from 1993 to 1996 and completed a 26-page self- administered questionnaire that included information regarding medical history, family history, diet, dietary supplements and medication use, and physical activity. Further details are provided elsewhere (Kolonel, Henderson et al. 2000). 63 Incident cancers in the MEC were identified (up to April 1, 2002) by cohort linkage to population-based Surveillance, Epidemiology and End Results (SEER) cancer registries covering Hawaii and California. Information on stage of disease at the time of diagnosis was also collected from the cancer registries. We defined localized and advanced disease based on tumor stage and differentiation (Gleason grade). Tumors confined to the prostate having a Gleason grade < 8 were defined as localized disease. Regional and metastatic tumors, or localized tumors with Gleason grade > 8 were defined as advanced disease. Controls were men without prostate cancer prior to entry into the cohort and without a prostate cancer diagnosis up to April 1, 2002. Controls were randomly selected from our random control pool of MEC participants, who provided blood specimens for genetic analyses in the cohort. The participation rate for blood collection was 74% for cases and 66% for controls. Controls were frequency matched to cases by age and ethnicity. This case-control study consisted of 2,320 prostate cancer cases and 2,290 controls. This study was approved by the Institutional Review Boards at the University of Hawaii and the University of Southern California. Sequencing To identify any missense SNPs not in public or private databases we attempted to sequence the four IGF1 exons in 95 cases of advanced prostate cancer (19 per racial/ethnic group). With three attempts Exon 2 did not meet our 64 sequencing criteria of more than 80% of the samples with phred scores > 20 for at least 80% of the target bases. Sequencing was performed by conventional dye- primer sequencing on ABI 3700 Sequencers. The PolyPhred program was used to identify polymorphisms with manual review by at least two observers, and all putative coding variants were validated by SNP genotyping. Further details are described elsewhere (Freedman, Pearce et al. 2005). SNP selection and genotyping for genetic characterization To characterize patterns of linkage disequilibrium across the gene, we genotyped single nucleotide polymorphisms (SNPs) spanning 156 kb across the IGF1 locus (from 23.4 kb upstream of exon 1 to 47.8 kb downstream of the transcribed region). We selected SNPs from the public (http://www.ncbi.nlm.nih.gov/SNP/) and private database (http://www.celera.com/). Our goal was to achieve a SNP density of 1 SNP every 2-5 kb across the locus. We preferentially selected SNPs located in coding and UTR regions as well as areas of mouse homology (80% identity in 200 bp sequence; http://www.dcode.org/). For genetic characterization, 154 SNPs were genotyped in a multiethnic panel of 349 individuals with no history of cancer: African-American (n = 70), Hawaiian (n = 69), Japanese (n = 70), Latina (n = 70), and White (n = 70) (Haiman, Stram et al. 2003). SNP genotyping was performed using the Sequenom mass spectrometry platform (Sequenom Inc. San Diego, CA). Of the 154 SNPs genotyped, 53 SNPs were identified as monomorphic and 65 37 SNPs displayed poor genotyping results (genotyped < 75% of samples or out of Hardy-Weinberg equilibrium (1-sided P < 0.01) in > 1 ethnic group); these SNPs were eliminated from further analysis. The remaining 64 SNPs were utilized for genetic characterization having an average density of 1 SNP every 2.4 kb (Table 7). Haplotype block determination and tSNP selection The D’ statistic was used to determine the pair-wise linkage disequilibrium (LD) between the 64 SNPs. Regions of strong LD, haplotype blocks, were defined following the methods of Gabriel et al. (Gabriel, Schaffner et al. 2002) To insure thorough coverage across each haplotype block, our goals were to genotype a minimum of 6 SNPs with minor allele frequencies (maf) > 10% within each block with the minimum distance of < 10 kb between neighboring blocks. Each ethnic group was evaluated for the extent of LD meeting block criteria. Hawaiians, Japanese, Latinos, and Whites shared similar LD patterns. We combined these populations to assess the LD structure of the locus. African-Americans, as expected, displayed a smaller extent of LD. In an effort to more fully characterize variation in the African-American population, we genotyped an additional 12 SNPs in regions where there were insufficient numbers of SNPs to formally meet our criteria of strong LD. This additional set consisted of all publicly available SNPs in dbSNP as of April 2004. Despite this 66 Table 7. Sixty-four SNPs utilized for IGF1 genetic analysis Nucleotide Minor † SNP# SNP Location Position * Change Allele AA HA JA LA WH 1 rs855228 5' 101400231 C/T C 63.2 26.8 31.4 30.4 18.8 2 rs7973621 5' 101399369 G/T T 28.5 -- -- 5.1 3.6 3 rs7978742 5' 101394219 A/G G 30.8 18.0 24.6 24.2 6.2 4 ‡ rs7965399 5' 101394153 C/T C 30.6 20.3 25.4 23.9 7.1 5 ‡ rs35765 5' 101384163 G/T T 27.9 6.5 5.7 7.1 10.9 6 rs1457601 5' 101383731 A/T T 1.5 17.6 26.1 19.1 2.9 7 hCV3061153 5' 101379786 A/C A 1.4 18.8 29.3 18.8 2.9 8 ‡ rs35767 5' 101378036 A/G A 46.4 26.5 33.6 30.7 18.1 9 ‡ rs2288377 5' 101377229 A/T T 0.7 21.6 29.4 19.6 2.9 10 rs2162679 I1 101373726 C/T C 44.2 25.0 34.1 27.5 16.7 11 ‡ hCV11752586 I2 101370134 A/G A 5.1 14.5 3.7 25.0 19.9 12 ‡ rs1019731 I2 101366892 A/C A 5.1 8.1 -- 19.6 12.9 13 hCV2801087 I2 101362446 A/G G 7.1 -- -- -- -- 14 ‡ hCV2801089 I2 101361295 C/G C 1.5 15.9 25.4 16.2 2.2 15 rs5742629 I2 101359730 C/T C 34.6 39.0 45.7 28.4 27.9 16 ‡ rs2195239 I2 101359169 C/G G 30.7 39.1 44.9 27.2 23.2 17 ‡ rs2195240 I2 101359114 A/G G 24.6 39.6 45.6 26.8 23.9 18 hCV2801091 I2 101355710 C/T T 5.1 -- -- 2.1 3.0 19 rs4764697 I2 101355639 C/T T 21.4 23.9 17.9 12.9 25.7 20 ‡ hCV2801093 I2 101351068 C/T C 6.5 -- -- -- -- 21 rs5742637 I2 101350809 C/T T 0.7 16.4 27.5 16.2 0.7 22 rs5742639 I2 101350713 A/G A 1.6 18.4 27.7 15.7 1.5 23 ‡ hCV346219 I2 101346703 A/G G 18.8 23.9 20.3 13.0 23.5 24 ‡ rs2373722 I2 101342924 A/G A 7.9 3.6 -- 3.6 5.0 25 ‡ rs5742657 I2 101337335 A/G G 11.2 17.6 26.9 16.2 1.4 26 hCV2801098 I2 101336724 A/G G 10.1 -- -- -- -- 27 hCV2801100 I2 101335776 A/G A 2.9 8.7 0.7 7.9 15.0 28 hCV2801101 I2 101334189 G/T T 34.1 41.9 47.9 29.0 22.9 29 rs2288378 I2 101332475 C/T T 22.8 23.9 -- 12.7 21.4 30 rs2373721 I2 101329512 C/G G 22.8 22.7 19.6 12.7 21.4 31 rs972936 I2 101327388 C/T T 34.1 43.4 48.6 29.7 23.2 32 ‡ hCV2801104 I2 101326017 C/G G 2.9 7.2 0.7 7.4 14.3 33 hCV2801105 I2 101323622 A/C C 22.1 23.9 20.3 13.2 20.6 34 hCV2801106 I2 101322854 A/G A 9.2 16.9 22.3 13.3 1.5 35 rs4764884 I2 101322066 C/T T 23.1 24.3 19.9 12.5 22.4 36 hCV2801109 I2 101316184 A/G G 9.3 -- -- -- -- 37 rs2072592 I2 101316099 C/T T 1.4 15.9 27.9 16.7 1.4 38 ‡ hCV2801111 I3 101306360 A/T A 49.3 42.6 47.8 32.4 22.9 39 ‡ rs4764882 I3 101305810 A/C A -- -- 1.5 12.5 0.7 40 hCV2801114 I3 101302065 A/G G 10.9 19.6 27.2 17.4 1.5 41 ‡ rs1549593 I3 101299258 G/T T 3.7 7.5 -- 20.3 12.1 42 ‡ rs1520220 I3 101298989 C/G G 34.3 39.9 47.9 23.9 16.7 43 rs6220 3'UTR 101296982 A/G G 37.5 43.4 48.5 29.9 25.4 44 ‡ rs6218 3'UTR 101296100 A/G G -- 15.0 27.3 11.9 0.8 45 rs6214 3'UTR 101296036 C/T T 57.7 44.1 51.5 42.7 45.6 46 rs6219 3'UTR 101292659 C/T T 6.0 16.7 20.6 3.0 9.9 47 ‡ rs5742723 3' 101290839 G/T T 1.5 15.6 27.1 17.7 2.2 48 ‡ rs2946834 3' 101290281 A/G A 54.4 47.1 50.0 40.0 33.6 49 rs2946831 3' 101289247 A/C C 17.2 3.7 -- 4.5 3.7 50 rs2971577 3' 101286022 A/G G 12.1 3.6 -- 5.0 5.7 51 rs2971575 3' 101283346 A/G G 54.3 52.2 50.0 44.8 42.5 52 ‡ rs2139573 3' 101281241 C/T C 32.1 36.0 25.4 47.8 48.6 53 ‡ rs4764880 3' 101270621 C/T T 0.7 12.1 26.1 19.5 2.1 54 ‡ rs2139572 3' 101270127 C/T T 2.4 9.2 8.3 -- 1.5 55 rs2139569 3' 101266871 C/T T 36.2 34.2 32.5 52.7 47.5 56 ‡ rs2139570 3' 101266684 G/T T 35.3 35.5 35.5 51.4 47.8 57 rs1106381 3' 101264923 G/T G 1.5 17.2 30.2 16.4 2.9 58 ‡ rs4764876 3' 101261169 C/G C 47.1 44.2 49.3 32.9 28.6 59 ‡ rs4764695 3' 101259580 A/G A 26.5 36.0 35.5 50.0 48.6 60 rs7132028 3' 101259202 C/T T 10.9 -- -- 1.5 -- 61 ‡ rs1520219 3' 101258042 C/T T 5.7 -- -- 0.7 0.7 62 ‡ rs1996656 3' 101254429 A/G G 8.8 20.6 19.6 8.2 17.1 63 rs2971578 3' 101246956 C/T C 2.9 8.1 -- 21.6 10.7 64 rs1520218 3' 101244363 A/C A 37.5 44.6 49.2 30.8 29.6 * SNP position based on July 2003 (UCSC version human genome16) † Minor allele based on all groups combined ‡ tSNP Allele Frequency (%) 67 extra effort, there were still an insufficient number of SNPs to completely fulfill our minimum 6 SNP criterion for this ethnic group. Genotype data for each ethnic group in the multiethnic panel was used to estimate haplotype frequencies within blocks using the Expectation-Maximization (E-M) algorithm (Excoffier and Slatkin 1995). The squared correlations (R 2 h ) between the true haplotypes (h) and their estimates from the E-M algorithm were estimated as described by Stram et al. (Stram, Haiman et al. 2003). For each ethnic group, we then selected the minimum set of tagging SNPs (tSNPs) within each block to assure an R 2 h > 0.7 for all haplotypes with an estimated frequency > 5% (Stram, Haiman et al. 2003). Genotyping in the case-control study The tSNPs were genotyped in the prostate case-control study using the 5’ nuclease Taqman allelic discrimination assay (Applied Biosystems, Foster City, CA, USA). All assays were performed blinded to case-control status. For quality control, 5% replicate samples were included. The concordance for replicate samples was 99.5%. The average successful genotyping percentage was 98.5%. Haplotype case-control analysis To utilize linkage disequilibrium to capture unmeasured variants, we evaluated the relationship between IGF1 haplotypes and prostate cancer risk. Haplotype frequencies among prostate cancer cases and controls were estimated 68 by using genotype data of the tSNPs as described by Stram et al. (Stram, Haiman et al. 2003). Haplotype dosage (i.e. an estimate of the number of copies of haplotype h) for each individual and each haplotype, h, was computed using that individual’s genotype data and haplotype frequency estimates obtained from the E-M algorithm (Zaykin, Westfall et al. 2002). A likelihood ratio test was performed to provide a global test for associations with the common haplotypes (haplotype frequency > 5%) in each block. Odds ratios (ORs) and 95% Confidence Intervals (CIs) for each common haplotype were estimated by unconditional logistic regression using all other observed haplotypes as the reference category. SNP case-control analysis To investigate the hypothesis that genetic susceptibility to prostate cancer risk may be associated with single causal variants, we evaluated the relationship between IGF1 genotypes and disease risk. Using the 29 tSNPs and genotype data obtained from the multiethnic panel, we estimated for each individual the allelic distribution of the remaining 35 SNPs that were not genotyped in the case-control study, which we refer herein as “unmeasured SNPs”. Within each region of strong LD using genotype data from either an individual tSNP or combination of tSNPs, we predicted each individual’s genotype for the unmeasured SNPs using the squared correlations (R 2 s ) between each SNP and their estimates obtained from the E-M algorithm. SNPs between regions of strong LD were predicted by 69 including them as part of a neighboring block if they met a minimum R 2 s > 0.7 for each ethnic group. ORs and 95% CIs were estimated by unconditional logistic regression for the association between IGF1 genotypes and prostate cancer risk. Stepwise logistic regression analysis was conducted to evaluate which combination of SNPs provided the best fit to prostate cancer risk. We conducted permutation testing to guide interpretation of nominally significant SNP associations. Case-control status within strata of racial/ethnic group was randomly permuted 5,000 times for the 29 tSNPs. The smallest P value for the associated SNP was examined in relation to the permutation distribution of minimal P values associated with the heterozygote and homozygote mutant variants (homozygote wild type as the reference). For example, if a nominal P value of 0.05 marked the 25 th percentile of this distribution, then the permutation P value (one-sided) would be 0.25. All P values quoted are 2-sided unless otherwise stated. Results Genetic characterization of the IGF1 locus We sequenced three of the four IGF1 exons in each of 95 individuals with prostate cancer (see Methods for full details) and found no missense polymorphisms. We genotyped 64 SNPs spanning 156 kb of the IGF1 locus (23.4 kb upstream, 47.8 kb downstream) in a multiethnic panel of 349 individuals (Table 7). We identified four regions of strong linkage disequilibrium (LD): block 70 1 (SNPs 1-9; size = 23 kb) spanned the upstream region of IGF1, block 2 (SNPs 11-17; size = 11 kb) included intron 2, block 3 (SNPs 19-44; size = 60 kb) spanned the majority of the gene from intron 2 through the 3’UTR, and block 4 (SNPs 46-62; size = 38 kb) covered predominantly the downstream region of IGF1 (Figure 1). The distances between adjacent blocks were < 7 kb. Common haplotypes were inferred within each region of LD and their corresponding frequencies are shown in Table 8. In blocks 1 and 2, we observed five and six common ( > 5%) haplotypes, respectively. For blocks 3 and 4, we observed 12 common haplotypes in each of these blocks. A total of 29 tagging SNPs (tSNPs) were able to predict the common haplotypes as well as 33 out of the 35 unmeasured SNPs across the IGF1 locus. The common haplotypes for each ethnic group accounted for 77-100% of the chromosomes in the panel population. Prostate case-control characteristics Characteristics of the 2,320 cases and 2,290 controls are presented in Table 9. The mean age at reference (age of diagnosis for cases and age at blood draw for controls) was 68.3 and 67.9 years, respectively. Approximately 20-30% of the study population were of each ethnic group, with the exception of Hawaiians (3%). As expected, cases were more likely than controls to report a family history of prostate cancer (either affected father or brother) (P < 0.001). All of the following results were only affected slightly when adjustment was 71 Figure 3. IGF1 locus spanning 156 kb on chromosome 12 (human genome assembly 16). LD blocks (n = 4) and haplotype patterns ( > 5%) are presented for all ethnic groups combined. 72 Table 8. Common haplotypes in blocks 1-4 of IGF1 estimated by tSNPs in the multiethnic panel * Haplotypes African- Americans Hawaiians Japanese Latinos Whites All Block 1: SNPs 1-9; tSNPs: 4,5,8, 9 1A TGGA 36 72 65 67 82 65 1B CGAT 20 25 18 13 1C TTAA 2867 11 11 1D CGAA 13 1E CGGA 18 Percentage of Chromosomes Observed (%) 95 97 90 92 93 89 R 2 h 0.91 0.82 0.99 0.90 1.00 0.94 Block 2: SNPs 11-17; tSNPs: 11, 12, 14, 16, 17 2A GCGCA 64 46 51 47 56 53 2B GCGGG 22 23 19 11 21 19 2C GCCGG 16 26 16 12 2D AAGCA 8 19 13 9 2E ACGCA 7775 2F GCGGA 7 Percentage of Chromosomes Observed (%) 93 100 96 100 98 98 R 2 h 0.89 1.00 1.00 0.92 0.84 0.91 Block 3: SNPs 19-44; tSNPs: 20, 23, 24, 25, 32, 38, 39, 41, 42, 44 3A TAGACTCGCA 45 44 52 42 53 47 3B TGGACACGGA 9 17 19 11 12 3C TAGGCACGGG 16 25 9 3D TAGACTCTCA 6 19 98 3E TAGAGTCGCA 56 13 6 3F TGAACACGGA 8 3G TGGACACGCA 5 3H TAGACACGCA 13 3I TAGGCAAGGG 12 3J TAGACACGGA 5 3K CAGGCACGGA 7 3L TAGGCACGGA 5 Percentage of Chromosomes Observed (%) 92 88 96 85 85 81 R 2 h 0.81 0.81 0.86 0.84 0.82 0.80 Block 4: SNPs 46-62; tSNPs: 47, 48, 52, 53, 54, 56, 58, 59, 61, 62 4A GGCCCTGACA 19 32 26 44 43 32 4B GGTCCGGGCA 20 15 14 11 16 16 4C GATCCGCGCA 29 8 10 12 12 4D TATTCGCGCA 12 25 18 11 4E GATCTGCGCG 15 14 6 4F GATCCGCGCG 75 4G GACCCTGACA 5 4H GGTCCGGGCG 4 4I TATCCGCGCA 5 4J GGTCCTGACA 9 4K GACCCTCGCA 5 4L GGTCCGGGTA 5 Percentage of Chromosomes Observed (%) 77 87 88 83 87 82 R 2 h 0.79 0.97 0.99 0.91 0.86 0.80 * Haplotypes > 5% in at least one ethnic group in the multiethnic panel Haplotype frequencies (%) 73 Table 9. Study characteristics of prostate cancer cases and controls in the Multiethnic Cohort Study Cases (N=2320) Controls (N=2290) Age (mean + SD) 68.3 + 6.5 67.9 + 6.8 Ethnicity N (%) African-American 683 (29) 650 (28) Hawaiian 71 (3) 68 (3) Japanese 462 (20) 471 (21) Latino 647 (28) 649 (28) White 457 (20) 452 (20) Family history of prostate cancer N (%)* 253 (12) 177 (8) Localized disease N (%) * 1406 (61) Advanced disease N (%) * 762 (33) * Numbers do not add to 100% due to missing data. made for the risk factor of family history of prostate cancer and the unadjusted values are given below. Case-control analysis We first searched for haplotype effects in global tests of differences in risk, finding statistically significant associations in block 1 (P = 0.021), block 2 (P = 0.044), but not in block 3 (P = 0.374) or block 4 (P = 0.278). At least one haplotype within each block was nominally associated with prostate cancer risk (Table 10). We observed positive associations between risk and haplotype 1B (OR = 1.21; 95% CI: 1.04-1.40), haplotype 2C (OR = 1.24; 95% CI: 1.06-1.44), haplotype 3C (OR = 1.25; 95%CI: 1.03-1.50) and haplotype 4D (OR = 1.19; 95% CI: 1.02-1.39). These haplotypes also demonstrated a consistent pattern with increased risk of prostate cancer across all racial/ethnic groups, with the exception 74 Table 10. Associations between common haplotypes in blocks 1-4 of IGF1 and prostate cancer risk * All groups combined All groups combined All groups combined All groups combined BLOCK 1 OR † (95% CI) BLOCK 2 OR † (95% CI) BLOCK 3 OR † (95% CI) BLOCK 4 OR † (95% CI) 1A TGGA 0.91 (0.83-0.99) ‡ 2A GCGCA 1.00 (0.93-1.09) 3A TAGACTCGCA 0.93 (0.86-1.01) 4A GGCCCTGACA 0.93 (0.85-1.02) 1B CGAT 1.21 (1.04-1.40) ‡ 2B GCGGG 0.97 (0.87-1.07) 3B TGGACACGGA 0.95 (0.84-1.08) 4B GGTCCGGGCA 0.94 (0.83-1.07) 1C TTAA 0.92 (0.81-1.04) 2C GCCGG 1.24 (1.06-1.44) § 3C TAGGCACGGG 1.25 (1.03-1.50) ‡ 4C GATCCGCGCA 0.99 (0.87-1.13) 1D CGAA 1.10 (0.91-1.34) 2D AAGCA 1.01 (0.87-1.18) 3D TAGACTCTCA 1.00 (0.85-1.16) 4D TATTCGCGCA 1.19 (1.02-1.39) ‡ 1E CGGA 1.19 (0.97-1.47) 2E ACGCA 0.83 (0.69-0.99) ‡ 3E TAGAGTCGCA 1.07 (0.89-1.28) 4E GATCTGCGCG 1.06 (0.90-1.24) 2F GCGGA 1.00 (0.75-1.34) 3F TGAACACGGA 1.08 (0.88-1.34) 4F GATCCGCGCG 1.03 (0.82-1.29) 3G TGGACACGCA 0.93 (0.73-1.19) 4G GACCCTGACA 1.03 (0.80-1.33) 3H TAGACACGCA 1.08 (0.86-1.36) 4H GGTCCGGGCG 0.90 (0.70-1.16) 3I TAGGCAAGGG 1.09 (0.84-1.43) 4I TATCCGCGCA -- 3J TAGACACGGA 1.14 (0.83-1.56) 4J GGTCCTGACA 1.00 (0.71-1.42) 3K CAGGCACGGA 1.13 (0.85-1.52) 4K GACCCTCGCA 0.94 (0.68-1.29) 3L TAGGCACGGA 1.23 (0.88-1.73) 4L GGTCCGGGTA 0.90 (0.69-1.17) * ORs are estimated by unconditional logistic regression for each haplotype versus all others. † ORs are estimated by unconditional logistic regression adjusted for age and ethnicity. ‡ P value < 0.05 § P value < 0.01 75 of Hawaiians, where power was limited due to a small sample size (Table 11). The haplotypes associated with risk (1B, 2C, 3C and 4D) seem to represent the same long range haplotype and therefore likely reflect the same underlying signal (Figure 2). Thirty-three of the 35 unmeasured SNPs in the case-control study were predicted by the set of 29 tSNPs with an average R 2 s = 0.91. SNP 18 (hCV2801091) and SNP 63 (rs2971578), which are both in areas of weaker LD, could not be predicted with an R 2 s > 0.7 in the five racial/ethnic groups (SNP 18, R 2 s = 0.04-0.38; SNP 63, R 2 s = 0.17-0.38). Figure 3 displays the association between prostate cancer risk and the sixty-three SNPs (29 tSNPs, 33 predicted SNPs and SNP 18). We were unable to genotype SNP 63 as a Taqman assay could not be manufactured. We observed nominally significant associations (P trend < 0.05) between prostate cancer risk and 17 SNPs (Table 12). Of the 17 SNPs, we identified SNP 3 (unmeasured SNP) and SNP 4 (tSNP) are perfectly correlated (R 2 s = 1.00 for all five racial/ethnic groups) and provided the best fit to prostate cancer risk through a stepwise logistic regression analysis. No other SNPs were statistically significant after adjustment for SNP 3 or SNP 4. SNP 3 and SNP 4 were also able to explain our haplotype findings in a similar stepwise logistic regression analysis of SNP 3, SNP 4, and associated haplotypes. SNP 3 (rs7978742) and SNP 4 (rs7965399), which are located ~17 kb upstream of the transcription start site, revealed the strongest nominal associations 76 with prostate cancer risk (P for trend = 0.002). Men carrying the CT genotype for Table 11. Odds ratios and 95% CI of prostate cancer associated IGF1 haplotypes by ethnic group * Haplotypes African Americans Hawaiians Japanese Latinos Whites OR † (95% CI) OR † (95% CI) OR † (95% CI) OR † (95% CI) OR † (95% CI) 1B CGAT 2.40 (0.99-5.84) ‡ 1.20 (0.68-2.11) 1.11 (0.89-1.37) 1.12 (0.86-1.46) 2.87 (1.41-5.88) § localized 2.87 (1.13-7.29) ‡ 1.13 (0.59-2.15) 1.22 (0.95-1.56) 1.09 (0.80-1.49) 2.81 (1.32-6.00) § advanced 0.53 (0.06-4.33) 1.48 (0.68-3.23) 1.04 (0.79-1.38) 1.21 (0.84-1.74) 3.24 (1.31-8.00) ‡ 2C GCCGG 3.51 (1.24-9.87) ‡ 0.77 (0.37-1.63) 1.19 (0.97-1.48) 1.15 (0.90-1.48) 2.74 (1.30-5.77) § localized 4.15 (1.43-12.08) § 1.02 (0.46-2.24) 1.31 (1.02-1.68) ‡ 1.06 (0.79-1.43) 2.49 (1.10-5.64) ‡ advanced 0.78 (0.09-6.90) 0.36 (0.09-1.55) 1.14 (0.85-1.51) 1.31 (0.94-1.83) 3.35 (1.36-8.24) § 3C TAGGCACGGG 1.69 (0.40-7.11) 0.77 (0.37-1.63) 1.25 (1.00-1.56) ‡ 1.28 (0.81-2.00) 5.21 (0.83-32.65) localized 2.06 (0.46-9.28) 1.02 (0.46-2.24) 1.33 (1.02-1.72) ‡ 1.04 (0.60-1.79) 6.12 (0.90-41.40) advanced -- 0.36 (0.09-1.55) 1.19 (0.89-1.59) 1.69 (0.97-2.97) 4.67 (0.49-44.25) 4D TATTCGCGCA 2.23 (0.68-7.28) 0.95 (0.43-2.13) 1.16 (0.93-1.45) 1.18 (0.92-1.51) 2.28 (0.73-7.14) localized 3.01 (0.90-10.09) 1.19 (0.50-2.83) 1.18 (0.90-1.53) 1.12 (0.84-1.50) 2.17 (0.62-7.64) advanced -- 0.58 (0.16-2.14) 1.17 (0.87-1.56) 1.31 (0.94-1.83) 2.79 (0.69-11.22) * ORs are estimated by unconditional logistic regression for each haplotype versus all others. † ORs are estimated by unconditional logistic regression adjusted for age. ‡ P value < 0.05 § P value < 0.01 Figure 4. Haplotype structure of blocks 1-4 of IGF1 using tSNPs in all ethnic groups combined. Haplotypes outlined in red correspond to haplotypes 1B, 2C, 3C, and 4D. Lines indicate haplotype frequency (bold=frequency > 5%) 1 3 4 2 77 0.00 0.50 1.00 1.50 2.00 2.50 3.00 snp1 snp2 snp3 snp4 snp5 snp6 snp7 snp8 snp9 snp10 snp11 snp12 snp13 snp14 snp15 snp16 snp17 snp18 snp19 snp20 snp21 snp22 snp23 snp24 snp25 snp26 snp27 snp28 snp29 snp30 snp31 snp32 snp33 snp34 snp35 snp36 snp37 snp38 snp39 snp40 snp41 snp42 snp43 snp44 snp45 snp46 snp47 snp48 snp49 snp50 snp51 snp52 snp53 snp54 snp55 snp56 snp57 snp58 snp59 snp60 snp61 snp62 snp64 -Log p-value Figure 5. -Log P value for trend of the association between IGF1 SNPs and prostate cancer risk. Red line denotes P value = 0.05. Blue denotes SNPs located in block 1, yellow = block 2, green = block 3, orange = block 4, and gray = inter-blocks. Bold indicates tSNP. 78 Table 12. Odds ratios and 95% CI of prostate cancer associated IGF1 SNPs All groups combined OR * (95% CI) SNP 3 AA 1.00 AG 1.24 (1.08-1.42) § GG 1.26 (0.95-1.67) SNP 4 † TT 1.00 CT 1.25 (1.09-1.43) § CC 1.26 (0.95-1.68) SNP 6 AA 1.00 AT 1.24 (1.04-1.49) ‡ TT 1.37 (0.84-2.24) SNP 7 CC 1.00 AC 1.17 (0.98-1.40) AA 1.48 (0.97-2.27) SNP 9 † AA 1.00 AT 1.17 (0.98-1.39) TT 1.53 (1.01-2.32) SNP 14 † GG 1.00 CG 1.19 (1.00-1.42) ‡ CC 1.78 (1.10-2.90) ‡ SNP 22 GG 1.00 AG 1.14 (0.95-1.37) AA 1.90 (1.14-3.18) ‡ SNP 25 † AA 1.00 AG 1.10 (0.95-1.27) GG 1.82 (1.21-2.73) § SNP 34 GG 1.00 AG 1.08 (0.92-1.27) AA 1.89 (1.19-3.00) § SNP 37 CC 1.00 CT 1.13 (0.94-1.37) TT 1.81 (1.08-3.03) ‡ SNP 40 AA 1.00 AG 1.08 (0.92-1.26) GG 1.85 (1.17-2.92) § SNP 42 † CC 1.00 CG 1.09 (0.96-1.24) GG 1.22 (1.00-1.49) ‡ SNP 44 † AA 1.00 AG 1.16 (0.97-1.40) GG 1.52 (0.97-2.40) SNP 47 † GG 1.00 GT 1.18 (0.98-1.41) TT 1.57 (1.00-2.47) ‡ SNP 48 † GG 1.00 AG 1.10 (0.97-1.26) AA 1.21 (1.02-1.44) ‡ SNP 57 TT 1.00 TG 1.18 (0.98-1.42) GG 1.83 (1.14-2.94) ‡ SNP 58 † GG 1.00 CG 1.02 (0.89-1.16) CC 1.33 (1.12-1.59) § * ORs are estimated by unconditional logistic regression adjusted for age and ethnicity. † tSNP ‡ P value < 0.05 § P value < 0.01 79 SNP 4 were at significantly increased risk of prostate cancer compared to men carrying the common homozygous TT genotype (OR = 1.25; 95% CI: 1.09-1.43; P = 0.001). This association was also statistically significant for localized disease (OR = 1.32; 95% CI: 1.13-1.55; P = 0.0005) and a positive relationship was observed with advanced disease (OR = 1.17; 95% CI: 0.96-1.42). Ethnic-stratified analysis revealed an overall consistent positive pattern across all racial/ethnic groups (Table 13). Men possessing the CC genotype for SNP4 were also positively associated with risk (OR = 1.26; 95% CI: 0.95-1.68). To measure how often such results would be expected by chance in a study of these 29 tag SNPs, we performed a permutation test where case/control status was randomized within each ethnic group, and a null distribution of P values obtained. Of the 29 tSNPs, SNP 4 (rs7965399) displayed the smallest observed P value and levels of significance as large as that of SNP 4 were observed in 5.6% of the simulated null datasets. Discussion A number of studies have found that men with the highest circulating levels of IGF-I have an increased risk of prostate cancer (Harman, Metter et al. 2000; Chan, Stampfer et al. 2002; Stattin, Rinaldi et al. 2004) and IGF-I levels in the circulation have been shown to be a highly heritable trait (Harrela, Koistinen et al. 1996; Hong, Pedersen et al. 1996). We undertook the first comprehensive evaluation of the relationship between genetic variants in IGF1 and prostate 80 Table 13. Odds ratios and 95% CI of prostate cancer associated IGF1 SNPs by ethnic group African Americans Hawaiians Japanese Latinos Whites OR * (95% CI) OR * (95% CI) OR * (95% CI) OR * (95% CI) OR * (95% CI) SNP 3 AA 1.00 1.00 1.00 1.00 1.00 AG 1.17 (0.93-1.48) 0.99 (0.42-2.29) 1.20 (0.91-1.57) 1.20 (0.92-1.57) 2.05 (1.27-3.33) § GG 1.30 (0.88-1.92) 1.52 (0.40-5.79) 1.04 (0.59-1.82) 1.10 (0.51-2.38) 2.71 (0.52-14.06) SNP 4 † TT 1.00 1.00 1.00 1.00 1.00 CT 1.20 (0.95-1.51) 0.92 (0.40-2.11) 1.19 (0.91-1.56) 1.22 (0.94-1.57) 2.06 (1.28-3.31) § CC 1.34 (0.91-1.99) 1.53 (0.40-5.80) 1.04 (0.59-1.83) 0.97 (0.45-2.10) 2.71 (0.52-14.08) SNP 6 AA 1.00 1.00 1.00 1.00 1.00 AT 2.39 (0.98-5.80) 1.00 (0.37-2.70) 1.19 (0.91-1.57) 1.05 (0.79-1.40) 2.85 (1.31-6.17) § TT -- 1.66 (0.28-9.88) 1.08 (0.62-1.89) 3.92 (0.72-21.35) -- SNP 7 CC 1.00 1.00 1.00 1.00 1.00 AC 2.39 (0.98-5.80) 1.01 (0.39-2.61) 1.08 (0.83-1.42) 1.04 (0.80-1.36) 2.90 (1.32-6.17) § AA -- 1.05 (0.21-5.33) 1.26 (0.77-2.06) 2.60 (0.81-8.38) -- SNP 9 † AA 1.00 1.00 1.00 1.00 1.00 AT 2.40 (0.99-5.82) 0.84 (0.37-1.90) 1.10 (0.84-1.44) 1.04 (0.80-1.36) 2.51 (1.23-5.12) ‡ TT -- 1.51 (0.40-5.77) 1.27 (0.78-2.08) 2.55 (0.79-8.17) -- SNP 14 † GG 1.00 1.00 1.00 1.00 1.00 CG 4.17 (1.40-12.48) ‡ 0.63 (0.22-1.77) 1.11 (0.84-1.45) 1.08 (0.82-1.42) 2.46 (1.20-5.04) ‡ CC -- 0.89 (0.12-6.58) 1.79 (1.01-3.17) ‡ 2.29 (0.70-7.49) -- SNP 22 GG 1.00 1.00 1.00 1.00 1.00 AG 0.71 (0.16-3.08) 0.82 (0.36-1.90) 1.12 (0.85-1.48) 1.08 (0.81-1.43) 2.76 (1.27-6.01) ‡ AA -- 1.95 (0.34-11.20) 1.82 (0.98-3.40) 2.05 (0.69-6.06) -- SNP 25 † AA 1.00 1.00 1.00 1.00 1.00 AG 0.90 (0.69-1.18) 0.97 (0.41-2.28) 1.12 (0.85-1.47) 1.11 (0.84-1.45) 2.84 (1.45-5.58) § GG 1.87 (0.79-4.41) 1.95 (0.34-11.18) 1.71 (0.98-2.97) 2.04 (0.76-5.50) -- SNP 34 GG 1.00 1.00 1.00 1.00 1.00 AG 0.92 (0.69-1.25) 0.82 (0.36-1.90) 1.12 (0.85-1.48) 1.08 (0.81-1.43) 2.76 (1.27-6.01) ‡ AA 1.93 (0.68-5.50) 1.95 (0.34-11.20) 1.82 (0.98-3.40) 2.05 (0.69-6.06) -- SNP 37 CC 1.00 1.00 1.00 1.00 1.00 CT 0.71 (0.16-3.08) 0.64 (0.24-1.71) 1.12 (0.85-1.48) 1.08 (0.81-1.43) 2.76 (1.27-6.01) ‡ TT -- 0.90 (0.12-6.70) 1.82 (0.98-3.40) 2.05 (0.69-6.06) -- SNP 40 AA 1.00 1.00 1.00 1.00 1.00 AG 0.93 (0.69-1.24) 0.82 (0.36-1.90) 1.12 (0.84-1.48) 1.07 (0.80-1.42) 2.76 (1.27-6.01) ‡ GG 1.79 (0.66-4.83) 1.95 (0.34-11.20) 1.80 (0.96-3.35) 2.06 (0.71-6.03) -- SNP 42 † GG 1.00 1.00 1.00 1.00 1.00 AG 1.05 (0.83-1.33) 0.63 (0.29-1.36) 1.06 (0.78-1.44) 1.15 (0.91-1.45) 1.16 (0.87-1.54) AA 1.16 (0.81-1.65) 0.74 (0.29-1.92) 1.22 (0.84-1.77) 1.67 (1.04-2.66) ‡ 0.87 (0.46-1.64) SNP 44 † AA 1.00 1.00 1.00 1.00 1.00 AG 2.25 (0.69-7.34) 0.73 (0.27-1.98) 1.11 (0.85-1.45) 1.13 (0.85-1.50) 3.23 (1.04-9.99) ‡ GG -- 0.90 (0.12-6.67) 1.61 (0.92-2.81) 1.55 (0.63-3.84) -- SNP 47 † GG 1.00 1.00 1.00 1.00 1.00 GT 3.14 (1.02-9.69) ‡ 0.71 (0.25-1.99) 1.14 (0.87-1.51) 1.01 (0.77-1.33) 2.83 (1.35-5.94) § TT -- 1.04 (0.14-7.75) 1.39 (0.82-2.36) 2.30 (0.79-6.67) -- SNP 48 † GG 1.00 1.00 1.00 1.00 1.00 AG 1.07 (0.83-1.38) 0.56 (0.25-1.24) 0.92 (0.66-1.27) 1.22 (0.96-1.54) 1.25 (0.94-1.64) AA 1.15 (0.85-1.55) 0.87 (0.33-2.32) 1.09 (0.75-1.57) 1.42 (1.01-2.00) ‡ 1.10 (0.70-1.73) SNP 57 TT -- 1.00 1.00 1.00 1.00 TG -- 0.86 (0.32-2.27) 1.17 (0.89-1.54) 1.01 (0.76-1.33) 2.59 (1.27-5.29) § GG -- 1.91 (0.17-21.85) 1.60 (0.94-2.72) 3.49 (0.95-12.75) -- SNP 58 † GG 1.00 1.00 1.00 1.00 1.00 CG 0.93 (0.72-1.19) 0.55 (0.25-1.20) 0.95 (0.69-1.30) 1.16 (0.92-1.47) 1.64 (0.93-2.89) CC 1.30 (0.96-1.76) 1.08 (0.41-2.82) 1.18 (0.82-1.70) 1.35 (0.93-1.97) 1.06 (0.81-1.39) * ORs are estimated by unconditional logistic regression adjusted for age. † tSNP ‡ P value < 0.05 § P value < 0.01 cancer risk in a large multiethnic case-control st udy. In this report, we identified a set of 29 tagging SNPs that captures the majority of common genetic diversity 81 across the locus in a multiethnic population. Our results suggest that inherited variation in IGF1 may play a role in prostate cancer susceptibility. We identified several haplotypes and SNPs associated with an increased risk of prostate cancer. These genetic associations were seen across the entire IGF1 locus (156 kb) with a similar magnitude of effect (OR ~1.2). Given the strong correlation observed between neighboring SNPs, the most parsimonious explanation is the existence of one signal that is detected at several sites due to underlying regional correlation. The strongest signal was located in block 1, which begins 438 bp upstream of the IGF1 transcription site and spans ~23 kb. It is interesting to note that this region also contains the (CA) n repeat polymorphism that has been previously suggested to be associated with prostate cancer risk, although with mixed results (Nam, Zhang et al. 2003; Li, Cicek et al. 2004; Neuhausen, Slattery et al. 2005; Schildkraut, Demark-Wahnefried et al. 2005; Tsuchiya, Wang et al. 2005). We were able to determine using data from a previous study (DeLellis, Ingles et al. 2003) that the less common (CA) n repeat length is in LD with SNP 4, but whether SNP 4 or another linked marker is the causal variant remains to be determined. Our