Variation in the Selenoenzyme Genes and Risk of Advanced Distal Colorectal Adenoma

Background: Epidemiologic and animal studies provide evidence for a chemopreventive effect of selenium on colorectal cancer, which may be mediated by the antioxidative and anti-inflammatory properties of selenoenzymes. We therefore investigated whether genetic variants in selenoenzymes abundantly expressed in the colon are associated with advanced colorectal adenoma, a cancer precursor.

Methods: Cases with a left-sided advanced adenoma (n = 772) and matched controls (n = 777) screen negative for polyps based on sigmoidoscopy examination were randomly selected from participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial. The underlying genetic variation was determined by resequencing. We genotyped 44 tagging single nucleotide polymorphisms (SNP) in six genes [glutathione peroxidase 1-4 (GPX1, GPX2, GPX3, and GPX4), selenoprotein P (SEPP1), and thioredoxin reductase 1 (TXNRD1)] to efficiently predict common variation across these genes.

Results: Four variants in SEPP1 were significantly associated with advanced adenoma risk. A rare variant in the 5′ region of SEPP1 (-4166C>G) was present in nine cases but in none of the controls (exact P = 0.002). Three SNPs located in the 3′ region of SEPP1, which is overlapping with the promoter region of an antisense transcript, were significantly associated with adenoma risk: homozygotes at two SEPP1 loci (31,174 bp 3′ of STP A>G and 43,881 bp 3′ of STP G>A) were associated with increased adenoma risk [odds ratio (OR), 1.48; 95% confidence interval (95% CI), 1.00-2.19 and OR, 1.53; 95% CI, 1.05-2.22, respectively] and the variant SEPP1 44,321 bp 3′ of STP C>T was associated with a reduced adenoma risk (CT versus CC OR, 0.85; 95% CI, 0.63-1.15). Furthermore, we observed a significant 80% reduction for advanced colorectal adenoma risk for carriers of the variant allele at TXNRD1 IVS1-181C>G (OR, 0.20; 95% CI, 0.07-0.55; P_trend = 0.004). Consistent with the individual SNP results, we observed a significant overall association with adenoma risk for SEPP1 and TXNRD1 (global P = 0.02 and 0.008, respectively) but not for the four GPX genes.

Conclusion: Our study suggests that genetic variants at or near the SEPP1 and TXNRD1 loci may be associated with advanced colorectal adenoma. As this is the first study to comprehensively investigate this hypothesis, confirmation in independent study populations is needed. (Cancer Epidemiol Biomarkers Prev 2008;17(5):1144–54)

An increasing number of studies suggests that the essential trace element selenium is a preventive agent for colorectal carcinogenesis. Evidence arises from the Nutritional Prevention of Cancer Trial, a randomized study to evaluate selenium supplementation and skin cancer prevention, which observed, as secondary endpoint, a 61% reduction in colorectal cancer (1). This association was further confirmed, although slightly attenuated, when based on 2 additional years of follow-up (2). Furthermore, several blood-based observational studies, including ours (3), support a beneficial effect of selenium on colorectal carcinogenesis (3-10).

Important biological activities of the essential trace element selenium are mediated through the function of selenoenzymes. Selenium is incorporated into the active center of selenoenzymes as selenocysteine, which was only recently discovered as the 21st naturally occurring amino acid (11). Compared with other amino acids, selenocysteine occurs infrequently in a small number of proteins (12); however, it is located in the active center of the selenoenzymes. The unique redox characteristics of selenocysteine confer important antioxidant properties to these selenoenzymes, which can reduce reactive oxygen species and thereby prevent damage of important biomolecules, including DNA, RNA, lipids, proteins, and membranes; reactive oxygen species–induced DNA damage is known to promote tumor progression (13-18). Because of the direct contact of the colonic epithelial cells with microbial- and food-derived reactive oxygen species, the gastrointestinal tract may be particularly susceptible to oxidative damage (19-23). In addition, selenium and selenoenzyme activity may reduce inflammatory processes, known triggers for colorectal carcinogenesis (24, 25).

