Many studies have established that loss of heterozygosity and/or altered expression of the fragile histidine triad (FHIT) gene is a common event in a number of tumor types including prostate carcinoma. Encompassing the most active fragile site in the human genome, FRA3B, FHIT has become the model fragile site–associated tumor suppressor gene. In a recent study, linkage and association between germline genetic variation in FHIT (specifically single nucleotide polymorphism rs760317) and prostate cancer were reported. We sought to confirm this finding in two independent samples: (a) a family-based sample of 817 men with (n = 434) and without (n = 383) prostate cancer from 323 Caucasian families, and (b) a community-based case-control sample of African American men with (n = 133) and without (n = 342) prostate cancer. Using a family-based association test, rs760317 was associated with prostate cancer in Caucasians (P = 0.031), with a reduction in the risk of prostate cancer among carriers of the minor allele (odds ratio, 0.66; 95% confidence interval, 0.42-1.04; P = 0.074). African American carriers experienced a similar risk reduction (odds ratio, 0.63; 95% confidence interval, 0.42-0.96; P = 0.032). These results are remarkably consistent across ethnic samples but are in opposition to results from the original study, which showed an association between the minor allele of rs760317 and an increased risk of prostate cancer. Taken together, the consistently significant but flipped association between single nucleotide polymorphism rs760317 and prostate cancer in three independent samples suggests that rs760317 may be in linkage disequilibrium with one or more prostate cancer susceptibility variants in or near FHIT. (Cancer Epidemiol Biomarkers Prev 2007;16(6):1294–7)

Since its discovery in 1996, the fragile histidine triad (FHIT) gene has been established as the model fragile site–associated tumor suppressor gene. This large gene (∼1.5 Mb) resides at chromosome 3p14.2 and encompasses the common fragile site FRA3B, overlapping exons 4 and 5. While there is evidence of loss of heterozygosity and/or protein in many tumor types, the function of this gene and the mechanism by which its loss leads to tumor initiation and/or progression are still unclear.

Studies of FHIT in prostate cancer have been sparse relative to cancers of the gastrointestinal tract, colon, cervix, lung, and breast (reviewed in ref. 1). However, among the few published reports, there is some consensus that FHIT protein expression is down-regulated in primary prostate carcinomas (2, 3) and that this decrease is not the result of loss of heterozygosity within the gene (3). In a recent study, Larson et al. (4) reported suggestive evidence for linkage between prostate cancer and a microsatellite marker within FHIT. Following up their linkage signal with a denser set of single-nucleotide polymorphisms (SNPs), these authors found a significant association between prostate cancer and SNP rs760317 (in intron 5 of FHIT) and a two-SNP haplotype (containing rs760317 and rs6791450). The present report examines these two FHIT SNPs in independent samples of Caucasians and African Americans.

Study Subjects

The first sample consisted of Caucasian families with at least one sibling pair discordant for prostate cancer. Men from these families were recruited as part of the Prostate Cancer Genetics Program at the University of Michigan (5). Prostate Cancer Genetics Program families were primarily recruited from the University of Michigan Comprehensive Cancer Center. Other sources included direct patient or physician referrals. Prostate Cancer Genetics Program enrollment was restricted to (a) families with two or more living members with prostate cancer in a first- or second-degree relationship or (b) men diagnosed with prostate cancer at ≤55 years of age without a family history of the disease. All participants were asked to provide a blood sample for DNA extraction, extended family history information, and access to medical records. For this sample, the oldest available unaffected brother from each family was preferentially enrolled to maximize the probability that unaffected men were truly unaffected and not simply unaffected by virtue of being younger than their affected brother(s). Additional male siblings and multiple sibships from the same family were included if DNA was available. For this analysis, 323 Caucasian families were genotyped.

