Metabolic Gene Variants and Risk of Non-Hodgkin's Lymphoma

Introduction

Several environmental and occupational exposures have been suspected as risk factors for non-Hodgkin's lymphoma. Increased non-Hodgkin's lymphoma incidence has been consistently observed among farmers (1, 2), and pesticide exposures underlying this association have been investigated, with inconsistent results. Polychlorinated biphenyls have been associated with increased non-Hodgkin's lymphoma incidence in several studies (3-5), and other types of persistent organochlorines, including pesticides and dioxins, have also been suspected. Benzene is a potent lymphomagen in experimental animals, and high-dose exposure in humans has been associated with both acute myeloid leukemia and non-Hodgkin's lymphoma (6). These observations have spurred investigation of other solvent exposures, including chlorinated solvents, thinner and white spirit, and mineral oil, which have been linked to non-Hodgkin's lymphoma risk in some but not all studies (7). Polycyclic aromatic hydrocarbons (PAH) are known to cause a variety of cancers primarily through evidence linking cigarette smoking, a potent source of PAHs, with various cancers. The evidence for cigarette smoking as a cause of non-Hodgkin's lymphoma is mixed, and several well-designed studies have not observed an association (8-11). However, an analysis of pooled case-control studies found increased risk to be limited to follicular lymphomas (12).

The potential of exogenous chemical exposures to cause non-Hodgkin's lymphoma is unclear probably because exposures have been difficult to evaluate and quantify over a period relevant to lymphomagenesis. By studying genes involved in biotransformation of these chemicals and their metabolites, associations between non-Hodgkin's lymphoma risk and variants in specific metabolic genes may provide clues about which of the many chemical exposures to pursue in future research. We studied variation in metabolic genes in a population-based case-control study in the United States. We selected several genes known to play a role in metabolizing a broad spectrum of substrates, including pesticides, organochlorines, solvents, and PAHs, such as the phase I cytochrome P450 enzymes (CYP1A1, CYP1B1, CYP2C9, and CYP2E1), the phase II glutathione S-transferases (GSTP1 and GSTM3), and epoxide hydrolase (EPHX1). Additional genes were selected based on involvement in metabolizing a specific substrate of interest, such as PON1 for organophosphate pesticide metabolism, AHR for its involvement in organochlorine metabolism through transcriptional regulation of CYP1A1 and other genes, and NQO1 for benzene and PAHs. Following gene selection, we studied known variants within those genes that were chosen based on their frequencies and functional consequence.

Materials and Methods

Study Population

The population used for this study has been described previously (13). Cases were incident non-Hodgkin's lymphoma patients without evidence of HIV infection, ages 20 to 74 years, who were identified in four Surveillance, Epidemiology, and End Results (SEER) registries (the state of Iowa, Los Angeles County, CA, and the metropolitan areas of Detroit, MI and Seattle, WA) between July 1998 and June 2000. In total, 1,321 (58%) of the eligible cases were interviewed, representing 77% of those who were approached. Major reasons for nonresponse were subject refusal, death, and physician refusal. Controls were identified by random-digit dialing if they were <65 years old or from Medicare eligibility files if they were ≥65 years old. To parallel the distributions among cases, all controls were stratified by age (in 5-year categories) and gender within each geographic region and by race in Detroit and Los Angeles County, where oversampling of African Americans occurred. In total, 1,057 (44%) of selected potential controls were interviewed, representing 52% of those approached. Major reasons for nonresponse were subject refusal and failure to locate. Written informed consent was obtained from each participant before interview. A computer-assisted personal interview was administered that contained questions about demographic characteristics, lifestyle factors, occupational history, and pesticide use history. All study participants were asked to provide a venous blood or mouthwash buccal cell sample, and these were shipped to the biological repository for processing and storage. Demographic characteristics (age, education, and gender) for individuals who provided blood versus buccal cells and compared with those who provided neither type of sample were equivalent within each study site. In this analysis of metabolic gene variants in relation to non-Hodgkin's lymphoma, we evaluated the 1,172 (89%) cases and 982 (93%) controls who provided biological samples from which DNA was successfully extracted (Table 1 ).