In our study, we focus on the six antioxidative selenoenzymes, glutathione peroxidase 1-4 (GPX1, GPX2, GPX3, and GPX4; refs. 1-4), selenoprotein P (SEPP1), and thioredoxin reductase 1 (TXNRD1), which are expressed in the gastrointestinal tract (26-30). Interestingly, the gastrointestinal tract expresses all four common GPX (GPX1-GPX4), which may suggest a significant role in colonic function (e.g., as a barrier for reactive oxygen species; refs. 26, 31, 32). Support for the importance of these selenoenzymes comes from an animal study showing that targeted disruption of both cytosolic and gastrointestinal GPX (GPX1 and GPX2, respectively) genes results in accumulation of lipid hydroperoxides (33) and high susceptibility to inflammation and colon cancer in mice (25, 34). Preliminary data suggest that genetic variants in some of these selenoenzymes may affect their function and cancer risk; however, only a limited number of the genetic variants in selenoenzymes has been identified or studied so far.

To investigate the association between polymorphisms in six selenoenzymes genes abundantly expressed in the colon (GPX1-GPX4, SEPP1, and TXNRD1) and advanced colorectal adenoma risk, we conducted a case-control study nested within the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial. In this large and well-characterized population, case and control status were identified following a standard protocol. To capture the underlying genetic variation across each of the six genes, we resequenced all functionally important regions and genotyped selected tagging single nucleotide polymorphisms (tagSNP). Gene-environmental interactions were explored for important factors, such as serum selenium concentration and smoking.

This case-control study was nested within the Prostate, Lung, Colorectal and Ovarian Trial, which was designed to evaluate selected methods for the early detection of these cancers and to investigate etiologic factors and early markers of cancer (35, 36). In brief, the Prostate, Lung, Colorectal and Ovarian Trial recruited 154,952 men and women ages 55 to 74 years at 10 centers in the United States (Birmingham, AL; Denver, CO; Detroit, MI; Honolulu, HI; Marshfield, WI; Minneapolis, MN; Pittsburgh, PA; Salt Lake City, UT; St. Louis, MO; and Washington, DC) between 1993 and 2001. Participants were randomized to routine care or screening for prostate cancer (prostate-specific antigen testing and digital rectal examination), lung cancer (chest X-ray), colorectal cancer (sigmoidoscopy), and ovarian cancer (CA125 testing and transvaginal ultrasound). If the sigmoidoscopy identified polyps or other suspect lesions, participants were advised to get further follow-up examination through their own medical care providers, which usually resulted in a full colonoscopy with polypectomy or surgical procedures, if indicated. All medical and pathologic reports of the follow-up examinations were obtained and coded by trained medical record abstractors. Written informed consent was obtained from participants and the trial received approval from the institutional review boards of the U.S. National Cancer Institute and the 10 study centers.

Cases and controls for this study were selected from participants randomized to the screening arm between September 1993 and September 1999, who had undergone a successful sigmoidoscopic examination at baseline (insertion to at least 50 cm with >90% of mucosa visible or a suspect lesion identified), completed a baseline risk factor questionnaire, donated a blood sample, and consented to participate in etiologic studies (n = 42,037). Of these participants, we excluded 4,834 with a self-reported history of cancer (except basal-cell skin cancer), ulcerative colitis, Crohn’s disease, familial polyposis, colorectal polyps, or Gardner’s syndrome. We randomly selected 772 cases for study from among 1,234 cases with advanced distal adenoma (≥10 mm in size, containing high-grade dysplasia, or villous characteristics) in the distal colon (descending colon and sigmoid or rectum) and 777 controls with a negative screening sigmoidoscopy (that is, no polyp or other suspect lesion; n = 26,651) frequency matched to cases by gender and self-reported ethnicity. Five cases and four controls either had no DNA or had discrepancies on repeated fingerprint analyses and were excluded from all analysis, leaving 767 cases and 773 controls for the study. Approximately 63% of the cases had at least one distal adenoma considered to be histologically aggressive (high-grade dysplasia or villous characteristics).