The second sample consisted of African American men with and without prostate cancer, who were recruited as part of the Flint Men's Health Study (6). Starting in 1996, 943 potentially eligible men were selected from a probability sample of African American men ages 40 to 79 years in Flint, Michigan, and neighboring Beecher Township (Genesee County, Michigan). Unaffected men were excluded if they were previously diagnosed with prostate cancer and/or had a previous operation involving the prostate gland. A total of 379 eligible unaffected men completed urologic and physical examinations in conjunction with prostate-specific antigen screening, a blood draw, and questionnaire, and 342 unaffected men had DNA available for this study. African American men with prostate cancer diagnosed between 1995 and 2002 were identified through the Genesee County Community-Wide Hospital Oncology program registry, which covers the three local hospitals servicing the community. Between 1999 and 2002, 138 men with prostate cancer agreed to participate in the study, and 133 had DNA available for this analysis.

Below, we refer to the Prostate Cancer Genetics Program sample as the “Caucasian sample” and the Flint Men's Health Study sample as the “African American sample.” The Institutional Review Board at the University of Michigan Medical School approved all aspects of both study protocols, and all participants gave written informed consent.

Genotyping Assays

Two SNPs in intron 5 of FHIT (rs760317 and hCV8351378/rs6791450) were genotyped using TaqMan SNP assays (Applied Biosystems). Genotyping call rates for rs760317 and rs6791450 were 98.9% and 97.9%, respectively, and the undetermined samples were sequenced to achieve a final call rate of 100% for both SNPs. A subset of genotypes was duplicated by TaqMan (5.5%) or direct sequencing (3.0%) for each SNP, and no discrepancies were observed.

To test for potential population substructure in the African American sample, 42 unlinked microsatellite markers were genotyped by deCODE Genetics in a separate collaborative project (7). These markers are located on the Marshfield genetic map and were selected to distinguish between European, African, and Asian ancestry.

Statistical Methods

Within each sample, observed genotype distributions were tested for departure from Hardy-Weinberg equilibrium in a subset of unrelated, unaffected men. For the Caucasian sample, this subset consisted of the oldest unaffected man from each family. SNP genotypes did not depart from Hardy-Weinberg equilibrium in either sample at a significance level of 0.05. Haplotype frequencies were estimated using the expectation-maximization algorithm and were used to calculate the linkage disequilibrium measure r2.

For the Caucasian sample, we used the family-based association method (ref. 8; implemented in the FBAT software, version 1.5.5) to test for association between single SNPs and prostate cancer. To maximize power, we analyzed the combined set of affected and unaffected men using the offset option to test the null hypothesis of no association and no linkage. To account for the possible misclassification of unaffected men, we analyzed only affected men using the empirical variance estimate to test the null hypothesis of no association in the presence of linkage. Conditional logistic regression, coupled with a robust variance estimate that incorporates familial correlations (9), was used to generate odds ratios (OR) and 95% confidence intervals (95% CIs). Two-SNP haplotypes were analyzed using the haplotype FBAT (HBAT) method (10).

For the African American sample, we used logistic regression to test for association between each SNP and prostate cancer and to estimate ORs and 95% CIs. Tests of association between two-SNP haplotypes and prostate cancer were conducted using the haplotype generalized linear model method proposed by Lake et al. (11). Individual haplotypes were evaluated using a model-specific Wald test. In all African American analyses, age and family history of prostate cancer in a first-degree relative were included as potential confounders.

To test for population substructure in the African American sample, we implemented the method of Pritchard and Rosenberg (12) using 42 unlinked microsatellite markers. The observed summary χ2 measure was 133.13 with 142 degrees of freedom (P = 0.96), suggesting that hidden population substructure is unlikely to generate false-positive evidence for association.

For both samples, we calculated single SNP and haplotype association tests under additive, dominant, and recessive models. For single SNPs, an additional genotype model (2 degree of freedom test) was used. All statistical tests were two sided, with the significance level set at 0.05. Conditional logistic regression was conducted using version 8.2 of the SAS programming language. All remaining analyses were carried out using the R-language.4

The Caucasian sample included 434 men with prostate cancer and 383 unaffected men from 323 families with at least one pair of brothers discordant for prostate cancer. Of these families, 221 included only a single discordant sibling pair (DSP). The remaining families included additional DSPs from the same sibship (e.g., two brothers with and one without prostate cancer or two DSPs) or from the same family but different sibships (e.g., a pair of DSPs related as first cousins), resulting in a total sample of 516 DSPs. The median age at diagnosis for Caucasian men with prostate cancer was 55 years (interquartile range, 50-63 years), and the median age at consent for unaffected men was 56 years (interquartile range, 50-63 years).