Histopathology

Each SEER registry provided non-Hodgkin's lymphoma pathology and subtype information derived from abstracted reports by the local diagnosing pathologist. All cases were histologically confirmed and coded according to the International Classification of Diseases for Oncology, Second Edition (ICD-O-2; ref. 14). We evaluated the following histologic outcomes: (a) non-Hodgkin's lymphoma overall (ICD-O-2 9590-01, 9595, 9670-73, 9675-76, 9680-88, 9690-91, 9695-98, 9700, 9702-03, 9705-11, 9713-15, 9823, and 9827), (b) B-cell lymphomas (ICD-O-2 9670-71, 9673, 9676, 9680-88, 9690-91, 9695-98, 9710-11, 9715, and 9823), (c) T-cell lymphomas (ICD-O-2 9700, 9702-03, 9705-09, 9713-14, and 9827), and four B-cell lymphoma subtypes: (d) diffuse large B-cell (ICD-O-2 9680-84 and 9688), (e) follicular (ICD-O-2 9676, 9690-91, and 9695-98), (f) marginal zone (ICD-O-2 9710-11 and 9715), and (g) small lymphocytic lymphoma/chronic lymphocytic leukemia (ICD-O-2 9670-71 and 9823).

Laboratory Methods

DNA Extraction. DNA was extracted from blood clots or buffy coats (from 10 mL blood) at BBI Biotech Repository (Gaithersburg, MD) using Puregene Autopure DNA Extraction kits (Gentra Systems, Minneapolis, MN). DNA was extracted from buccal cell samples by phenol chloroform extraction methods (15).

Genotyping. Genotyping methods have been described previously (16). For this study, we selected 15 variants in 11 metabolic genes (Table 2 ) based on expected population prevalence >5% (for homozygous variant and heterozygote combined), with evidence of functional consequence and/or evidence of an association with lymphoma-related risk conditions in animal or human studies. All genotyping was conducted at the National Cancer Institute (NCI) Core Genotyping Facility (Advanced Technology Corp., Gaithersburg, MD) using the Taqman platform. Sequence data and assay conditions can be found at http://snp500cancer.nci.nih.gov. Blood samples were genotyped first, and buccal samples were subsequently analyzed when there was sufficient buccal cell DNA. In all, we had data from blood and buccal samples for seven variants and data from only blood samples for eight variants. Successful genotyping was achieved for 96% to 100% of DNA samples for all variants. Completion rates were generally slightly higher among blood-based samples, significantly so for the EPHX1 H139R variant. A small percentage of samples (2.7% of controls and 1.9% of cases; P > 0.05) showed chromosomal sex status that differed from gender status for the subject; these subjects were excluded from risk analyses of genotype-non-Hodgkin's lymphoma associations.

Quality Control. Forty replicate samples from each of 2 blood donors and duplicate samples from 100 study subjects that had been processed in an identical fashion were interspersed for all genotyping assays and blinded from the laboratory. Agreement for quality-control replicates and duplicates was >99% for all assays. In addition, for each plate of 368 samples, genotype-specific quality-control samples were included by the laboratory that comprised four common homozygote, four heterozygote, and four variant homozygote samples and four DNA negative controls. For each genotype, we tested Hardy-Weinberg equilibrium among Black and non-Hispanic White controls to assess whether the genotypes were distributed as expected (17).

Statistical Analyses

For each genotype and non-Hodgkin's lymphoma outcome (all non-Hodgkin's lymphoma, all B-cell non-Hodgkin's lymphoma, and all T-cell non-Hodgkin's lymphoma), we estimated odds ratios (OR) and 95% confidence intervals (95% CI) by logistic regression using the common homozygous genotype as the reference group. We calculated the P_trend based on a three-level ordinal variable (0, 1, 2) of the common homozygote, heterozygote, and variant homozygote. Genotype-non-Hodgkin's lymphoma associations for mutually exclusive specific B-cell histologic subtypes were modeled using polytomous logistic regression. All genotype risk estimates were adjusted for age (indicator variables for ages <35, 35-44, 45-54, and 55-64, with age ≥65 as the reference), gender (male or female), and race (indicator variables for Black and White races, with other race as the reference), as these were all study design variables. We evaluated the risk of false-positive findings by estimating values which account for the false discovery rate (FDR, the Benjamini-Hochberg adjustment; ref. 18). The false discovery rate reflects the expected ratio of false-positive findings to the total number of significant findings. FDR values were estimated for the genotype effects (three-level trend test variable or dichotomous genotype) for all NHL, for B- and T-cell lymphomas, and for B-cell histologic subtypes.

We conducted additional analyses of associations with all non-Hodgkin's lymphoma stratified by age (<50 and ≥50 years), gender (male or female), and race (non-Hispanic Whites or Blacks) to further explore consistency of associations observed in analyses of all cases. Interactions of metabolic gene variants with the smoking status were examined in an exploratory analysis first by evaluating the gene variant effect stratified by smoking status (never, former, or current) and then by estimating the statistical significance of a multiplicative interaction term for the dichotomous gene variant (heterozygotes and variant homozygotes combined) with smoking (one interaction term for current smoking and one interaction term for former smoking, with never smoking as the reference). However, statistical power for examining these interactions was limited, as information on smoking was by design obtained from approximately half of the study participants.