To determine the underlying genetic variation in each gene (GPX1-GPX4, SEPP1, and TXNRD1), we resequenced the promoter regions, 5′ and 3′ untranslated regions, including the selenocysteine insertion sequence, as well as all exons and intron-exon boundaries in a multiethnic panel of 102 subjects, including 31 Caucasians and 24 African Americans (details are described elsewhere; ref. 37). To efficiently predict the common variation across each gene, we selected tagSNPs based on the resequencing data using the method developed by Clayton.⁷

The selection criteria were minimum haplotype r² = 0.9 and minor allele frequency > 5% in Caucasians or African Americans, the predominant ethnic groups in this study population. We preselected any nonsynonymous SNPs and those located in the selenocysteine insertion sequence as tagSNPs. TagSNPs were determined separately within the subgroup of Caucasians and African Americans, with the goal to maximize the overlap between tagSNPs to minimize the total number of tagSNPs.

In total, 44 tagSNPs (GPX1, n = 6; GPX2, n = 6; GPX3, n = 11; GPX4, n = 7; SEPP1, n = 6; and TXNRD1; n = 8) were successfully genotyped. These 44 tagSNPs resulted in substantial coverage of the common SNPs (minor allele frequency > 5%). The average r² and minimal r² are as follows: GPX1, 0.73 and 0.33; GPX2, 0.81 and 0.22; GPX3, 0.83 and 0.62; GPX4, 0.92 and 0.53; SEPP1, 0.73 and 0.38; and TXNRD1, 0.72 and 0.20, respectively. The fraction of common SNPs captured by the tagSNPs with r² ≥ 0.8 and r² ≥ 0.5 is 0.75 and 0.75 for GPX1, 0.59 and 0.73 for GPX2, 0.67 and 1.0 for GPX3, 0.88 and 1.0 for GPX4, 0.40 and 0.90 for SEPP1, and 0.43 and 0.71 for TXNRD1, respectively.

DNA was extracted from whole-blood or buffy-coat samples using QIAamp DNA Blood Midi or Maxi kits (Qiagen). Genotyping was done using SNPlex or TaqMan assays at the National Cancer Institute Core Genotyping Facility (38). Assays were validated and optimized as described in the SNP500 Cancer Web page.⁸

Laboratory personnel were blinded to case-control status, and replicate quality-control samples (46 individuals assayed two to six times per polymorphism) were interspersed in the plates. Replicate samples displayed >99% concordance for all SNPs, except for GPX1 1,553 bp 3′ of STP (97%), GPX2 2,680 bp 3′ of STP (97%), GPX2 IVS1-604 (98%), GPX3 IVS1+2268 (96%), GPX3 IVS4-14 (95%), GPX3 Ex5+314 (96%), and TXNRD1 IVS1-181 (98%). Given the high concordance rates, all SNPs were included in the analysis. Depending on the batch, 0.5% to 1.7% of the subjects had insufficient DNA for genotyping and a small percentage of participants (<1%) was found to have discrepancies on repeated DNA fingerprint analysis and hence were excluded from the analysis. Of those with sufficient DNA, genotyping was successfully completed for 96.5% to 99.5% of the participants depending on the genotype. Genotype frequency among Caucasian controls were consistent with Hardy-Weinberg proportions (P > 0.05), except for SEPP1 44,321 bp 3′ of STP (P = 0.03), TXNRD1 Ex15+410 (P = 0.0008), and TXNRD1 -27129 (P = 0.04).

