Effect modification of endocrine disruptors and testicular germ cell tumor risk by hormone-metabolizing genes

Abstract

It has been hypothesized that the increased prevalence of testicular germ cell tumors (TGCT) may be attributable to endocrine disrupting chemicals, such as persistent organic pollutants (POPs); these may be modulated by hormone-metabolizing enzymes. Using data from 568 cases and 698 controls enrolled in the U.S. Servicemen’s Testicular Tumor Environmental and Endocrine Determinants Study, we examined associations between TGCT and POPs, including p,p′-DDE, chlordane-related compounds, and polychlorinated biphenyls (PCBs), modified by polymorphisms in 5 hormone-metabolizing genes (CYP17A1, CYP1A1, HSD17B1, HSD17B4, and AR). Odds ratios (OR) and 95% confidence intervals (CI) were estimated using logistic regression models that stratified associations of POP exposure and TGCT risk by genotype. Two polymorphisms in CYP1A1, rs1456432 and rs7495708, modified the association between trans-nonachlor and total chlordanes and TGCT risk. Among men with a minor allele for rs1456432, those with the highest quartiles had an increased risk of TGCT (OR=1.90, 95% CI, 1.01–3.56) compare to those with the lowest; there were no increased risk among men with the homozygous major allele genotype (p-interaction=0.024). Similar results were seen for rs7495708. HSD17B4 rs384346 modified the associations between TGCT risk and PCB-118 and PCB-138 concentrations: the 45–55% reductions in TGCT risk for men with the highest quartiles compared to the lowest quartiles were only present in those who had a major homozygous allele genotype (p-interactions<0.04). Thus, there are suggestions that certain CYP1A1 and HSD17B4 polymorphisms may modify the associations between POPs and TGCT risk. With false discovery rate values >0.2, however, caution is advisable when interpreting the findings of this study.

Introduction

Testicular germ cell tumors (TGCT) are the most common tumors among U.S. men aged 15–35 years. With the exception of cryptorchism, family history of TGCT, and prior diagnosis of TGCT, risk factors for TGCT are not well-identified (Garner, et al., 2005; McGlynn, 2001), however, the increasing incidence of TGCT in Western countries (Purdue, et al., 2005) suggests that, in addition to possible genetic factors, environmental exposures may play a role. It has been hypothesized that the increased prevalence of a number of male reproductive tract disorders (cryptorchism, hypospadias, impaired spermatogenesis, and TGCT), known as testicular dysgenesis syndrome, may be attributable to endocrine disrupting chemicals (Skakkebaek, 2002), such as persistent organic pollutants (POPs), including pesticides and polychlorinated biphenyls (PCB) (Skakkebaek, 2002). Although there is some epidemiologic evidence that suggests p,p′-dichlorodiphenyldichloroethylene (p,p′-DDE), chlordane-related compounds, and several PCBs may be associated with risk of testicular germ cell tumors (TGCT) (Hardell, et al., 2003; McGlynn, et al., 2008; McGlynn, et al., 2009), the data are limited and inconsistent (Biggs, et al., 2008; Hardell, et al., 2004).

Endocrine disruptors can bind to estrogen or androgen receptors to act as weak estrogen agonists or androgen antagonists (Toppari, et al., 1996). Thus, the effects of endocrine disruptors may be modulated by hormone-metabolizing enzymes. The CYP1A1, CYP17A1, HSD17B1, HSD17B4, and androgen receptor (AR) genes are involved in binding and metabolizing sex hormones, and have been heavily studied with regards to risk of several hormonally-related cancers (Bugano, et al., 2008; Mononen & Schleutker, 2009). Genetic variation in these genes and risk of TGCT has been previously studied by our group (Figueroa, et al., 2008); however, there was only a suggestion of an association between a CYP1A1 polymorphism and nonseminomatous testicular germ cell tumors. It is plausible though that POPs may differentially affect TGCT risk based on genetic predisposition. Thus, as a follow-up to our prior studies that observed associations between TGCT risk and several POPs, including p,p′-DDE, chlordane-related compounds, and PCB congeners (McGlynn, et al., 2008; McGlynn, et al., 2009), we examined the risk of TGCT and the effect of interactions between hormone-metabolizing genes and POPs that were statistically significantly associated with TGCT risk. In a secondary analysis, we also examined interactions for the other POPs that had not been previously associated with TGCT risk.