Results

Cases and controls included in the analyses were similar in terms of gender, study site, education, body mass index, family history of non-Hodgkin's lymphoma, and DNA source (Table 1). Cases were more likely to be younger and White.

Genes and variants evaluated for this study are listed in Tables 2 and 3 (18). Prevalences of variant genotypes in the control group were similar to previously reported values (http://snp500cancer.nci.nih.gov/snp.cfm; ref. 20). One variant (AHR R554K) was not in Hardy-Weinberg Equilibrium in Black controls; quality-control inquiries showed no discrepancies for this variant.

Subjects who were homozygous for the cytochrome P450 gene variant CYP1B1 V432L were at slightly increased risk of non-Hodgkin's lymphoma (OR, 1.27; 95% CI, 0.97-1.65) than the nonvariant homozygotes (Table 3). The association was consistently observed across B-cell subtypes but not T-cell lymphoma (Fig. 1 ); in particular, diffuse B-cell lymphoma was associated with the CYP1B1 variant (heterozygotes OR, 1.36; 95% CI, 1.01-1.81; variant homozygotes OR, 1.48; 95% CI, 1.02-2.15; P_trend = 0.02 FDR value = 0.42). The increase was observed among both non-Hispanic White (homozygotes OR, 1.33; 95% CI, 0.99-1.78) and Black (homozygotes OR, 1.23; 95% CI, 0.41-3.65) subjects.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Associations between metabolic gene variants and non-Hodgkin's lymphoma for non-Hodgkin's lymphoma overall and for specific histologic subtypes (frequencies, ORs, and 95% CIs). NHL, non-Hodgkin's lymphoma; SLL, small lymphocytic lymphoma.

The CYP2E1 −1054T allele was associated with decreased risk of non-Hodgkin's lymphoma (CT and TT genotypes combined OR, 0.59; 95% CI, 0.37-0.93 FDR value = 0.34); this association was consistent for all B- and T-cell histologic types we examined, although not significantly so (Fig. 1). The inverse association with the CYP2E1 −1054T allele seemed most pronounced among men, older subjects (>50), and non-Hispanic Whites, with statistically significant 49% to 51% decreases in risk; however, some decreased risk was observed in each gender, age, and race group examined (see results in Supplementary Tables).

The PON1 L55M variant AA genotype was associated with increased risk of non-Hodgkin's lymphoma (OR, 1.36; 95% CI, 0.96-1.95 FDR value = 0.37). This association was observed in non-Hispanic White subjects (OR, 1.44; 95% CI, 1.00-2.08; only one Black control and no Black case subject carried the variant homozygote genotype) and was fairly consistent among age and gender groups examined. The PON L55M variant was associated with follicular non-Hodgkin's lymphoma (heterozygotes OR, 1.31; 95% CI, 0.90-1.92; variant homozygotes OR, 2.12; 95% CI, 1.27-3.52; P_trend = 0.003 FDR value = 0.20) and T-cell lymphoma (heterozygotes OR, 1.64; 95% CI, 0.81-3.33; variant homozygotes OR, 2.93; 95% CI, 1.21-7.08; P_trend = 0.02 FDR value = 0.38) but not with other histologic subtypes.

There was no overall association with non-Hodgkin's lymphoma for the other gene variants we examined in AHR, CYP1A1, CYP1A2, EPHX1, GSTM3, GSTP1, NQO1, or PON1 (Q192R) and no trends by numbers of variant alleles. In addition, the analyses of specific histologic subtypes and analyses stratified by age, race, and gender did not reveal any notable associations (Supplementary Tables are available online).

There were no readily interpretable interactions between metabolic gene variants and smoking (Supplementary Table is available online). Significant interactions between former smoking and EPHX1 H139R (P = 0.03) and PON1 Q192R (P = 0.01) did not have clear biologically meaningful interpretation, as the interactions resulted from inverse associations of the variant genotype among former smokers and positive associations among nonsmokers.

Discussion

Although studying genetic variants can provide clues about exposures involved in disease etiology, we did not find any strong associations in the metabolic pathways we studied, with the variants we selected in genes that have either broad or more specific xenobiotic substrates. We did observe associations with some cytochrome P450 gene variants that seemed reasonably consistent across histologic subtypes of non-Hodgkin's lymphoma and demographic subgroups.