study of 2,300 cases and 2,290 controls had substantial power to detect modest genetic effects. For the observed genotype effect of 1.25 and allele frequency of 20% (total population), we had 99% and 95% power ( = 0.05, two- sided) under a log-additive and dominant model, respectively. The similar genetic effects observed for both the heterozygote (OR = 1.25) and homozygote (OR = 82 1.26) classes of SNP 4 suggests it may operate under a dominant model, although other models may also be consistent with these data. Any claim that inherited variation in a region is associated with a disease must be carefully scrutinized (Lander and Schork 1994; Page, George et al. 2003). Follow-up studies that attempt to replicate initial positive reports are often costly and underpowered to detect effects of modest magnitude. Failing to observe consistent associations across studies generates uncertainty about the role of a locus in disease pathogenesis (Freedman, Pearce et al. 2005). For genetic association studies, the traditional threshold of significance (P value = 0.05) may not be sufficiently stringent given the low prior probability that each of the 22,000 genes in the human genome plays a role in disease. Overly liberal thresholds can lead to false positive results when the number of tests, i.e. SNPs, is large (thousands to millions) in relation to the true underlying signal (more likely on the scale of tens to hundreds). This has likely contributed to difficulties in replication among subsequent studies (Ioannidis, Ntzani et al. 2001; Hirschhorn, Lohmueller et al. 2002). Attempts to distinguish between true positive or false positive results are usually predicated on a combination of the prior likelihood of the result being true and the magnitude of the observed P value. Individual studies are rarely large enough to detect with a high degree of statistical confidence the often modest risks associated with complex disease traits (Altshuler, Hirschhorn et al. 2000; 2004). Therefore, in evaluating a nominally significant result, it is essential to 83 address the possible sources of false positives. False positives can arise from statistical fluctuations, issues surrounding multiple testing, and population stratification. Attempts to address these issues can help guide the interpretation of suggestive findings. For our study, we evaluated the significance of the 29 tSNPs by subjecting our nominally significant associations to permutation testing (Hirschhorn and Daly 2005). Permutation testing is generally used to empirically assess the significance of a nominally significant P value as well as to guide the interpretation of results when multiple hypotheses are tested. The permutation P value for our best SNP was 0.056. This indicates that when considering IGF1 as an experimental unit, a similarly strong result would be seen by chance only 5.6% of the time. In addition, we also applied the false positive report probability (FPRP) proposed by Wacholder et al. to our results (Wacholder, Chanock et al. 2004). The FPRP estimates the likelihood that a result is a false positive by incorporating the prior probability that the locus is involved in the disease and the estimated magnitude of the effect size of the variant (Wacholder, Chanock et al. 2004). Using this framework, the model then evaluates the likelihood of a false positive given the observed odds ratio and confidence interval. While there is an inherent subjectivity to this process, a range of likelihoods (and therefore a range of the probability of a false positive result) may help to guide the investigator’s priorities regarding further pursuit of the variant(s) under study. Based on the published 84 data on IGF-I levels and prostate cancer, we believe an odds ratio of 1.50 is a reasonable magnitude of effect and for a prior probability of 1:100, 1:1,000, and 1:10,000, the likelihood that the observed effect of SNP 4 is a false positive is estimated to be 10%, 54%, and 92%, respectively. Population stratification is unlikely to be operative in our study given that we observed a consistent trend across all racial/ethnic groups (with the exception of Hawaiians where the sample size was small) for the SNPs that demonstrated a positive association with prostate cancer risk. The likelihood that stratification would occur in the same direction across all groups is low. Furthermore, the strongest effect observed was among Whites, a group which may have a lower likelihood of stratification than other groups (Wacholder, Rothman et al. 2000). Prior evidence implicating the product of this locus in prostate cancer pathogenesis coupled with the results from this study provide a solid foundation for attempting to replicate these findings in other cohorts such as the NCI sponsored Cohort Consortium (http://epi.grants.cancer.gov/BPC3/). Replication is critical to confirm the role of any variant suspected to play a role in disease. Moreover, it will be of particular interest to examine whether inherited variation at the IGF1 locus can account for some of the variation in circulating levels of IGF-I, given that IGF-I levels have been found to be highly heritable and related to prostate cancer risk (Harrela, Koistinen et al. 1996; Hong, Pedersen et al. 1996; Harman, Metter et al. 2000; Chan, Stampfer et al. 2002; Pollak, Schernhammer et al. 2004; Stattin, Rinaldi et al. 2004). 85 PART II. DATA ANALYSIS Section 2. Haplotype analysis of IGFBP1 and IGFBP3 in relation to prostate cancer risk: the Multiethnic Cohort Iona Cheng 1 , Kathryn L Penney 2,3,4,6 , Daniel O Stram 1 , Malcolm Pike 1 , Loic Le Marchand 9 , Laurence N. Kolonel 9 , Joel Hirschhorn 2,3,5,10 , David Altshuler 2,3,4,6,7 , Brian E Henderson 1 , Matthew L Freedman 2,3,4,6,8 1 Department of Preventive Medicine, Norris Comprehensive cancer center, Keck School of Medicine, University of Southern California, Los Angeles, CA 90089, USA 2 Program in Medical and Population Genetics, Broad Institute of Harvard and Massachusetts Institute of Technology, Cambridge, MA 02139, USA 3 Department of Genetics, 4 Medicine, and 5 Pediatrics, Harvard Medical School, Boston, MA 02115, USA 6 Department of Molecular Biology, 7 Diabetes Unit, and 8 Hematology-Oncology, Massachusetts General Hospital, MA 02214, USA 9 Cancer Etiology Program, Cancer Research Center of Hawaii, University of Hawaii, Honolulu, HI 96813, USA 10 Division of Genetics and Endocrinology, Children’s Hospital and Department of Pediatrics, Boston, MA 02115, USA Address for correspondence: Matthew Freedman, Program in Medical and Population Genetics, Broad Institute of Harvard and MIT, Cambridge, MA 02139, USA (e-mail: freedman@broad.mit.edu) 86 Abstract Background: Collective evidence suggests that the insulin-like growth factor (IGF) system plays a role in prostate cancer etiology. Insulin like growth factor binding proteins (IGFBPs) are important regulatory molecules that modulate the bioavailability of IGF-I in the circulation and tissues. Methods: To examine whether inherited differences in IGFBP1 and IGFBP3 impact prostate cancer susceptibility, we conducted a large case-control study of prostate cancer risk nested within the Multiethnic Cohort. We undertook a two-prong approach to systematically assess the genetic variation across the two loci: 1) we sequenced the IGFBP1 and IGFBP3 exons in 95 advanced prostate cancer cases to identify any novel missense variants at appreciable frequencies and 2) characterized the linkage disequilibrium patterns and common haplotypes of IGFBP1 and IGFBP3 by genotyping 34 SNPs spanning 71 kb in a multiethnic panel of 349 controls of five racial/ethnic groups. Next, we tested IGFBP1 and IGFBP3 genotypes and haplotypes for their association with prostate cancer risk (cases/controls = 2,320/2,290). Results: No missense SNPs were found after sequencing 95 prostate cases. We identified three blocks of strong linkage disequilibrium and selected 20 tagging SNPs to predict both the common IGFBP1 and IGFBP3 haplotypes. We observed no significant association between single IGFBP1 and IGFBP3 SNPs and prostate cancer risk. Haplotype analysis of prostate cancer risk identified a significant global effect with block1 (P = 0.048), whereby heterozygous carriers of haplotype 1F (~6%) had a nominally significant increased risk compared to 87 noncarriers (OR = 1.23; 95% CI: 1.03-1.47). Conclusion: Our data suggests that common genetic variation in IGFBP1 and IGFBP3 does not play a major role in prostate cancer susceptibility. 88 Introduction Multiple lines of evidence from experimental and epidemiologic studies provide strong support for the role of the insulin-like growth factor (IGF) system in prostate tumorigenesis. Insulin-like growth factor I (IGF-I) is a mitogen that regulates cell proliferation and also inhibits apoptosis (Jones and Clemmons 1995). The bioavailability of IGF-I in the circulation and tissues is regulated by insulin-like growth factor binding proteins (IGFBPs) that bind IGF-I with high affinity and specificity. IGFBPs have complex functions involved with either suppressing or enhancing the effects of IGF-I. All IGFBPs are able to suppress IGF-mediated mitogenic effects by competitively binding to IGF-I and preventing its interaction with the insulin-like growth factor 1 receptor (IGF1R). IGFBP-1 and IGFBP-3 are also able to enhance IGF-I effects via steady and slow release of unbound IGF-I, while protecting its receptor from downregulation in response to high IGF-I levels (Conover and Powell 1991; Grimberg and Cohen 2000). IGFBP-3 is the principal binding protein of IGF-I in the circulation. Prospective studies of prostate cancer risk comparing men at the highest to lowest quartile of circulating IGFBP-3 have had mixed results (OR meta = 0.90; 95% CI: 0.67-1.20) (Harman, Metter et al. 2000; Chan, Stampfer et al. 2002; Woodson, Tangrea et al. 2003). Although, in the largest study of 530 prostate cancer cases from the Physicians’ Health Study, men at the top quartile of IGFBP-3 levels had a significant 80% reduction in risk of advanced prostate cancer in comparison to those at the lowest quartile of IGFBP-3 levels (Chan, Stampfer et al. 2002). 89 Circulating IGFBP-1 levels and prostate cancer risk has been examined in only one prospective study and no association was observed (Stattin, Bylund et al. 2000). Few studies have examined whether inherited differences in IGFBP1 and IGFBP3 influence prostate cancer risk. These studies have focused on only a small handful of single nucleotide polymorphisms (SNPs) in IGFBP3 (Nam, Zhang et al. 2003; Wang, Habuchi et al. 2003; Schildkraut, Demark-Wahnefried et al. 2005). The IGFBP3 (A-202C) has been most often studied and is suggested to be associated with circulating levels of IGFBP-3 (Deal, Ma et al. 2001; Schernhammer, Hankinson et al. 2003; Ren, Cai et al. 2004; Le Marchand, Kolonel et al. 2005). Four studies reported no association between IGFBP3 (A- 202C) and prostate cancer risk (Nam, Zhang et al. 2003; Wang, Habuchi et al. 2003; Li, Cicek et al. 2004; Schildkraut, Demark-Wahnefried et al. 2005). No studies to date have examined whether variants in IGFBP1 relate to risk of prostate cancer risk. IGFBP1 and IGFBP are located in close proximity on chromosome 7p13- p12, separated by a 19 kb distance and transcription oriented in a tail-to-tail configuration. Recent efforts to characterize the genetic variation in the human genome have lead to the identification of close to a hundred single nucleotide polymorphisms across the two genes. To thoroughly assess how inherited differences in IGFBP1 and IGFBP3 contribute to prostate cancer risk, we utilized a haplotype-based approach to conduct a large nested case-control study of 2,320 90 prostate cancer cases and 2,290 controls from the Multiethnic Cohort Study (MEC). Methods The Multiethnic Cohort is a large population-based cohort study of over 215,000 men and women from Hawaii and Los Angeles. The cohort is comprised of predominantly five racial/ethnic groups: African-Americans, Native Hawaiians, Japanese, Latinos, and Whites. Participants between the ages of 45 and 75 years were recruited from 1993 to 1996 and completed a 26-page self- administered questionnaire that included information regarding medical history, family history, diet, dietary supplements and medication use, and physical activity. Further details are provided elsewhere (Kolonel, Henderson et al. 2000). Incident cancers in the MEC were identified (up to April 1, 2002) by cohort linkage to population-based Surveillance, Epidemiology and End Results (SEER) cancer registries covering Hawaii and California. Information on stage of disease at the time of diagnosis was also collected from the cancer registries. We defined localized and advanced disease based on tumor stage and differentiation (Gleason grade). Tumors confined to the prostate having a Gleason grade < 8 were defined as localized disease. Regional and metastatic tumors, or localized tumors with Gleason grade > 8 were defined as advanced disease. Controls were men without prostate cancer prior to entry into the cohort and without a prostate cancer diagnoses up to April 1, 2002. Controls were randomly selected from our random control pool of MEC participants, who 91 provided blood specimens for genetic analyses in the cohort. The participation rate for blood collection was 74% for cases and 66% for controls. Controls were frequency matched to cases by age and ethnicity. The prostate cancer case-control study consists of 2,320 cases and 2,290 controls. This study was approved by the Institutional Review Boards at the University of Hawaii and the University of Southern California. Sequencing To identify missense SNPs not in public or private databases we attempted to sequence the four exons of IGFBP1 and IGFBP3, respectively, in 95 cases of advanced prostate and breast cancers (n = 19 per racial/ethnic group). Exon 1 of IGFBP1 and exon 4 of IGFBP3 did not meet our sequencing criteria of more than 80% of the samples with phred scores > 20 for at least 80% of the target bases. Sequencing was performed by conventional dye-primer sequencing on ABI 3700 Sequencers. The PolyPhred program was used to identify polymorphisms with manual review by at least two observers, and all putative coding variants were validated by SNP genotyping. Further details are described elsewhere (Freedman, Pearce et al. 2005). SNP selection and genotyping for genetic characterization We genotyped single nucleotide polymorphisms spanning 71 kb across the IGFBP1 and IGBP3 locus (Figure 1). We evaluated 19.4 kb and 18.8 kb upstream 92 Figure 6. IGFBP1 and IGFBP3 loci spanning 71 kb on chromosome 7 (human genome assembly 16). SNPS (n=34), LD blocks (n=3) and haplotype patterns (>5%) are presented for all ethnic groups combined. of the first exons of IGFBP1 and IGFBP3, respectively, and 46.3 kb and 43.3 kb downstream of the transcribed region for each gene, respectively. We selected SNPs from the public (http://www.ncbi.nlm.nih.gov/SNP/) and private database (http://www.celera.com/). Our goal was to achieve a SNP density of 1 SNP every 2-5 kb across the locus. We preferentially selected SNPs located in coding and 93 UTR regions as well as areas of mouse homology (80% identity in 200 bp sequence; http://www.dcode.org/). For genetic characterization, 86 SNPs were genotyped in a multiethnic panel of 349 individuals with no history of cancer: African-American (n = 70), Hawaiian (n = 69), Japanese (n = 70), Latina (n = 70), and White (n = 70) (Haiman, Stram et al. 2003). SNP genotyping was performed using the Sequenom mass spectrometry platform (Sequenom Inc. San Diego, CA). Of the 86 SNPs, 39 SNPs were identified as monomorphic and 13 SNPs displayed poor genotyping results (genotyped < 75% of samples or out of Hardy- Weinberg equilibrium (1-sided P < 0.01) in > 1 ethnic group) and were eliminated from further analysis. A total of 34 SNPs was used for genetic characterization having an average density of 1 SNP every 2.1 kb. Haplotype block determination and tSNP selection The D’ statistic was used to determine the pair-wise linkage disequilibrium (LD) between the 34 SNPs. Regions of strong LD, haplotype blocks, were defined following the methods of Gabriel et al. (Gabriel, Schaffner et al. 2002). We aimed to genotype a minimum of 6 SNPs with minor allele frequencies (maf) > 10% within each block and a maximum distance of 10 kb between adjacent blocks. Each ethnic group was evaluated for the extent of LD meeting block criteria. Hawaiians, Japanese, Latinos, and Whites shared similar LD patterns. We combined these populations to assess the LD structure of the locus. African-Americans displayed a smaller extent of LD as expected. We 94 genotyped an additional 15 SNPs in regions where there were insufficient number of SNPs to meet our criteria of strong LD among this group. These additional SNPs were comprised of all publicly available SNPs in dbSNP as well as SNPs from the private Celera database as of April 2004. With this additional effort, block 3 among African-Americans was similar to the other groups, yet, there were still an insufficient number of SNPs for the remaining regions to meet our 6 SNP criterion for this ethnic group. Genotype data for each ethnic group in the multiethnic panel was used to estimate haplotype frequencies within blocks using the Expectation-Maximization (E-M) algorithm (Excoffier and Slatkin 1995). The squared correlations (R 2 h ) between the true haplotypes (h) and their estimates from the E-M algorithm were estimated as described by Stram et al. (Stram, Haiman et al. 2003). For each ethnic group, we selected the minimum set of tagging SNPs (tSNPs) within each block for each ethnic group to assure an R 2 h > 0.7 for all haplotypes with an estimated frequency > 5% (Stram, Haiman et al. 2003) . Genotyping in the case-control study The tSNPs were genotyped in the prostate case-control study using the 5’ nuclease Taqman allelic discrimination assay (Applied Biosystems, Foster City, CA, USA). All assays were performed blinded to case-control status. For quality control, 5% replicate samples were included. The concordance for replicate samples was 99.8%. The average successful genotyping percentage was 98.8%. 95 Case-control analysis To investigate whether prostate cancer susceptibility can be attributed to a single variant, we examined the relationship between IGFBP1 and IGFBP3 genotypes and prostate cancer risk. We evaluated a total of 24 SNPs: 20 tSNPs, two SNPs located within the IGFBP3 gene that fell outside the regions of strong LD, one IGFBP3 missense SNP (rs2854746), and the IGFBP3 (A-202C) polymorphism (rs2854744) (Table 14). Odds ratios (ORs) and 95% CIs were estimated by unconditional logistic regression for the association between IGFBP1 and IGFBP3 genotypes and risk of prostate cancer. To utilize linkage disequilibrium to capture unmeasured variants, we evaluated the relationship between common IGFBP1 and IGFBP3 haplotypes and prostate cancer risk. Haplotype frequencies among prostate cancer cases and controls were estimated by using genotype data of the tSNPs as described by Stram et al. (Stram, Haiman et al. 2003). Haplotype dosage (i.e. an estimate of the number of copies of haplotype h) for each individual and each haplotype, h, was computed using that individual’s genotype data and haplotype frequency 96 Table 14. IGFBP1 and IGFBP3 SNPs utilized for genetic analysis Nucleotide Minor b SNP# SNP Location Position a Change Allele AA HA JA LA WH 1 rs1995050 5' 45649469 A/G G 46.4 47.1 39.7 46.3 31.4 2 c hCV395979 5' 45649695 A/G G 36.4 33.1 33.3 34.1 36.2 3 c rs1553009 5' 45649774 A/G A 9.3 26.1 22.5 21.7 19.3 4 rs1027438 5' 45650778 A/G G 47.8 41.2 44.0 43.4 44.2 5 c hCV395975 5' 45653244 C/G G 27.4 19.7 32.6 14.8 20.0 6 rs2331390 5' 45656853 A/G A 47.1 41.2 44.3 43.6 44.2 7 c rs2201638 5' 45663690 A/G A 21.7 13.8 3.6 5.7 5.1 8 rs4724445 5' 45663980 A/G A 8.0 26.5 21.7 17.9 17.9 9 c rs1065780 5' 45668457 A/G A 39.0 39.8 46.2 36.1 45.4 10 c rs3793344 I1 45669675 A/G G 39.0 44.9 45.6 39.7 45.7 11 c rs1065781 I2 45672177 C/T T 7.9 -- -- 1.6 -- 12 c rs1874479 I3 45673006 A/G G 2.9 13.8 18.7 8.5 22.8 13 hCV26836723 I3 45673050 A/T A 4.3 14.0 17.4 7.4 22.9 14 hCV1842780 I3 45673077 A/C C 4.3 2.9 1.4 8.0 3.6 15 c rs4988515 E4 45673380 C/T T 11.6 3.6 0.7 8.6 4.3 16 c rs4619 E4 45673449 A/G G 40.0 44.7 44.8 37.7 41.8 17 rs7454 E4 45673786 C/G C 3.8 3.1 0.7 8.3 3.0 18 c rs1908751 3' 45676299 C/T T 33.3 36.2 35.7 22.9 29.3 19 c rs1496495 3' 45678041 C/T C 11.4 16.7 18.7 15.9 25.8 20 rs1496496 3' 45681683 A/G G 39.1 44.2 44.9 38.4 43.6 21 c rs1496497 3' 45681823 G/T G 23.6 16.7 18.6 17.9 26.8 22 c rs2270628 3' 45690350 C/T T 30.9 11.0 17.9 17.4 27.2 23 rs6670 UTR 45693034 A/T A 13.8 12.3 -- 10.9 17.1 24 rs2453839 I4 45694353 C/T C 37.5 20.6 9.6 10.1 24.3 25 rs2453840 I4 45694592 G/T T 7.4 8.2 -- 8.0 22.5 26 c rs3110697 I3 45695809 A/T A 42.9 36.2 20.3 36.4 43.6 27 c rs6953668 I3 45696655 A/G A 7.1 11.6 10.0 1.4 0.7 28 rs2471551 I1 45697835 C/G G 15.0 7.1 -- 11.2 14.9 29 hCV1842667 I1 45698082 A/G G 0.7 0.7 -- 3.6 4.3 30 rs3793345 I1 45698458 C/T C 21.4 11.8 -- 18.4 19.3 31 c rs2854747 I1 45700697 A/G G 40.4 35.3 19.3 37.0 44.2 32 rs2854746 E1 45701425 G/C C 32.6 57.4 78.1 31.2 39.6 33 rs2854744 5' 45701855 A/C C 41.9 39.6 22.2 64.9 55.5 34 c rs2132570 5' 45703243 G/T T 17.2 24.6 18.8 43.2 25.4 35 c rs2960436 5' 45718062 A/G G 43.8 41.3 18.2 32.6 40.4 36 c rs2471554 5' 45720347 G/T G 15.0 11.6 -- 19.3 19.6 a SNP position based on July 2003 (UCSC version human genome16) b Minor allele based on all groups combined c tSNP Allele Frequency (%) estimates obtained from the E-M algorithm (Zaykin, Westfall et al. 2002). To test whether common haplotypes within each block were associated with risk, we conducted a global likelihood ratio test. Application of the global test corrects for multiple testing of individual haplotypes effects within each block. Odds ratios (ORs) and 95% Confidence Intervals (CIs) for each common haplotype were estimated by unconditional logistic regression. 97 Al l of the following results were only affected slightly when adjustment was made for the risk factor of family history of prostate cancer and the unadjusted values are given below. Results Genetic characterization of the IGFBP1 and IGFBP3 locus We found no missense polymorphisms in the remaining three exons of IGFBP1 and four exons of IGFBP3 (see Methods for further details). We genotyped 34 SNPs in the multiethnic panel spanning 71 kb across the IGFBP1 and IGFBP3 locus. We identified three regions of strong linkage disequilibrium (LD): block 1 (SNPs 2-9; size = 19 kb) spanned the upstream region of IGFBP1, block 2 (SNPs 10-22; size = 21 kb) included the region of intron 1 to 3’ downstream of IGFBP1, block 3 (SNPs 19-36; size = 60 kb) spanned intron 3 to 5’ upstream of IGFBP3 (Figure 1). The distances between adjacent blocks were < 5 kb. Common haplotypes within each region of LD and their corresponding frequencies are presented in Table 15. We observed six common ( > 5%) haplotypes in block 1, four common haplotypes in block 2, and five common haplotypes in block 3. A total of 20 tagging (tSNPs) were able to predict the common haplotypes across the IGFBP1 and IGFBP3 locus. The common haplotypes for each ethnic group accounted for 76-100% of the chromosomes in the panel population. 98 Table 15. Common haplotypes in blocks 1-3 of IGFBP1 and IGFBP3 estimated by tSNPs in the multiethnic panel a Haplotypes African- Americans Hawaiians Japanese Latinos Whites All Block 1: SNPs 2-10; tSNPs: 2,3,5,7,9 1A AGCGG 0.13 0.21 0.13 0.30 0.24 0.24 1B GGCGA 0.15 0.13 0.22 0.17 0.26 0.26 1C AGGGG 0.26 0.20 0.31 0.15 0.20 0.20 1D AACGA 0.07 0.26 0.22 0.21 0.19 0.19 1E GGCAG 0.18 0.14 0.06 0.05 0.05 1F GGCGG 0.06 0.07 0.12 0.05 0.05 1G AGCGA 0.11 Percentage of Chromosomes Observed (%) 90 100 95 100 100 100 R 2 h 0.90 1.00 0.97 1.00 1.00 1.00 Block 2: SNPs 10-22; tSNPs: 10,11,12,15,16,18,19,21,22 2A ACACATTTC 0.30 0.36 0.34 0.21 0.26 0.29 2B ACACACTTC 0.23 0.20 0.19 0.39 0.27 0.25 2C GCACGCTTC 0.10 0.28 0.27 0.21 0.15 0.20 2D GCGCGCCGT 0.08 0.16 0.07 0.23 0.11 2E GTACGCTGT 0.07 2F GCATGCTTT 0.06 2G GCGCGCCGC 0.05 2H GCATGCCGT 0.08 Percentage of Chromosomes Observed (%) 76 96 96 95 90 86 R 2 h 0.98 1.00 0.99 1.00 0.99 0.99 Block 3: SNPs 26-36; tSNPs: 26,27,31,34,35,36 3A TAGAGT 0.50 0.57 0.80 0.29 0.38 0.51 3B TGTGGA 0.08 0.13 0.09 0.44 0.23 0.19 3C GGGGGA 0.14 0.11 0.19 0.18 0.12 3D TGTGAA 0.11 0.09 0.06 3E TGGAGT 0.05 0.07 0.05 0.15 0.06 3F TGGGGA 0.08 Percentage of Chromosomes Observed (%) 84 98 99 97 94 94 R 2 h 0.94 1.00 1.00 1.00 0.99 0.98 a Haplotypes > 5% in at least one ethnic group in the multiethnic panel Haplotype frequencies (%) Prostate cancer association study Study subject characteristics of our prostate cancer nested case-control study have been described earlier (Cheng, et al submitted). We conducted a SNP analysis of the 24 SNPs spanning the IGFBP1 and IGFBP3 locus to examine whether prostate cancer risk could be attributed to a single variant. Table 16 presents the findings for two missense SNPs, rs4619 (IGFBP1) and rs2854746 (IGFBP3). The third SNP is the IGFBP3 A-202C polymorphism (rs2854744) located in the promoter region. Both rs2854746 and IGFBP3 A-202C have been 99 Table 16. Odds ratios and 95% CI of prostate cancer for IGFBP1 and IGFBP3 missense SNPs and IGFBP3 (-202) Cases N (%) Controls N (%) OR a (95% CI) rs4619 AA 826 (36.2) 793 (35.3) 1.00 AG 1083 (47.5) 1080 (48.0) 0.97 (0.85-1.10) GG 370 (16.2) 376 (16.7) 0.95 (0.79-1.13) rs2854746 GG 816 (36.0) 826 (36.2) 1.00 CG 923 (40.7) 959 (42.0) 1.03 (0.90-1.18) CC 529 (23.2) 500 (21.9) 0.94 (0.79-1.13) rs2854744 AA 720 (31.9) 687 (30.1) 1.00 (A-202C) AC 954 (42.3) 1002 (44.0) 1.09 (0.95-1.26) CC 581 (25.8) 590 (25.9) 1.06 (0.90-1.26) a ORs are estimated by unconditional logistic regression adjusted for age and ethnicity. Prostate Analysis associated with circulating levels of IGFBP3. For these three polymorphisms as well as the remaining 21 SNPs (data not shown), we observed no significant associations with prostate cancer risk. We utilized a haplotype analysis to identify chromosomal regions that may harbor unmeasured disease associated variants. Associations between common IGFBP1 and IGFBP3 haplotypes and prostate cancer risk are presented in Table 17. We observed a statistically significant haplotype-effect in a global test of differences in risk in block 1 (P = 0.048). Heterozygous carriers of haplotype 1F had a nominally significant increased risk of prostate cancer (OR = 1.23; 95% CI: 1.03-1.47) compared to noncarriers. This association was also significant in localized disease (OR = 1.25; 95% CI: 1.02-1.53) and a positive relationship was observed with advanced disease (OR = 1.14; 95 CI: 0.95-1.36). Ethnic-stratified analysis revealed an overall consistent positive pattern across all five racial/ethnic groups among heterozygous carriers (Table 18). Rare homozygous carriers of 100 Table 17. Associations between common haplotypes in blocks 1-3 of IGFBP1 and IGFBP3 and prostate cancer risk Haplotype % All groups All groups African-Americans Hawaiians Japanese Latinos Whites heterzygous versus homozygousrare homozygous versus homozygous N = 683/650 a N = 71/68 a N = 462/471 a N = 647/649 a N = 457/452 a OR b (95% CI) OR b (95% CI) BLOCK 1 1A AGCGG 11/10 19/20 14/16 22/26 25/27 0.89 (0.78-1.01) 0.85 (0.64-1.15) 1B GGCGA 20/20 16/17 15/17 16/17 20/18 0.99 (0.87-1.13) 0.92 (0.67-1.27) 1C AGGGG 28/27 18/25 30/25 25/22 24/24 1.11 (0.97-1.25) 1.18 (0.93-1.48) 1D AACGA 7/9 30/21 24/24 21/20 17/19 0.97 (0.85-1.11) 1.02 (0.74-1.41) 1E GGCAG 15/15 11/12 6/7 4/4 3/2 0.99 (0.83-1.19) 0.80 (0.40-1.59) 1F GGCGG 4/4 4/2 9/7 10/9 9/9 1.23 (1.03-1.47) c 0.95 (0.44-2.03) 1G AGCGA 8/8 -- -- -- -- 0.99 (0.72-1.35) 4.97 (0.57-43.23) BLOCK 2 2A ACACATTTC 30/28 30/38 32/35 26/23 27/25 1.09 (0.96-1.23) 1.23 (0.99-1.53) 2B ACACACTTC 24/23 14/18 22/23 32/34 33/33 0.95 (0.84-1.09) 0.90 (0.72-1.12) 2C GCACGCTTC 6/6 36/23 26/27 19/18 15/16 1.03 (0.90-1.19) 1.07 (0.76-1.50) 2D GCGCGCCGT 5/3 9/12 12/13 7/9 14/13 1.07 (0.91-1.25) 0.48 (0.25-0.94) c 2E GTACGCTGT -- -- -- -- -- -- 2F GCATGCTTT 14/15 -- -- -- -- 0.93 (0.73-1.19) 0.87 (0.42-1.78) 2G GCGCGCCGC -- -- -- -- -- -- 2H GCATGCCGT - -- -- 7/7 -- 0.99 (0.78-1.27) 0.26 (0.05-1.28) BLOCK 3 3A TAGAGT 54/55 56/59 75/76 34/34 46/42 1.02 (0.88-1.18) 1.00 (0.84-1.18) 3B TGTGGA 7/5 13/10 11/10 41/41 20/24 1.07 (0.92-1.23) 0.93 (0.72-1.20) 3C GGGGGA 12/14 13/17 -- 16/16 20/21 0.90 (0.78-1.04) 0.98 (0.65-1.46) 3D TGTGAA 5/4 14/11 12/11 -- -- 1.08 (0.86-1.36) 1.48 (0.59-3.71) 3E TGGAGT 7/6 2/3 -- 6/6 11/10 1.09 (0.90-1.32) 1.83 (0.72-4.63) 3F TGGGGA 6/7 -- -- -- -- 1.05 (0.76-1.45) 0.46 (0.04-5.20) a N = cases/controls b ORs are estimated by unconditional logistic regression adjusted for age and ethnicity. c P value < 0.05 101 Table 18. Associations between haplotype 1F and prostate cancer risk by racial/ethnic group All groups All groups heterzygous versus homozygous rare homozygous versus homozygous OR a (95% CI) OR a (95% CI) African-Americans 1.31 (0.83-2.07) 0.86 (0.05-13.75) Hawaiians 1.50 (0.34-6.64) -- Japanese 1.28 (0.89-1.85) 3.18 (0.34-29.40) Latinos 1.18 (0.87-1.60) 0.92 (0.33-2.54) Whites 1.17 (0.82-1.68) 0.48 (0.09-2.67) a ORs are estimated by unconditional logistic regression adjusted for age and ethnicity. haplotype 1F displayed no overall association with prostate cancer risk and ethnic stratified results were inconsistent across groups. No significant global haplotype- effects were observed in block 2 (P = 0.376) or block 3 (P = 0.675) and within these blocks no associations were observed with individual haplotypes. Discussion In the present study, we sought to characterize the haplotype structure of the IGFBP1 and IGFBP3 and conduct a large multiethnic case-control study to assess whether inherited differences in these genes may relate to prostate cancer risk. We identified three regions of strong LD across the locus and 20 tagging SNPs that capture the majority of common variation in IGFBP1 and IGFBP3. We found no associations with individual IGFBP1 and IGFBP3 SNPs. Although a modest positive effect was observed with a ~6% haplotype located 5’ upstream of IGFBP1, our data suggest that that common variation at this locus does not substantially influence prostate cancer risk. Our results are consistent with previous studies that found no association between the IGFBP3 (A-202C) polymorphism and prostate cancer risk (Nam, 102 Zhang et al. 2003; Wang, Habuchi et al. 2003; Li, Cicek et al. 2004; Schildkraut, Demark-Wahnefried et al. 2005). One study of 307 Japanese cases observed that the C allele was correlated with advanced prostate cancer in comparison to localized disease (Wang, Habuchi et al. 2003). Among 2,130 cases with genotype and prostate tumor information, we found no association between this polymorphism and risk of advanced disease (P for trend = 0.742) in comparison to localized disease and similar findings were observed in ethnic-stratified analysis. Although the IGFBP3 (A-202C) has been associated with circulating levels of IGFBP-3 (Deal, Ma et al. 2001; Schernhammer, Hankinson et al. 2003; Ren, Cai et al. 2004; Le Marchand, Kolonel et al. 2005), our data suggests there is no evidence to support the involvement of this polymorphism with prostate cancer risk. In addition, we observed no association with the missense IGFBP3 (rs2854746) variant that has also been found to be related to IGFBP-3 levels (Le Marchand, Kolonel et al. 2005). Our findings of a haplotype effect in the absence of any SNP effects suggests that the signal for the association of IGFBP1 and prostate cancer risk originates from an unmeasured polymorphism within the 19 kb region of block 1. One approach to further investigate these findings would be to resequence this region among individuals carrying the suggestive haplotype then assess the association between novel polymorphisms and cancer risk. Before undertaking such extensive efforts, it is essential that additional studies confirm our observed association, and careful consideration should be given to the number of subjects 103 that would be needed to replicate these weak findings. Given the relatively low frequency of haplotype 1B (~6%) and the current OR of 1.23 for prostate cancer risk, we estimate that at least 3,000 cases/control pairs would be needed to achieve an 80% power (R 2 h = 0.9, =0.05, two-sided). Subsequent studies must have substantial power to detect such weak effects with statistical confidence (Hirschhorn, Lohmueller et al. 2002). Our study alone had moderate power (70%) to detect the observed haplotype effect with 2,320 prostate cancer cases. Association studies may be subject to the problem of population stratification, whereby spurious associations can occur among ethnically mixed populations as a result of differences in allele frequencies and baseline disease risks. Given our well-matched study design and the consistent positive association among heterozygous carriers of haplotype 1F across all five racial/ethnic groups, we find it unlikely that population stratification could have biased our estimates. If stratification occurred, it would have to be operative across all five groups. The lack of a consistent pattern among rare homozygous carriers may be explained by its low frequency whereby only 1% of each population possessed two copies of this haplotype. Issues of reproducibility is a principal concern of association studies (Ioannidis, Ntzani et al. 2001; Hirschhorn, Lohmueller et al. 2002) and weak effects such as those observed are left in a state of uncertainty. A system as advocated by Hirschhorn et al. to curate all association data until a complete meta-analysis can be conducted would serve as useful tool to assess the 104 worthiness of these types of associations waiting further judgment (Hirschhorn, Lohmueller et al. 2002). Future work that investigates whether inherited differences in IGFBP1 and IGFBP3 influence circulating levels of these binding proteins will be of particular interest. 