Materials and Methods

Study population

Study participants, enrolled from April 2002 to January 2005 in the U.S. Servicemen’s Testicular Tumor Environmental and Endocrine Determinants (STEED) Study, a matched case-control study, have been previously described in detail (McGlynn, et al., 2007). Briefly, men who developed TGCT while on active duty were eligible to be enrolled as cases if they had at least one serum sample stored in the Department of Defense Serum Repository (Silver Spring, MD), were ≤ 45 years of age at diagnosis, and were diagnosed at a later date than the date of serum donation. TGCT diagnoses, confirmed by pathology reports or by a pathologist if the pathology report was missing (5%), were limited to classic seminomas or nonseminomas (embryonal carcinoma, yolk sac carcinoma, choriocarcinoma, teratomas, and mixed germ cell tumors). Men who developed spermatocytic seminomas were not eligible for inclusion. Eligible controls, men who did not develop TGCT, were pair-matched to cases based on age at diagnosis (<1 year), race/ethnicity (white, black, other), and date serum sample was donated (<30 days).

Of the possible case participants (n=853), 22 were in the process of being contacted when the study closed, leaving 831 men who were contacted. Of those, 77 refused to participate and 754 cases (91%) agreed to participate in and completed the study. Of the possible control participants (n=1,182), 32 were in the process of being contacted when the studyclosed, leaving 1,150 men who were contacted. Of those, 222 refused to participate and 928 (81%) controls agreed to participate in and completed the study. Of the enrolled cases and controls, 720 were matched case-control pairs.

Each participant completed a study questionnaire through a computer-assisted telephone interview. Questionnaires elicited information in regard to a reference date on demographic factors, such as height and weight, and on known or suspected risk factors for TGCT, such as cryptorchism and family history of testicular cancer. For the cases, the reference date was the date of TGCT diagnosis. For the controls, the reference date was the date of TGCT diagnosis of the matched case. The median time from serum blood draw to diagnosis date was 4.4 years, ranging from 15 days to 13.3 years prior to diagnosis; however, only 16 participants had a blood draw less than 1 year prior to diagnosis date. All participants were also asked to donate a buccal cell sample collected in mouthwash. Buccal cell samples were provided by 590 cases and 712 controls. The study was approved by the Institutional Review Boards of the National Cancer Institute and the Walter Reed Army Institute for Research. All participants provided written informed consent prior to enrolling in the study.

Laboratory Methods

Serum measurements of persistent organic pollutants

Laboratory assays for POP compounds that were measured at the Human Toxicology Laboratory of the Institut National de Santé Publique du Québec have been previously described in detail (McGlynn, et al., 2008; McGlynn, et al., 2009). Briefly, serum samples were enriched with isotopically labeled internal standard and were denatured using formic acid, and all analytes were then extracted from the matrix by solid-phase extraction. Extracts were cleaned on a florisil column and analyzed by gas chromatography-mass spectrometry. The mean limits of detections were ~0.005 μg/L. Average within-day variability ranged from 2–5%, and the average recovery was 80%. In order to adjust POP concentrations for lipids (Phillips, et al., 1989), serum triglycerides, free and total cholesterol, and phospholipids were measured with enzyme bioassays using reagents produced by Randox Laboratories (Antrim, UK). Each POP level was then dividedby the total lipid level. POPs measured in our previous studies (McGlynn, et al., 2008; McGlynn, et al., 2009), included pesticides, such as cis-nonachlor, trans-nonachlor, oxychlordane, p,p′-DDE, p,p′-dichlorodiphenyltrichloroethylene (p,p′-DDT), β-hexachlorocyclohexane, and mirex, and the PCB congers, PCB-99, PCB-101, PCB-118, PCB-138, PCB-153, PCB-156, PCB-163, PCB-170, PCB-180, PCB-183, and PCB-187.

Genotyping

The selection of SNPs in 5 candidate genes (CYP1A1, CYP17A1, HSD17B1, HSD17B4, and AR) in the hormone-metabolizing pathway has been previously described in detail (Figueroa, et al., 2008). In brief, DNA was extracted from buccal cell specimens using the Gentra AutoPure system (Gentra Systems, Inc., Minneapolis, MN), and samples were genotyped at the National Cancer Institute Core Genotyping Facility (Gaithersburg, MD). Assay methods and descriptions can be found at http://snp500cancer.nci.nih.gov (Packer, et al., 2006). Thirty-three SNPs were genotyped, including 3 in CYP17A1, 13 in CYP1A1, 2 in HSD17B1, 12 in HSD17B4, and 3 in AR. There was >98% concordance for all SNPs in quality controls samples, except for CYP1A1 rs24272299 (97%), rs4646903 (97%), HSD17B1 rs2830 (97%), and HSD17B4 rs28943585 (94%). All genotypes in the autosomal chromosomes (CYP1A1, CYP17A1, HSD17B1, and HSD17B4) were in Hardy-Weinberg equilibrium among the control population except for HSD17B1 rs2830 (p<0.001). Haploview (Barrett, et al., 2005) was used to determine linkage disequilibrium between SNPs.