There have been few previous studies of metabolic gene variants in relation to non-Hodgkin's lymphoma, and several variants were investigated for the first time in our study. Sarmanova et al. conducted a study in Prague, Czech Republic, of 143 non-Hodgkin's lymphoma cases and 455 population-based controls (21, 22). They did not observe any association with the risk of non-Hodgkin's lymphoma for the CYP2E1 −1054C>T variant. Similar to our study, they found no overall associations between risk of non-Hodgkin's lymphoma and EPHX1 variants Y113H or H139R or NQO1 P187S, although subgroup associations according to gender and clinicopathologic characteristics were reported. In a hospital-based study of 169 cases and 205 controls in Australia, Kerridge et al. (23) observed significantly increased non-Hodgkin's lymphoma risk associated with the PON1 Q192R GG genotype; this finding was not replicated in our study. Finally, a study of 389 non-Hodgkin's lymphoma cases and 535 controls in Nebraska found no association between GSTP1 I105V variant and non-Hodgkin's lymphoma risk, similar to our findings (24).

No previous study has evaluated variants in CYP1B1 with risk of non-Hodgkin's lymphoma or other lymphohematopoietic cancers. CYP1B1, like other cytochrome P450 enzymes, plays an important role in phase I metabolism by activating chemicals, such as PAHs and dioxins, to create oxidized, reactive intermediates. CYP1B1 also plays an important role in estrogen metabolism by catalyzing the formation of 4-hydroxyestradiol, a carcinogenic metabolite that retains significant estrogenic activity. The V432L G (Val) allele has been associated with higher activity, resulting in increased oxidation of benzo(a)pyrene and formation of 4-hydroxyestradiol than the C (Leu) allele (25, 26). Furthermore, induction of CYP1B1 by PAHs and/or dioxins has been reported (26, 27), and higher induction has been associated with the G versus C allele of the V432L variant (27). We observed increased non-Hodgkin's lymphoma risk associated with the G allele, which fits our hypothesized biological mechanism that increased CYP1B1 activity leading to increased oxidation and formation of toxic intermediates would contribute to increased cancer risk. It is possible that if increased CYP1B1 activity is truly a risk factor for non-Hodgkin's lymphoma, the effect we observed was weak in the context of our general population sample, which likely had relatively low exposures to PAH and other relevant substrates (e.g., as opposed to a highly exposed occupational cohort). At present, we cannot link PAH exposure to the genotype data because PAH exposure modeling is still under way in our study; this is a topic for future analyses.

We observed decreases in non-Hodgkin's lymphoma risk associated with a variant in CYP2E1. CYP2E1 metabolizes and activates solvent carcinogens that also act to induce its expression, including benzene, styrene, carbon tetrachloride, ethylene glycol, and ethanol (20). In an oral clearance study, the −1054C>T (also called RsaI) variant was associated with decreased metabolic activity (29); however, other functional studies have shown higher expression of the variant allele (30). More information on the functional consequence of the variant is necessary to inform the meaning of our observed association with CYP2E1 −1054C>T.

PON1, a high-density lipoprotein-associated enzyme that is synthesized in the liver and secreted into plasma, is a phase II enzyme involved in hydrolyzing various phase I intermediates, such as the highly toxic oxon metabolites of several organophosphate pesticides, including chlorpyrifos, diazinon, and parathion (31). The PON1 L55M methionine allele has shown lower paraoxonase activity than the more common allele (32). The modest association we observed was in the hypothesized direction of increased non-Hodgkin's lymphoma risk associated with lower phase II activity that might make a person more susceptible to toxicity from exogenous chemical exposures, including organophosphate insecticides. However, this variant was only associated with follicular B-cell lymphoma and T-cell non-Hodgkin's lymphoma, and there was no a priori reason to suspect specific associations by histologic subtype. The other variant we studied in PON1 was not associated with the risk of non-Hodgkin's lymphoma, and further studies are needed to completely characterize the role of these variants in PON1 function.

The essentially null effects we observed for most of the metabolic gene variants in relation to non-Hodgkin's lymphoma may reflect the context of exposures within the general population represented in our study. For example, the AHR gene may not have great importance in the context of relatively low levels of exposure to dioxins and dioxin-like compounds in the population (33). It is also possible that the variants we selected within candidate metabolic genes did not adequately characterize variability in gene function. Nevertheless, our study adds to the literature about the potential contribution of variation in metabolic genes to non-Hodgkin's lymphoma. These data provide some evidence for CYP1B1, CYP2E1, and PON1 in non-Hodgkin's lymphoma etiology but do not indicate importance of most of the other genes we examined. The modest associations we observed could be false-positive results due to chance variation, as the FDR values for our key findings were relatively large. These results are therefore considered suggestive and require replication in other study populations and in large, pooled analyses.