105 PART II. DATA ANALYSIS Section 3. Association study of common genetic variation in insulin-like growth factor 1 (IGF1), height, and circulating IGF-I levels: The Multiethnic Cohort Iona Cheng 1 , Johannah Butler 2,3,4,10 , Daniel O Stram 1 , Malcolm Pike 1 , Loic Le Marchand 9 , Laurence N. Kolonel 9 , David Altshuler 2,3,4,6,7 , Brian E Henderson 1 , Joel Hirschhorn 2,3,5,10 ,Matthew L Freedman 2,3,4,6,8 1 Department of Preventive Medicine, Norris Comprehensive cancer center, Keck School of Medicine, University of Southern California, Los Angeles, CA 90089, USA 2 Program in Medical and Population Genetics, Broad Institute of Massachusetts Institute of Technology and Harvard, Cambridge, MA 02139, USA 3 Department of Genetics, 4 Medicine, and 5 Pediatrics, Harvard Medical School, Boston, MA 02115, USA 6 Department of Molecular Biology, 7 Diabetes Unit, and 8 Hematology-Oncology, Massachusetts General Hospital, MA 02214, USA 9 Cancer Etiology Program, Cancer Research Center of Hawaii, University of Hawaii, Honolulu, HI 96813, USA 10 Division of Genetics and Endocrinology, Children’s Hospital and Department of Pediatrics, Boston, MA 02115, USA Address for correspondence: Matthew Freedman, Program in Medical and Population Genetics, Broad Institute of Harvard and MIT, Cambridge, MA 02139, USA (e-mail: freedman@broad.mit.edu) 106 Abstract Background: Adult height is a highly heritable trait. Genomewide linkage studies have identified several chromosomal regions associated with height, although the specific loci involved have yet to be determined. Insulin-like growth factor 1 (IGF1) is important in human growth as it regulates cellular proliferation and mediates the actions of growth hormone. We investigated whether inherited differences in IGF1 are associated with variation in height and circulating levels of IGF-I. Methods: Twenty-nine IGF1 tagging SNPs (tSNPs) were genotyped in the Multiethnic Cohort (MEC) (N = 9,980) and 14 tSNPs specific for Whites were genotyped in two independent studies of stature (U.S. subjects; N = 2,189 and Polish subjects; N = 1,018). In the MEC, we examined the relationship between IGF1 and height among subjects at the extremes of the height, short (N = 1,029) and tall (N = 1,104), the lower and upper 10% of the height distribution. We also evaluated SNPs nominally associated with height for their effect in the entire study population of 9,980 MEC subjects. Plasma IGF-I levels were measured among 1,000 MEC control subjects and tested for their association with common IGF1 genetic variation. Results: In the MEC, we observed significant global associations with height in block 1 (P = 0.016), block 2 (P = 0.017), block 3 (P = 0.003), and block 4 (P = 0.001). Eleven of the 29 tSNPs were associated with height (P = 0.0001 – P = 0.049). Permutation testing to correct for multiple hypothesis testing revealed a gene-wise empirical P value < 0.01. In the U.S. stature study, similar genetic associations between IGF1 and height was observed 107 as seen in the MEC, although no association was observed among the Polish population. Haplotype 1C was nominally associated with circulating levels of IGF-I. Conclusion: Our findings suggest that inherited differences in IGF1 influence variation in height. 108 Introduction Adult height is a complex trait that is largely regulated by genetic control. Twin studies estimate the heritability of height to range from 68% to 90% (Phillips and Matheny 1990; Carmichael and McGue 1995; Silventoinen, Sammalisto et al. 2003). Recently, genomewide linkage studies have identified several potential regions of linkage on chromosomes 3p, 6q, 7q, 12p, and 13q (Perola, Ohman et al. 2001; Hirschhorn, Lohmueller et al. 2002; Wiltshire, Frayling et al. 2002; Xu, Bleecker et al. 2002). Studies of candidate genes have reported loci such as ESR1, CYP19, and VDR to be associated with height (Ellis, Stebbing et al. 2001; Schuit, van Meurs et al. 2004; Xiong, Xu et al. 2005), although confirmatory studies are still needed. Identifying the genomic regions that influence stature has important clinical implications, as variation in height has been associated with common diseases such as cancer, cardiovascular disease, and hip fractures (Hemenway, Feskanich et al. 1995; Forsen, Eriksson et al. 2000; Gunnell, Okasha et al. 2001; Engeland, Tretli et al. 2003). Furthermore, as height is an easily measured trait, it may serve as a useful intermediate phenotype to investigate the underlying physiology of such diseases. The Insulin-like growth factor 1 gene (IGF1) is of particular interest as it plays a key role in regulating somatic growth and mediating the effects of growth hormone (GH). Animal studies of gene knockouts and transgenics have demonstrated the importance of IGF1 in fetal and postnatal growth (Liu, Baker et al. 1993; Woods, Camacho-Hubner et al. 1996; Yakar, Liu et al. 1999). IGF1 109 polymorphisms have been associated with small birth size among French and Swedish infants (Johnston, Dahlgren et al. 2003). The IGF1 dinucleotide (CA) n repeat polymorphism has been associated with adult height in a large population- based study of the Netherlands (Vaessen, Heutink et al. 2001; Rietveld, Janssen et al. 2004). Among 9,278 Dutch subjects, homozygous carriers of the common (CA) 19 repeat (167.3 + 0.1 cm) were found to be significantly taller compared to noncarriers (166.6 + 0.2 cm) (P < 0.01) (Rietveld, Janssen et al. 2004). Circulating levels of IGF-I have been correlated with height and cancer. There is a strong relationship between IGF-I levels and height during childhood (Blum, Albertsson-Wikland et al. 1993; Juul, Bang et al. 1994), yet, the relationship among adults remains unclear (Signorello, Kuper et al. 2000; Chang, Wu et al. 2002; Teramukai, Rohan et al. 2002; Gapstur, Kopp et al. 2004; Gunnell, Oliver et al. 2004; Sandhu, Gibson et al. 2004; Wolk, Larsson et al. 2004). IGF-I levels steadily increase during childhood to a peak pubertal level then slowly decline through adulthood (Cara, Rosenfield et al. 1987; Daughaday and Rotwein 1989). Elevated circulating levels of IGF-I have been associated with increased risk of premenopausal breast, colon, and prostate cancers (Giovannucci, Pollak et al. 2000; Chan, Stampfer et al. 2002; Schernhammer, Holly et al. 2005). Recently, a study of the Multiethnic Cohort (MEC) reported that common genetic variation in the IGF1 is associated with prostate cancer susceptibility (Cheng to be published Jul 2005). Building on this work, we investigated in the MEC whether inherited variation at the IGF1 locus influences 110 IGF-I bioactivity by examining the relationship between common IGF1 genetic variation, height, and circulating IGF-I levels. We attempted to replicate any findings from our height analyses in two additional study populations of U.S. and Polish subjects. Methods Study Populations MEC The Multiethnic Cohort (MEC) is a large population-based cohort study of over 215,000 men and women from Hawaii and Los Angeles. The cohort is comprised of predominantly five racial/ethnic groups: African-Americans, Native Hawaiians, Japanese, Latinos, and Whites. Participants between the ages of 45 and 75 years were recruited from 1993 to 1996 and completed a 26-page self- administered questionnaire that included information regarding height, weight, medical history, family history, diet, dietary supplements and medication use, and physical activity. Further details are provided elsewhere (Kolonel, Henderson et al. 2000). Subjects for this study were drawn from our genetic association studies of breast, colorectal, and prostate cancers and IGF1. We included both cases and controls from each cancer site as we observed no association between height and breast (n = 1,716), colorectal (n = 1,060), and prostate (n =2,320) cancers. The total study population consisted of 9,980 subjects. 111 U.S. and Polish Stature Studies The U.S. and Polish stature studies are case-control panels of cancer, asthma, and diabetes from Genomics Collaboratvive, Cambridge, MA. Height was not associated with any of these diseases. Individuals who were diagnosed with type 1 diabetes, rheumatoid arthritis, and osteoporosis were excluded from these studies. The U.S. study is comprised of 2,189 subjects that were categorized into two groups: 1,057 short (5 th -10 th percentile in height) and 1,132 tall (90 th -95 th percentile in height). The Polish study is comprised of 1,018 subjects that were categorized into 512 short individuals and 506 tall individuals according to the same percentile distribution as used in the U.S. study. Short and tall subjects for each study were matched on ten year age groups, sex, and disease status. All study subjects were Caucasian. IGF-I Plasma Measurement In a previous study of the MEC, we measured plasma levels of IGF-I among a random sample of 1,000 MEC control participants (100 subjects in each sex and racial/ethnic group) (DeLellis, Rinaldi et al. 2004). IGF-I was measured by ELISAs from Diagnostic System Laboratories (Webster, TX). Nine-hundred and fifty-five subjects had complete IGF-I measurements, of whom 928 subjects had genotype information. Further details are described elsewhere (DeLellis, Rinaldi et al. 2004). 112 IGF1 genotyping IGF1 genotyping of the MEC subjects was conducted as part of the Multiethnic Cohort’s genetic association studies of breast, colorectal, and prostate cancer. Twenty-nine IGF1 tagging SNPs (tSNPs) that were previously identified to capture the common genetic variation of the IGF1 locus were genotyped by Taqman allelic discrimination assay (Applied Biosystems, Foster City, CA, USA) (Cheng 2005). The concordance for replicate samples was 99.7%. The average successful genotyping was 97.4%. Genotyping of the U.S. and Polish stature subjects was performed using the Sequenom mass spectrometry platform (Sequenom Inc. San Diego, CA). Fourteen tSNPs specific for Whites were genotyped (tSNPs: 1, 3, 5, 6, 8, 11, 12, 14, 17, 18, 21, 26, 27, 29). Two SNPs, tSNP 12 (rs2373722) and tSNP 14 (hCV2801104) did not meet genotyping criteria (genotyped < 75% of samples or out of Hardy-Weinberg equilibrium (1-sided P < 0.01) in > 1 ethnic group); these SNPs were eliminated from analysis. The concordance for replicate samples was 100% for both the U.S. and Polish populations. The average genotyping success rate was 97.8% and 94.3%, respectively, for the U.S. and Polish samples. Statistical Analysis In the MEC, haplotype frequencies were estimated by using genotype data of the tSNPs as described by Stram et al. (Stram, Haiman et al. 2003). Haplotype dosage (i.e. an estimate of the number of copies of haplotype h) for each 113 individual and each haplotype, h, was computed using that individual’s genotype data and haplotype frequency estimates obtained from the E-M algorithm (Zaykin, Westfall et al. 2002). To efficiently examine the relationship between common genetic variation and height, we tested the association between IGF1 haplotypes/genotypes and the extremes of the height distribution. We defined subjects as “short” (N = 1,029) and “tall” (N = 1,104) based on the lower and upper 10% of the height distribution for each sex and racial/ethnic group. Odds ratios and 95% confidence intervals were estimated by unconditional logistic regression, where the probability of being tall versus short was modeled in relation to IGF1 haplotypes and genotypes. To maximize power, we first evaluated all groups together while adjusting for sex and ethnicity; then in the presence of positive findings, we conducted sex- and ethnicity-stratified analyses. To determine which of the nominally significant SNP(s) provided the best fit of our model of height, we conducted a stepwise logistic regression analysis. We also examined using the entire height distribution (N = 9,980) whether the associated SNPs displayed differences in mean height across each genotype category (homozygote wild, heterzygote, homozygote mutant). Analysis of covariance was conducted on logarithmically transformed values of height. To correct for multiple hypothesis testing, we conducted permutation testing to assess how often the most significant observation would occur by chance if a similar study was repeated under a null distribution (i.e., no variant was associated with height) (Hirschhorn and Daly 2005). Height status (short vs. 114 tall) within strata of age, sex, and racial/ethnic group was randomly permuted 10,000 times for the 33 common haplotypes and 29 tSNPs. The smallest P value was examined in relation to the distribution of minimal P values generated from 10,000 permutations. For example, if a nominal P value of 0.05 marked the 25 th percentile of this distribution, then the permutation P value (one-sided) would be 0.25, indicating that a result as strong as the one observed would be seen by chance 25% of the time. To test for differences in mean IGF-I levels by haplotypes/genotypes, we conducted an analysis of covariance among 928 control subjects with IGF-I levels and genotype data. Plasma IGF-I values were logarithmically transformed to obtain the best approximate normal distribution. All reported P values are two-sided unless otherwise noted. Results Study Subject Characteristics MEC Characteristics of the 9,980 MEC subjects are presented in Table 19. The mean age at baseline ranged from 58 to 62 years across the five racial/ethnic groups. Approximately 20-30% of the study population was comprised of each ethnic group, with the exception of Hawaiians (9% females, 3% males). For both females and males, mean height significantly differed across racial/ethnic groups (P < 0.001). A similar racial/ethnic pattern was observed among both sexes, 115 Table 19.Study Characteristics for Height Analysis African-American Hawaiian Japanese Latino White All subjects (N = 9,980) Age (mean + SD) a 61.8 + 8.1 57.7 + 8.5 62.9 + 8 61.4 + 7.4 61.7 + 8.2 Females N (%) 1193 (25) 455 (9) 1013 (21) 1174 (24) 1012 (21) Height in cm (mean + SD) a 164.0 + 7.2 162.0 + 6.3 155.3 + 5.9 159.0 + 6.4 163.4 + 6.7 Males N (%) 1399 (27) 174 (3) 1140 (22) 1416 (28) 1004 (20) Height in cm (mean + SD) a 177.8 + 7.3 174.6 + 6.4 167.5 + 6.1 172.1 + 7.6 177.1 + 6.9 African-American Hawaiian Japanese Latino White U.S. Study Polish Study Extremes of height (N = 2,133) Females Short N (%) 139 (26) 68 (13) 101 (19) 96 (18) 130 (24) 550 (50) 236 (47) Height in cm (mean + SD) 152.5 + 3.5 152.6 + 4.4 145.7 + 3.5 147.4 + 4.3 152.9 + 2.7 153.2 + 1.5 153.4 + 1.1 Tall N (%) 155 (27) 63 (11) 124 (22) 123 (22) 107 (19) 545 (50) 268 (53) Height in cm (mean + SD) 176.2 + 5.2 172.1 + 2.3 165.4 + 5.8 170.3 + 4.6 175.3 + 4.7 172.0 + 1.8 169.6 + 0.9 Males Short N (%) 90 (18) 20 (4) 172 (35) 136 (27) 77 (16) 507 (46) 276 (54) Height in cm (mean + SD) 164.7 + 3.3 165.2 + 2.1 158.7 + 2.0 159.2 + 4.9 163.9 + 3.3 167.1 + 1.4 164.7 + 1.9 Tall N (%) 153 (29) 29 (5) 148 (29) 123 (23) 79 (15) 587 (54) 238 (46) Height in cm (mean + SD) 190.3 + 4.4 184.7 + 3.0 178.0 + 4.6 186.4 + 6.2 189.8 + 2.8 186.9 + 2.0 180.9 + 1.3 a P value <0.001 116 whereby African-Americans were taller than Whites, Hawaiians, Latinos, and Japanese. A total of 2,133 subjects were identified at the extremes of the distribution of height: 1,029 short subjects (lower 10% of height distribution) and 1,104 tall subjects (upper 10%). Approximately, 15-30% of the extremes of height were comprised of each ethnic group with the exception of Hawaiians (4-13%). U.S. and Polsih Stature Studies Study characteristics of the 2,189 U.S. and 1,018 Polish individuals are presented in Table 19. Approximately 50% of each stature panel was categorized into short (5 th to 10 th percentile of height distribution) and tall (90 th to 95 th percentile of height distribution) categories. For the U.S. stature study, there were 1,057 short and 1,132 tall subjects. For the Polish stature study, there were 512 short and 506 tall individuals. Genetic Analysis of IGF1 and Height In the MEC, using the 29 tSNPs we estimated the common haplotypes of the four haplotype blocks of IGF1. We observed 5 common haplotypes in block 1, 6 in block 2, 11 in block 3, and 11 in block 4. These haplotypes were consistent as previously reported (Cheng 2005), with the exception of a 5% Hawaiian- specific haplotype (4I) that was not observed in this study. We first investigated haplotype effects by global tests of associations with height (short vs. tall) and observed statistically significant associations in each of the four blocks: block 1 117 (P = 0.016), block 2 (P = 0.017), block 3 (P = 0.003), and block 4 (P = 0.001). Within each block at least one haplotype was nominally associated with height (Table 20). We observed short stature to be associated with haplotype 1B (OR = 0.70; 95% CI: 0.56-0.86), haplotype 2C (OR = 0.74; 95% CI: 0.59-0.93), haplotype 3C (OR = 0.72; 95% CI: 0.55-0.95), haplotype 3I (OR = 0.63; 95% CI: 0.41-0.96), and haplotype 4D (OR = 0.61; 95% CI: 0.47-0.78). Haplotype 3C and haplotype 3I are closely ancestrally related as they have identical alleles at all of the 10 SNPs within the block, with exception of tSNP 16, which delineates haplotype 3I as a common Latino-specific haplotype. Haplotypes 1B, 2C, 3C, and 4D demonstrated consistent associations with height when stratified by sex (Table 21). These haplotypes also display a similar pattern across racial/ethnic groups, with the exception of haplotype 3C among Latinos (Table 21). Haplotype 3C is rare (~2%) among Latinos, while the similar Latino-specific haplotype 3I (~7%) displays a similar association (OR = 0.69) with height as seen in the ethnic- stratified analysis. Haplotype 3B and 3F were associated with tall stature with borderline nominal significance. Haplotype 4D was strongest haplotype effect seen (P < 0.0001), and the permutation p-value that corrects for multiple hypothesis testing of the common haplotypes was statistically significant (P = 0.003; one-sided). Figure 7 displays the association between height and 29 tSNPs. We observed nominally significant associations (P < 0.05) between height and 11 SNPs (Table 22). These SNPs displayed a consistent pattern when stratified by 118 Table 20. Association between IGF1 common haplotypes and height: short versus tall OR (95% CI) b African-Americans Hawaiians Japanese Latinos Whites All N =229/308 a N =88/92 a N = 390/272 a N =503/356 a N =207/244 a (N=1029/1104) BLOCK 1 1A TGGA 47/44 66/75 65/66 77/80 76/80 1.00 1B CGAT 1/0 24/13 26/23 12/7 5/3 0.70 (0.56-0.86) d 1C TTAA 24/24 6/9 5/7 7/9 13/14 1.10 (0.91-1.34) 1D CGAA 13/14 1/1 0/0 2/1 3/3 1.03 (0.75-1.41) 1E CGGA 12/14 1/0 0/0 1/1 1/1 1.19 (0.84-1.68) BLOCK 2 2A GCGCA 58/59 49/56 58/52 48/50 53/54 1.00 2B GCGGG 26/28 25/28 14/20 11/13 21/19 1.11 (0.94-1.30) 2C GCCGG 1/1 17/7 23/20 12/9 4/2 0.74 (0.59-0.93) c 2D AAGCA 5/3 3/4 0/0 22/18 13/14 0.90 (0.71-1.13) 2E ACGCA 3/2 5/4 4/6 6/8 9/11 1.18 (0.90-1.56) 2F GCGGA 7/7 0/0 0/0 1/0 0/0 1.03 (0.65-1.62) BLOCK 3 3A TAGACTCGCA 46/45 47/50 56/51 42/44 45/49 1.00 3B TGGACACGGA 8/10 20/24 18/25 5/8 11/9 1.23 (1.00-1.50) c 3C TAGGCACGGG 0/0 17/6 22/19 2/3 2/0 0.72 (0.55-0.95) c 3D TAGACTCTCA 5/2 4/4 0/0 21/17 11/11 0.85 (0.66-1.09) 3E TAGAGTCGCA 3/2 2/4 1/2 9/10 12/10 0.97 (0.73-1.30) 3F TGAACACGGA 4/5 2/4 0/0 4/6 8/9 1.39 (1.00-1.94) c 3G TGGACACGCA 3/2 2/3 0/0 4/3 9/7 0.84 (0.57-1.23) 3H TAGACACGCA 11/12 1/0 0/0 0/1 1/0 1.04 (0.71-1.52) 3I TAGGCAAGGG 0/0 0/1 1/1 9/6 1/0 0.63 (0.41-0.96) c 3J TAGACACGGA 8/6 0/1 0/0 0/0 0/1 0.68 (0.41-1.11) 3K TAGGCACGGA 6/7 0/0 0/0 0/0 0/0 1.19 (0.73-1.93) BLOCK 4 4A GGCCCTGACA 24/25 35/35 26/26 49/45 37/42 1.00 4B GGTCCGGGCA 12/13 16/14 19/19 7/10 14/14 1.04 (0.85-1.28) 4C GATCCGCGCA 27/23 4/8 1/2 11/11 12/10 0.87 (0.70-1.09) 4D TATTCGCGCA 0/0 14/6 22/15 13/11 4/0 0.61 (0.47-0.78) e 4E GATCTGCGCG 5/7 19/19 18/23 1/1 2/1 1.16 (0.91-1.49) 4F GATCCGCGCG 2/2 1/3 0/0 4/7 8/8 1.22 (0.85-1.77) 4G GACCCTGACA 1/2 2/3 0/0 3/4 7/9 1.43 (0.96-2.15) 4H GGTCCGGGCG 2/2 2/3 0/0 4/3 9/7 0.79 (0.53-1.19) 4J GGTCCTGACA 0/0 1/2 7/5 0/1 0/1 0.93 (0.57-1.52) 4K GACCCTCGCA 5/5 0/0 0/0 1/1 1/0 0.89 (0.52-1.52) 4L GGTCCGGGTA 10/10 0/0 0/0 1/1 0/0 0.90 (0.59-1.36) a N = short/tall b Adjusted for age, sex, ethnicity c P < 0.05, d P < 0.001, e P < 0.0001 Haplotype Frequencies 119 Table 21. Odds Ratios and 95% CIs for IGF1 common haplotypes associated with height by sex and ethnic group a Haplotype 1B Haplotype 2C Haplotype 3C Haplotype 4D OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) Sex b Females 0.63 (0.46-0.85) e 0.57 (0.40-0.83) e 0.65 (0.43-0.98) d 0.55 (0.38-0.80) e Males 0.76 (0.57-1.02) 0.92 (0.68-1.26) 0.82 (0.56-1.20) 0.67 (0.47-0.95) d Ethnicity c African Americans 0.56 (0.12-2.67) 0.54 (0.12-2.57) -- -- Hawaiians 0.51 (0.30-0.88) d 0.33 (0.16-0.71) e 0.32 (0.14-0.71) e 0.38 (0.16-0.91) d Japanese 0.89 (0.66-1.20) 0.96 (0.67-1.37) 0.92 (0.66-1.29) 0.65 (0.45-0.96) d Latinos 0.57 (0.36-0.89) d 0.81 (0.52-1.27) 1.25 (0.51-3.03) 0.91 (0.60-1.38) Whites 0.56 (0.26-1.21) 0.40 (0.15-1.02) 0.18 (0.03-1.27) -- a Reference most common haplotype within each block b Adjusted for age and ethnicity c Adjusted for age and sex d P < 0.05, e P < 0.01 120 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 SNP -Log P value for trend Figure 7. -Log P value for trend for 29 IGF1 tSNPs and height. Block 1: snps 1-4, block 2: snps 5-9, block 3: snps 10-19, block 4: snps 20-29. Dotted line indicates P value = 0.05. 121 Table 22. Odds Ratio and 95% CI for IGF1 SNPs associated with height a All Short N (%) Tall N (%) OR (95% CI) SNP1 TT 635 (63.3) 729 (67.8) 1.00 CT 324 (32.3) 290 (27.0) 0.73 (0.60-0.89) c CC 44 (4.4) 56 (5.2) 0.97 (0.64-1.49) SNP4 AA 759 (75.3) 893 (82.2) 1.00 AT 222 (22.0) 169 (15.6) 0.65 (0.51-0.83) d TT 27 (2.7) 24 (2.2) 0.69 (0.38-1.24) SNP7 GG 778 (79.3) 888 (84.2) 1.00 CG 189 (19.3) 154 (14.6) 0.71 (0.55-0.92) c CC 14 (1.4) 13 (1.2) 0.81 (0.37-1.78) SNP11 AA 659 (65.8) 642 (60.2) 1.00 AG 304 (30.3) 370 (34.7) 1.30 (1.07-1.58) d GG 39 (3.9) 55 (5.2) 1.50 (0.97-2.32) SNP12 GG 931 (92.6) 973 (90.3) 1.00 AG 73 (7.3) 101 (9.4) 1.41 (1.01-1.95) b AA 1 (0.1) 3 (0.3) 4.12 (0.41-41.44) SNP13 AA 742 (74.1) 845 (78.2) 1.00 AG 238 (23.8) 212 (19.6) 0.74 (0.60-0.93) c GG 22 (2.2) 24 (2.2) 0.87 (0.48-1.60) SNP16 CC 946 (94.2) 1034 (96.1) 1.00 AC 55 (5.5) 41 (3.8) 0.64 (0.41-0.99) b AA 3 (0.3) 1 (0.1) 0.31 (0.03-2.98) SNP19 AA 795 (79.0) 921 (85.5) 1.00 AG 194 (19.3) 142 (13.2) 0.60 (0.46-0.78) c GG 18 (1.8) 14 (1.3) 0.62 (0.30-1.29) SNP20 GG 759 (76.5) 896 (83.4) 1.00 GT 206 (20.8) 163 (15.2) 0.66 (0.52-0.85) c TT 27 (2.7) 15 (1.4) 0.47 (0.24-0.91) b SNP23 CC 789 (78.9) 927 (86.2) 1.00 CT 190 (19.0) 134 (12.5) 0.59 (0.45-0.76) e TT 21 (2.1) 14 (1.3) 0.58 (0.29-1.17) SNP24 CC 857 (84.7) 889 (81.6) 1.00 CT 141 (13.9) 179 (16.4) 1.29 (0.99-1.69) TT 14 (1.4) 21 (1.9) 1.57 (0.77-3.19) a Adjusted for age, sex, ethnicity b P < 0.05, c P < 0.01, d P < 0.001, e P < 0.0001 122 sex and racial/ethnic group (Table 23). Tests of heterogeneity effects across ethnic groups were not statistically significant (P > 0.16). As height decreases with age, we also conducted stratified analysis by age-group, younger (<50 years) versus older age (> 50 years); a consistent pattern was observed in both age groups (data not shown). Of the 11 associated SNPs, we identified two non-correlated SNPs (tSNP 11 and tSNP 23; r 2 s = 0.03) that could account for the genetic effects observed. In comparison to individuals carrying the homozygous AA genotype of tSNP 11, both the AG and GG genotypes were associated with tall stature, OR = 1.30; 95% CI: 1.07-1.58 and OR =1.50; 95% CI: 0.97-2.23, respectively. Individuals carrying the CT genotype for tSNP 23 were more likely to be short compared to individuals carrying the common homozygous CC genotype (OR = 0.59; 95% CI: 0.45-0.76), and TT genotype as well was associated with short stature (OR = 0.58; 95% CI: 0.29-1.17). The mean height for tSNP 11 did not differ across genotype class (P = 0.203), while tSNP 23 was higher among individuals with the CC genotype than the CT and TT genotypes (P = 0.001) (Table 24). Tagging SNP 23, which uniquely marks haplotype 4D, was the most strongly associated SNP out of 29 tSNPs (P < 0.0001), and the permutation P value was statistically significant (P = 0.012; one-sided). This indicates a similar strong effect would occur by chance 1.2% of the time. To further assess the significance of these findings, we genotyped the tSNPs specific for Whites in two replication samples of U.S. and Polish populations. In the U.S. study, two SNPs were associated with height. Tagging 123 Table 23. Odds Ratio and 95% CI for IGF1 SNPs associated with height by sex a and ethnic group b Females Males AA HA JA LA WH OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) SNP1 TT 1.00 1.00 1.00 1.00 1.00 1.00 1.00 CT 0.66 (0.50-0.87) d 0.80 (0.59-1.07) 1.00 (0.69-1.45) 0.50 (0.25-0.99) c 0.76 (0.52-1.10) 0.55 (0.35-0.86) d 0.76 (0.42-1.40) CC 0.96 (0.50-1.85) 0.96 (0.55-1.68) 1.70 (0.83-3.46) 0.28 (0.07-1.16) 1.08 (0.50-2.31) 0.67 (0.18-2.56) 0.21 (0.02-1.89) SNP4 AA 1.00 1.00 1.00 1.00 1.00 1.00 1.00 AT 0.50 (0.35-0.72) e 0.82 (0.58-1.16) 0.56 (0.12-2.65) 0.45 (0.22-0.89) c 0.79 (0.55-1.14) 0.59 (0.37-0.94) c 0.64 (0.28-1.48) TT 0.94 (0.37-2.42) 0.59 (0.27-1.26) -- 0.31 (0.07-1.34) 1.01 (0.49-2.09) 0.60 (0.10-3.65) -- SNP7 GG 1.00 1.00 1.00 1.00 1.00 1.00 1.00 CG 0.60 (0.41-0.87) d 0.88 (0.62-1.25) 0.56 (0.12-2.65) 0.39 (0.18-0.83) d 0.89 (0.61-1.30) 0.79 (0.49-1.27) 0.42 (0.16-1.06) CC 0.76 (0.15-3.91) 0.94 (0.38-2.31) -- -- 1.19 (0.45-3.13) 0.94 (0.19-4.71) -- SNP11 AA 1.00 1.00 1.00 1.00 1.00 1.00 1.00 AG 1.34 (1.02-1.74) c 1.28 (0.97-1.69) 1.64 (1.08-2.48) c 1.10 (0.58-2.09) 1.44 (0.99-2.11) 1.43 (0.94-2.16) 0.86 (0.56-1.32) GG 1.64 (0.91-2.97) 1.45 (0.76-2.74) 1.31 (0.49-3.46) 1.72 (0.52-5.72) 1.73 (0.72-4.17) 1.57 (0.43-5.72) 1.19 (0.53-2.68) SNP12 GG 1.00 1.00 1.00 1.00 1.00 1.00 1.00 AG 1.41 (0.90-2.21) 1.41 (0.86-2.29) 1.57 (0.85-2.91) 1.01 (0.42-2.43) -- 1.46 (0.79-2.71) 1.11 (0.64-1.94) AA 4.21 (0.43-41.69) -- -- 2.00 (0.56-7.23) -- -- 2.97 (0.25-35.17) SNP13 AA 1.00 1.00 1.00 1.00 1.00 1.00 1.00 AG 0.74 (0.54-1.01) 0.80 (0.59-1.10) 0.90 (0.57-1.42) 0.44 (0.21-0.89) c 0.83 (0.58-1.21) 0.72 (0.46-1.14) 0.51 (0.20-1.34) GG 0.66 (0.25-1.76) 1.12 (0.52-2.44) 2.31 (0.62-8.71) 0.16 (0.02-1.50) 0.80 (0.32-1.98) 0.92 (0.18-4.64) -- SNP16 CC 1.00 1.00 1.00 1.00 AC 0.38 (0.19-0.77) d 0.97 (0.55-1.70) -- -- 1.11 (0.39-3.10) 0.67 (0.40-1.12) -- AA 0.96 (0.06-15.64) -- -- -- -- 0.30 (0.03-2.95) -- SNP19 AA 1.00 1.00 1.00 1.00 1.00 1.00 AG 0.54 (0.37-0.80) d 0.71 (0.50-1.02) -- 0.35 (0.16-0.75) d 0.79 (0.55-1.15) 0.66 (0.41-1.06) 0.15 (0.03-0.66) c GG 0.28 (0.07-1.13) 0.96 (0.40-2.31) -- -- 0.83 (0.33-2.09) 1.25 (0.28-5.66) -- SNP20 GG 1.00 1.00 1.00 1.00 1.00 1.00 1.00 GT 0.58 (0.40-0.83) d 0.80 (0.56-1.12) 0.69 (0.13-3.61) 0.45 (0.21-0.95) c 0.72 (0.50-1.05) 0.82 (0.53-1.28) 0.38 (0.15-0.93) c TT 0.32 (0.11-0.97) c 0.63 (0.28-1.44) -- -- 0.53 (0.23-1.25) 0.90 (0.26-3.20) -- SNP23 CC 1.00 1.00 1.00 1.00 1.00 1.00 1.00 CT 0.47 (0.32-0.70) e 0.76 (0.53-1.08) -- 0.29 (0.12-0.67) d 0.73 (0.50-1.08) 0.83 (0.53-1.29) -- TT 0.66 (0.22-1.93) 0.55 (0.21-1.41) -- -- 0.67 (0.27-1.67) 0.91 (0.26-3.21) -- SNP24 CC 1.00 1.00 1.00 1.00 1.00 1.00 1.00 CT 1.41 (0.97-2.04) 1.17 (0.80-1.71) 1.19 (0.69-2.05) 0.90 (0.46-1.75) 1.45 (1.00-2.12) 0.85 (0.17-4.32) 1.81 (0.62-5.31) TT 2.15 (0.78-5.94) 1.25 (0.46-3.40) -- 1.67 (0.38-7.37) 1.42 (0.59-3.41) -- -- b Adjusted for age and ethnicity b Adjusted for age and sex c P < 0.05, d P < 0.01, e P < 0.001, f P < 0.0001 124 Table 24. Age, sex, and ethnicity adjusted geometric mean height by IGF1 SNPs among entire MEC study a N Height (cm) 95 % CI SNP1 TT 6508 167.15 (166.96, 167.33) CT 2806 166.65 (166.38, 166.92) CC 434 166.94 (166.29, 167.59) P b 0.007 SNP3 GG 4896 167.15 (166.94, 167.35) AG 3716 166.78 (166.55, 167.02) AA 910 166.87 (166.41, 167.32) P b 0.048 SNP4 AA 7907 167.14 (166.96, 167.31) AT 1668 166.55 (166.20, 166.89) TT 223 166.07 (165.17, 166.98) P b 0.003 SNP7 GG 7918 166.91 (166.91, 167.25) CG 1492 166.65 (166.28, 167.02) CC 163 166.09 (165.04, 167.14) P b 0.033 SNP11 AA 6140 166.90 (166.71, 167.09) AG 3111 167.09 (166.84, 167.34) GG 458 167.37 (166.75, 167.99) P b 0.203 SNP13 AA 7441 167.10 (166.93, 167.27) AG 2085 166.73 (166.42, 167.04) GG 226 166.37 (165.48, 167.26) P b 0.043 SNP16 CC 9280 167.01 (166.86, 167.17) AC 457 166.25 (165.60, 166.90) AA 23 165.92 (163.19, 168.69) P b 0.057 SNP19 AA 8154 167.11 (166.94, 167.28) AG 1448 166.42 (166.04, 166.80) GG 176 166.05 (165.04, 167.06) P b 0.001 SNP20 GG 7815 167.13 (166.96, 167.31) GT 1669 166.57 (166.22, 166.93) TT 208 165.74 (164.81, 166.68) P b 0.001 SNP23 CC 8095 167.11 (166.94, 167.28) CT 1497 166.38 (166.01, 166.75) TT 170 166.30 (165.28, 167.33) P b 0.001 SNP24 CC 8403 166.98 (166.81,167.15) CT 1330 167.02 (166.64, 167.40) TT 133 167.12 (165.96, 168.28) P b 0.962 SNP29 AA 6938 166.96 (166.77, 167.14) AG 2442 167.06 (166.78, 167.33) GG 225 167.29 (166.41, 168.18) P b 0.662 a Numbers do not add up to 9,980 due to missing genotype data b P value for trend 125 SNP 1 was associated with short stature (P = 0.027) and tSNP 6 was associated with tall stature (P = 0.006). The results of tSNP 1 in this population supported the findings of the MEC, where tSNP 1 was also associated with short stature (Table 23). The association observed with tSNP 6 in the U.S. study was not seen the MEC. No association was observed between tSNP 11 and height in the U.S. study as was seen in MEC. Tagging SNP 23 which was significantly associated with height in the MEC was not genotyped in the replication studies as it was not selected as a tSNP for Whites. Among the Polish subjects, none of the IGF1 tSNPs for Whites were associated with height. In this study, we also examined the relationship between common genetic variation in IGF1 and circulating IGF-I levels among 928 control subjects (Figure 8). We observed a nominally significant association with haplotype 1C (11%) and circulating levels (P = 0.013). Ethnic-stratified analysis demonstrated only a marginal association between halpotype 1C among African-Americans (P = 0.054) and no association was observed in the other groups. This haplotype effect could be attributed to the genotype effect of tSNP 2 (P = 0.024) in African- Americans. No other associations were observed between IGF1 haplotypes and circulating levels. Similarly, no other associations were observed between IGF-I levels and the remaining tSNPs (data not shown). 