Statistical Analysis

Of the 590 cases and 712 controls that provided buccal samples, 577 cases and 707 controls were successfully genotyped. Of the 754 cases and 928 controls for whom we had serum samples, 568 cases and 698 controls (434 of whom were matched pairs) were successfully genotyped and had organochlorine pesticides and polychlorinated biphenyls measured. Data analyses were conducted for all TGCT together and for each histologic type (seminoma, nonseminoma), separately. Of the 568 cases, 251 were seminomas and 317 were nonseminomas.

For our primary hypothesis, the POPs of interest included those that were found to be associated with TGCT risk (McGlynn, et al., 2008; McGlynn, et al., 2009): cis-nonachlor, trans-nonachlor, p,p′-DDE, PCB-118, PCB-138, PCB-153, PCB-156, PCB-163, PCB-170, PCB-180, and PCB-187 (Table 1). A secondary, exploratory hypothesis examined interactions between hormone-metabolizing SNPs and POPs that were not associated with TGCT risk (oxychlordane, p,p′-DDT, β-hexachlorocyclohexane, mirex, PCB-99, PCB-101, and PCB-183). Results for our primary and secondary hypotheses will be presented separately. POP concentrations in ng/g lipid were not normally distributed and were log-transformed for analysis. Total chlordane exposure was calculated by adding standardized log-transformed concentrations [(concentration-mean concentration)/SD] of lipid-adjusted cis-nonachlor, trans-nonachlor, and oxychlordane. Quartiles of serum POP concentrations were calculated using only the controls.

Table 1.

List of persistent organochlorine pesticides and polymoprhisms in hormone receptor and metabolizing genes

Persistent organochlorine pesticides associated with testicular germ cell tumor risk (ref. 6, 7)	Persistent organochlorine pesticides not associated with testicular germ cell tumor risk (ref. 6, 7)
Cis-nonachlor	Oxychlordane
Trans-nonachlor	p,p′-DDT
p,p′-DDE	β-hexachlorocyclohexane
PCB-118	Mirex
PCB-138	PCB-99
PCB-153	PCB-101
PCB-156	PCB-183
PCB-163
PCB-170
PCB-180
PCB-187

Gene	Gene name	Chromosomal location	Nucleotide change	dbSNP ID	Minor allele frequency*
CYP17A1	cytochrome P450, family 17, subfamily A, polypeptide 1	10q24.3	Ex1+27T>C	rs743572	0.37
			Ex1−160C>T	rs6162	0.39
			Ex1 −103G>T	rs6163	0.37
CYP1A1	cytochrome P450, family 1, subfamily A, polypeptide 1	15q22-q24	−17961T>C	rs2472299	0.27
			−16488T>C	rs11632547	0.02
			−10549G>A	rs4886605	0.17
			−10375G>A	rs12441817	0.09
			−9893G>A	rs17861115	0.04
			−4404G>A	rs7495708	0.18
			−4010A>G	rs2470893	0.33
			IVS1+606G>T	rs2606345	0.34
			IVS1 −728C>T	rs4646421	0.11
			Ex7+131A>G	rs1048943	0.04
			1188bp 3′ of STP C>T	rs4646903	0.11
			9778bp 3′ of STP A>G	rs1456432	0.17
			11599bp 3′ of STP C>G	rs2198843	0.17
HSD17B1	hydroxysteroid (17-beta) dehydrogenase 1	17q11–q21	Ex1 −486G>A	rs2830	0.46
			IVS4 −150C>A	rs676387	0.26
HSD17B4	hydroxysteroid (17-beta) dehydrogenase 4	5q21	−27855G>T	rs2451818	0.46
			−18796A>T	rs384346	0.15
			−2124A>T	rs28943585	0.13
			−601A>C	rs28943586	0.13
			IVS1 −178G>A	rs32651	0.47
			Ex6+15G>A	rs25640	0.46
			IVS8+4959C>G	rs2455463	0.45

^*Minor allele frequency in control population for whites only

Among controls, multiple linear regression models were used to determine if the hormone-metabolizing SNPs were associated with continuous POP concentrations, adjusting for the matching variables, age at reference date, race/ethnicity, and date of serum sample collection. Genotypes were included as the independent predictor and categorized as 0, 1, and 2 for the homozygous major allele, heterozygous, and homozygous minor allele groups, respectively. Geometric mean concentrations of POPs were calculated for each genotype, and the test for trend assessed mean levels when the genotypes were included in the model as an ordinal variable.