At initial screening, all participants were asked to complete a questionnaire about sociodemographic factors, height, weight, medical history (including current and former use of aspirin and other nonsteroidal anti-inflammatory drugs), and risk factors for cancer (including smoking and physical activity). Usual dietary intake over the 12 months before enrollment was assessed with a 137-item food frequency questionnaire, including an additional 14 questions about intake of vitamin and mineral supplements and 10 questions about meat cooking practice (39). Daily dietary nutrient intake was calculated by multiplying the daily frequency of each consumed food item by the nutrient value of the sex-specific portion size (40) using the nutrient database from the U.S. Department of Agriculture (41). Serum selenium concentrations were determined using inductively coupled plasma mass spectrometry (for details, see refs. 3, 42).

To estimate the association between individual SNPs and colorectal adenoma, we calculated odds ratios (OR) and 95% confidence intervals (95% CI) using unconditional logistic regression analysis. Genotypes were evaluated using indicator variables with the common homozygote as reference. Test for a linear trend was assessed by including a single variable for each SNP, coded as the number of variant alleles, in the regression model. For genotypes with small cell numbers (<5), we used exact test statistics.

All ORs were adjusted for the matching factors, sex, and race as well as for age at randomization, study center, and ethnicity. Adjusting for additional potential risk factors of colorectal tumors, including physical activity, body mass index, smoking, alcohol intake, aspirin use, ibuprofen use, educational attainment, energy intake, red meat intake, folate intake, fiber intake, and calcium intake, did not materially affect the results (data not shown); hence, these variables were not included in the final analysis.

To adjust for multiple comparisons, we conducted global tests of association by simultaneously including all of the SNPs in a given gene in the logistic regression model and then comparing it with a null model that includes none of the SNPs (43). In this multivariate analysis, each SNP was coded by two dummy variables corresponding to homozygous and heterozygous variant genotypes. The resulting likelihood ratio χ² value had 2k df, with k denoting the number of SNPs in a given gene. This multiloci global test automatically adjusts for multiple testing based on the df of the corresponding χ² test. Moreover, the multiloci test can efficiently capture the multivariate linkage disequilibrium pattern within a gene and hence can be more efficient than tests based on SNPs for detecting associations when the true causal variant in the region may not have been genotyped.

We explored interaction of any significant SNPs with serum selenium, age, and smoking. We coded selenium and age as continuous variables. For smoking, we combined current smokers and those that quit smoking <10 years ago as recent smokers to avoid small cell numbers and to account for the relevant period of advanced adenoma development. Former smokers were defined as those that quit smoking ≥10 years ago. To test for statistical significance of multiplicative interaction, we compared models with and without the cross-product terms using likelihood ratio test.

We used polytomous logistic regression analysis to examine the association between individual SNPs and adenoma subtypes, defined by histologic characteristics (tubular, tubular/villous, and villous) and multiplicity (single adenoma, more than one adenoma), and to test for heterogeneity to assess differences between these subtypes.

We used Haploview to assess pairwise linkage disequilibrium measurements (D and r²) and define haplotype blocks based on Gabriel’s algorithm (44). As the linkage disequilibrium pattern may differ between ethnic groups, we explore linkage disequilibrium and conducted haplotype analysis within Caucasian only (sample size of any other ethnic group was too small). Within each haplotype block, haplotype frequency and ORs and 95% CI were estimated using HaploStats (version 3.32), assuming an additive model based on the generalized linear model (45, 46). Haplotypes were estimated using expectation-maximization algorithm and overall difference between cases and controls was evaluated using the global score test statistic adjusted for age, sex, and screening center (46). We used the most common haplotype as reference.

Advanced adenoma cases were slightly older at the time of colorectal cancer screening than controls (average, 63.1 and 61.8 years, respectively; Table 1). The distributions by gender (∼31% were female) and ethnic origin (∼94% were Caucasian) were similar for cases and controls because we matched on these two variables. Cases were less likely to have graduated from college and were more likely to be smokers, to be obese, and to have a family history of colorectal cancer. Serum selenium concentrations were lower in cases than controls.