126 0 50 100 150 200 250 300 350 2A 2B 2C 2D 2E 2F Block 2 Haplotypes Geometric Mean IGF-I (ng/ml) No Copy 1 Copy 2 Copies 0 100 200 300 400 500 600 700 3A 3B 3C 3D 3E 3F 3G 3H 3I 3K 3L Block 3 Haplotypes Geometric Mean IGF-I (ng/ml) No Copy 1 Copy 2 Copies 0 50 100 150 200 250 300 350 400 450 500 4A 4B 4C 4D 4E 4F 4G 4J Block 4 Haplotypes Geometric Mean IGF-I (ng/ml) No Copy 1 Copy 2 Copies 0 50 100 150 200 250 300 350 1A 1B 1C 1D 1E Block 1 Haplotypes Geometric Mean IGF-I (ng/ml) No Copy 1 Copy 2 Copies *P = 0.013 * Figure 8. Geometric mean IGF-I levels (ng/ml) and 95% CIs for IGF1 blocks 1-4. 127 Discussion It is well-established that IGF1 plays a principal role in somatic growth and production of IGF-I is believed to be an important determinant in various aspects of human physiology including height. In this report, we examined whether inherited differences in IGF1 relates to variation in height and circulating levels of IGF-I in a large multiethnic study of five racial/ethnic groups. We also followed-up our investigation of height in two independent study populations. Our data demonstrate that inherited variation in IGF1 influences height. Overall, the common genetic variation in IGF1 was predominantly associated with short stature, whereby an approximately 2 cm decrease in mean height was observed for each additional copy of the minor allele of the associated polymorphisms. We estimated the genetic effects observed explained approximately 55% of the variation in height. Several haplotypes and SNPs were associated with height across the entire 156 kb of the IGF1 locus. It is of particular interest that the haplotypes associated with height, haplotypes 1B, 2C, 3C, and 4D, were also associated with increased risk of prostate cancer in a previous study of the MEC (Cheng to be published Jul 2005). Similarly, 7 of the 11 SNPs associated with height were also associated with increased prostate cancer risk (Cheng 2005). These shared genetic effects on height and prostate cancer suggest that the genetic variation in IGF1 may have functional consequences. Previous studies have reported increasing height to be associated with higher risk of prostate cancer (Giovannucci, Rimm et al. 1997; 128 Rodriguez, Patel et al. 2001; Engeland, Tretli et al. 2003), although within the MEC we observed no association (Table 25). With reports of a positive association between height and prostate cancer, we would presume that the genetic variation in IGF1 would be associated with tall stature; yet, we observed the majority of inherited differences in IGF1 to be associated with short stature. Our findings draw attention to the complexity of the GH-IGF axis (GH/IGFs, receptors, and associated binding proteins) and that much remains to be learned about the mechanisms that regulate growth and disease. Table 25. Association between prostate cancer and height OR (95% CI) a All b 1.00 (0.99-1.01) African-American c 1.01 (0.99-1.02) Hawaiian c 1.01 (0.95-1.06) Japanese c 1.00 (0.98-1.02) Latino c 0.99 (0.98-1.01) White c 0.99 (0.97-1.01) a OR estimated by unconditional logisitic regression for risk associated with cm increase in height b Adjusted for age and ethnicity In two reports of the Rotterdam Study of the Netherlands, the IGF1 (CA) n repeat polymorphism was found to be associated with height (Vaessen, Heutink et al. 2001; Rietveld, Janssen et al. 2004). In the initial report of 900 subjects, individuals who were non-carriers of the most common (CA) 19 repeat were on average 2.7 cm shorter than homozygous (CA) 19 carriers (P = 0.004) (Vaessen, Heutink et al. 2001). A follow-up study of 9,278 subjects confirmed this initial observation, whereby individuals who did not possess the (CA) 19 polymorphism were significantly shorter (166.6 + 0.2 cm) in comparison individuals who were 129 homozygous for the common (CA) 19 repeat (167.3 + 0.1 cm) (P < 0.01) (Rietveld, Janssen et al. 2004). Our results are consistent with these reports as the less common IGF1 variants were predominantly associated with short stature. Using data from a previous study (DeLellis, Ingles et al. 2003), we were able to determine that the (CA) n repeat, which is located ~1 kb upstream of the transcription start site in block 1 is in linkage disequilbrium (LD) with neighboring SNPs whereby non-(CA) 19 alleles are linked with the less frequent alleles in block 1. Two population-based studies have examined the role of IGF1 genetic variation and circulating IGF-I levels; these studies focused solely on the (CA) n repeat polymorphism (Allen, Davey et al. 2002; Rietveld, Janssen et al. 2004). In the EPIC study of 696 individuals from U.K., the (CA) n repeat was not associated with circulating levels (Allen, Davey et al. 2002); while in the Rotterdam study of 189 subjects, homozygous carriers of the (CA) 19 repeat had the highest IGF-I levels in comparison to subjects with either shorter or longer (CA) n repeats (Rietveld, Janssen et al. 2004). Although haplotype 1C was associated with circulating levels, we observed no other genetic effects. It is likely that this association is due to chance given its weak effect and the number of hypotheses tested. Inconsistencies between reports may be due to differences in study design, population characteristics, and accuracy of IGF-I measurements. Furthermore, trans-acting variants at other loci may also exist that control circulating levels in vivo (Morley, Molony et al. 2004). 130 We observed in our study a stronger association between IGF1 and height among women than men. During adolescence normal growth and development are dependent on the critical interactions between sex hormones and the GH-IGF axis (Juul 2001). It is possible that minor perturbations in the hormonal milieu in conjunction with variation in IGF1 may have differential effects on growth for women and men. We examined whether early age at menarche confounded the relationship between IGF1 and height given that the rise in estrogen levels during menarche stimulates fusion of the epiphyseal growth plates and antagonizes the growth effects of GH and IGF-I (Stanhope, Pringle et al. 1988; Caufriez 1997; Eastell 2005). We observed similar results with adjustment for age at menarche, suggesting that an earlier age at menarche is an unlikely explanation for the association between IGF1 and short stature. One important concern of genetic association studies, especially for studies of height, is the presence of population stratification due to recent admixture. Given that Japanese subjects tend to be shorter than the other racial/ethnic groups, it is possible that spurious associations may have occurred due to an overrepresentation of Japanese alleles among short individuals in comparison to those who are tall. In MEC study, we matched the short and tall groups on self-reported ethnicity, which is believed to minimize false positive associations from stratification (Ardlie, Lunetta et al. 2002), and our findings were consistent when stratified by ethnicity. This argues against large-scale 131 stratification, although mild stratification due to imbalances in population substructure cannot be excluded (Freedman, Reich et al. 2004). Another concern of genetic association studies is the low rate of replication of subsequent studies (Hirschhorn, Lohmueller et al. 2002). Encouragingly, we observed tSNP 1 to be associated with height in both the MEC and U.S. stature studies. Tagging SNP 1 (rs7965399) is of particular interest as it was identified to be associated with prostate cancer risk in a previous study of the MEC (Cheng to be published Jul 2005). Yet, in the Polish study, we did not observe an association between IGF1 genetic variation and height. Heterogeneity in allele frequencies is one possible explanation for the lack of replication. For example, in the Polish study, the frequency of the minor allele of tSNP 1 was 2%, while in the U.S. study and among Whites in the MEC, the minor allele frequency was 5%. It is possible that a causal variant may be common in one population but rare in others (Hirschhorn, Lindgren et al. 2001). Limited study power is another possible explanation as individuals studies are rarely large enough to detect modest genetic effects (Altshuler, Hirschhorn et al. 2000; 2004). Our findings of an association between IGF1 and height were robust when evaluated by permutation testing. Permutation testing corrects for multiple hypothesis testing by empirically assessing the probability of having observed a given association by chance (Hirschhorn and Daly 2005). We determined that the haplotype and SNP effects observed would have occurred by chance less than 1% of the time. This demonstrates that IGF1 is a strong candidate gene to explain the 132 variation in height and increases the likelihood that subsequent studies will observe a similar association. Confirmation of our findings by additional studies is critical in establishing a relationship between IGF1 and height. Moreover, the causal IGF1 variant(s) responsible for differences in height remain to be elucidated. Identifying the genetic determinants of stature will further our insight into the physiology of human growth as well as our understanding of diseases in which height is believed to play a role. 133 PART III. GRANT PROPOSAL Section 1. A Comprehensive genomic approach to characterize the role of IGF Receptor genes in relation to breast cancer risk: The Multiethnic Cohort The Susan G. Komen Breast Cancer Foundation Request for Application Forms for Basic, Clinical and Translational Breast Cancer Research Principal Investigator: Christopher A. Haiman, Sc.D. Institution: Keck School of Medicine, USC Address: 1441 Eastlake Avenue, Room 4441 City, State, Zip: Los Angeles, CA 90033-0804 Phone: 323-865-0429 Fax: 323-865-0127 Email: haiman@usc.edu Total Amount Requested: 250,000.00 134 Abstract Background: Breast cancer is characterized by uncontrolled cell growth and the spread of abnormal cells. Insulin-like growth factors (IGFs) are potent stimulators of breast cell proliferation and inhibitors of cell death. Multiple lines of human and experimental evidence support the role of the IGF system in breast cancer development. As the effects of IGFs are mediated through interactions with their receptors, the Insulin-like Growth Factor-1 receptor (IGF1R) and Insulin-like Growth-Factor-2 receptor (IGF2R) are attractive genes in studying breast cancer etiology. Objective/Aims: Our central question is whether common genetic variation in IGF1R and IGF2R genes are associated with breast cancer risk. In addition, we plan to investigate whether racial/ethnic differences in breast cancer risk among African-American, Native Hawaiian, Japanese, Latina, and white women may be related to differences in the genetic sequences of the IGF receptors. Methods: Our study population will consist of a large multiethnic population of African-American, Native Hawaiian, Japanese, Latina, and white participants in a multiethnic cohort study. We will utilize high-throughput genotyping technology to comprehensively evaluate genetic variation in the IGF1R and IGF2R genes. We will investigate whether genetic differences in IGF1R and IGF2R are associated with breast cancer risk among 1,715 breast cancer cases and 2,502 controls. 135 Outcomes/Benefits: This proposal integrates a large multiethnic population-based study, cutting edge genomic resources, and a novel methodologic approach to investigate the genetic contributions of the IGF receptors to breast cancer risk. This will be the first study to extensively evaluate genetic germ-line variations in IGF1R and IGF2R as they relate to breast cancer susceptibility. In addition, our study population of African-Americans, Native Hawaiian, Japanese, Latinas, and whites provides us the ability to examine genetic differences among a diversity of racial/ethnic backgrounds. 136 Scientific Abstract The Insulin-like Growth Factor (IGF) system has an established role in cellular growth, and accumulating evidence support the involvement of the IGF system in breast cancer development and progression. IGFs are mitogens that play a critical role in cell proliferation and apoptosis. The Insulin-like Growth Factor-1 Receptor (IGF1R) and Insulin-like Growth Factor-2 Receptor (IGF2R) genes are of particular interest given their role in mediating IGF activity; it has yet to be determined whether genetic variations in these genes contribute to breast cancer risk. With recent innovations in genomic technology, large scale studies have advanced our understanding of human genetic variation and have lead to the development of promising approaches to evaluate the inherited contribution of candidate genes to breast cancer susceptibility. The hypothesis of this proposal is that common genetic variations in IGF1R and IGF2R are associated with breast cancer risk. Our specific aims are: 1. To characterize the haplotype block structures and identify a set of haplotype-tagging single nucleotide polymorphisms (htSNPs) that capture the common haplotype diversity of IGF1R and IGF2R among African- American, Native Hawaiian, Japanese, Latina, and white women. 2. To examine the association between common haplotypes of IGF1R and IGF2R and breast cancer risk. 137 3. To evaluate gene-gene haplotype interactions between the IGF1R and IGF2R genes and previously characterized haplotypes of IGF-I, IGFBP1, and IGFBP3. A haplotype-based approach using linkage disequilibrium mapping will be used to identify chromosomal regions of IGF1R and IGF2R that may harbor putative disease-associated variants. A high density of SNPs spanning IGF1R and IGF2R will be genotyped using high-throughput genotyping methods in a multiethnic panel of 350 controls. We will estimate the common haplotypes and identify htSNPs that predict the common haplotypes in each population. Associations between common haplotypes of IGF1R and IGF2R and breast cancer risk will be examined in a case-control study (1715 cases/2502 controls) within a multiethnic cohort study. The overall goal of this proposal is to assess the genetic contribution of two candidate breast cancer susceptibility genes in the development of breast cancer, with a particular focus on understanding ethnic differences in risk. Such detailed knowledge may be beneficial to the development and design of therapies for breast cancer, and aid in the identification of women at greater risk of disease who may benefit from early detection and prevention strategies. 138 Breast cancer is characterized by uncontrolled cellular growth and the spread of aberrant cells. This course of pathogenesis is facilitated by the disruption of regulatory systems that mediate normal cell proliferation and death. Multiple lines of evidence from population and experimental studies provide strong support for the role of the Insulin-like Growth Factor (IGF) system in breast cancer tumorigenesis. IGFs serve as regulatory peptides involved in cell proliferation, differentiation, and apoptosis. As the actions of the IGF system are mediated through their receptors, the Insulin-like Growth Factor-1 Receptor (IGF1R) and Insulin-like Growth Factor-2 Receptor (IGF2R) are attractive candidate genes in understanding breast cancer etiology. Currently, no studies have investigated the association between germ-line genetic variation in the IGF receptors and breast cancer risk. Objective/Hypothesis In this study, we hypothesize that genetic variation in IGF1R and IGF2R play a role in breast cancer susceptibility. We propose a haplotype-based approach to evaluate the association between common genetic variation in IGF receptors and breast cancer risk in a large multiethnic population. This study will include African-American, Native Hawaiian, Japanese, Latina, and white participants from the Multiethnic Cohort Study (MEC), a large prospective cohort study established in 1993 to investigate environmental and genetic factors in relation to cancer etiology. 139 The specific aims of this research proposal are: 1) To characterize the haplotype block structures of IGF1R and IGF2R among African-Americans, Native Hawaiian, Japanese, Latina, and whites in the MEC. 2) To identify a set of haplotype-tagging single nucleotide polymorphisms (htSNPs) that capture the common haplotype diversity within haplotype blocks of IGF1R and IGF2R in the five racial/ethnic groups. 3) To examine the association between common haplotypes of IGF1R and IGF2R and breast cancer risk in a large case-control study within the MEC. 4) To evaluate breast cancer risk respective to the interactions between the common haplotypes of IGF1R and IGF2R and previously characterized common haplotypes of IGF-I, IGFBP-1, and IGFBP-3. The overall goal of this proposal, as reflected in its specific aims, is to obtain a comprehensive understanding of the genetic differences of the IGF receptors as they relate to breast cancer risk, particularly with regard to racial/ethnic differences in risk. Such detailed knowledge may be highly useful in breast cancer prevention and the development of targeted treatment strategies. 140 Background and Signficance The IGF System The IGF system is comprised of several molecules that form a highly regulated network: two ligands (IGF-I and IGF-II), two types of cell membrane receptors (IGF1R and IGF2R), six binding proteins (IGFBP-1 through IGFBP-6), and several binding proteases (1). Most circulating IGFs are produced by the liver in response to growth hormone stimulation (2) and exert their actions by binding to the IGF receptors in target tissues. The bioavailability of free IGFs in the circulation and tissues is regulated by IGFBPs, which bind IGFs with high affinity and specificity. More than 90% of circulating IGFs are bound to IGFBP-3 and a liver-derived glycoprotein, acid-labile subunit (ALS), to form a 150 kd ternary complex. The IGF System and Breast Cancer Prospective case-control studies have demonstrated a positive relationship between circulating IGF-I levels and breast cancer risk in premenopausal women (3-5). Treatment of breast cancer by antiestrogens (6-8) or retinoids (9, 10) have been found to decrease circulating IGF-I levels. Such findings have prompted the hypothesis that higher circulating levels of IGFs may relate to elevated IGF tissue bioactivity and corresponding increases in epithelial turnover; thus more opportunities for genetic alterations that eventually result in cellular 141 transformation. In addition, IGFs may affect the progression of neoplasms through the enhanced survival and proliferation of early neoplastic cells (11). IGF-I and IGF-II are potent mitogens in breast cancer cell lines as well as breast tumors (12, 13), and both IGFs and their receptors are expressed in breast tissue. In the breast, IGF-I is produced primarily by stromal cells (14). Malignant breast epithelial cells have been found to induce IGF-I expression in breast stroma in vitro (15). IGF-I has also been proposed to act as a paracrine stimulator of adjacent epithelial cells or an autocrine stimulator of stromal cells in the breast (14). IGF-II is highly expressed in the breast stroma (16), but is also found in malignant epithelial cells (17). A high level of IGF-II expression has been found to correlate with enhanced tumor growth (18). In addition, IGF-II is also considered to have important paracrine effects on breast epithelial cell growth (16). The potential role of IGFBPs in cancer development is unclear, although recent studies have shown that IGFBPs can modulate cell proliferation in the breast epithelium through both IGF-dependent and IGF-independent mechanisms (19). IGF1R The insulin-like growth factor-I receptor (IGF1R) is a transmembrane tyrosine kinase receptor that mediates the biological actions of both IGF-I and IGF-II (20). It is composed of two alpha chains that comprise the extracellular 142 ligand binding domain, and two beta chains that comprise the intracellular tyrosine kinase domain and extracellular region (21). Activation of the receptor by IGF-I or IGF-II directly phosphorylates various substrates such as insulin receptor substrate-1 (IRS-1) and src-homology 2/collagen-alpha proteins (Shc), linking this receptor to downstream signaling mechanisms such as the ras/MAPK pathway (through Grb-2/Sos) and the PI3-K/Akt pathway (through the p85 regulatory subunit) (22). Ras/MAPK signaling leads to cellular growth and activation of the PI3-K/Akt pathway is implicated in governing anti - apoptotic effects (23, 24). Breast cancer cell lines and tumor specimens have been shown to express mRNA encoding IGF1R (25). Several studies have reported an elevated expression of IGF1R in malignant breast tissue compared to normal or benign tissue (26, 27). Results from studies of breast cancer treatment agents such as the antiestrogens, Tamoxifen and ICI 182,780, indicate that these compounds modulate IGF1R signaling by down-regulating IGF1R and subsequent downstream signaling components (28-30). In addition, a study investigating trastuzumab (Herceptin), a breast cancer therapeutic that targets the HER2/neu receptor, has shown that IGF1R activation may be a mechanism in which tumors become resistant to trastuzumab treatment (31). Expression of IGFIR in breast tumors is highly correlated with estrogen receptor (ER) expression (32-34). Estrogen has been shown to mediate the up- regulation of IGF signaling system components such as IGF1R, insulin-receptor 143 substrate-I (IRS1), and the p85 regulatory subunit of the PI - 3K pathway (33-35). Estrogen also modulates the down-regulation of IGF2R and IGFBP-3 (36, 37). IGF-I and estrogen have been shown to act synergistically in stimulating the proliferation of breast cancer cell lines as well as breast tumors (38). In addition, the proliferation response can be inhibited by using ER-blocking compounds or IGF1R antibodies (28, 39), which suggests that both hormonal signals may be necessary for breast cancer cell proliferation. Clinical studies have implicated IGF1R activity in breast cancer prognosis. Elevated IGF1R levels are reported to be highly correlated with breast tumor recurrence after lumpectomy and radiation therapy (40). IGF1R-positive patients have also been reported to have a significantly lower proportion of disease-free survival compared to IGF1R-negative patients (40% and 75%, respectively) (32). Interestingly, when stratified by ER status, patients with ER-/IGF1R+ had worse prognosis in comparison to ER-/IGF1R- patients (32). Although the precise role of IGF1R as a prognostic factor remains to be determined, IGF1R appears to be an important component in breast cancer progression. The IGF1R gene is located on chromosome 15q26.3. It spans approximately 309 kb and is comprised of 21 exons. Alternate splicing of the mRNA transcript at the 5’ end of exon 14 results in replacement of an arginine for a threonine-glycine in the extracellular portion of the beta subunit (41). The arginine form of the receptor displays a 50% decrease in ligand-mediated 144 endocytosis and two-fold higher autophosphorylation activity (41). Both transcripts are expressed at similar levels in various human tissues and cell lines (42). IGF2R The insulin-like growth factor-II receptor (IGF2R), also known as the Mannose-6-Phosphate (M6P) receptor, is a single-chain transmembrane receptor that binds IGF-II with a hundred-fold higher affinity than IGF-I. In contrast to IGF1R, IGF2R has no transmembrane signaling activity (43). Binding of IGF-II to IGF2R results in degradation of IGF-II by internalization and transport to the lysosomes. By removing IGF-II from the extracellular environment and precluding its activation of the IGF1R, IGF2R is believed to reduce the biological effects of IGF-II (44). IGF2R also binds and activates ligands containing mannose-6-phosphate, including factors implicated in breast carcinogenesis such as latent transforming growth factor beta-1 (TGF beta-1), a potent growth factor inhibitor that has been positively associated with breast cancer (45, 46). In addition, IGF2R binds cathepsin D, a lysosomal protease that has been proposed as a prognostic marker in breast cancer (47, 48). Breast cancer cells transfected with IGF2R cDNA have been shown to produce smaller tumors and a markedly lower tumor growth rate in nude mice (49). Over-expression of IGF2R has also been seen to be associated with tumor regression in a rodent model of mammary carcinoma (50). In a recent study of 145 102 human breast tumor specimens, 24% (10/41) of invasive cases displayed significantly lower IGF2R expression in comparison to adjacent normal breast tissue, in contrast to the 10% (6/61) observed in carcinomas in situ (51). This approximately two-fold decrease in IGF2R level indicates that the receptor may be an important component of tumor progression. The IGF2R gene is located on chromosome 6q26-27, spans approximately 137 kb, and is comprised of 48 exons. The receptor is highly conserved among different species, with 80% homology shared between bovine, rat, mouse, and human receptors (33). The loss of heterozygosity (LOH) at the IGF2R locus has been reported in breast carcinomas (52, 53). In one study of 40 breast cancer specimens, a deletion of one IGF2R allele was observed in 33% invasive cancers and 26% of in situ cancers (52). In addition, analysis of the remaining allele revealed somatic missense mutations (52) that also have been shown to alter ligand binding of IGF2R (54). Given the antagonist role of IGF2R and evidence of LOH and loss-of-function mutations, the receptor has been proposed to serve as a tumor suppressor in the breast. Germ-line Genetic Variation in IGF Receptors With recent efforts to characterize the genetic variation in the human genome, over nine million single nucleotide variants known as single nucleotide polymorphisms (SNPs) have been catalogued (http://www.ncbi.nlm.nih.gov/SNP/). A multitude of SNPs have been identified 146 in both IGF receptors. Across 309 kb of IGF1R, 790 SNPs have been reported with an average distance of 1 SNP every 391 bp (http://genome.ucsc.edu). Within IGF1R many SNPs have been identified in potentially relevant regions: 6 SNPs in exons, 10 SNPs in the promoter, and 28 SNPs in the untranslated region (UTR). For IGF2R, 638 SNPs have been reported spanning 137 kb, averaging approximately 1 SNP every 215 bp (http://genome.ucsc.edu). Similarly, many SNPs have been identified within relevant regions of IGF2R: 53 SNPs in exons, 21 SNPs in the promoter, and 14 SNPs in the UTR. The large genetic variation in these genes has yet to be comprehensively examined; and their genetic contribution to breast cancer susceptibility is unknown. In this proposal, we plan to employ a haplotype-based approach to examine how inherited differences in these genes contribute to breast cancer risk within and between a diversity of racial/ethnic groups. Haplotype Analysis Individuals are 99.9% identical at the nucleotide level. The majority of the 0.1% differences consist of single nucleotide polymorphisms (SNPs). Since SNPs comprise the vast majority of human genetic variation, these single nucleotide variations likely regulate the majority of phenotypic variation in the population including disease risk. Recent studies have examined a dense set of SNPs across genomic regions and have identified “haplotype blocks”, which are regions of strong linkage disequilibrium (LD) that are flanked by areas of historical 147 recombination (55). Studies have further elucidated that common (>=5%) haplotypes typically capture more than 80% of the genetic diversity in a haplotype block (55). As many SNPs are informationally equivalent, only a subset of SNPs, known as haplotype-tagging SNPs (htSNPs), is necessary to describe and predict the full complement of haplotypes. Once identified, htSNPs may be tested in association studies to examine the relationship between haplotypes and disease risk. This haplotype-based approach has been shown as a powerful and comprehensive strategy to identify chromosomal regions that may contain susceptible variants associated