Using logistic regression, adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were used to estimate associations of TGCT risk and interactions between the endocrine disruptors and hormone-metabolizing SNPs. Risks of TGCT associated with quartiles of serum POP concentrations were estimated within a binary genotype category (homozygous major allele genotype versus genotypes with any minor allele). Tests of interaction were conducted by including a cross-product of the binary genotype variable and the continuous POP concentration. Tests for trend within each genotype category were evaluated by including the POP as a continuous variable in the regression model. All logistic regression models adjusted for the matching factors (age at reference date, race/ethnicity, and date of serum sample collection), and other confounders chosen a priori (body mass index in kg/m2, history of cryptorchism, and family history of testicular cancer). All analyses were also evaluated using conditional logistic regression models with matched case-control pairs; because the associations were similar to those found when the matches were broken, in order to maximize power, logistic regression models adjusting for the matched variables are presented. Additional adjustment for height did not substantially alter estimates. Separate analyses were also conducted for each of the two histology-specific case groups, seminoma and nonseminoma, compared to controls, but because numbers were small, power to run these models was limited and these models will not be presented. All analyses were conducted using the SAS program (version 9.1, SAS Institute, Cary, NC), unless otherwise specified. All tests of significance were two-sided, and p-values <0.05 were considered statistically significant. Because a number of interaction tests were conducted, values for the false discovery rate (FDR), a method by Benjamini and Hochberg (Benjamini & Hochberg, 1995) that evaluates the potential for finding false-positives, were calculated using the “multtest” package in R (http://www.r-project.org/).

Results

Selected characteristics of cases and controls are presented in Table 2. Cases and controls had a mean age of 28 years and more than 85% were white. Compared to controls, cases were more likely to have had a prior diagnosis of cryptorchism and a family history of testicular cancer. Distribution of body mass index in kg/m² was similar between cases and controls.

Table 2.

Characteristics of study participants, U.S. Servicemen’s Testicular Tumor Environmental and Endocrine Determinants Study, 2002–2005

	Controls (n=698)		All testicular germ cell tumors (n=568)		Seminomas (n=251)		Nonseminomas (n=317)
	n	%	n	%	n	%	n	%
Age (years)
<25	221	31.7	195	34.3	45	17.9	150	47.3
25–29	198	28.4	155	27.3	72	28.7	83	26.2
30–34	138	19.8	102	18.0	57	22.7	45	14.2
35–39	106	15.2	85	15.0	56	22.3	29	9.2
≥40	35	5.0	31	5.5	21	8.4	10	3.2
Race/Ethnicity
White	603	86.4	503	88.6	211	84.1	292	92.1
Black	24	3.4	13	2.3	8	3.2	5	1.6
Other	71	10.2	52	9.2	32	12.8	20	6.3
Cryptorchidism
No	683	97.9	536	94.4	242	96.4	294	92.7
Yes	15	2.2	32	5.6	9	3.6	23	7.3
Family history of testicular cancer*
Absent	685	98.1	547	96.3	240	95.6	307	96.9
Present	13	1.9	21	3.7	11	4.4	10	3.2
Body mass index (kg/m2)
<23.4	171	24.5	147	25.9	61	24.4	86	27.1
23.4–25.1	172	24.6	125	22.1	55	22.0	70	22.1
25.2–27.1	179	25.6	137	24.2	51	20.4	86	27.1
>27.2	176	25.2	158	27.9	83	33.2	75	23.7