Individual analysis of polymorphisms in the GPX genes revealed a borderline nonsignificant association between T variant located in the 3 region (273 bp 3′ of STP) of GPX4 and advanced colorectal adenoma (P_trend = 0.07; Table 2), which was significant when restricted to Caucasians (P_trend = 0.04). Haplotype analysis provided similar results: those carrying the haplotype with the rare allele at 273 bp 3′ of STP were associated with a borderline significant inverse adenoma risk (OR, 0.81; 95% CI, 0.67-0.99; P = 0.04; Table 3). Adjusting for all other polymorphisms in GPX4 had little effect on 273 bp 3′ of STP. No associations with adenoma risk were observed for any of the other polymorphisms in the four GPX genes and the global test for association was not significant for any of the four GPX genes (Tables 2 and 3).

Within SEPP1, four SNPs were significantly associated with risk of advanced colorectal adenoma. In the promoter region, the variant SEPP1 -4166G was present in nine cases (of which eight were Caucasian) but none of the control subjects (Fisher’s exact test P = 0.002; Table 2). Furthermore, significant associations were observed between colorectal adenoma risk and variants at two loci in the 3 region of SEPP1 (31,174 bp 3′ of STP and 43,881 bp 3′ of STP) with an OR of 1.48 (95% CI, 1.00-2.19) for GG versus AA and OR of 1.53 (95% CI, 1.05-2.22) for AA versus GG, respectively. Results were very similar for Caucasians only (data not shown). In addition, the variant at a third loci in the 3 region of SEPP1 (44,321 bp 3′ of STP) was inversely associated with risk of colorectal adenoma (OR for CT versus CC, 0.73; 95% CI, 0.57-0.92), which remained the same when restricted to Caucasian (data not shown). SEPP1 31,174 bp 3′ of STP and 43,881 bp 3′ of STP were strongly correlated (r² = 0.90) among Caucasians, so that it was not possible to differentiate the effects of one polymorphism from the other statistically. The correlation of these two SNPs with the third significant SNP in the 3 region (44,321 bp 3′ of STP) was moderate (r² ≤ 0.37). Haplotype analysis containing the three significant polymorphisms in the 3 region did not reveal significant results, which might be expected because we observed association for two SNPs (31,174 bp 3′ of STP and 43,881 bp 3′ of STP) under a recessive model, whereas the haplotype analysis assumes an additive genetic model. Adjustment for the other SEPP1 polymorphisms did not significantly alter the ORs observed for any of the four SEPP1 SNPs. The global P value testing for an overall association within the gene was significant (global P = 0.02).

We also found a significant global test for TXNRD1 (global P = 0.008) due to a strong association for IVS1-181C>G: carriers of the variant allele had a statistically significant 80% reduction in advanced colorectal adenoma risk (OR for GG+CG versus CC, 0.20; 95% CI, 0.07-0.55), with a strong linear trend (P = 0.004; Table 2). Results for Caucasian only were very similar (data not shown). This polymorphism was not correlated (r² < 0.01) with any other genotyped SNPs in this gene. Adjustment for the other TXNRD1 polymorphisms did not significantly alter the ORs observed for TXNRD1 IVS1-181. No other genotype or haplotype in TXNRD1 was significantly associated with adenoma risk.

As adenoma subtypes differ in their tendency to undergo malignant transformation, we assessed differences by histologic characteristics (villous, tubular villous, and tubular) and number of tumors (single and multiple adenoma); however, we did not observe any meaningful differences. Furthermore, associations between tagSNPs in any of the selenoenzymes and advanced adenoma risk did not vary by serum selenium concentrations, age, or smoking: the number of observed interactions significant at P < 0.05 (n = 7) did not exceed the expected number of interactions significant at P < 0.05 by chance alone given the total number of tests (n = 132; expected number: 0.05 × 132 = 6.6). Of these seven interactions with P < 0.05, we observed two with serum selenium (GPX1 2,616 bp 3′ of STP, P = 0.04; RAB15 3,306 bp 3′ of STP, P = 0.01), three with age (GPX1 2,616 bp 3′ of STP, P = 0.04; GPX3 IVS1+1494, P = 0.04; GPX4 273 bp 3′ of STP, P = 0.02), and two with smoking (GPX1 2,616 bp 3′ of STP, P = 0.03; GPX3 IVS1-1961, P = 0.02).