The minor allele frequency of rs760317 was 5% greater in unaffected men compared with affected men (P = 0.047; Table 1). Consistent with this difference, we also detected significant overtransmission of the minor allele of rs760317 to unaffected men compared with affected men in our family-based analysis. In the combined sample of affected and unaffected men, both additive and dominant models for rs760317 showed significant evidence for prostate cancer association (Table 2). Before estimating ORs, we excluded 18 men who were not brothers of the index case from seven multisibship families, resulting in a reduced sample size of 799 men and 506 DSPs. Conditional logistic regression results are presented in Table 3. The OR associated with each minor allele at rs760317 was 0.77 (95% CI, 0.57-1.03; P = 0.073).

Table 1.

Minor allele frequencies in affected and unaffected men

Sample (no. affected/no. unaffected)dbSNP IDMinor allele frequency
AffectedUnaffectedP*
Caucasian (434/383) rs760317 0.45 0.50 0.047 
 rs6791450 0.32 0.33 0.524 
African American (133/342) rs760317 0.23 0.29 0.105 
 rs6791450 0.47 0.47 0.995 
Sample (no. affected/no. unaffected)dbSNP IDMinor allele frequency
AffectedUnaffectedP*
Caucasian (434/383) rs760317 0.45 0.50 0.047 
 rs6791450 0.32 0.33 0.524 
African American (133/342) rs760317 0.23 0.29 0.105 
 rs6791450 0.47 0.47 0.995 
*

P value from the Z test of proportions assuming independence of all individuals.

rs760317 (G > A) is located at base pair 60,074,196 on chromosome 3.

rs6791450 (T > C) is located at base pair 60,057,979 on chromosome 3 and is recorded as hCV8351378 by Larson et al.

Table 2.

Family-based association test results from the Caucasian sample

dbSNP IDModel*Affecteds and unaffecteds
Affecteds only
nZ scorePnZ scoreP
rs760317 Additive 162 −2.22 0.026 152 −2.31 0.021 
 Dominant 96 −2.15 0.031 92 −2.04 0.041 
rs6791450 Additive 152 −0.85 0.396 141 −0.91 0.363 
 Dominant 121 −1.11 0.266 123 −1.09 0.276 
dbSNP IDModel*Affecteds and unaffecteds
Affecteds only
nZ scorePnZ scoreP
rs760317 Additive 162 −2.22 0.026 152 −2.31 0.021 
 Dominant 96 −2.15 0.031 92 −2.04 0.041 
rs6791450 Additive 152 −0.85 0.396 141 −0.91 0.363 
 Dominant 121 −1.11 0.266 123 −1.09 0.276 
*

Both models are with respect to the minor allele, which is “A” for rs760317 and “C” for rs6791450.

Number of informative families.

Table 3.

Estimated ORs from logistic regression

dbSNP IDModel*Sample
Caucasian
African American
OR (95% CI)POR (95% CI)P
rs760317 Additive 0.77 (0.57-1.03) 0.073 0.71 (0.51-1.00) 0.050 
 Dominant 0.66 (0.42-1.04) 0.074 0.63 (0.42-0.96) 0.032 
rs6791450 Additive 0.91 (0.69-1.19) 0.483 1.00 (0.75-1.33) 0.997 
 Dominant 0.81 (0.56-1.17) 0.267 0.96 (0.61-1.51) 0.862 
dbSNP IDModel*Sample
Caucasian
African American
OR (95% CI)POR (95% CI)P
rs760317 Additive 0.77 (0.57-1.03) 0.073 0.71 (0.51-1.00) 0.050 
 Dominant 0.66 (0.42-1.04) 0.074 0.63 (0.42-0.96) 0.032 
rs6791450 Additive 0.91 (0.69-1.19) 0.483 1.00 (0.75-1.33) 0.997 
 Dominant 0.81 (0.56-1.17) 0.267 0.96 (0.61-1.51) 0.862 
*

Both models are with respect to the minor allele, which is “A” for rs760317 and “C” for rs6791450.