^*Among first- and second-degree relatives

Genotype-phenotype associations between hormone-metabolizing genes and POP concentrations were assessed (data not shown). Among controls, a CYP1A1 SNP was associated with oxychlordane concentrations, and several HSD17B4 SNPs were associated with PCB concentrations. Men with the CYP1A1 rs4886605 AA genotype had lower oxychlordane concentrations than those with the GG genotype (9.20 ng/g lipid for AA versus 11.04 ng/g lipid for GG, p-trend=0.043). This SNP was not associated with concentrations for any of the other chlordane-related compounds, or with total chlordanes. There were also trends in PCB-118, PCB-138, and PCB-187 concentrations associated with HSD17B4 genotype. Men with the HSD17B4 rs2455463 GG genotype had lower PCB118 concentrations than men with the CC genotype (9.80 versus 11.24 ng/g lipid, respectively, p-trend=0.035). Men had higher PCB-138 concentrations if they had the AA genotype for HSD17B4 rs25640 compared to the GG genotype (p-trend=0.049), and the GG genotype for HSD17B4 rs246899 compared to the AA genotype (p-trend=0.041). Men also had higher PCB187 concentrations if they had the TT genotype for HSD17B4 rs28943585 compared to the CC genotype (p-trend=0.047), and the GG genotype for HSD17B4 rs246899 compared to the AA genotype (p-trend=0.015). No SNPs were associated with any of the other POP concentrations.

HapMap indicated that the CYP1A1 SNPs rs2198843 (11,599bp 3′ of STP C>G) and rs1456432 (9,778bp 3′ of STP A>G) were highly correlated (r²=1.00) and had a high level of linkage disequilibrium (D′=1.0). Similarly, HapMap found that the CYP1A1 SNPs rs7495708 (−4404G>A) and rs4886605 (−10549G>A) were highly correlated (r²=0.92) and had a high level of linkage disequilibrium (D′=1.0). As the associations with TGCT risk were similar for these correlated pairs of SNPs, only POP associations with TGCT risk, modified by rs1456432 and rs7495708, are presented in Table 3. We observed that the CYP1A1 genotype modified the effects of trans-nonachlor and total chlordanes on TGCT risk. For both the CYP1A1 rs1456432 and rs7495708 SNPs, among men with the homozygous major allele genotype, there were no associations between quartiles of trans-nonachlor and total chlordanes and TGCT risk. Among men with any minor allele for the CYP1A1 rs1456432 SNP (AG or GG genotype), those with the highest quartile of trans-nonachlor had a 1.90 times (95% CI, 1.01–3.56) the risk of TGCT compared with men with the lowest (p-interaction=0.024). Similar results were seen for total chlordanes. There was no effect modification by the CYP1A1 rs1456432 SNP for cis-nonachlor (data not shown). Among men with any minor allele for the CYP1A1 rs7495708 SNP (GA or AA genotype), those with the highest quartile of trans-nonachlor had an almost 2-fold (OR=1.92, 95% CI, 1.03–3.58) increased risk of TGCT compared with men with the lowest (p-interaction=0.014). Again, similar results were seen for total chlordanes, and there was no effect modification by genotype for cis-nonachlor. No SNPs modified the associations between TGCT risk and p,p′-DDE (data not shown).

Table 3.

Risk of testicular germ cell tumors by quartile of lipid-adjusted serum concentrations of persistent organochlorine pesticides, stratified by hormone-metabolizing genotype