Selenoenzymes have antioxidative properties important to prevent oxidative stress, which can lead to DNA damage and neoplastic progression. Several selenoenzymes are expressed in the large intestine, which is exposed to microbial- and food-borne oxidative stress (19-23). Furthermore, selenoenzymes may reduce inflammatory processes. Despite the biological rational for an effect of these genes on colorectal cancer, this hypothesis was previously only investigated to a very limited extent. In our comprehensive analysis of the common variation in six selenoenzymes, genetic variants in SEPP1 and TXNRD1 were significantly associated with advanced colorectal adenoma risk after adjusting for multiple comparisons. Furthermore, our findings may suggest an association with advanced adenomas and a variant in the 3 region of GPX4. However, this SNP did not result in an overall significant association of the GPX4 gene as observed for SEPP1 and TXNRD1.

SEPP1 contains 10 selenocysteines and is the only selenoprotein with more than one selenocysteine. SEPP1 binds ∼40% to 60% of the selenium in plasma and is found in several tissues, including the colon, where it becomes down-regulated during tumor development (26, 47-49). SEPP1 has three important functions: transport of selenium to target tissues, intracellular selenium binding and supply, and protection against oxidative stress (47, 50-54). Our finding of an association with a SNP in the promoter region may point toward an effect on gene expression. Interestingly, the promoter is cytokine responsive, suggesting a role of SEPP1 on inflammation, a risk factor for colorectal cancer (55, 56). Another group (48, 57) examined SEPP1, suggesting a potential effect of a complex repeat structure in the promoter region; however, the results are difficult to interpret due to limited information on the study population and controls were selected from different sources than cases. As our study was limited to SNPs, we have no data on this complex repeat polymorphism.

We further found significant associations for three polymorphisms in the 3 region of SEPP1. We note that one of these three SNPs (44,321 bp 3′ of STP) was out of Hardy-Weinberg equilibrium (P = 0.03), which may explain the significant association for this one SNP. All three SNPs are located in the promoter region of an uncharacterized antisense transcript overlapping SEPP1 (BC039102, we are calling it asSEPP137). The function of asSEPP1 is unknown, but this unique genomic organization may be clinically relevant because some antisense transcripts regulate the expression of the overlapping gene post-transcriptionally; furthermore, the expression of SEPP1 is impaired in some prostate tumors (58). Interestingly, the SNP 44,321 bp 3′ of STP is predicted to disrupt a putative STAF transcription factor binding site of asSEPP1 based on MatInspector (59). STAF is a zinc finger protein originally identified as activating the RNA polymerase III promoter of the selenocysteine tRNA gene (60). Altogether, these data suggest that STAF and asSEPP1 might be part of a complex regulatory mechanism regulating the production of SEPP1. Additional studies will be necessary to define the functional importance of the identified variants.

TXNRD1 is involved in antioxidant defense as part of the thioredoxin system, which exists in nearly all cells (61). Its high reactivity is attributed to selenocysteine (62). TXNRD1 is also involved in the regulation of transcription factors, protein-DNA interaction, and growth control and it reduces nucleotides for DNA synthesis (61, 63-65) TXNRD1 is highly expressed in various tumor tissues likely in direct response to stress (64-66). We observed a highly significant association for advanced adenoma with TXNRD1 IVS1-181, which remained significant after adjusting for multiple comparisons within the gene. Only one other group investigated TXNRD1 in relation to breast cancer risk and survival and reported no association for any of the four genotyped tagSNPs, not including IVS1-181 (67-69). TXNRD1 IVS1-181 is located -30 bp from the ATG of E1A-like inhibitor of differentiation 3 (EID3), a small gene nested within the intron of TXNRD1. Little is know about this gene, indicating that EID3 inhibits transcription and blocks cellular differentiation of cultured muscle cells. Importantly, the C-to-G polymorphism at IVS1-181 is predicted to change a SRY to NkX-2 transcription factor binding site identified via MatInspector (59). Accordingly, this SNP may not indicate an association with TXNRD1 but instead with EID3. Replication and functional studies are needed to provide additional evidence.