with common diseases (56, 57). Preliminary Results In a corresponding project in the MEC, we are investigating IGF-I and its associated binding proteins as candidate genes for breast, prostate, and colorectal cancer. We have completed haplotype characterization and htSNP selection for the genes, IGF-I, IGFBP-1, and IGFBP-3. For IGF-I, we genotyped 64 SNPs spanning 156 kb of the locus and identified four haplotype blocks: Block 1 (23 kb), Block 2 (11 kb), Block 3 (60 kb), and Block 4 (38 kb) (Figure 1, Appendix). Twenty-nine htSNPs were selected that captured the common haplotype diversity across the IGF-I locus among the five racial/ethnic groups in the MEC: Block 1 (4 htSNPs), Block 2 (5 htSNPs), Block 3 (10 htSNPs), Block 4 (10 htSNPs). For IGFBP-1, we genotyped 20 SNPs spanning 32 kb across the locus and identified two blocks: Block 1 (19 kb) and Block 2 (12 kb). Twelve htSNPs were selected to 148 capture the diversity of common haplotypes across the IGFBP-1 locus: Block 1 (5 htSNPs) and Block 2 (7 htSNPs). For IGFBP-3, we genotyped 14 SNPs spanning 38 kb across the locus and identified one block, Block 1 (25 kb), and selected 6 htSNPs to capture the common haplotype diversity. Within each block of IGF-I, IGFBP-1, and IGFBP-3 common haplotypes (>=5%) account for more than 80% of the diversity across loci in all populations. We are currently genotyping the selected htSNPs in IGF-I, IGFBP-1, and IGFBP-3 in our case-control studies in the MEC (cases/controls: breast 1,715/2,502; prostate: 2,320/2,300). Research Design and Methods 1. Study Population a. The Multiethnic Cohort (MEC) From 1993 to 1996, over 215,000 men and women between the ages of 45-75 years, comprised of African-Americans, Japanese, Native Hawaiians, Latinos, whites, and a small number of other racial/ethnic groups were enrolled in the MEC study (58). Driver’s license files in Hawaii and California were used to establish a cohort that would be ethnically and socioeconomically diverse. Two additional sources of participants were included to reach enrollment goals. In Hawaii, the voter’s registration file was used to identify older Japanese women, and in California, the Health Care Financing Administration was used to identify African-Americans of ages 65 and older. All cohort members completed a 26- page questionnaire at baseline that included information regarding medical 149 history, family history, diet, dietary supplements and medication use, physical activity, and for women, reproductive history and use of exogenous hormones. Incident breast cancer cases in the MEC are identified quarterly by linkage with the Los Angeles County Cancer Surveillance Program (CSP) and annually by linkage with the State of California Cancer Registry (CCR). Incident cases in Hawaii are identified annually by the Hawaii Tumor Registry (HTR). The Los Angeles and Hawaii tumor registries are participants of the NIH Surveillance, Epidemiology, End Results (SEER) registry. Case ascertainment in the SEER program is 98% (http://seer.cancer.gov/about/quality.html). Information on stage of disease is also collected from the cancer registries. Women are classified as having advanced, high stage disease if they had non-localized breast cancer. b. Biological Specimen Collection Collection of biological specimens in the MEC began in July 1995 in Los Angeles and in February 1997 in Hawaii and California counties outside the Los Angeles area. Incident breast cancer cases and a random sample of cohort memb ers were contacted by phone and asked to provide a blood and urine specimen. As of May 1, 2004, we had collected blood from over 1,715 breast cancer cases and 2,502 controls. The participation rate of specimen collection was 74% for cases and 66% for controls; the difference in participation rates between cases with high and low stage disease was <10%. The automated specimen component dispensing machine (the Cryo-Bio System) was used to aliquot 0.5 ml 150 serum, plasma, red blood cells, and white blood cells into 0.5 cc straw containers. Processed samples were pre-frozen to -80°C for 30 minutes and then stored in liquid nitrogen at -180°C until needed. The Multiethnic Cohort Study is currently funded to prospectively collect biospecimens for an additional 45,000 cohort participants in Los Angeles and 40,000 in Hawaii by the end of 2006. c. Multiethnic Panel for Haplotype Discovery A multiethnic panel of 350 women (70 African-Americans, Native Hawaiians, Japanese, Latinas, and whites) with no history of breast cancer will be selected for haplotype characterization. Single nucleotide polymorphisms (SNPs) will be genotyped in this panel of subjects to assess haplotype block structures of the IGF receptors. d. Breast Cancer Case-Control Study Study subjects for the case-control study include breast cancer cases with blood collected up to May 1, 2004. Control subjects are selected by frequency matching to cases on age and self-reported ethnicity. This case-control study is comprised of 1,715 breast cancer cases and 2,502 controls: (cases/controls: African-Americans, 362/658, Native Hawaiians, 115/296, Japanese 453/422, Latinas 359/680, and whites 426/446). 151 2. Laboratory Methods a. DNA extraction and storage DNA was previously extracted from white blood cell fractions using the Qiagen Blood Kit (Qiagen, Chatsworth, CA) and stored in 96-well plates at - 20°C. For genotyping, 384-well dilution plates (DNA=2.5 ng/ml) will be constructed from 96-well DNA stock plates by automated liquid handling robotics. The Multimec 94 and Tomtec Quadra 384 will be used to accurately transfer and aliquot DNA between plates. One 384-well plate of control individuals will be used for haplotype characterization (n=70/ethnic group). Eleven 384-plates comprised of the breast cancer cases and control subjects will be used for the association study. All plates will include 5% masked replicates for quality control purposes. b. SNP Genotyping Two stages of genotyping of IGF1R and IGF2R will be conducted: 1) genotyping of SNPs to densely survey the genetic variation and identify the haplotype structure of each locus, and 2) genotyping of a reduced set of haplotype-tagging SNPs for each gene for the breast cancer case-control analysis. 1. Haplotype Characterization The haplotype structure will be determined by genotyping a dense set of common SNPs across each locus in the multiethnic panel. A SNP map will be 152 created that begins 20 kb upstream of the transcription start site and ends 10 kb downstream of the last exon or untranslated region (UTR) for each locus. SNPs will be selected with a goal of one common SNP (>=5% in at least one ethnic group) spaced every 3-5 kb across each locus to ensure a high density of markers at moderate allele frequency. This is in order to adequately define the regions of linkage disequilibrium and characterize the haplotype diversity. For IGF1R and IGF2R, the total region covered will be 340 kb and 170 kb, respectively. SNPs will be selected from both the public databases (dbSNP; www.ncbi.nlm.nih.gov/SNP) and private databases (Celera; www.celera.com). To minimize the genotyping of monomorphic or rare polymorphisms, SNPs with “double hit” entries in the public database will be considered first. All SNPs located in the exons and UTRs will be tested. Also, SNPs found in regions that display >=80% sequence homology with the mouse over 200 bp will be selected to evaluate potentially important regulatory regions (ECR browser; www.dcode.org). Given the existence of many low frequency and monomorphic SNPs, multiple iterations of SNP selection and genotyping will be required in order to obtain adequate coverage (i.e. one common SNP every 3-5 kb). We project that approximately 170 SNPs (~ 1 SNP every 3 kb) will be needed to survey the 510 kb region of both genes. All genotyping for haplotype characterization will be conducted at the USC Genomic Center using the Illumnia BeadArray™ genotyping platform. The Illumina system is an automated high- throughput genotyping system that allows for 1,536-plexing in a 96-well format. 153 The system uses an oligo-ligation assay followed by PCR amplification to determine the genotype at a specific locus by a bead array containing locus specific oligonucleotide codes. Automated genotype calling tools, data storage, and analysis technology developed by Illumina will be employed. Software programs developed at USC will be utilized to assess genotyping quality, allele frequency, and Hardy-Weinberg equilibrium and linkage disequilibrium statistics (i.e., D’ and pairwise r2). 2. Genotyping of htSNPs and missense SNPs Haplotype-tagging SNPs will be genotyped using the Taqman Allelic Discrimination assay at the USC Genomic Center using the ABI 7900 Sequence Detection System from Applied Biosystems (Foster City, CA). We estimate that IGF1R will be comprised of 6-8 haplotype blocks, and that IGF2R will be comprised of 4-6 haplotype blocks. Based on our previous experience with IGF-I and IGFBPs, we estimate that 4-6 htSNPs per block will be required to predict the common haplotypes (>=5%) across all populations. This includes approximate totals of 35 htSNPs for IGF1R and 25 htSNPs for IGF2R (total n=60). In addition, all validated missense SNPs from dbSNP will be genotyped to capture any putatively functional SNPs. Ten validated missense SNPs in IGF2R have been identified in dbSNP (http://snpper.chip.org/). There are no missense snps in IGF1R. These htSNPs and missense SNPs will genotyped in 4,428 individuals: 1,715 breast cancer cases, 2,502 controls, and 5% quality control repeats (n~211). 154 3. Statistical Analysis a. Haplotype Block Definition, Haplotype Estimation, and htSNP Selection The linkage disequilibrium statistics, D’ and r2, will be used for haplotype analysis to measure pair-wise linkage disequilibrium between SNPs across each locus. SNPs with an allele frequency >=10% will be used to define haplotype blocks using the criteria of Gabriel et al. (2002), which uses the 90% confidence bounds of D’ to define historical recombination between SNPs (55). Haplotype frequencies within each block will be estimated from genotype data from the multiethnic panel of control subjects using the expectation-maximization (EM) algorithm (59). We will include SNPs with allele frequencies as low as 5% to enumerate all common haplotypes (>=5%) in each population. Haplotype-tagging SNPs (htSNPs) for the case-control study will be selected by the methods of Stram et al. (2003) (60). A computer program for the calculation of Rh2 is available at D. Stram’s website (www-rcf.usc.edu/~stram). Haplotype-tagging SNPs will be chosen by identifying a minimal set of SNPs within each block that have Rh2 >= 0.8 for all haplotypes with an estimated allele frequency >= 5%. Rh2 is the squared correlation coefficient between the true haplotypes (h’s) and their estimates (60). b. Haplotype Imputation for Case-Control Analysis Haplotype-specific risks will be estimated following the method of Zaykin et al. (2002) (59). For each individual and each haplotype h, an estimate of the 155 number of copies of haplotype h (haplotype dosage) will be computed using each individual’s genotype data and haplotype frequency estimates obtained from the entire case-control dataset. The estimate of haplotype dosage serves as a proxy for the true haplotype. Under the null hypothesis of no haplotype-specific effects on risk, the usual score test from logistic regression will correspond to the test described by Zaykin et al. (2002) (59). From our experience, this approach gives accurate estimates of the statistical significance (p-values) and confidence intervals (CI’s). Odds ratios and 95% CIs for each haplotype will be estimated using the most common haplotype observed among all ethnic groups in each block as the reference category. In addition, risk estimates will be measured separately for each common haplotype versus all other haplotypes. We will perform permutation testing to guide interpretation of nominally statistically significant associations observed with individual SNPs and haplotypes. Case- control status within racial/ethnic group will be randomly permuted 1,000 times for each SNP/haplotype. The nominal P-values associated with independent SNPs/haplotypes will then be re-evaluated in relation to the distribution of minimal P-values associated with each genotype or haplotype class that was generated by permutation (e.g., if a nominal P-value of 0.05 marks the 25th percentile of this distribution, then the permutation P-value is 0.25). In addition, for both genes we will assess the likelihood of detecting a false-positive signal by implementing the false-positive report probability criteria as described by Wacholder et al. (61). We will also evaluate haplotype-haplotype (gene 1 – gene 156 2) interactions by including interaction terms in logistic regression models. The likelihood ratio test will be used to assess the statistical significance of these interactions. c. Power to Detect Common Haplotypes For haplotype characterization in the multiethnic panel (350 subjects, 700 chromosomes), we estimate having 99% power to detect common haplotypes >=5%. With 70 subjects (140 chromosomes) in each of the five racial/ethnic groups, we have 97% power to detect ethnic-specific haplotypes with frequencies as low as 5%. d. Power for Main Genotype, Haplotype, and Gene-Gene Effects For the genotype case-control analysis, we will have substantial power to detect the main effects of genotypes (e.g. missense SNPs) among 1,715 cases and 2,502 controls. The power to detect main genotype effects relies on the model of penetrance (i.e. dominant, co-dominant, or recessive). Assuming a dominant model for an allele with a 5% frequency, we estimate having 80% statistical power (alpha=0.05; two-sided) to detect a relative risk of 1.40, and 97% power to detect a relative risk of 1.55. Under a recessive model with an allele frequency of 10%, we estimate having 80% statistical power (alpha=0.05; two-sided) to detect a relative risk of 2.20. For each ethnic group, we estimate that in a dominant model for an allele with a frequency of 10%, we will have 90% power to detect a 157 relative risk of 1.80 in all racial/ethnic groups with the exception of Native Hawaiians (minimum detectable relative risk of 2.70). For the haplotype case-control analysis, the power to detect haplotype effect relies on the ability of the selected htSNPs to predict the common haplotypes, the frequency of the haplotype, and the model of penetrance. Thus, a higher uncertainty associated with estimating common haplotypes will result in a corresponding decrease in power to detect haplotype effects. The increase in sample size needed to detect haplotype-specific effects is approximately proportional to 1/Rh2. Given our large sample size of cases and controls (n=4,217), we estimate that for those haplotypes predicted with an Rh2 >= 0.8, we will have more than sufficient statistical power to detect modest risks associated with haplotypes with frequencies as low as 5%. For example, assuming a dominant effect, we will have 80% power (alpha=0.05; two-sided) to detect a relative risk of 1.40 for a haplotype frequency as low as 5% and 97% power to detect a relative risk of 1.60. For a recessive effect, we will have 80% power (alpha=0.05; two-sided) to detect a relative risk of 2.40 for haplotypes of frequencies as low as 10%. We recognize our power to detect small effects (relative risks < 1.5) in ethnic-stratified analyses is limited. However, assuming a dominant model, for haplotypes of 10% frequency that are predicted with an Rh2 of 0.9, we will have 78-82% power to detect a relative risk of 1.6 in each ethnic group (except Native Hawaiians, relative risk of 2.0 and power, 74%). 158 The power to detect statistically significant gene-gene interactions relies on the level of certainty we are able to predict the common haplotypes, haplotype frequency, model of penetrance, as well as the main effects. Assuming independence of two genes, the inflation in sample size will be approximately proportional to 1/Rh2 x 1/Rh2. For an interaction effect on a multiplicative scale, we will have 80% power (alpha=0.05; two-sided) to detect a relative risk of 2.00 and 95% power to detect a relative risk of 2.40 using a dominant model with 10% haplotype frequency and main effect of 1.5 for each gene, respectively. Potential outcomes and benefits of research By investigating the association between genetic differences in candidate genes and breast cancer risk, we may discover genetic markers that may be useful in identifying those women at high risk of developing disease. New opportunities for prevention regimens may then be applied to reduce breast cancer morbidity and mortality. In addition, with further knowledge of the molecular mechanisms of breast carcinogenesis, new intervention strategies may be developed that target specific genetic variants. As this project will be conduced in a large multiethnic population, our findings will expand our understanding of the reasons for racial/ethnic disparities in breast cancer risk. Through this study of the genetic contributions to breast cancer susceptibility we will be better positioned to reduce the burden of disease in the population. 159 Timeline This study will be conducted in 2 years, according to the timeline outlined below. Given the iterative nature of haplotype characterization and htSNP analysis, several stages will overlap throughout the study period (Table 25). Table 26. Grant proposal timeline Year 1 Year 2 SNP genotyping and haplotype characterization X Haplotype tag SNP selection X X Haplotype tag SNP genotyping X X Case-control analysis X Manuscript preparation X Dissemination Plan The results obtained from the proposed study will be disseminated across the scientific community by presenting at national conferences and publication in leading scientific journals. We plan to present our research at the annual AACR conferences (2006/2007) and the Susan G. Komen conference. Our findings will be submitted to the Journal of the National Cancer Institute and Cancer Research for publication. In addition, information on IGF1R/IGF2R haplotype-tagging SNPs will be made publicly available through posting on the MEC website (http://www.uscnorris.com/MECGenetics). This will allow other investigators and groups, such as the National Cancer Institute’s Cohort Consortium (http://plan2004.cancer.gov/discovery/genes.htm) to examine these genes in relation to other disease phenotypes where the IGF system is believed to play a role. 160 Personnel Principal Investigator (Christopher, Haiman ScD): Dr. Haiman (15% effort) will be directly responsible for the proposed research project. Dr. Haiman will direct the scientific research and provide expertise in genetic epidemiology and breast cancer research. He will oversee the study design and supervise the genotyping and haplotype analyses. He will also contribute to the preparation of manuscripts for publication. Year 1: $14,277 + 34/34.75% fringe = $19,221 Year 2: $14,987 + 34.75% fringe = $20,195 Research Assistant (Iona Cheng, MPH): Ms. Cheng (50% effort) will work closely with Dr. Haiman and assist in all aspects of the proposed research. She will participate with the initial selection of SNPs for constructing high-density SNP maps, conducting the genotyping assays, and haplotype analyses. She will also contribute to the preparation of manuscripts for publication. Year 1: $19,207 + 34/34.75% fringe = $25,857 Year 2: $20,160 + 34.75% fringe = $27,166 161 A 5% inflation factor has been applied to salaries. The fringe benefit for the period of 07/01/04 to 06/30/05 is 34/34.75% and from 07/01/05 and beyond is estimated to be 34.75%. The total personnel cost will be $92,439. Suplies and Expenses USC Genomics Center Laboratory Supplies: Haplotype Characterization: Year 1 We estimate genotyping 170 SNPs at approximately 1 SNP every 3 kb across the two IGF receptor genes (510 kb) using the Illumina System. These 170 SNPs will be genotyped among the multiethnic panel of 368 individuals (350 controls and 5% quality control repeats). Genotyping cost of $0.22 per reaction includes costs of the oligo pooled assays, buffers, enzymes, beads and arrays. The total genotyping cost for haplotype characterization for Year 1 will be $13,763 (170 SNPs x 368 samples x $0.22). Genotyping of htSNPs and missense SNPs: Year 1 (6 mos) and Year 2 We estimate genotyping 60 htSNPs (12 blocks x 5 htSNPs per block) and 10 missense SNPs for the two IGF receptor genes These SNPs will be genotyped among 4,428 individuals (1,715 breast cancer cases, 2,502 controls, and 5% quality control repeats). Genotyping cost at USC of $0.30 per reaction includes 162 primers, fluorescently labeled probes, PCR polymerase and buffer, and PCR reaction plates and covers. The total genotyping cost for htSNPs and missense SNPs will be $92,988 (70 htSNPs & missense SNPs x 4,428 samples x $0.30): Year 1 ($30,996) and Year 2 ($61,992). Equipment None Travel Travel expenses are requested for the Principal Investigator and Research Assistant to attend the annual Susan G. Komen meeting ($750=$375 x 2 years). These expenses include round-trip airfare, lodging and transportation. Other expenses Additional funds of $100 are requested for manuscript publication during Year 2. 163 PART IV. CONCLUSION Previous studies of the genetic determinants of prostate cancer have examined only a small number of polymorphisms within candidate genes, and have had limited success in identifying alleles associated with disease. A haplotype-based strategy offers a promising approach to identify genomic regions that are associated with disease by utilizing the properties of linkage disequilibriumthe correlation between polymorphisms. For these studies, identifying the linkage disequilibrium relationship between many SNPs across a locus ensures that the common genetic variation is captured. This may not be the case when only a handful of SNPs are selected for study. Additionally, the selection of tagging SNPs to predict the common haplotypes as well as other polymorphisms is an efficient method that substantially reduces genotyping costs. Evidence in support of this approach is rapidly emerging; however, it has yet to reach complete maturity. This dissertation utilizes this comprehensive genetic strategy to investigate three candidate genes in the Insulin-like Growth Factor family, IGF1, IGFBP1, and IGFBP3, and makes a fundamental contribution to our understanding of the genetic susceptibility to prostate cancer risk. As demonstrated by this work, inherited differences in IGF1 influence prostate cancer risk while genetic variation in IGFBP1 and IGFBP3 do not appear to do so. This is in line with studies of circulating levels of IGFs in which levels of IGF-I have been most consistently associated with prostate cancer risk in 164 comparison to the other IGF proteins. We have identified a polymorphism located in block 1 of IGF1 (rs7965399) that was associated with a statistically significant 1.2-fold risk of prostate cancer. This polymorphism was also associated with height, an indicator of IGF-I bioactivity, giving support that this variant may have functional consequences on prostate cancer biology. Approximately 10% of the prostate cancer cases in our study could be attributed to possessing either one or two copies of this variant allele. In other words, if the population possessed solely the wild type allele, the prevalence of prostate cancer would be 10% lower. It is believed that most of the genetic liability of cancer is due to variants that modestly elevate risk and are present in a large proportion of the population (Balmain, Gray et al. 2003), this is in contrast to Mendelian cancer syndromes (such as BRCA1 and BRCA2) in which inherited mutations confer a very high risk of disease in only a small fraction of the population (1-2%) (Easton 1999). A strength of this work is the ability to test genetic variation across large and diverse populations: 500 to 700 African-American, Japanese, Latino, and White prostate cancer cases. The consistent genetic effects of IGF1 across the different populations suggest that inherited variation in IGF1 behaves similarly among ancestral groups and shares an overall biological effect. While ethnicity is an established risk factor for prostate cancer with African-Am ericans having the highest disease rates followed by Whites, Latinos, and Asian/Pacific Islanders (Ries LAG 2005), genetic differences in IGF1 cannot account for the variation in risk across groups. This is consistent with studies of circulating levels of IGF-I, 165 where the variation in IGF-I levels do not mirror the racial/ethnic patterns of disease (Platz, Pollak et al. 1999; Winter, Hanlon et al. 2001; DeLellis, Rinaldi et al. 2004). The similar genetic effects across populations do not preclude the possibility that environmental factors and other genetic variants may modify the biological effects of IGF1 genetic variation. For example, nutritional factors are important regulators of IGF-I levels in the circulation (Thissen, Ketelslegers et al. 1994). Recent studies have demonstrated that dietary factors such as protein and dairy intake are associated with higher circulating levels of IGF-I (Heaney, McCarron et al. 1999; Holmes, Pollak et al. 2002; Giovannucci, Pollak et al. 2003). It is possible that the genetic effects of IGF1 may be stronger in the presence of dietary factors that promote higher levels of IGF-I and weaker in the absence of such nutrients. Additional studies are needed to investigate these potential interactive effects as differences in risk across groups and overall susceptibility is likely due to the culmination of gene-environment and gene-gene effects. Our findings implicating IGF1 as an important contributor to prostate cancer risk warrants further investigation of the genetic contribution of the IGF receptors to prostate cancer susceptibility; IGF1R and IGF2R are the key mediators of IGF-I biological activity. Our proposal to characterize the haplotype structure of both receptors will allow us to expand the investigation of the IGF 166 system for its influence on prostate cancer risk as well as the examination of the possible interactive effects between IGF1 and its receptors. Genetic association studies such as the ones presented are faced with three primary challenges: 1) distinguishing between true and false positive associations; 2) identifying the predisposing causal allele(s); and 3) determining the functional significance of these causal allele(s). Firstly, in order to disentangle true and false positive findings, follow-up studies must clearly confirm observed associations. These studies must be sufficiently large to have the necessary statistical power to detect the modest genetic effects usually associated with complex diseases (Altshuler, Hirschhorn et al. 2000; 2004). For example, confirmation of our findings of an association between IGF1 and prostate cancer risk among the 8,850 prostate cancer cases of the NCI Cohort Consortium, a consortium of cohort studies from several institutions and countries, will be essential in establishing the role of IGF1 in prostate cancer susceptibility. Failure to confirm our findings may be a result of the “winner’s curse’” phenomenon, whereby the first report of an association often overestimates the true effect of the polymorphism (Hirschhorn, Lohmueller et al. 2002). Secondly, genomic regions associated with disease can range from small intervals spanning a few kilobases (e.g. the 23 kb IGF1 haplotype and prostate cancer) to large distances covering multiple genes (e.g. the 250 kb haplotype on chromosome 5q13 and Chron’s disease (Rioux, Daly et al. 2001)). Identifying the causal allele(s) within such regions may require additional sequencing to isolate 167 and test other polymorphisms for their association with disease, and moreover, calls for functional studies to decipher their biological significance. Lastly, genetic association studies must link susceptibility alleles to their molecular effects on disease such as influencing gene transcription, gene expression, and/or protein structure. This is critical in advancing our understanding of the key disease pathways involved and the molecular mechanisms by which these variants operate. Ultimately, this information may be used for disease prevention, diagnosis, and treatment. With the rapid advancements in genome databases, cost-effective genotyping capabilities, and analytic methods, future prospects in prostate cancer genetic research may extend past evaluating single candidate genes to examining entire biological pathways and eventually conducting genome-wide scans. Currently, a large international effort is underway to characterize the common genetic variation of the entire human genome, The International HapMap Project. This project will provide investigators with the means to utilize a haplotype-based strategy to select markers and efficiently test chromosomal regions for their association with disease. Also, the public SNP database, dbSNP, now contains over 10 million SNPs, which can supply future studies with the necessary information to comprehensively examine the genetic diversity of the human genome. In recent years, genotyping technology has vastly improved and it has become increasingly feasible for researchers to genotype a great number of SNPs in large study populations. Admixture mapping is another promising approach for 168 future genome-wide studies of prostate cancer as it is particularly useful in identifying disease susceptibility genes that impact admixed populations that have a higher risk of disease such as African-Americans. This approach identifies genomic regions among African-American cases of prostate cancer that have enhanced African ancestry, which may be indicative of a disease causing variant. Although we are at the early stage of unraveling the genetic basis of prostate cancer, we have made substantial progress in evaluating the genetic contribution of candidate genes such as the Insulin-like Growth Factor family of genes. Future research holds great promise in developing molecular models that incorporate both genetic and environmental effects on prostate cancer susceptibility, and with the significant progress in genomic research, technological and methodological advancements will accelerate our capabilities of identifying disease susceptibility alleles that will extend our understanding of prostate cancer biology, prevention, and treatment of the disease. 