	AA*						AB/BB*
	Cases		Controls				Cases		Controls
	n	%	n	%	OR†	95% CI	n	%	n	%	OR†	95% CI
CYP1A1 9778bp 3′ of STP A>G (AA vs. AG/GG)
rs1456432
Trans -nonachlor (ng/g lipid)
Quartile 1 (≤10.43)	101	26.6	99	22.5	1.00		40	22.2	71	29.2	1.00
Quartile 2 (10.44–16.58)	87	22.9	115	26.1	0.86	0.56–1.32	49	27.2	57	23.5	1.62	0.89–2.98
Quartile 3 (16.59–25.37)	89	23.4	111	25.2	0.90	0.58–1.39	36	20.0	60	24.7	1.12	0.59–2.13
Quartile 4 (>25.37)	103	27.1	116	26.3	1.03	0.66–1.61	55	30.6	55	22.6	1.90	1.01–3.56
p-trend (continuous)					0.738						0.027
p-interaction											0.024
Total Chlordanes‡
Quartile 1	98	25.9	102	23.1	1.00		38	21.2	68	28.0	1.00
Quartile 2	89	23.5	118	26.8	0.91	0.59–1.38	46	25.7	53	21.8	1.70	0.92–3.14
Quartile 3	80	21.1	103	23.4	0.93	0.60–1.45	38	21.2	68	28.0	1.05	0.56–1.96
Quartile 4	112	29.6	118	26.8	1.21	0.78–1.89	57	31.8	54	22.2	2.07	1.09–3.92
p-trend (continuous)					0.538						0.038
p-interaction											0.014
CYP1A1 -4404 G>A (GG vs. GA/AA)
rs7495708
Trans -nonachlor (ng/g lipid)
Quartile 1 (≤10.43)	101	26.9	100	23.0	1.00		41	22.0	73	29.1	1.00
Quartile 2 (10.44–16.58)	85	22.7	115	26.4	0.86	0.56–1.31	50	26.9	56	22.3	1.70	0.93–3.10
Quartile 3 (16.59–25.37)	88	23.5	108	24.8	0.92	0.59–1.43	36	19.4	64	25.5	1.05	0.56–1.99
Quartile 4 (>25.37)	101	26.9	112	25.8	1.06	0.68–1.65	59	31.7	58	23.1	1.92	1.03–3.58
p-trend (continuous)					0.783						0.016
p-interaction											0.014
Total Chlordanes‡
Quartile 1	98	26.2	102	23.5	1.00		39	21.1	71	28.3	1.00
Quartile 2	89	23.8	118	27.1	0.91	0.60–1.38	46	24.9	54	21.5	1.73	0.94–3.18
Quartile 3	79	21.1	100	23.0	0.94	0.60–1.46	39	21.1	70	27.9	1.08	0.58–2.02
Quartile 4	108	28.9	115	26.4	1.19	0.76–1.85	61	33.0	56	22.3	2.21	1.17–4.15
p-trend (continuous)					0.593						0.022
p-interaction											0.008

^*AA=homozygous for major allele, AB/BB=any minor allele

^†Adjusted for age at reference date, race/ethnicity, date of serum draw, cryptorchism, family history of testicular cancer, and body mass index in kg/m²

^‡Total chlordanes = cis -nonachlor + trans -nonachlor + oxychlordane

Effect modification by a HSD17B4 SNP of the associations between two PCB congeners and TGCT risk is presented in Table 4. HSD17B4 rs384346 was the only SNP to modify the association of TGCT risk with PCB-118 and PCB-138 concentrations. Among men with the homozygous major allele genotype (AA), there was a statistically significant reduction in risk with the highest PCB-118 and PCB-138 quartiles (p-trends<0.001); men with the highest quartile of PCB-118 had an almost 50% reduction in TGCT risk (OR=0.46, 95% CI, 0.31–0.70) compared to men with the lowest. Similar results were seen for PCB-138. Among men with any minor allele for HSD17B4 rs384346, there were no associations between PCB-118 and PCB-138 concentrations and TGCT risk (p-interaction<0.001 for PCB-118 and p-interaction=0.034 for PCB-138). No other interactions between other PCB congeners of interest (PCB-153, PCB-156, PCB-163, PCB170, PCB-180. PCB-187) and SNPs were observed (data not shown). After adjusting for multiple comparisons, FDR values for all interaction tests in the primary hypothesis were >0.20.

Table 4.

Risk of testicular germ cell tumors by quartile of lipid-adjusted serum concentrations of polychlorinated biphenyls (PCB), stratified by hormone-metabolizing genotype

HSD17B4-18796 A>T	AA						AT/TT
rs384346	Cases		Controls				Cases		Controls
	n	%	n	%	OR*	95% CI	n	%	n	%	OR*	95% CI
PCB-118 (ng/g lipid)
Quartile 1 (≤7.00)	138	34.2	125	24.3	1.00		41	26.8	49	28.2	1.00
Quartile 2 (7.01–10.40)	100	24.8	129	25.1	0.66	0.46–0.96	38	24.8	42	24.1	1.27	0.66–2.41
Quartile 3 (10.41–15.56)	92	22.8	128	24.9	0.59	0.40–0.87	31	20.3	44	25.3	1.06	0.54–2.08
Quartile 4 (>15.57)	74	18.3	133	25.8	0.46	0.31–0.70	43	28.1	39	22.4	1.69	0.85–3.38
p-trend (continuous)					<0.001						0.019
p-interaction											<0.001
PCB-138 (ng/g lipid)
Quartile 1 (≤15.84)	134	33.2	128	24.9	1.00		47	30.7	45	25.9	1.00
Quartile 2 (15.85–25.00)	95	23.5	114	22.1	0.72	0.49–1.07	27	17.7	52	29.9	0.61	0.31–1.20
Quartile 3 (25.01–38.53)	96	23.8	140	27.2	0.57	0.38–0.85	36	23.5	40	23.0	1.10	0.54–2.25
Quartile 4 (>38.53)	79	19.6	133	25.8	0.46	0.30–0.72	43	28.1	37	21.3	1.61	0.76–3.41
p-trend (continuous)					<0.001						0.287
p-interaction											0.034