All logistic regression models for the African American sample were adjusted for age and family history of prostate cancer in a first-degree relative.

The African American sample included 133 affected and 342 unaffected men. The median age at diagnosis for African American men with prostate cancer was 63 years (interquartile range, 56-69 years) and the median age at consent for unaffected men was 55 years (interquartile range, 49-63 years). Similar to the Caucasian sample, the rs760317 minor allele frequency was 6% greater in unaffected men compared with affected men (Table 1). Using logistic regression (Table 3), the OR associated with each minor allele at rs760317 was 0.71 after adjustment for age and family history of prostate cancer (95% CI, 0.51-1.00; P = 0.050). Under a dominant model, the effect of the minor allele was also significant (P = 0.032).

SNP rs6791450 was not associated with prostate cancer in either sample (Tables 2 and 3). Notably, rs6791450 is located ∼16 kb from rs760317 and was not in strong linkage disequilibrium with rs760317 in either the Caucasian (r2 = 0.18) or African American (r2 < 0.01) sample. In the Caucasian sample, the haplotype defined by the major alleles of both SNPs was overtransmitted to affected men under additive (P = 0.041) and recessive models (P = 0.045), consistent with the single SNP result for rs760317. In the African American sample, there was a reduction in risk associated with the haplotype defined by the minor allele of rs760317 and the major allele of rs6791450 under additive (P = 0.003) and dominant (P = 0.005) models.

In summary, our results showed association between genetic variation in FHIT (specifically rs760317) and prostate cancer in two independent samples. The association between rs760317 and prostate cancer was remarkably similar in direction and magnitude in Caucasian and African American samples. Although our data indicated a protective effect associated with the minor allele of rs760317, Larson et al. (4) found the opposite effect. In their study, men homozygous for the minor allele showed an ∼2-fold increased risk of prostate cancer in comparison with carriers of at least one copy of the major allele.5

5

Personal communication.

We were able exclude the possibility that genotyping error was the source of this allelic reversal through a mutual exchange of 12 anonymous DNA samples with Larson et al. (i.e., there were no discrepancies; data not shown).

This pattern of allelic reversal has been noted in replication studies of other candidate SNPs (13, 14), and several such discrepancies have been shown to differ beyond what would occur by chance alone (14). Further, in a recent study investigating the potential causes of this “flip-flop” phenomenon, Lin et al. (15) suggested that a genotyped SNP interacting with a nongenotyped causal SNP may show a flipped association when the minor allele frequency of the genotyped SNP is high (∼0.5), the pair is in relatively low linkage disequilibrium (r2 < 0.3), and the interaction of the two is not accounted for in the model. Given the relatively high minor allele frequency of rs760317, this explanation of the observed results is plausible. Of note, rs760317 was not genotyped in the International HapMap project6

or the recent prostate cancer genome-wide association study conducted by the Cancer Genetic Markers of Susceptibility initiative.7

While a functional relationship between FHIT and tumorigenesis and/or progression is still unknown, data from the mouse suggest that FHIT haploinsufficiency predisposes to a wide range of tumors (16). In addition, alternatively spliced FHIT transcripts have been shown to occur in non-neoplastic tissue (17), some of which lead to loss of a functional protein product. Although rs760317 does not directly alter a known splice site (18), it could be in linkage disequilibrium with another SNP that influences alternative splicing of the gene, potentially reducing the amount of the functional protein product. Further, rs760317 resides in a region of intron 5 that is commonly deleted in tumor cell lines (19), suggesting an important role for sequence variation in this region. Additional resequencing and functional work will be required to evaluate the direct or indirect influence of rs760317 on the integrity of normal FHIT expression. In view of the data presented here, this additional work seems justified.

Grant support: NIH Specialized Program of Research Excellence in Prostate Cancer grant P50 CA69568, NIH grant R01 CA79596 (K.A. Cooney), the Department of Urology at the University of Michigan Medical School, and the University of Michigan Comprehensive Cancer Center.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Joe Washburn and the Microarray Core Facility at the University of Michigan Comprehensive Cancer Center for assistance with genotyping assays, and Drs. Theodore G. Krontiris and Yan Ding (City of Hope National Medical Center) for openly discussing and sharing data from their ongoing investigation of FHIT.