169 ALPHABETIZED BIBLIOGRAPHY (2004). "CHEK2*1100delC and susceptibility to breast cancer: a collaborative analysis involving 10,860 breast cancer cases and 9,065 controls from 10 studies." Am J Hum Genet 74(6): 1175-82. Ahmad, S., N. Singh, et al. (1999). "Role of AKT1 in 17beta-estradiol- and insulin-like growth factor I (IGF-I)-dependent proliferation and prevention of apoptosis in MCF-7 breast carcinoma cells." Biochem Pharmacol 58(3): 425-30. Allander, S. V., S. Bajalica, et al. (1993). "Structure and chromosomal localization of human insulin-like growth factor-binding protein genes." Growth Regul 3(1): 3-5. Allen, N. E., P. N. Appleby, et al. (2000). "Hormones and diet: low insulin-like growth factor-I but normal bioavailable androgens in vegan men." Br J Cancer 83(1): 95-7. Allen, N. E., G. K. Davey, et al. (2002). "Serum insulin-like growth factor I (IGF- I) concentration in men is not associated with the cytosine-adenosine repeat polymorphism of the IGF-I gene." Cancer Epidemiol Biomarkers Prev 11(3): 319-20. Altshuler, D., J. N. Hirschhorn, et al. (2000). "The common PPARgamma Pro12Ala polymorphism is associated with decreased risk of type 2 diabetes." Nat Genet 26(1): 76-80. Ardlie, K. G., L. Kruglyak, et al. (2002). "Patterns of linkage disequilibrium in the human genome." Nat Rev Genet 3(4): 299-309. Ardlie, K. G., K. L. Lunetta, et al. (2002). "Testing for population subdivision and association in four case-control studies." Am J Hum Genet 71(2): 304-11. Baibas, N., C. Bamia, et al. (2003). "Dietary and lifestyle factors in relation to plasma insulin-like growth factor I in a general population sample." Eur J Cancer Prev 12(3): 229-34. Balmain, A., J. Gray, et al. (2003). "The genetics and genomics of cancer." Nat Genet 33 Suppl: 238-44. Baxter, R. C. (1994). "Insulin-like growth factor binding proteins in the human circulation: a review." Horm Res 42(4-5): 140-4. Baxter, R. C., M. Binoux, et al. (1998). "Recommendations for nomenclature of the insulin-like growth factor binding protein (IGFBP) superfamily." Growth Horm IGF Res 8(3): 273-4. Baxter, R. C., N. Hizuka, et al. (1993). "Responses of insulin-like growth factor binding protein-1 (IGFBP-1) and the IGFBP-3 complex to administration of insulin-like growth factor-I." Acta Endocrinol (Copenh) 128(2): 101-8. Baxter, R. C. and H. Saunders (1992). "Radioimmunoassay of insulin-like growth factor-binding protein-6 in human serum and other body fluids." J Endocrinol 134(1): 133-9. 170 Benbassat, C. A., K. C. Maki, et al. (1997). "Circulating levels of insulin-like growth factor (IGF) binding protein-1 and -3 in aging men: relationships to insulin, glucose, IGF, and dehydroepiandrosterone sulfate levels and anthropometric measures." J Clin Endocrinol Metab 82(5): 1484-91. Blum, W. F., K. Albertsson-Wikland, et al. (1993). "Serum levels of insulin-like growth factor I (IGF-I) and IGF binding protein 3 reflect spontaneous growth hormone secretion." J Clin Endocrinol Metab 76(6): 1610-6. Bork, P. (1993). "The modular architecture of a new family of growth regulators related to connective tissue growth factor." FEBS Lett 327(2): 125-30. Borst, S. E., D. V. De Hoyos, et al. (2001). "Effects of resistance training on insulin-like growth factor-I and IGF binding proteins." Med Sci Sports Exerc 33(4): 648-53. Borst, S. E., K. R. Vincent, et al. (2002). "Effects of resistance training on insulin- like growth factor and its binding proteins in men and women aged 60 to 85." J Am Geriatr Soc 50(5): 884-8. Bratt, O. (2002). "Hereditary prostate cancer: clinical aspects." J Urol 168(3): 906-13. Braulke, T. (1999). "Type-2 IGF receptor: a multi-ligand binding protein." Horm Metab Res 31(2-3): 242-6. Braulke, T., M. Claussen, et al. (1995). "Proteolysis of IGFBPs by cathepsin D in vitro and in cathepsin D-deficient mice." Prog Growth Factor Res 6(2-4): 265-71. Bunker, C. H., A. L. Patrick, et al. (2002). "High prevalence of screening-detected prostate cancer among Afro-Caribbeans: the Tobago Prostate Cancer Survey." Cancer Epidemiol Biomarkers Prev 11(8): 726-9. Calle, E. E., C. Rodriguez, et al. (2003). "Overweight, obesity, and mortality from cancer in a prospectively studied cohort of U.S. adults." N Engl J Med 348(17): 1625-38. Cara, J. F., R. L. Rosenfield, et al. (1987). "A longitudinal study of the relationship of plasma somatomedin-C concentration to the pubertal growth spurt." Am J Dis Child 141(5): 562-4. Cardillo, M. R., S. Monti, et al. (2003). "Insulin-like growth factor (IGF)-I, IGF-II and IGF type I receptor (IGFR-I) expression in prostatic cancer." Anticancer Res 23(5A): 3825-35. Carmichael, C. M. and M. McGue (1995). "A cross-sectional examination of height, weight, and body mass index in adult twins." J Gerontol A Biol Sci Med Sci 50(4): B237-44. Carter, B. S., T. H. Beaty, et al. (1992). "Mendelian inheritance of familial prostate cancer." Proc Natl Acad Sci U S A 89(8): 3367-71. Caufriez, A. (1997). "The pubertal spurt: effects of sex steroids on growth hormone and insulin-like growth factor I." Eur J Obstet Gynecol Reprod Biol 71(2): 215-7. 171 Chambery, D., B. de Galle, et al. (1998). "Retinoic acid stimulates IGF binding protein (IGFBP)-6 and depresses IGFBP - 2 and IGFBP-4 in SK-N-SH human neuroblastoma cells." J Endocrinol 159(2): 227-32. Chan, J. M., M. J. Stampfer, et al. (1998). "Plasma insulin-like growth factor-I and prostate cancer risk: a prospective study." Science 279(5350): 563-6. Chan, J. M., M. J. Stampfer, et al. (1998). "Plasma insulin-like growth factor-I and prostate cancer risk: a prospective study.[comment]." Science. 279(5350): 563-6. Chan, J. M., M. J. Stampfer, et al. (2002). "Insulin-like growth factor-I (IGF-I) and IGF binding protein-3 as predictors of advanced-stage prostate cancer.[comment]." Journal of the National Cancer Institute. 94(14): 1099- 106. Chang, S., X. Wu, et al. (2002). "Plasma concentrations of insulin-like growth factors among healthy adult men and postmenopausal women: associations with body composition, lifestyle, and reproductive factors." Cancer Epidemiol Biomarkers Prev 11(8): 758-66. Chen, C., S. K. Lewis, et al. (2005). "Prostate carcinoma incidence in relation to prediagnostic circulating levels of insulin-like growth factor I, insulin-like growth factor binding protein 3, and insulin." Cancer 103(1): 76-84. Cheng, I., Stram, DO, Penney, KL, Pike, MC, Marchand, L, Kolonel, LN, Hirschhorn, J, Altshuler, D, Henderson, BE, Freedman, ML (2005). "Association study of common genetic variation in IGF1 and prostate cancer risk in the Multiethnic Cohort." JNCI in press. Cheng, I., Yu, MC, Koh, WP, Pike, MC, Kolonel, LN, Henderson, BE, Stram, DO (to be published Jul 2005). "Comparison of prostate-specific antigen and hormone levels among men in Singapore and the United States." Cancer Epidemiol Biomarkers Prev. Chestnut, R. E. and V. Quarmby (2002). "Evaluation of total IGF-I assay methods using samples from Type I and Type II diabetic patients." J Immunol Methods 259(1-2): 11-24. Chokkalingam, A. P., M. Pollak, et al. (2001). "Insulin-like growth factors and prostate cancer: a population-based case-control study in China." Cancer Epidemiology, Biomarkers & Prevention. 10(5): 421-7. Claussen, M., B. Kubler, et al. (1997). "Proteolysis of insulin-like growth factors (IGF) and IGF binding proteins by cathepsin D." Endocrinology 138(9): 3797-803. Cohen, P., H. C. Graves, et al. (1992). "Prostate-specific antigen (PSA) is an insulin-like growth factor binding protein-3 protease found in seminal plasma." J Clin Endocrinol Metab 75(4): 1046-53. Cohen, P., D. M. Peehl, et al. (1991). "Insulin-like growth factors (IGFs), IGF receptors, and IGF-binding proteins in primary cultures of prostate epithelial cells." J Clin Endocrinol Metab 73(2): 401-7. Cohen, P., D. M. Peehl, et al. (1994). "The IGF axis in the prostate." Horm Metab Res 26(2): 81-4. 172 Collett-Solberg, P. F. and P. Cohen (1996). "The role of the insulin-like growth factor binding proteins and the IGFBP proteases in modulating IGF action." Endocrinol Metab Clin North Am 25(3): 591-614. Conover, C. A. and D. R. Powell (1991). "Insulin-like growth factor (IGF)- binding protein-3 blocks IGF-I-induced receptor down-regulation and cell desensitization in cultured bovine fibroblasts." Endocrinology 129(2): 710-6. Cook, P. J., R. Doll, et al. (1969). "A mathematical model for the age distribution of cancer in man." Int J Cancer 4(1): 93-112. Costello, M., R. C. Baxter, et al. (1999). "Regulation of soluble insulin-like growth factor II/mannose 6-phosphate receptor in human serum: measurement by enzyme-linked immunosorbent assay." J Clin Endocrinol Metab 84(2): 611-7. Coverley, J. A. and R. C. Baxter (1995). "Regulation of insulin-like growth factor (IGF) binding protein-3 phosphorylation by IGF-I." Endocrinology 136(12): 5778-81. Culig, Z., A. Hobisch, et al. (1994). "Androgen receptor activation in prostatic tumor cell lines by insulin-like growth factor-I, keratinocyte growth factor, and epidermal growth factor." Cancer Res 54(20): 5474-8. DaCosta, S. A., L. M. Schumaker, et al. (2000). "Mannose 6-phosphate/insulin- like growth factor 2 receptor, a bona fide tumor suppressor gene or just a promising candidate?" J Mammary Gland Biol Neoplasia 5(1): 85-94. Daly, M. J., J. D. Rioux, et al. (2001). "High-resolution haplotype structure in the human genome." Nat Genet 29(2): 229-32. Damon, S. E., K. L. Haugk, et al. (1997). "Developmental regulation of Mac25/insulin-like growth factor-binding protein-7 expression in skeletal myogenesis." Exp Cell Res 237(1): 192-5. Daughaday, W. H., M. Kapadia, et al. (1987). "Serum somatomedin binding proteins: physiologic significance and interference in radioligand assay." J Lab Clin Med 109(3): 355-63. Daughaday, W. H. and P. Rotwein (1989). "Insulin-like growth factors I and II. Peptide, messenger ribonucleic acid and gene structures, serum, and tissue concentrations." Endocr Rev 10(1): 68-91. Dawson, E., G. R. Abecasis, et al. (2002). "A first-generation linkage disequilibrium map of human chromosome 22." Nature 418(6897): 544-8. Deal, C., J. Ma, et al. (2001). "Novel promoter polymorphism in insulin-like growth factor-binding protein-3: correlation with serum levels and interaction with known regulators." J Clin Endocrinol Metab 86(3): 1274- 80. DeChiara, T. M., E. J. Robertson, et al. (1991). "Parental imprinting of the mouse insulin-like growth factor II gene." Cell 64(4): 849-59. DeLellis, K., S. Ingles, et al. (2003). "IGF1 genotype, mean plasma level and breast cancer risk in the Hawaii/Los Angeles multiethnic cohort." Br J Cancer 88(2): 277-82. 173 DeLellis, K., S. Rinaldi, et al. (2004). "Dietary and Lifestyle Correlates of Plasma Insulin-Like Growth Factor-I (IGF-I) and IGF Binding Protein-3 (IGFBP- 3): The Multiethnic Cohort." Cancer Epidemiol Biomarkers Prev 13(9): 1444-51. Delfino, R. J., R. L. Ferrini, et al. (1998). "Demographic differences in prostate cancer incidence and stage: an examination of population diversity in California." Am J Prev Med 14(2): 96-102. Demori, I., S. Balocco, et al. (2004). "Retinoic acid increases insulin-like growth factor-binding protein-4 expression in cultured rat hepatocytes." Horm Metab Res 36(1): 7-13. D'Ercole, A. J. (1996). "Insulin-like growth factors and their receptors in growth." Endocrinol Metab Clin North Am 25(3): 573-90. DiGiovanni, J., K. Kiguchi, et al. (2000). "Deregulated expression of insulin-like growth factor 1 in prostate epithelium leads to neoplasia in transgenic mice." Proceedings of the National Academy of Sciences of the United States of America. 97(7): 3455-60. Dufourny, B., J. Alblas, et al. (1997). "Mitogenic signaling of insulin-like growth factor I in MCF-7 human breast cancer cells requires phosphatidylinositol 3-kinase and is independent of mitogen-activated protein kinase." J Biol Chem 272(49): 31163-71. Eastell, R. (2005). "Role of oestrogen in the regulation of bone turnover at the menarche." J Endocrinol 185(2): 223-34. Easton, D. F. (1999). "How many more breast cancer predisposition genes are there?" Breast Cancer Res 1(1): 14-7. Echimane, A. K., A. A. Ahnoux, et al. (2000). "Cancer incidence in Abidjan, Ivory Coast: first results from the cancer registry, 1995-1997." Cancer 89(3): 653-63. Ekstrand, J., E. Ehrenborg, et al. (1990). "The gene for insulin-like growth factor- binding protein-1 is localized to human chromosomal region 7p14-p12." Genomics 6(3): 413-8. Ellis, J. A., M. Stebbing, et al. (2001). "Significant population variation in adult male height associated with the Y chromosome and the aromatase gene." J Clin Endocrinol Metab 86(9): 4147-50. Engeland, A., S. Tretli, et al. (2003). "Height, body mass index, and prostate cancer: a follow-up of 950000 Norwegian men." Br J Cancer 89(7): 1237- 42. Erfurth, E. M., L. E. Hagmar, et al. (1996). "Serum levels of insulin-like growth factor I and insulin-like growth factor-binding protein 1 correlate with serum free testosterone and sex hormone binding globulin levels in healthy young and middle-aged men." Clin Endocrinol (Oxf) 44(6): 659- 64. Excoffier, L. and M. Slatkin (1995). "Maximum-likelihood estimation of molecular haplotype frequencies in a diploid population." Mol Biol Evol 12(5): 921-7. 174 Ferry, R. J., Jr., R. W. Cerri, et al. (1999). "Insulin-like growth factor binding proteins: new proteins, new functions." Horm Res 51(2): 53-67. Fleshner, N., P. S. Bagnell, et al. (2004). "Dietary fat and prostate cancer." J Urol 171(2 Pt 2): S19-24. Florez, J. C., N. Burtt, et al. (2004). "Haplotype structure and genotype-phenotype correlations of the sulfonylurea receptor and the islet ATP-sensitive potassium channel gene region." Diabetes 53(5): 1360-8. Forsen, T., J. Eriksson, et al. (2000). "Short stature and coronary heart disease: a 35-year follow-up of the Finnish cohorts of The Seven Countries Study." J Intern Med 248(4): 326-32. Fowlkes, J. L., J. J. Enghild, et al. (1994). "Matrix metalloproteinases degrade insulin-like growth factor-binding protein-3 in dermal fibroblast cultures." J Biol Chem 269(41): 25742-6. Frasca, F., G. Pandini, et al. (1999). "Insulin receptor isoform A, a newly recognized, high-affinity insulin-like growth factor II receptor in fetal and cancer cells." Mol Cell Biol 19(5): 3278-88. Freedman, M. L., C. L. Pearce, et al. (2005). "Systematic Evaluation of Genetic Variation at the Androgen Receptor Locus and Risk of Prostate Cancer in a Multiethnic Cohort Study." Am J Hum Genet 76(1): 82-90. Freedman, M. L., D. Reich, et al. (2004). "Assessing the impact of population stratification on genetic association studies." Nat Genet 36(4): 388-93. Frisse, L., R. R. Hudson, et al. (2001). "Gene conversion and different population histories may explain the contrast between polymorphism and linkage disequilibrium levels." Am J Hum Genet 69(4): 831-43. Funk, B., U. Kessler, et al. (1992). "Expression of the insulin-like growth factor- II/mannose-6-phosphate receptor in multiple human tissues during fetal life and early infancy." J Clin Endocrinol Metab 75(2): 424-31. Gabbitas, B. and E. Canalis (1997). "Growth factor regulation of insulin-like growth factor binding protein-6 expression in osteoblasts." J Cell Biochem 66(1): 77-86. Gabriel, S. B., S. F. Schaffner, et al. (2002). "The structure of haplotype blocks in the human genome." Science 296(5576): 2225-9. Gapstur, S. M., P. Kopp, et al. (2004). "Longitudinal associations of age, anthropometric and lifestyle factors with serum total insulin-like growth factor-I and IGF binding protein-3 levels in Black and White men: the CARDIA Male Hormone Study." Cancer Epidemiol Biomarkers Prev 13(12): 2208-16. Ghosh, P., N. M. Dahms, et al. (2003). "Mannose 6-phosphate receptors: new twists in the tale." Nat Rev Mol Cell Biol 4(3): 202-12. Giovannucci, E., M. Pollak, et al. (2003). "Nutritional predictors of insulin-like growth factor I and their relationships to cancer in men." Cancer Epidemiol Biomarkers Prev 12(2): 84-9. 175 Giovannucci, E., M. N. Pollak, et al. (2000). "A prospective study of plasma insulin-like growth factor-1 and binding protein-3 and risk of colorectal neoplasia in women." Cancer Epidemiol Biomarkers Prev 9(4): 345-9. Giovannucci, E., E. B. Rimm, et al. (1997). "Height, body weight, and risk of prostate cancer." Cancer Epidemiol Biomarkers Prev 6(8): 557-63. Glover, F. E., Jr., D. S. Coffey, et al. (1998). "The epidemiology of prostate cancer in Jamaica." J Urol 159(6): 1984-6; discussion 1986-7. Godar, S., V. Horejsi, et al. (1999). "M6P/IGFII-receptor complexes urokinase receptor and plasminogen for activation of transforming growth factor- beta1." Eur J Immunol 29(3): 1004-13. Gomez, J. M., F. J. Maravall, et al. (2004). "The IGF-I system component concentrations that decrease with ageing are lower in obesity in relationship to body mass index and body fat." Growth Horm IGF Res 14(2): 91-6. Gooch, J. L., C. L. Van Den Berg, et al. (1999). "Insulin-like growth factor (IGF)- I rescues breast cancer cells from chemotherapy-induced cell death-- proliferative and anti-apoptotic effects." Breast Cancer Res Treat 56(1): 1- 10. Goodman-Gruen, D. and E. Barrett-Connor (1997). "Epidemiology of insulin-like growth factor-I in elderly men and women. The Rancho Bernardo Study." Am J Epidemiol 145(11): 970-6. Gray, S. G., I. Stenfeldt Mathiasen, et al. (2003). "The insulin-like growth factors and insulin-signalling systems: an appealing target for breast cancer therapy?" Horm Metab Res 35(11-12): 857-71. Grimberg, A. and P. Cohen (2000). "Role of insulin-like growth factors and their binding proteins in growth control and carcinogenesis." Journal of Cellular Physiology. 183(1): 1-9. Groskopf, J. C., L. J. Syu, et al. (1997). "Proliferin induces endothelial cell chemotaxis through a G protein-coupled, mitogen-activated protein kinase-dependent pathway." Endocrinology 138(7): 2835-40. Grulich, A. E., M. McCredie, et al. (1995). "Cancer incidence in Asian migrants to New South Wales, Australia." Br J Cancer 71(2): 400-8. Grzmil, M., B. Hemmerlein, et al. (2004). "Blockade of the type I IGF receptor expression in human prostate cancer cells inhibits proliferation and invasion, up-regulates IGF binding protein - 3, and suppresses MMP-2 expression." J Pathol 202(1): 50-9. Gucev, Z. S., Y. Oh, et al. (1997). "Evidence for insulin-like growth factor (IGF)- independent transcriptional regulation of IGF binding protein-3 by growth hormone in SKHEP-1 human hepatocarcinoma cells." Endocrinology 138(4): 1464-70. Guler, H. P., J. Zapf, et al. (1989). "Insulin-like growth factors I and II in healthy man. Estimations of half-lives and production rates." Acta Endocrinol (Copenh) 121(6): 753-8. 176 Gunnell, D., M. Okasha, et al. (2001). "Height, leg length, and cancer risk: a systematic review." Epidemiol Rev 23(2): 313-42. Gunnell, D., S. E. Oliver, et al. (2004). "Do height-related variations in insulin- like growth factors underlie the associations of stature with adult chronic disease?" J Clin Endocrinol Metab 89(1): 213-8. Gunnell, D., S. E. Oliver, et al. (2003). "Are diet-prostate cancer associations mediated by the IGF axis? A cross-sectional analysis of diet, IGF-I and IGFBP-3 in healthy middle-aged men." Br J Cancer 88(11): 1682-6. Haiman, C. A., D. O. Stram, et al. (2003). "A comprehensive haplotype analysis of CYP19 and breast cancer risk: the Multiethnic Cohort." Hum Mol Genet 12(20): 2679-92. Harman, S. M., E. J. Metter, et al. (2000). "Serum levels of insulin-like growth factor I (IGF-I), IGF-II, IGF-binding protein-3, and prostate-specific antigen as predictors of clinical prostate cancer." Journal of Clinical Endocrinology & Metabolism. 85(11): 4258-65. Harrela, M., H. Koistinen, et al. (1996). "Genetic and environmental components of interindividual variation in circulating levels of IGF-I, IGF-II, IGFBP- 1, and IGFBP-3." J Clin Invest 98(11): 2612-5. Hasegawa, T., P. Cohen, et al. (1995). "Characterization of the insulin-like growth factors (IGF) axis in a cultured mouse Leydig cell line (TM-3)." Growth Regul 5(3): 151-9. Hassan, A. B. (2003). "Keys to the Hidden Treasures of the Mannose 6- Phosphate/Insulin-Like Growth Factor 2 Receptor." Am J Pathol 162(1): 3-6. Heald, A. H., J. E. Cade, et al. (2003). "The influence of dietary intake on the insulin-like growth factor (IGF) system across three ethnic groups: a population-based study." Public Health Nutr 6(2): 175-80. Heaney, R. P., D. A. McCarron, et al. (1999). "Dietary changes favorably affect bone remodeling in older adults." J Am Diet Assoc 99(10): 1228-33. Hellawell, G. O., G. D. Turner, et al. (2002). "Expression of the type 1 insulin- like growth factor receptor is up-regulated in primary prostate cancer and commonly persists in metastatic disease." Cancer Res 62(10): 2942-50. Hemenway, D., D. Feskanich, et al. (1995). "Body height and hip fracture: a cohort study of 90,000 women." Int J Epidemiol 24(4): 783-6. Hernandez-Sanchez, C., H. Werner, et al. (1997). "Differential regulation of insulin-like growth factor-I (IGF-I) receptor gene expression by IGF-I and basic fibroblastic growth factor." J Biol Chem 272(8): 4663-70. Hirschhorn, J. N. and M. J. Daly (2005). "Genome-wide association studies for common diseases and complex traits." Nat Rev Genet 6(2): 95-108. Hirschhorn, J. N., C. M. Lindgren, et al. (2001). "Genomewide linkage analysis of stature in multiple populations reveals several regions with evidence of linkage to adult height." Am J Hum Genet 69(1): 106-16. Hirschhorn, J. N., K. Lohmueller, et al. (2002). "A comprehensive review of genetic association studies." Genet Med 4(2): 45-61. 177 Hoeck, W. G. and V. R. Mukku (1994). "Identification of the major sites of phosphorylation in IGF binding protein-3." J Cell Biochem 56(2): 262-73. Hoffman, R. M., F. D. Gilliland, et al. (2001). "Racial and ethnic differences in advanced-stage prostate cancer: the Prostate Cancer Outcomes Study." J Natl Cancer Inst 93(5): 388-95. Holmes, M. D., M. N. Pollak, et al. (2002). "Lifestyle correlates of plasma insulin-like growth factor I and insulin-like growth factor binding protein 3 concentrations." Cancer Epidemiol Biomarkers Prev 11(9): 862-7. Holmes, M. D., M. N. Pollak, et al. (2002). "Dietary correlates of plasma insulin- like growth factor I and insulin-like growth factor binding protein 3 concentrations." Cancer Epidemiol Biomarkers Prev 11(9): 852-61. Hong, Y., N. L. Pedersen, et al. (1996). "Quantitative genetic analyses of insulin- like growth factor I (IGF-I), IGF-binding protein-1, and insulin levels in middle-aged and elderly twins." J Clin Endocrinol Metab 81(5): 1791-7. Huynh, H., R. M. Seyam, et al. (1998). "Reduction of ventral prostate weight by finasteride is associated with suppression of insulin-like growth factor I (IGF-I) and IGF-I receptor genes and with an increase in IGF binding protein 3." Cancer Research. 58(2): 215-8. Hwa, V., Y. Oh, et al. (1999). "Insulin-like growth factor binding proteins: a proposed superfamily." Acta Paediatr Suppl 88(428): 37-45. Hwa, V., Y. Oh, et al. (1999). "The insulin - like growth factor-binding protein (IGFBP) superfamily." Endocr Rev 20(6): 761-87. Hwa, V., C. Tomasini-Sprenger, et al. (1998). "Characterization of insulin-like growth factor-binding protein-related protein-1 in prostate cells." J Clin Endocrinol Metab 83(12): 4355-62. Ikezu, T., T. Okamoto, et al. (1995). "In vivo coupling of insulin-like growth factor II/mannose 6-phosphate receptor to heteromeric G proteins. Distinct roles of cytoplasmic domains and signal sequestration by the receptor." J Biol Chem 270(49): 29224-8. Ikushima, H., Y. Munakata, et al. (2000). "Internalization of CD26 by mannose 6- phosphate/insulin-like growth factor II receptor contributes to T cell activation." Proc Natl Acad Sci U S A 97(15): 8439-44. Ingermann, A. R., H. Kim, et al. (2000). "Characterization of a functional receptor for insulin-like growth factor binding protein 3." Growth Hormone IGF research 10:A27. Ioannidis, J. P., E. E. Ntzani, et al. (2001). "Replication validity of genetic association studies." Nat Genet 29(3): 306-9. Janssen, J. A., R. P. Stolk, et al. (1998). "Serum free IGF-I, total IGF-I, IGFBP-1 and IGFBP-3 levels in an elderly population: relation to age and sex steroid levels." Clin Endocrinol (Oxf) 48(4): 471-8. Johnston, L. B., J. Dahlgren, et al. (2003). "Association between insulin-like growth factor I (IGF-I) polymorphisms, circulating IGF-I, and pre- and postnatal growth in two European small for gestational age populations." J Clin Endocrinol Metab 88(10): 4805-10. 178 Jones, J. I. and D. R. Clemmons (1995). "Insulin-like growth factors and their binding proteins: biological actions." Endocr Rev 16(1): 3-34. Jones, J. I., A. J. D'Ercole, et al. (1991). "Phosphorylation of insulin-like growth factor (IGF)-binding protein 1 in cell culture and in vivo: effects on affinity for IGF-I." Proc Natl Acad Sci U S A 88(17): 7481-5. Jones, J. I., A. Gockerman, et al. (1993). "Extracellular matrix contains insulin- like growth factor binding protein-5: potentiation of the effects of IGF-I." J Cell Biol 121(3): 679-87. Jones, J. I., A. Gockerman, et al. (1993). "Insulin-like growth factor binding protein 1 stimulates cell migration and binds to the alpha 5 beta 1 integrin by means of its Arg-Gly - Asp sequence." Proc Natl Acad Sci U S A 90(22): 10553-7. Juul, A. (2001). "The effects of oestrogens on linear bone growth." Hum Reprod Update 7(3): 303-13. Juul, A., P. Bang, et al. (1994). "Serum insulin-like growth factor-I in 1030 healthy children, adolescents, and adults: relation to age, sex, stage of puberty, testicular size, and body mass index." J Clin Endocrinol Metab 78(3): 744-52. Juul, A., P. Dalgaard, et al. (1995). "Serum levels of insulin-like growth factor (IGF)-binding protein-3 (IGFBP-3) in healthy infants, children, and adolescents: the relation to IGF-I, IGF-II, IGFBP-1, IGFBP-2, age, sex, body mass index, and pubertal maturation." J Clin Endocrinol Metab 80(8): 2534-42. Juul, A., S. Moller, et al. (1998). "The acid-labile subunit of human ternary insulin-like growth factor binding protein complex in serum: hepatosplanchnic release, diurnal variation, circulating concentrations in healthy subjects, and diagnostic use in patients with growth hormone deficiency." J Clin Endocrinol Metab 83(12): 4408-15. Kaaks, R., A. Lukanova, et al. (2003). "Interrelationships between plasma testosterone, SHBG, IGF-I, insulin and leptin in prostate cancer cases and controls." Eur J Cancer Prev 12(4): 309-15. Kaaks, R., P. Toniolo, et al. (2000). "Serum C-peptide, insulin-like growth factor (IGF)-I, IGF-binding proteins, and colorectal cancer risk in women." J Natl Cancer Inst 92(19): 1592-600. Kaklamani, V. G., A. Linos, et al. (1999). "Dietary fat and carbohydrates are independently associated with circulating insulin-like growth factor 1 and insulin-like growth factor-binding protein 3 concentrations in healthy adults." J Clin Oncol 17(10): 3291-8. Kaklamani, V. G., A. Linos, et al. (1999). "Age, sex, and smoking are predictors of circulating insulin-like growth factor 1 and insulin-like growth factor- binding protein 3." J Clin Oncol 17(3): 813-7. Kang, J. X., Y. Li, et al. (1997). "Mannose-6-phosphate/insulin-like growth factor-II receptor is a receptor for retinoic acid." Proc Natl Acad Sci U S A 94(25): 13671-6. 179 Kaplan, P. J., S. Mohan, et al. (1999). "The insulin-like growth factor axis and prostate cancer: lessons from the transgenic adenocarcinoma of mouse prostate (TRAMP) model." Cancer Research. 59(9): 2203-9. Kaplan, P. J., S. Mohan, et al. (1999). "The insulin-like growth factor axis and prostate cancer: lessons from the transgenic adenocarcinoma of mouse prostate (TRAMP) model." Cancer Res 59(9): 2203-9. Kato, I., J. Eastham, et al. (2003). "Genotype -phenotype analysis for the polymorphic CA repeat in the insulin-like growth factor-I (IGF-I) gene." Eur J Epidemiol 18(3): 203-9. Kelley, K. M., Y. Oh, et al. (1996). "Insulin-like growth factor-binding proteins (IGFBPs) and their regulatory dynamics." Int J Biochem Cell Biol 28(6): 619-37. Kelley, K. M., K. E. Schmidt, et al. (2002). "Comparative endocrinology of the insulin-like growth factor-binding protein." J Endocrinol 175(1): 3-18. Khandwala, H. M., I. E. McCutcheon, et al. (2000). "The effects of insulin-like growth factors on tumorigenesis and neoplastic growth." Endocr Rev 21(3): 215-44. Kim, H. S., A. R. Ingermann, et al. (2004). "Insulin-like growth factor-binding protein 3 induces caspase-dependent apoptosis through a death receptor- mediated pathway in MCF-7 human breast cancer cells." Cancer Res 64(6): 2229-37. Kolonel, L. N., D. Altshuler, et al. (2004). "The multiethnic cohort study: exploring genes, lifestyle and cancer risk." Nat Rev Cancer 4(7): 519-27. Kolonel, L. N., J. Hankin, et al. (1986). Multiethnic studies of diet, nutrition, and cancer in Hawaii. Diet, Nutrition, and Cancer: Proceedings of the 16th International Symposium of the Princess Takamatsu Cancer Research Fund, Tokyo, Tokyko, Japan, Japan Scientific Societies Press. Kolonel, L. N., B. E. Henderson, et al. (2000). "A multiethnic cohort in Hawaii and Los Angeles: baseline characteristics." Am J Epidemiol 151(4): 346- 57. Korner, C., B. Nurnberg, et al. (1995). "Mannose 6-phosphate/insulin-like growth factor II receptor fails to interact with G-proteins. Analysis of mutant cytoplasmic receptor domains." J Biol Chem 270(1): 287-95. Kornfeld, S. (1992). "Structure and function of the mannose 6- phosphate/insulinlike growth factor II receptors." Annu Rev Biochem 61: 307-30. Kraemer, R. R., R. J. Durand, et al. (2004). "Rigorous running increases growth hormone and insulin-like growth factor-I without altering ghrelin." Exp Biol Med (Maywood) 229(3): 240-6. Kubler, B., S. Cowell, et al. (1998). "Proteolysis of insulin-like growth factor binding proteins by a novel 50-kilodalton metalloproteinase in human pregnancy serum." Endocrinology 139(4): 1556-63. 180 Kunitomi, M., J. Wada, et al. (2002). "Relationship between reduced serum IGF-I levels and accumulation of visceral fat in Japanese men." Int J Obes Relat Metab Disord 26(3): 361-9. Lacey, J. V., Jr., A. W. Hsing, et al. (2001). "Null association between insulin-like growth factors, insulin-like growth factor-binding proteins, and prostate cancer in a prospective study." Cancer Epidemiology, Biomarkers & Prevention. 