^*Adjusted for age at reference date, race/ethnicity, date of serum draw, cryptorchism, family history of testicular cancer, and body mass index in kg/m²

Our secondary aim to evaluate effect modification among POPs that were not associated with TGCT risk yielded three statistically significant interactions. HSD17B4 rs28943585 and rs28943586 modified the association between PCB-99 concentrations and TGCT risk (p-interactions<0.003). Among men with a minor allele for rs28943585, those in the highest quartiles of PCB-99 had a suggestion, although not statistically significant, of a reduced TGCT risk (OR=0.65, 95% CI, 0.31–1.36, p-trend=0.004) compared to men in the lowest quartile; there was no association between PCB-99 concentrations and TGCT risk among men with the homozygote major allele genotype. Similar associations were seen for rs28943586, which has been shown by HapMap data to be in LD with rs28943585 (r²=1.0, D′=1.0). The association between PCB-183 and TGCT risk was also modified by HSD17B4 rs426899. Among men with the GG genotype, those in the highest quartiles of PCB-183 had a reduced TGCT risk (OR=0.47, 95% CI, 0.26–0.84, p-trend=0.014) compared to men in the lowest quartile; there was no association between PCB-183 concentrations and TGCT risk among men with the GA and AA genotypes. No effect modification was observed for TGCT risk associated with oxychlordane, p,p′-DDT, β-hexachlorocyclohexane, mirex, or PCB-101 (data not shown). After adjusting for multiple comparisons, FDR values for all interactions tests in the secondary hypothesis were also >0.20.

Discussion

This study follows-up findings from our previously published studies investigating exposures to organochlorines, genetic polymorphisms in metabolizing genes and TGCT risk (Figueroa, et al., 2008; McGlynn, et al., 2008; McGlynn, et al., 2009). In our population, we previously reported that among various organochlorine pesticide exposures, men with the highest quartiles of trans-nonachlor and total chlordanes had an approximate 1.5-fold increased risk of TGCT compared to men with the lowest quartiles (McGlynn, et al., 2008). In this study, we observed that these increased risks were only present in men who had a minor allele for CYP1A1 rs1456432 and rs7495708. No interactions were seen for the other hormone-metabolizing gene SNPs and other chlordane-related compounds, p,p′-DDE, p,p′-DDT, or mirex. There were suggestions that polymorphisms in CYP1A1 and HSD17B4 may be associated with concentrations of oxychlordane and PCBs, respectively; thus, it may be plausible that these polymorphisms may also affect TGCT risk by altering the metabolism of oxychlordane and PCB PCB-118, PCB-138, and PCB-187.

Examining polychlorinated biphenyl exposures, we previously found that men with the highest quartiles of PCB-118 and PCB-138 concentrations had an approximate 45–55% reduction in risk of TGCT compared to men with the lowest quartiles (McGlynn, et al., 2009). In this study, we observed that these decreased risks were only present in men who had a major homozygous allele genotype for HSD17B4 rs384346. There was no reduction in risk among men with a minor allele. No interactions were seen for the other hormone-metabolizing gene SNPs and other PCB congeners.

Although the evidence is limited, previous studies have suggested that POPs are associated with TGCT risk. A small study (n_cases=58) by Hardell et al. of Sweden men (Hardell, et al., 2003) observed that of the 8 POPs measured, only cis-nonachlordane was associated with an increased risk of testicular cancer (OR=2.6, 95% CI, 1.2–5.7). This study also found increased levels of PCBs, hexachlorbenzene, and trans- and cis-nonachlordane in mothers’ of cases compared to those in mothers’ of controls, however the exposure period may not have been relevant as serum from mothers was collected at the time of testicular cancer diagnosis in the sons. Using the same population, Hardell et al. followed up these finding by examining 37 additional PCB congeners; this study again found no differences in PCB levels between testicular cancer cases and controls (Hardell, et al., 2004). In a U.S. study of 247 TGCT cases, serum concentrations of 11 organochlorine pesticide residues were measured: β-hexachlorocyclohexane, γ-hexachlorocyclohexane, dieldrin, hexachlorobenzene, heptachlor epoxide, mirex, p,p′-DDT, o,p-DDT, p,p′-DDE, oxychlordane, and trans-nonachlor (Biggs, et al., 2008). The study also genotyped microsatellite repeat polymorphisms in the AR gene (CAG genotype: <23 versus ≥23 repeats, GGN genotype: <17 versus ≥ 17 repeats). No associations between any of the organochlorines and TGCT risk were identified, and no effect modification by the AR polymorphisms was evident. In the current study, we also found no associations between polymorphisms in the AR gene and serum p,p′-DDE concentrations, although we genotyped SNPs versus their microsatellite repeat polymorphisms.