1
Huebner K, Croce CM. Cancer and the FRA3B/FHIT fragile locus: it's a HIT.
Br J Cancer
2003
;
88
:
1501
–6.
2
Fouts RL, Sandusky GE, Zhang S, et al. Down-regulation of fragile histidine triad expression in prostate carcinoma.
Cancer
2003
;
97
:
1447
–52.
3
Guo Z, Johansson SL, Rhim JS, Vishwanatha JK. Fragile histidine triad gene expression in primary prostate cancer and in an in vitro model.
Prostate
2000
;
43
:
101
–10.
4
Larson GP, Ding Y, Cheng LSC, et al. Genetic linkage of prostate cancer risk to the chromosome 3 region bearing FHIT.
Cancer Res
2005
;
65
:
805
–14.
5
Douglas JA, Zuhlke KA, Beebe-Dimmer J, et al. Identifying susceptibility genes for prostate cancer—a family-based association study of polymorphisms in CYP17, CYP19, CYP11A1, and LH-β.
Cancer Epidemiol Biomarkers Prev
2005
;
14
:
2035
–9.
6
Cooney KA, Strawderman MS, Wojno KJ, et al. Age-specific distribution of serum prostate-specific antigen in a community-based study of African-American men.
Urology
2001
;
57
:
91
–6.
7
Amundadottir LT, Sulem P, Gudmundsson J, et al. A common variant associated with prostate cancer in European and African populations.
Nat Genet
2006
;
38
:
652
–8.
8
Rabinowitz D, Laird N. A unified approach to adjusting association tests for population admixture with arbitrary pedigree structure and arbitrary missing marker information.
Hum Hered
2000
;
50
:
211
–23.
9
Siegmund KD, Langholz B, Kraft, P, Thomas DC. Testing linkage disequilibrium in sibships.
Am J Hum Genet
2000
;
67
:
244
–8.
10
Horvath S, Xu X, Lake SL, Silverman EK, Weiss ST, Laird NM. Family-based tests for associating haplotypes with general phenotype data: application to asthma genetics.
Genet Epidemiol
2004
;
26
:
61
–9.
11
Lake SL, Lyon H, Tantisira K, et al. Estimation and tests of haplotype-environment interaction when linkage phase is ambiguous.
Hum Hered
2003
;
55
:
56
–65.
12
Pritchard JK, Rosenberg NA. Use of unlinked genetic markers to detect population stratification in association studies.
Am J Hum Genet
1999
;
65
:
220
–8.
13
Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN. Meta-analysis of genetic association studies supports a contribution of common variants to susceptibility to common disease.
Nat Genet
2003
;
33
:
177
–82.
14
Ioannidis JP, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG. Replication validity of genetic association studies.
Nat Genet
2001
;
29
:
306
–9.
15
Lin PI, Vance JM, Pericak-Vance MA, Martin ER. No gene is an island: the flip-flop phenomenon.
Am J Hum Genet
2007
;
80
:
531
–8.
16
Zanesi N, Fidanza V, Fong LY, et al. The tumor spectrum in FHIT-deficient mice.
Proc Natl Acad Sci U S A
2001
;
98
:
10250
–5.
17
Panagopoulos I, Thelin S, Mertens F, Mitelman F, Aman P. Variable FHIT transcripts in non-neoplastic tissues.
Genes Chromosomes Cancer
1997
;
19
:
215
–9.
18
Ohta M, Inoue H, Cotticelli MG, et al. The FHIT gene, spanning the chromosome 3p14.2 fragile site and renal carcinoma-associated t(3;8) breakpoint, is abnormal in digestive tract cancers.
Cell
1996
;
84
:
587
–97.
19
Iliopoulos D, Guler G, Han SY, et al. Roles of FHIT and WWOX fragile genes in cancer.
Cancer Lett
2006
;
232
:
27
–36.