We observed a borderline significantly reduced risk for one polymorphism in GPX4 (273 bp 3′ of STP) among Caucasians. This SNP is only ∼150 bp 3 of the selenocysteine insertion sequence, which is required for selenocysteine insertion into the active center of the protein, a highly complex mechanism (11). GPX4, a ubiquitously expressed enzyme, and compared with other GPX, has the unique ability to directly reduce destructive phospholipid hydroperoxide within the cell and membrane, thus promoting membrane integrity (70). Besides antioxidative properties, GPX4 is involved in regulating inflammatory processes (71); GPX4 reduces lipid hydroperoxides, which activate lipoxygenase and cyclooxygenase (72). Compared with the other isozymes of the GPX family, the resistance of GPX4 to selenium deprivation may suggest its vital role, which is further suggested by the finding that mice with disrupted GPX4 genes die in utero (73, 74). One group investigated the association between tagSNPs in GPX4 and breast cancer and reported a significant association for two SNP (Ex7+77 and GPX4 -1928) with survival but not risk of breast cancer (67-69). These two SNPs were not significantly associated with adenoma risk in our study; however, the nonsignificant associations were in the same direction seen for breast cancer survival. Based on these limited data, further follow-up studies may be warranted given the biological relevance of GPX4.

Besides the suggested association for one GPX4 polymorphism, we observed no significant association for any other SNP in the four GPX genes. The redundant expression of the GPX genes in the gastrointestinal tract may limit the effect of any single genetic variant on cancer risk and may explain the limited evidence for an association between the four GPX genes and colorectal adenoma risk. Similar GPX knockout studies showed that colon cancer risk only increased when both GPX1 and GPX2 were knocked out simultaneously but not if only one gene was (25, 34, 75-77).

The study design enabled us to randomly select cases and controls derived from the same source population and screened following a standardized procedure; in particular, cases were not screened based on symptoms. The large study population allowed us to confine the analysis to cases with advanced adenoma, which have a higher potential for malignant transformation and are a particularly meaningful intermediate outcome for studying factors related to colorectal cancer. Cases were identified from the initial baseline screening, so we cannot distinguish the effect of genetic variants on the temporal component of adenoma formation. We investigated potentially important interaction with selenium based on serum measurements, which are preferred over dietary assessment due to a large variation of selenium concentrations in the same foods. However, only a few participants had low selenium concentrations. It may be possible that specific variants in selenoenzymes only affect the function among individuals with selenium deficiency, which would result in limited power to investigate interactions with selenium. TagSNPs selection was based on resequencing data to allow a comprehensive analysis of the common genetic variation in all six genes. However, except for GPX1, we did not have the resources to resequence the entire gene, making it possible that we missed some of the genetic variation. Nonetheless, our resequencing was focused on the most functionally relevant regions of the genes (promoter, coding regions, intron, exon boundaries, and selenocysteine insertion sequence). Due to the use of sigmoidoscopy as screening tool, we focused our analysis on left-sided adenoma. Because differences in the associations with right-sided and left-sided colorectal neoplasia have been reported, the extrapolation of our findings with all adenomas (right-sided and left-sided) or right-sided adenoma should be done with caution.

In summary, we observed significant associations for advanced colorectal adenoma and genetic variants in SEPP1 and TXNRD1, which remained significant after adjusting for multiple comparisons. However, as this is the first study to comprehensively analyze the genetic variation in selenoenzymes and risk of colorectal adenoma, further studies are needed to replicate these interesting findings.

No potential conflicts of interest were disclosed.

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.