It is biologically plausible that POPs may be associated with TGCT risk, and that certain populations may be susceptible based on their genetic predisposition. An in vitro study (Bonefeld-Jorgensen, et al., 2001) found that PCB-138 and PCB-153 can bind to estrogen- and androgen-receptors, thus giving these congeners the ability to modulate sex hormone processes. Although the estrogen- and androgen-receptors bind these endocrine disruptors at a lower affinity compared with that for the sex hormones, there may be higher free concentrations of POPs in the blood due to their lower affinity for sex hormone-binding globulin (Hodgert Jury, et al., 2000).

There is evidence from animal studies indicating that PCBs can induce CYP1A1 expression (Bandiera, et al., 1997; Drahushuk, et al., 1997; Li, 2007). In addition, epidemiologic studies in women have observed interactions between CYP1A1 polymorphisms and PCB exposure on breast cancer risk (Laden, et al., 2002; Li, et al., 2005; Moysich, et al., 1999; Zhang, et al., 2004). However, we found no evidence of interactions between PCBs and CYP1A1 polymorphisms on TGCT risk. These studies all included two polymorphisms that we also examined, CYP1A1 M1 (also known as CYP1A1*2A, rs4646903) and CYP1A1 M2 (also known as CYP1A1*2C, rs1048943), and consistently observed that CYP1A1 M2 modified the association between PCB exposure and breast cancer risk. Risk of breast cancer was highest among women with any minor allele and high PCB concentrations; women with a homozygous major allele genotype had no increase in risk even with high PCB concentrations (Laden, et al., 2002; Moysich, et al., 1999; Zhang, et al., 2004). Based on HapMap data, the CYP1A1 M1 and M2 polymorphisms are in strong linkage disequilibrium with CYP1A1 rs1456432, which we did find to interact with the pesticide, trans-nonachlor. The mechanism behind these interactions is still unknown, though, and one reason why no interactions between PCBs and CYP1A1 polymorphisms were observed may be due to differences in hormonal actions and profiles between men and women.

There are several strengths and limitations that must be considered when evaluating the results of this study. The sample size is relatively large and there was sufficient power to detect interactions of genotype by POP exposure on TGCT risk; however, some participants could not be contacted due to deployment and missing or insufficient buccal and serum samples may have introduced selection bias. But, because most young men in the military are healthy, it is unlikely that deployment status introducedsubstantial bias. This study also has detailed histological information, however, we were not able to examine differences in associations observed by histology, seminomas and nonseminomas (Figueroa, et al., 2008; McGlynn, et al., 2008; McGlynn, et al., 2009), because of limited case numbers. This is the first study to assess risk of TGCT and interactions between polymorphisms in several common hormone-metabolizing genes and pre-diagnostic serum concentrations of POPs. Because serum was collected prior to TGCT diagnosis, our study does not have the concerns that other studies with serum collected post-diagnosis may have: that chemotherapy (Baris, et al., 2000) and changes in body weight (Hagmar, et al., 2006) due to disease and treatment may affect organochlorine concentrations. Sensitivity analyses were conducted to see if age of serum affected risk estimates, exclusion of men with serum samples taken within one year of diagnosis, within two years of diagnosis, or more than 10 years before diagnosis yielded similar estimates of effect. Because we examined interactions across a number of SNPs and POPs, the probability of false positive interactions is not negligible, and after adjustment for multiple comparisons, the FDR values were >0.20. As a result, caution is warranted in interpreting these findings.

In conclusion, the current study suggests that some CYP1A1 and HSD17B4 polymorphisms may modify the associations between chlordanes and PCBs, respectively, and risk of TGCT. Further studies of the relationships, however, are necessary in order to confirm these findings.