10(10): 1101-2. Lamson, G., L. C. Giudice, et al. (1991). "Insulin-like growth factor binding proteins: structural and molecular relationships." Growth Factors 5(1): 19- 28. Lander, E. S. and N. J. Schork (1994). "Genetic dissection of complex traits." Science 265(5181): 2037-48. Landin-Wilhelmsen, K., L. Wilhelmsen, et al. (1994). "Serum insulin-like growth factor I in a random population sample of men and women: relation to age, sex, smoking habits, coffee consumption and physical activity, blood pressure and concentrations of plasma lipids, fibrinogen, parathyroid hormone and osteocalcin." Clin Endocrinol (Oxf) 41(3): 351-7. Larsson, S. C., K. Wolk, et al. (2005). "Association of diet with serum insulin-like growth factor I in middle-aged and elderly men." Am J Clin Nutr 81(5): 1163-7. Lavigne, J. A., D. J. Baer, et al. (2005). "Effects of alcohol on insulin-like growth factor I and insulin-like growth factor binding protein 3 in postmenopausal women." Am J Clin Nutr 81(2): 503-7. Le Marchand, L., L. N. Kolonel, et al. (2005). "Association of an Exon 1 Polymorphism in the IGFBP3 Gene with Circulating IGFBP-3 Levels and Colorectal Cancer Risk: The Multiethnic Cohort Study." Cancer Epidemiol Biomarkers Prev 14(5): 1319-1321. Lee, P. D., L. C. Giudice, et al. (1997). "Insulin-like growth factor binding protein-1: recent findings and new directions." Proc Soc Exp Biol Med 216(3): 319-57. Leifke, E., V. Gorenoi, et al. (2000). "Age-related changes of serum sex hormones, insulin-like growth factor-1 and sex-hormone binding globulin levels in men: cross-sectional data from a healthy male cohort." Clin Endocrinol (Oxf) 53(6): 689-95. LeRoith, D., V. M. Kavsan, et al. (1993). "Phylogeny of the insulin-like growth factors (IGFs) and receptors: a molecular approach." Mol Reprod Dev 35(4): 332-6; discussion 337-8. LeRoith, D. and C. T. Roberts, Jr. (1991). "Insulin-like growth factor I (IGF-I): a molecular basis for endocrine versus local action?" Mol Cell Endocrinol 77(1-3): C57-61. LeRoith, D. and C. T. Roberts, Jr. (2003). "The insulin-like growth factor system and cancer." Cancer Lett 195(2): 127-37. 181 Li, L., M. S. Cicek, et al. (2004). "No association between genetic polymorphisms in IGF-I and IGFBP-3 and prostate cancer." Cancer Epidemiol Biomarkers Prev 13(3): 497-8. Lighten, A. D., K. Hardy, et al. (1997). "Expression of mRNA for the insulin-like growth factors and their receptors in human preimplantation embryos." Mol Reprod Dev 47(2): 134-9. Liu, B., H. Y. Lee, et al. (2000). "Direct functional interactions between insulin- like growth factor-binding protein-3 and retinoid X receptor-alpha regulate transcriptional signaling and apoptosis." J Biol Chem 275(43): 33607-13. Liu, J. P., J. Baker, et al. (1993). "Mice carrying null mutations of the genes encoding insulin-like growth factor I (Igf-1) and type 1 IGF receptor (Igf1r)." Cell 75(1): 59-72. Lukanova, A., E. Lundin, et al. (2004). "Body mass index, circulating levels of sex-steroid hormones, IGF-I and IGF-binding protein-3: a cross-sectional study in healthy women." Eur J Endocrinol 150(2): 161-71. Lukanova, A., S. Soderberg, et al. (2002). "Nonlinear relationship of insulin-like growth factor (IGF)-I and IGF-I/IGF-binding protein-3 ratio with indices of adiposity and plasma insulin concentrations (Sweden)." Cancer Causes Control 13(6): 509-16. Lukanova, A., P. Toniolo, et al. (2001). "A cross-sectional study of IGF-I determinants in women." Eur J Cancer Prev 10(5): 443-52. Ma, J., E. Giovannucci, et al. (2001). "Milk intake, circulating levels of insulin- like growth factor-I, and risk of colorectal cancer in men." J Natl Cancer Inst 93(17): 1330-6. Majeed, N., M. J. Blouin, et al. (2005). "A germ line mutation that delays prostate cancer progression and prolongs survival in a murine prostate cancer model." Oncogene. Manes, S., M. Llorente, et al. (1999). "The matrix metalloproteinase-9 regulates the insulin-like growth factor-triggered autocrine response in DU-145 carcinoma cells." J Biol Chem 274(11): 6935-45. Manetta, J., J. F. Brun, et al. (2003). "Serum levels of insulin-like growth factor-I (IGF-I), and IGF-binding proteins-1 and -3 in middle-aged and young athletes versus sedentary men: relationship with glucose disposal." Metabolism 52(7): 821-6. Marinaro, J. A., E. C. Hendrich, et al. (1999). "HaCaT human keratinocytes express IGF-II, IGFBP-6, and an acid-activated protease with activity against IGFBP-6." Am J Physiol 276(3 Pt 1): E536-42. Martin, J. L., J. A. Coverley, et al. (1995). "Insulin-like growth factor-binding protein-3 production by MCF-7 breast cancer cells: stimulation by retinoic acid and cyclic adenosine monophosphate and differential effects of estradiol." Endocrinology 136(3): 1219-26. Maskarinec, G., Y. Takata, et al. (2005). "The relation between nutritional factors and insulin-like growth factor-I in premenopausal women of different ethnicity." Eur J Nutr 44(2): 105-13. 182 Massague, J. and M. P. Czech (1982). "The subunit structures of two distinct receptors for insulin-like growth factors I and II and their relationship to the insulin receptor." J Biol Chem 257(9): 5038-45. Mathieu, M., F. Vignon, et al. (1991). "Estradiol down-regulates the mannose-6- phosphate/insulin-like growth factor-II receptor gene and induces cathepsin-D in breast cancer cells: a receptor saturation mechanism to increase the secretion of lysosomal proenzymes." Mol Endocrinol 5(6): 815-22. McKinnon, T., C. Chakraborty, et al. (2001). "Stimulation of human extravillous trophoblast migration by IGF-II is mediated by IGF type 2 receptor involving inhibitory G protein(s) and phosphorylation of MAPK." J Clin Endocrinol Metab 86(8): 3665-74. Mettlin, C. J., G. P. Murphy, et al. (1997). "The National Cancer Data Base report on race, age, and region variations in prostate cancer treatment." Cancer 80(7): 1261-6. Minniti, C. P., E. C. Kohn, et al. (1992). "The insulin-like growth factor II (IGF- II)/mannose 6-phosphate receptor mediates IGF-II-induced motility in human rhabdomyosarcoma cells." J Biol Chem 267(13): 9000-4. Minshall, C., S. Arkins, et al. (1997). "IL-4 and insulin-like growth factor-I inhibit the decline in Bcl-2 and promote the survival of IL-3-deprived myeloid progenitors." J Immunol 159(3): 1225-32. Mohan, S., J. R. Farley, et al. (1995). "Age-related changes in IGFBP-4 and IGFBP-5 levels in human serum and bone: implications for bone loss with aging." Prog Growth Factor Res 6(2-4): 465-73. Morley, M., C. M. Molony, et al. (2004). "Genetic analysis of genome-wide variation in human gene expression." Nature 430(7001): 743-7. Morton, R. A., Jr. (1994). "Racial differences in adenocarcinoma of the prostate in North American men." Urology 44(5): 637-45. Nam, R. K., W. W. Zhang, et al. (2003). "Comprehensive Assessment of Candidate Genes and Serological Markers for the Detection of Prostate Cancer." Cancer Epidemiol Biomarkers Prev 12(12): 1429-1437. Naranjo, W. M., S. Yakar, et al. (2002). "Protein calorie restriction affects nonhepatic IGF-I production and the lymphoid system: studies using the liver-specific IGF-I gene-deleted mouse model." Endocrinology 143(6): 2233-41. Neuhausen, S. L., M. L. Slattery, et al. (2005). "Prostate cancer risk and IRS1, IRS2, IGF1, and INS polymorphisms: Strong association of IRS1 G972R variant and cancer risk." Prostate. Ngo, T. H., R. J. Barnard, et al. (2003). "Insulin-like growth factor I (IGF-I) and IGF binding protein-1 modulate prostate cancer cell growth and apoptosis: possible mediators for the effects of diet and exercise on cancer cell survival." Endocrinology 144(6): 2319-24. 183 Nickerson, T., F. Chang, et al. (2001). "In vivo progression of LAPC-9 and LNCaP prostate cancer models to androgen independence is associated with increased expression of insulin-like growth factor I (IGF-I) and IGF-I receptor (IGF-IR)." Cancer Research. 61(16): 6276-80. Nishimoto, I. (1993). "The IGF-II receptor system: a G protein-linked mechanism." Mol Reprod Dev 35(4): 398-406; discussion 406-7. Noble, R. L. (1977). "The development of prostatic adenocarcinoma in Nb rats following prolonged sex hormone administration." Cancer Res 37(6): 1929-33. Ntais, C., A. Polycarpou, et al. (2003). "Molecular epidemiology of prostate cancer: androgens and polymorphisms in androgen-related genes." Eur J Endocrinol 149(6): 469-77. Nystrom, F. H., P. K. Ohman, et al. (1997). "Population-based reference values for IGF-I and IGF-binding protein-1: relations with metabolic and anthropometric variables." Eur J Endocrinol 136(2): 165-72. Oakley-Girvan, I., L. N. Kolonel, et al. (2003). "Stage at diagnosis and survival in a multie thnic cohort of prostate cancer patients." Am J Public Health 93(10): 1753-9. Oates, A. J., L. M. Schumaker, et al. (1998). "The mannose 6-phosphate/insulin- like growth factor 2 receptor (M6P/IGF2R), a putative breast tumor suppressor gene." Breast Cancer Res Treat 47(3): 269-81. O'Connor, K. G., J. D. Tobin, et al. (1998). "Serum levels of insulin-like growth factor-I are related to age and not to body composition in healthy women and men." J Gerontol A Biol Sci Med Sci 53(3): M176-82. Ogunbiyi, J. O. and O. B. Shittu (1999). "Increased incidence of prostate cancer in Nigerians." J Natl Med Assoc 91(3): 159-64. Oh, Y., H. L. Muller, et al. (1993). "Insulin-like growth factor (IGF)-independent action of IGF-binding protein-3 in Hs578T human breast cancer cells. Cell surface binding and growth inhibition." J Biol Chem 268(20): 14964-71. Oh, Y., H. L. Muller, et al. (1995). "Transforming growth factor-beta-induced cell growth inhibition in human breast cancer cells is mediated through insulin-like growth factor-binding protein-3 action." J Biol Chem 270(23): 13589-92. Oh, Y., S. R. Nagalla, et al. (1996). "Synthesis and characterization of insulin-like growth factor-binding protein (IGFBP)-7. Recombinant human mac25 protein specifically binds IGF-I and -II." J Biol Chem 271(48): 30322-5. Ohlsson, R., A. Nystrom, et al. (1993). "IGF2 is parentally imprinted during human embryogenesis and in the Beckwith-Wiedemann syndrome." Nat Genet 4(1): 94-7. Olivier, M. (2003). "A haplotype map of the human genome." Physiol Genomics 13(1): 3-9. 184 Olney, R. C., D. M. Wilson, et al. (1995). "Interleukin-1 and tumor necrosis factor-alpha increase insulin-like growth factor-binding protein-3 (IGFBP- 3) production and IGFBP-3 protease activity in human articular chondrocytes." J Endocrinol 146(2): 279-86. Page, G. P., V. George, et al. (2003). ""Are we there yet?": Deciding when one has demonstrated specific genetic causation in complex diseases and quantitative traits." Am J Hum Genet 73(4): 711-9. Pandini, G., F. Frasca, et al. (2002). "Insulin/insulin-like growth factor I hybrid receptors have different biological characteristics depending on the insulin receptor isoform involved." J Biol Chem 277(42): 39684-95. Pandini, G., R. Mineo, et al. (2005). "Androgens up-regulate the insulin-like growth factor-I receptor in prostate cancer cells." Cancer Res 65(5): 1849- 57. Parkin, D. M., S. L. Whelan, et al. (2002). Cancer Incidence in Five Continents, Vol. VIII. Lyon, France, International Agency for Research on Cancer. Patil, N., A. J. Berno, et al. (2001). "Blocks of limited haplotype diversity revealed by high-resolution scanning of human chromosome 21." Science 294(5547): 1719-23. Pearce, C. L., J. N. Hirschhorn, et al. (2005). "Clarifying the PROGINS allele association in ovarian and breast cancer risk: a haplotype-based analysis." J Natl Cancer Inst 97(1): 51-9. Perola, M., M. Ohman, et al. (2001). "Quantitative-trait-locus analysis of body- mass index and of stature, by combined analysis of genome scans of five Finnish study groups." Am J Hum Genet 69(1): 117-23. Pfeilschifter, J., C. Scheidt-Nave, et al. (1996). "Relationship between circulating insulin-like growth factor components and sex hormones in a population- based sample of 50- to 80-year-old men and women." J Clin Endocrinol Metab 81(7): 2534-40. Phillips, K. and A. P. Matheny, Jr. (1990). "Quantitative genetic analysis of longitudinal trends in height: preliminary results from the Louisville Twin Study." Acta Genet Med Gemellol (Roma) 39(2): 143-63. Pietrzkowski, Z., G. Mulholland, et al. (1993). "Inhibition of growth of prostatic cancer cell lines by peptide analogues of insulin-like growth factor 1." Cancer Res 53(5): 1102-6. Platz, E. A., M. N. Pollak, et al. (1999). "Racial variation in insulin-like growth factor-1 and binding protein-3 concentrations in middle-aged men." Cancer Epidemiology, Biomarkers & Prevention. 8(12): 1107-10. Polascik, T. J., J. E. Oesterling, et al. (1999). "Prostate specific antigen: a decade of discovery--what we have learned and where we are going." J Urol 162(2): 293-306. Pollak, M. (2000). "Insulin-like growth factor physiology and cancer risk." European Journal of Cancer. 36(10): 1224-8. 185 Pollak, M. (2003). IGF-I, Insulin, and Cancer Risk: Linking the Biology and Epidemiology. Insulin-like Growth Factors. D. LeRoith, W. Zumkeller and R. C. Baxter. New York, NY, Kluwer Academic/Plenum Publishers: 343. Pollak, M. N., E. S. Schernhammer, et al. (2004). "Insulin-like growth factors and neoplasia." Nat Rev Cancer 4(7): 505-18. Posner, B. I., H. J. Guyda, et al. (1978). "Partial purification, characterization, and assay of a slightly acidic insulin-like peptide (ILAs) from human plasma." J Clin Endocrinol Metab 47(6): 1240-50. Powell-Braxton, L., P. Hollingshead, et al. (1993). "Inactivation of the IGF-I gene in mice results in perinatal lethality." Ann N Y Acad Sci 692: 300-1. Powell-Braxton, L., P. Hollingshead, et al. (1993). "IGF-I is required for normal embryonic growth in mice." Genes Dev 7(12B): 2609-17. Probst-Hensch, N. M., H. Wang, et al. (2003). "Determinants of circulating insulin-like growth factor I and insulin-like growth factor binding protein 3 concentrations in a cohort of Singapore men and women." Cancer Epidemiol Biomarkers Prev 12(8): 739-46. Rajaram, S., D. J. Baylink, et al. (1997). "Insulin-like growth factor-binding proteins in serum and other biological fluids: regulation and functions." Endocr Rev 18(6): 801-31. Reich, D. E., M. Cargill, et al. (2001). "Linkage disequilibrium in the human genome." Nature 411(6834): 199-204. Ren, Z., Q. Cai, et al. (2004). "Genetic polymorphisms in the IGFBP3 gene: association with breast cancer risk and blood IGFBP-3 protein levels among Chinese women." Cancer Epidemiol Biomarkers Prev 13(8): 1290- 5. Ries, L. A. G., M. P. Eisner, et al. (2003). SE ER Cancer Statistics Review, 1975 - 2000. Bethesda, MD, National Cancer Institute. Ries LAG, E. M., Kosary CL, Hankey BF, Miller BA, Clegg L, Mariotto A, Feuer EJ, Edwards BK (2005). Cancer Statistics Review, 1975-2002. National Cancer Institute. Bethesda, MD, http://seer.cancer.gov/csr/1975_2002/, based on November 2004 SEER data submission, posted to the SEER web site 2005. Rietveld, I., J. A. Janssen, et al. (2003). "A polymorphism in the IGF-I gene influences the age-related decline in circulating total IGF-I levels." Eur J Endocrinol 148(2): 171-5. Rietveld, I., J. A. Janssen, et al. (2004). "A polymorphic CA repeat in the IGF-I gene is associated with gender-specific differences in body height, but has no effect on the secular trend in body height." Clin Endocrinol (Oxf) 61(2): 195-203. Rioux, J. D., M. J. Daly, et al. (2001). "Genetic variation in the 5q31 cytokine gene cluster confers susceptibility to Crohn disease." Nat Genet 29(2): 223-8. 186 Robbins, A. S., A. S. Whittemore, et al. (2000). "Differences in socioeconomic status and survival among white and black men with prostate cancer." Am J Epidemiol 151(4): 409-16. Rodriguez, C., A. V. Patel, et al. (2001). "Body mass index, height, and prostate cancer mortality in two large cohorts of adult men in the United States." Cancer Epidemiol Biomarkers Prev 10(4): 345-53. Rosenfeld, R. G., H. Pham, et al. (1994). "Insulin-like growth factor binding proteins and their regulation." Acta Paediatr Suppl 399: 154-8. Rosenfield, R. I., R. Furlanetto, et al. (1983). "Relationship of somatomedin-C concentrations to pubertal changes." J Pediatr 103(5): 723-8. Rotwein, P. (1991). "Structure, evolution, expression and regulation of insulin- like growth factors I and II." Growth Factors 5(1): 3-18. Rotwein, P. (1999). Molecular Biology of IGF-I and IGF-II. The IGF System. R. G. Rosenfeld. Totowa, NJ, Human Press Inc.: 40-41. Ruan, W., L. Powell-Braxton, et al. (1999). "Evidence that insulin-like growth factor I and growth hormone are required for prostate gland development." Endocrinology. 140(5): 1984-9. Rubin, J., L. W. Chung, et al. (2004). "Prostate carcinoma cells that have resided in bone have an upregulated IGF-I axis." Prostate 58(1): 41-9. Rudman, D., M. H. Kutner, et al. (1985). "The short child with subnormal plasma somatomedin C." Pediatr Res 19(10): 975-80. Sachidanandam, R., D. Weissman, et al. (2001). "A map of human genome sequence variation containing 1.42 million single nucleotide polymorphisms." Nature 409(6822): 928-33. Sakano, K., T. Enjoh, et al. (1991). "The design, expression, and characterization of human insulin-like growth factor II (IGF-II) mutants specific for either the IGF-II/cation-independent mannose 6-phosphate receptor or IGF-I receptor." J Biol Chem 266(31): 20626-35. Salmon, W. D., Jr. and W. H. Daughaday (1957). "A hormonally controlled serum factor which stimulates sulfate incorporation by cartilage in vitro." J Lab Clin Med 49(6): 825-36. Sandhu, M. S., J. M. Gibson, et al. (2004). "Association between insulin-like growth factor-I: insulin-like growth factor-binding protein-1 ratio and metabolic and anthropometric factors in men and women." Cancer Epidemiol Biomarkers Prev 13(1): 166-70. Schaefer, C., Friedman, GD, Quesenberry, CP (1998). "IGF-I and prostate cancer." Science 282: 199a. Schaffer, B. S., M. F. Lin, et al. (2003). "Opposing roles for the insulin-like growth factor (IGF)-II and mannose 6-phosphate (Man-6-P) binding activities of the IGF-II/Man-6-P receptor in the growth of prostate cancer cells." Endocrinology 144(3): 955-66. 187 Schernhammer, E. S., S. E. Hankinson, et al. (2003). "Polymorphic variation at the -202 locus in IGFBP3: Influence on serum levels of insulin-like growth factors, interaction with plasma retinol and vitamin D and breast cancer risk." Int J Cancer 107(1): 60-4. Schernhammer, E. S., J. M. Holly, et al. (2005). "Circulating levels of insulin-like growth factors, their binding proteins, and breast cancer risk." Cancer Epidemiol Biomarkers Prev 14(3): 699-704. Schildkraut, J. M., W. Demark-Wahnefried, et al. (2005). "IGF1 (CA)19 repeat and IGFBP3 -202 A/C genotypes and the risk of prostate cancer in Black and White men." Cancer Epidemiol Biomarkers Prev 14(2): 403-8. Schmitz, K. H., R. L. Ahmed, et al. (2002). "Effects of a 9-month strength training intervention on insulin, insulin-like growth factor (IGF)-I, IGF- binding protein (IGFBP)-1, and IGFBP-3 in 30-50-year-old women." Cancer Epidemiol Biomarkers Prev 11(12): 1597-604. Schneider, M. R., E. Wolf, et al. (2002). "IGF-binding protein-5: flexible player in the IGF system and effector on its own." J Endocrinol 172(3): 423-40. Schuit, S. C., J. B. van Meurs, et al. (2004). "Height in pre- and postmenopausal women is influenced by estrogen receptor alpha gene polymorphisms." J Clin Endocrinol Metab 89(1): 303-9. Schultz, G. A., A. Hahnel, et al. (1993). "Expression of IGF ligand and receptor genes during preimplantation mammalian development." Mol Reprod Dev 35(4): 414-20. Shibata, A., A. S. Whittemore, et al. (1997). "Serum levels of prostate-specific antigen among Japanese-American and native Japanese men." J Natl Cancer Inst 89(22): 1716-20. Shimasaki, S. and N. Ling (1991). "Identification and molecular characterization of insulin-like growth factor binding proteins (IGFBP-1, -2, -3, -4, -5 and - 6)." Prog Growth Factor Res 3(4): 243-66. Shimizu, H., R. K. Ross, et al. (1991). "Possible underestimation of the incidence rate of prostate cancer in Japan." Jpn J Cancer Res 82(5): 483-5. Signorello, L. B., H. Kuper, et al. (2000). "Lifestyle factors and insulin-like growth factor 1 levels among elderly men." Eur J Cancer Prev 9(3): 173-8. Silventoinen, K., S. Sammalisto, et al. (2003). "Heritability of adult body height: a comparative study of twin cohorts in eight countries." Twin Res 6(5): 399- 408. Simard, J., M. Dumont, et al. (2002). "Perspective: prostate cancer susceptibility genes." Endocrinology 143(6): 2029-40. Smith, D. J. and A. J. Lusis (2002). "The allelic structure of common disease." Hum Mol Genet 11(20): 2455-61. Smith, W. J., L. E. Underwood, et al. (1995). "Effects of caloric or protein restriction on insulin-like growth factor-I (IGF-I) and IGF-binding proteins in children and adults." J Clin Endocrinol Metab 80(2): 443-9. Society, A. C. (2004). Cancer facts and figures. Society, A. C. (2005). Cancer Facts and Figures. 188 Stanford, J. L., R. A. Stephenson, et al. (1999). Prostate Cancer Trends 1973- 1995. Bethesda, MA, SEER Program, National Cancer Insitute. NIH Pub. No. 99-543. Stanhope, R., P. J. Pringle, et al. (1988). "Growth, growth hormone and sex steroid secretion in girls with central precocious puberty treated with a gonadotrophin releasing hormone (GnRH) analogue." Acta Paediatr Scand 77(4): 525-30. Stattin, P., A. Bylund, et al. (2000). "Plasma insulin-like growth factor-I, insulin- like growth factor-binding proteins, and prostate cancer risk: a prospective study." J Natl Cancer Inst 92(23): 1910-7. Stattin, P., A. Bylund, et al. (2000). "Plasma insulin-like growth factor-I, insulin- like growth factor-binding proteins, and prostate cancer risk: a prospective study.[comment]." Journal of the National Cancer Institute. 92(23): 1910- 7. Stattin, P., S. Rinaldi, et al. (2004). "High levels of circulating insulin-like growth factor-I increase prostate cancer risk: a prospective study in a population- based nonscreened cohort." J Clin Oncol 22(15): 3104-12. Stattin, P., U. H. Stenman, et al. (2001). "Ratios of IGF-I, IGF binding protein-3, and prostate-specific antigen in prostate cancer detection." Journal of Clinical Endocrinology & Metabolism. 86(12): 5745-8. Steinberg, G. D., B. S. Carter, et al. (1990). "Family history and the risk of prostate cancer." Prostate 17(4): 337-47. Stewart, C. E. and P. Rotwein (1996). "Growth, differentiation, and survival: multiple physiological functions for insulin-like growth factors." Physiol Rev 76(4): 1005-26. Stram, D. O., C. A. Haiman, et al. (2003). "Choosing haplotype-tagging SNPS based on unphased genotype data using a preliminary sample of unrelated subjects with an example from the Multiethnic Cohort Study." Hum Hered 55(1): 27-36. Teramukai, S., T. Rohan, et al. (2002). "Anthropometric and behavioral correlates of insulin-like growth factor I and insulin-like growth factor binding protein 3 in middle-aged Japanese men." Am J Epidemiol 156(4): 344-8. Thissen, J. P., J. M. Ketelslegers, et al. (1994). "Nutritional regulation of the insulin-like growth factors." Endocr Rev 15(1): 80-101. Tishkoff, S. A., E. Dietzsch, et al. (1996). "Global patterns of linkage disequilibrium at the CD4 locus and modern human origins." Science 271(5254): 1380-7. Torring, N., L. Vinter-Jensen, et al. (1997). "Systemic administration of insulin- like growth factor I (IGF-I) causes growth of the rat prostate." Journal of Urology. 158(1): 222-7. Tran, N., B. S. Bharaj, et al. (2004). "Short tandem repeat polymorphism and cancer risk: influence of laboratory analysis on epidemiologic findings." Cancer Epidemiol Biomarkers Prev 13(12): 2133-40. 189 Tsuchiya, N., L. Wang, et al. (2005). "CA repeat polymorphism in the insulin-like growth factor-I gene is associated with increased risk of prostate cancer and benign prostatic hyperplasia." Int J Oncol 26(1): 225-31. Tsuruta, J. K., E. M. Eddy, et al. (2000). "Insulin-like growth factor-II/cation- independent mannose 6-phosphate receptor mediates paracrine interactions during spermatogonial development." Biol Reprod 63(4): 1006-13. Underwood, L. E., J. P. Thissen, et al. (1994). "Hormonal and nutritional regulation of IGF-I and its binding proteins." Horm Res 42(4-5): 145-51. Vaessen, N., P. Heutink, et al. (2001). "A polymorphism in the gene for IGF-I: functional properties and risk for type 2 diabetes and myocardial infarction." Diabetes 50(3): 637-42. Valentinis, B., A. Bhala, et al. (1995). "The human insulin-like growth factor (IGF) binding protein-3 inhibits the growth of fibroblasts with a targeted disruption of the IGF-I receptor gene." Mol Endocrinol 9(3): 361-7. Voskuil, D. W., H. B. Bueno de Mesquita, et al. (2001). "Determinants of circulating insulin-like growth factor (IGF)-I and IGF binding proteins 1-3 in premenopausal women: physical activity and anthropometry (Netherlands)." Cancer Causes Control 12(10): 951-8. Vrieling, A., D. W. Voskuil, et al. (2004). "Dietary determinants of circulating insulin-like growth factor (IGF)-I and IGF binding proteins 1, -2 and -3 in women in the Netherlands." Cancer Causes Control 15(8): 787-96. Wabinga, H. R., D. M. Parkin, et al. (2000). "Trends in cancer incidence in Kyadondo County, Uganda, 1960-1997." Br J Cancer 82(9): 1585-92. Wacholder, S., S. Chanock, et al. (2004). "Assessing the probability that a positive report is false: an approach for molecular epidemiology studies." J Natl Cancer Inst 96(6): 434-42. Wacholder, S., N. Rothman, et al. (2000). "Population stratification in epidemiologic studies of common genetic variants and cancer: quantification of bias." J Natl Cancer Inst 92(14): 1151-8. Wall, J. D. (2001). "Insights from linked single nucleotide polymorphisms: what we can learn from linkage disequilibrium." Curr Opin Genet Dev 11(6): 647-51. Wang, L., T. Habuchi, et al. (2003). "Insulin-like growth factor-binding protein-3 gene -202 A/C polymorphism is correlated with advanced disease status in prostate cancer." Cancer Research. 63(15): 4407-11. Wang, L., W. Ma, et al. (1998). "Insulin-like growth factor I modulates induction of apoptotic signaling in H9C2 cardiac muscle cells." Endocrinology 139(3): 1354-60. Wang, Z. Q., M. R. Fung, et al. (1994). "Regulation of embryonic growth and lysosomal targeting by the imprinted Igf2/Mpr gene." Nature 372(6505): 464-7. 190 Weiner, J. A., A. Chen, et al. (2000). "Platelet-derived growth factor is a principal inductive factormodulating mannose 6-phosphate/insulin-like growth factor-II receptorgene expression via a distal E-box in activated hepatic stellate cells." Biochem J 345 Pt 2: 225-31. Weiss, K. M. and A. G. Clark (2002). "Linkage disequilibrium and the mapping of complex human traits." Trends Genet 18(1): 19-24. Weksberg, R., A. C. Smith, et al. (2003). "Beckwith-Wiedemann syndrome demonstrates a role for epigenetic control of normal development." Hum Mol Genet 12 Spec No 1: R61-8. West, C. A., T. R. Arnett, et al. (1996). "Expression of insulin-like growth factor I (IGF-I) mRNA variants in rat bone." Bone 19(1): 41-6. Wetterau, L. A., M. G. Moore, et al. (1999). "Novel aspects of the insulin-like growth factor binding proteins." Mol Genet Metab 68(2): 161-81. Whittemore, A. S., L. N. Kolonel, et al. (1995). "Prostate cancer in relation to diet, physical activity, and body size in blacks, whites, and Asians in the United States and Canada." J Natl Cancer Inst 87(9): 652-61. Wiltshire, S., T. M. Frayling, et al. (2002). "Evidence for linkage of stature to chromosome 3p26 in a large U.K. Family data set ascertained for type 2 diabetes." Am J Hum Genet 70(2): 543-6. Winter, D. L., A. L. Hanlon, et al. (2001). "Plasma levels of IGF-1, IGF-2, and IGFBP-3 in white and African-American men at increased risk of prostate cancer." Urology. 58(4): 614-8. Woelfle, J., J. Billiard, et al. (2003). "Acute control of insulin-like growth factor-I gene transcription by growth hormone through Stat5b." J Biol Chem 278(25): 22696-702. Wolk, K., S. C. Larsson, et al. (2004). "Metabolic, anthropometric, and nutritional factors as predictors of circulating insulin-like growth factor binding protein-1 levels in middle-aged and elderly men." J Clin Endocrinol Metab 89(4): 1879-84. Woods, K. A., C. Camacho-Hubner, et al. (1996). "Intrauterine growth retardation and postnatal growth failure associated with deletion of the insulin-like growth factor I gene." N Engl J Med 335(18): 1363-7. Woodson, K., J. A. Tangrea, et al. (2003). "Serum insulin-like growth factor I: tumor marker or etiologic factor? A prospective study of prostate cancer among Finnish men." Cancer Research. 63(14): 3991-4. Xiong, D. H., F. H. Xu, et al. (2005). "Vitamin D receptor gene polymorphisms are linked to and associated with adult height." J Med Genet 42(3): 228- 34. Xu, J., E. R. Bleecker, et al. (2002). "Major recessive gene(s) with considerable residual polygenic effect regulating adult height: confirmation of genomewide scan results for chromosomes 6, 9, and 12." Am J Hum Genet 71(3): 646-50. 191 Yakar, S., J. L. Liu, et al. (1999). "Normal growth and development in the absence of hepatic insulin-like growth factor I." Proc Natl Acad Sci U S A 96(13): 7324-9. Yamamoto, H., M. Sohmiya, et al. (1991). "Effects of aging and sex on plasma insulin-like growth factor I (IGF-I) levels in normal adults." Acta Endocrinol (Copenh) 124(5): 497-500. Yang, D. H., H. S. Kim, et al. (1998). "Identification of glycosylated 38-kDa connective tissue growth factor (IGFBP-related protein 2) and proteolytic fragments in human biological fluids, and up-regulation of IGFBP-rP2 expression by TGF-beta in Hs578T human breast cancer cells." J Clin Endocrinol Metab 83(7): 2593-6. Yang, Y., A. Hoeflich, et al. (1996). "Opposite regulation of IGF-I and IGF-I receptor mRNA and concomitant changes of GH receptor and IGF-II/M6P receptor mRNA in human IM-9 lymphoblasts." Biochim Biophys Acta 1310(3): 317-24. Yu, H., J. Mistry, et al. (1999). "Insulin-like growth factors (IGF-I, free IGF-I and IGF-II) and insulin-like growth factor binding proteins (IGFBP-2, IGFBP- 3, IGFBP-6, and ALS) in blood circulation." J Clin Lab Anal 13(4): 166- 72. Yu, H. and T. Rohan (2000). "Role of the insulin-like growth factor family in cancer development and progression." J Natl Cancer Inst 92(18): 1472-89. Zadeh, S. M. and M. Binoux (1997). "The 16-kDa proteolytic fragment of insulin- like growth factor (IGF) binding protein-3 inhibits the mitogenic action of fibroblast growth factor on mouse fibroblasts with a targeted disruption of the type 1 IGF receptor gene." Endocrinology 138(7): 3069-72. Zapf, J., E. Schoenle, et al. (1978). "Insulin-like growth factors I and II: some biological actions and receptor binding characteristics of two purified constituents of nonsuppressible insulin-like activity of human serum." Eur J Biochem 87(2): 285-96. Zaykin, D. V., P. H. Westfall, et al. (2002). "Testing association of statistically inferred haplotypes with discrete and continuous traits in samples of unrelated individuals." Hum Hered 53(2): 79-91. Zhang, Q., M. Tally, et al. (1997). "Insulin-like growth factor II signaling through the insulin-like growth factor II/mannose-6-phosphate receptor promotes exocytosis in insulin-secreting cells." Proc Natl Acad Sci U S A 94(12): 6232-7. Zhao, H., J. Dupont, et al. (2004). "PTEN inhibits cell proliferation and induces apoptosis by downregulating cell surface IGF-IR expression in prostate cancer cells." Oncogene 23(3): 786-94. Zhou, R., D. Diehl, et al. (2003). "IGF-binding protein-4: biochemical characteristics and functional consequences." J Endocrinol 178(2): 177-93.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Genetic risk factors in breast cancer susceptibility: The multiethnic cohort

PDF

Breast cancer in the multiethnic cohort study: Genetic (prolactin pathway genes) and environmental (hormone therapy) factors

PDF

Insulin-like growth factor 1 genotype, phenotype and breast cancer risk, by racial/ethnic group

PDF

Single-nucleotide polymorphisms in 17beta-hydroxysteroid dehydrogenase type III and prostate cancer risk

PDF

Colorectal cancer: genomic variations in insulin-like growth factor-1 and -2

PDF

Risk factors, sequellae and intervention studies for the treatment of diabetes mellitus in Native American populations

PDF

Colorectal cancer risks in Singapore Chinese: Polymorphisms in the insulin-like growth factor-1 and the vitamin D receptor

PDF

Monoamine oxidase A: (1) As one of many candidate genes for preeclampsia and (2) as a candidate gene for tobacco addiction

PDF

Determinants of the age at natural menopause: The multiethnic cohort

PDF

The effects of dietary and other lifestyle behaviors on the risk of colorectal cancer among the Mexican -origin Latino population

PDF

A large-scale genetic association study of prostate cancer in a multi-ethnic population

PDF

Androgen receptor gene and prostate -specific antigen gene in breast cancer

PDF

Physical activity and adenomatous polyps: Measures of association and impact

PDF

Molecular and genetic epidemiology of Chlamydia trachomatis in the United States

PDF

Androgens and breast cancer

PDF

Recreational physical activity and risk of breast cancer

PDF

A case-control study of passive smoking and bladder cancer risk in Los Angeles

PDF

Progression of subclinical atherosclerosis in type 2 diabetes: Therapeutic and metabolic influences

PDF

Recreational physical activity and risk of breast cancer: The California Teachers Study

PDF

Effect of genetic factors in the development of childhood lymphocytic leukemia (ALL)

Asset Metadata

Creator Cheng, Iona (author)

Core Title The role of genetic variation in insulin -like growth factors in prostate cancer risk: The multiethnic cohort

Degree Doctor of Philosophy

Degree Program Epidemiology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag biology, genetics,OAI-PMH Harvest

Language English

Contributor Digitized by ProQuest (provenance)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c16-468760

Unique identifier UC11336585

Identifier 3219809.pdf (filename),usctheses-c16-468760 (legacy record id)

Legacy Identifier 3219809.pdf

Dmrecord 468760

Document Type Dissertation

Rights Cheng, Iona

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA