Risk Modification of Colorectal Adenoma by CYP7A1 Polymorphisms and the Role of Bile Acid Metabolism in Carcinogenesis

Bile acids, such as the primary bile acid chenodeoxycholic acid (CDCA) and secondary bile acid deoxycholic acid (DCA), are implicated in the etiology of colorectal cancer (CRC; ref. 1). Ecological (2–5) and case–control studies (6–8), including a meta-analysis (9), support a positive association between fecal bile acids and CRC. In addition, elevated DCA in serum has been positively associated with rectal mucosal proliferation (10, 11) and colorectal adenoma (CRA) in men (12–14). Conversely, ursodeoxycholic acid (UDCA), a low abundance bile acid in humans with potent antineoplastic properties (15–18) has been shown to protect against colorectal neoplasia (CRN), especially in ulcerative colitis patients (19, 20). Our own randomized, placebo-controlled chemoprevention trial showed that 3-year treatment with UDCA showed no overall effect on CRA; however, we detected activity against high-grade dysplasia (21), consistent with studies in ulcerative colitis patients.

To maintain cholesterol homeostasis, most intestinal bile acids are reabsorbed via an active transport mechanism in the terminal ileum. In a healthy state, less than 5% of bile acids pass into the colon (22). The pool size of the primary bile acids is subject to wide interindividual variability and is strongly influenced by the rate-limiting enzyme cholesterol 7α-hydroxylase (CYP7A1), which catalyzes the initial step in the classical pathway of bile acid synthesis in the liver (23). Evidence from population studies supports direct effects of specific CYP7A1 polymorphisms on the metabolic intermediates of enterohepatic circulation, with more limited evidence linking variation in CYP7A1 with CRC. Early studies of a common A to C substitution in the CYP7A1 promoter (–204 from the transcription start site, −278 from the translation start site; designated rs3808607) provided the first evidence for an association between allelic variation in CYP7A1 and plasma total and low density lipoprotein (LDL) cholesterol concentrations, with carriers of the C allele postulated to have reduced CYP7A1 activity (24, 25). Using the ratio of serum 7α-hydroxycholest-4-en-3-one (C4) to either total or nonhigh-density lipoprotein cholesterol to estimate CYP7A1 activity, Leníček and colleagues (26) showed significantly higher CYP7A1 activity in patients with the −204A allele than with C carriers who underwent resection of the distal ileum, a procedure that results in upregulation of CYP7A1. In contrast, no variant related differences were detected in healthy controls. In case–control studies, carriage of the −204CC genotype was associated with reduced risk of both CRC and CRA arising in the proximal colon, suggesting a protective gene effect resulting from reduced bile acid synthesis in CC carriers (27, 28). More recently, a genome-wide association study (GWAS) has provided confirmatory evidence linking allelic variation in CYP7A1 to total cholesterol and LDL (29), with findings from an independent study showing that such variation modifies individual response to the lipid-lowering effects of atorvastatin (30).

Not previously studied are the effects of CYP7A1 variants on fecal bile acid concentrations and the potential effect modification of CYP7A1 on UDCA given therapeutically for CRA prevention. UDCA has been shown to have antineoplastic properties, but its mechanism of action in vivo remains unclear, particularly as to whether or not the partial effects of UDCA are mediated through action on CYP7A1 levels and synthesis of bile acids in humans. In recent animal studies, feeding UDCA was associated with dramatic reductions in mRNA levels of Cyp7a1 in both male and female mice that were accompanied by dramatic changes in the hepatic bile acid composition. Similar studies in humans are not feasible; therefore, to gain insight into the role of CYP7A1 in potential mechanisms of UDCA activity, we investigated the impact of CYP7A1 polymorphisms on fecal bile acid levels, risk of metachronous CRA, and UDCA efficacy in our completed, randomized, controlled trial of UDCA for CRA prevention.

The UDCA chemoprevention trial was conducted at the Arizona Cancer Center (21). Briefly, this double-blinded, placebo-controlled, randomized trial tested the efficacy of UDCA to prevent metachronous CRA. A total of 1,285 participants were randomized to UDCA (n = 661) or placebo (n = 624), of which 1,192 (613 UDCA, 579 placebo) completed the trial. Participants with unknown baseline adenoma characteristics were excluded from the current analysis (n = 2, both in the UDCA arm). Of the remaining 1,190 patients, 703 (355 UDCA, 348 placebo) supplied DNA samples and had complete genotype data. Among these participants with known genotypes, 495 (242 UDCA, 253 placebo) supplied fecal samples and had measured baseline bile acid concentrations (Fig. 1).

A participant was classified as having an advanced baseline CRA if the qualifying colonoscopy detected multiple (≥3) adenomas or at least 1 adenoma with high-grade dysplasia, villous or tubulovillous histology, or a diameter of 1 cm or more. The primary UDCA trial endpoint of metachronous CRA was defined as any CRN detected at follow-up colonoscopies at least 6 months after randomization. The mean follow-up time for participants with complete genetic information was 35.4 ± 11.5 months. Information on CRN status was obtained from review of endoscopy and pathology reports that included abstracted data on size, histology, number, and anatomic location.

Seven single-nucleotide polymorphisms (SNP) in CYP7A1 (rs10957057, rs8192879, rs8192877, rs11786580, rs8192871, rs13251096, and rs181273) were selected for genotyping. We applied a bin-tagging approach to select these SNPs using available information on a European Caucasian population. Briefly, initial tag SNPs and linkage disequilibrium blocks were identified from HapMap data release 16c.1, June 2005, on National Center for Biotechnology Information B34 assembly, dbSNP b124. Tag SNPs were identified by the following criteria: minor allele frequency >5%, pairwise r² > 0.75, and 60 or more base pairs between neighboring SNPs to reduce assay failure. Genotyping was carried out under contract at BioServe. One of the 7 initial SNPs, rs1801273, showed no variation in our study population and was excluded from further analyses. The remaining 6 SNPs passed all quality checks and met Hardy–Weinberg equilibrium. At the time of tagging, these 6 SNPs captured 18 of 18 alleles at r² > 0.75; with more recent updates to HapMap, they capture 27 of 36 alleles at r² > 0.75. For the rs2081687 tag identified in the GWAS meta-analysis (29), rs8192871 was most strongly linked, with D′ = 0.946 and r² = 0.631. In addition, rs8192871 has high linkage (D′ = 1.0 and r² = 0.89) with the putative functional A-204C promoter variant, rs3808607. We also evaluated the linkage between the rs3808607 and rs2081687 and found D′ = 0.949 and r² = 0.721. Collectively, these results show that our approach captured similar information to that derived from the GWAS (29) and candidate studies (25, 26).

Samples that failed to generate genotype data for all 7 SNPs were excluded from our analyses. In sum, data for 6 polymorphic SNPs in CYP7A1 were available for 703 subjects. For these 6 SNPs, laboratory blinded replicates showed 98% or more concordance. In addition, CYP7A1 haplotypes for these 6 loci (rs10957057, rs8192879, rs8192877, rs11786580, rs8192871, and rs13251096) were generated with PHASE (version 2.1.1) software (31, 32).

Briefly, participants provided 72-hour pooled stool samples in containers that were transported on dry ice to the central laboratory at the University of Arizona. Samples were processed by homogenization with equal weight of water for 15 minutes. After ultracentrifugation, the aqueous phase (i.e., fecal water) was removed and stored in aliquots at −80°C until analyzed. Bile acid concentrations (μg/mL) were determined for lithocholic, DCA, CDCA, cholic, UDCA, ursocholic, and other (isodeoxycholic, isoursodeoxycholic, 7-ketolithocholic, and 12-ketolithocholic) acids by gas chromatography, as previously described (21, 33). Values of undetectable concentrations were set to zero. Each subject's total primary bile acid concentration was calculated by summing together the concentrations of CDCA and cholic acid, and the sum of the other 8 bile acids comprised the total secondary bile acid concentration. Primary and secondary bile acids were analyzed as binary outcomes, with the median level as the cut point between low versus high levels, to maximize statistical power by generating equally sized groups and to avoid the influence of extreme values. For primary bile acids, concentrations less than 7.8 μg/mL were considered low; likewise, secondary bile acid levels less than 189 μg/mL were considered low.

Patient characteristics at baseline were compared between the treatment arms with 2 sample t tests (age) and the Fisher exact tests (sex and genotypes) to ensure adequate randomization. Associations between genotypes, bile acids, and metachronous adenoma were tested in the placebo group only to eliminate any modifying effect of UDCA treatment. Bile acid levels were compared between genotypes by Wilcoxon rank-sum tests. An OR and 95% CI was calculated to test the association between each SNP and bile acid levels using logistic regression with a codominant model because we had no prior assumptions about the mode of inheritance for each SNP. To assess the independent effects of each SNP and bile acid levels on CRA in the postpolypectomy subject, logistic regression models were adjusted for age (continuous), gender (male/female), baseline advanced adenoma status (yes/no), and follow-up time (continuous) as known risk factors for metachronous adenoma in this population. Associations between each SNP and metachronous adenoma were tested using similar adjusted logistic regression models.

Of the 17 total CYP7A1 haplotypes created, 6 (each with frequencies >5%) were deemed “major” haplotypes and included in our analyses. These major haplotypes were ranked in order of frequency and numbered accordingly as haplotypes 1 to 6. The most frequent haplotype (CTACAG, 39.3%) was designated as the reference group for testing associations between haplotype and bile acids or metachronous adenoma in the placebo arm.

Potential modification of UDCA efficacy was examined for bile acid level (low versus high), genotype, and haplotype using stratified analyses. Statistical interactions between bile acids or genotypes/haplotypes and UDCA were tested by likelihood ratio tests. All statistical analyses were conducted using Stata 11.2 (StataCorp), and all statistical tests were 2-sided.

UDCA trial participants with complete genotype data were 68.2% male with a mean age of 66.4 ± 8.2 years, and 367 patients (52.2%) had advanced adenoma status at baseline. There were no significant differences in age, sex, or genotypes between the 2 treatment arms (Table 1). Baseline adenoma characteristics, including size, anatomic location, multiplicity, and advanced status, did not significantly differ by individual genotype with only one exception: carriers of 2 copies of the minor variant for rs10957057 were more likely to have proximal adenoma than participants with 2 copies of the major allele for this locus (Supplementary Table S1). Furthermore, the characteristics of individuals with genetic data did not differ substantially from those without known genotypes (data not shown).

The main effects of individual SNPs on bile acid levels (categorized as high or low, with equally sized groups, to maximize statistical power) and metachronous adenoma were tested assuming codominant inheritance, adjusted for age, gender, baseline advanced adenoma status, and follow-up time. Two SNPs (rs8192877 and rs13251096) were significantly associated with primary bile acid levels, and a third (rs8192871) was significantly associated with secondary bile acid levels (Table 2). Carriers of the GG genotype at rs8192871 had 74% reduced odds for high (i.e., >189 μg/mL) secondary bile acid levels (OR = 0.26, 95% CI: 0.10–0.69) compared with carriers of the AA genotype; actual unadjusted levels (mean ± SD) in GG versus AA patients were 158.6 ± 108.8 versus 235.5 ± 164.2 μg/mL, respectively (Wilcoxon rank-sum test, P = 0.016). Furthermore, among the 6 CYP7A1 variants, only rs8192871 showed a significant association with CRA development, with GG carriers (i.e., the genotype with lower bile acid levels) showing 59% lower odds of metachronous adenoma (OR = 0.41, 95% CI: 0.19–0.89) than carriers of the AA genotype.

In light of the suggested SNP-specific effects on bile acid concentrations and risk of CRA, we next investigated the effects of CYP7A1 SNPs as they occur in the population as haplotypes (i.e., SNP–SNP interactions). We constructed 6 CYP7A1 haplotypes that occur at more than 5% frequency in this study population (see Methods). Analysis of these haplotypes (order of loci: rs10957057, rs8192879, rs8192877, rs11786580, rs8192871, and rs13251096) revealed a significant association between allele structure and primary bile acid levels. Participants with haplotype 3 (CCGTAG) were 2.34 times more likely to have high (i.e., >7.8 μg/mL) primary bile acid concentrations than those with the common haplotype 1 (CTACAG; Table 3). Also, individuals with this same haplotype 3 (CCGTAG) showed 89% increased odds of developing metachronous CRA, compared with haplotype 1 (CTACAG).

Therapeutic UDCA has been postulated to influence bile acid composition and to regulate the expression of CYP7A1. Thus, we next assessed whether bile acid levels or genetic variation in CYP7A1 modified the chemopreventive activity of UDCA for metachronous CRA in the subset of trial participants with both genotype and bile acid measures. As previously shown in the parent study (21), the overall main effect of UDCA treatment was nonsignificantly beneficial for preventing metachronous CRA: OR = 0.90, 95% CI: 0.67–1.22 (Table 4). The chemopreventive effect was greater in patients with low primary (OR = 0.62, 95% CI: 0.37–1.04) or low secondary (OR = 0.67, 95% CI: 0.40–1.12) bile acid levels, but these effects were not statistically significant (likelihood ratio tests for UDCA–bile acid interactions, P = 0.085 and 0.220 for primary and secondary bile acids, respectively).

In contrast, 3 of 6 CYP7A1 SNPs showed significant statistical interaction with UDCA treatment on metachronous CRA (Table 5). Subjects with the AA genotype at rs8192871 or the GG genotype at rs13251096 received significant benefit from UDCA treatment (OR = 0.50, 95% CI: 0.30–0.83; and OR = 0.57, 95% CI: 0.35–0.94, respectively), but patients with one or 2 minor alleles at either locus achieved no benefit (OR = 1.26, 95% CI: 0.86–1.85; and OR = 1.18, 95% CI: 0.80–1.72; respectively). Both alleles showed significant genotype–UDCA interaction (likelihood ratio test, P = 0.004 and 0.023, respectively). Furthermore, minor allele carriers of rs8192877 achieved benefit from UDCA treatment (OR = 0.53, 95% CI: 0.30–0.96), whereas carriers of 2 major alleles of this polymorphism did not (OR = 1.09, 95% CI: 0.76–1.55; likelihood ratio test for genotype–UDCA interaction, P = 0.042).

Consistent with the SNP level effects, carriers of CYP7A1 haplotype 3 (CCGTAG), who experienced the highest recurrence rate in the placebo group (61.2%), obtained the greatest UDCA benefit, with the recurrence rate reduced dramatically to 31.6% (Table 6). Thus, UDCA treatment was highly efficacious in this group (OR = 0.29, 95% CI: 0.13–0.65) but not in any of the other haplotypes (likelihood ratio test for haplotype–UDCA interaction, P = 0.020).

We previously reported no main effect of UDCA for CRA prevention, with marginally significant effects for the prevention of high-grade dysplasia (21). Here, we show that allelic variation in CYP7A1, which has been shown in mice to be a gene target for repression by UDCA (34), modified response to UDCA treatment. These results support similar findings showing that CYP7A1 alleles modify the effectiveness of atorvastatin for lowering blood lipids (35).

A considerable body of evidence, including recent findings from GWAS, links genetic variation in CYP7A1, particularly the A-204C promoter variant (rs3808607), with several intermediate metabolites of cholesterol metabolism, including circulating levels of total cholesterol, LDL, and triglycerides (35). In subjects who received placebo in our UDCA chemoprevention trial, we show a significant main effect of CYP7A1 allelic variation on primary and secondary fecal bile acid concentrations as well as increased odds of CRA, even after controlling for known CRA risk factors (i.e., adenoma characteristics, age, and sex). Our results corroborate the findings of Tabata and colleagues (28) and Hagiwara and colleagues (27), who reported an association between the CYP7A1 promoter variant and risk of CRA and CRC, respectively. Collectively, these results strengthen the link between genetic variation in CYP7A1 and CRC risk and provide additional support favoring the role of bile acid metabolism in the biology of colorectal carcinogenesis.

We propose 2 explanations for our findings. UDCA has been postulated to repress CYP7A1 through direct effects on the farnesoid X receptor and promoter inhibition. Our data suggest that the low activity variant may be resistant to this signaling effect and therefore nonresponsive. Alternatively, but not mutually exclusively, individuals at risk for CRN through the bile acid pathway, because of CYP7A1 high activity conversion of cholesterol to bile acids, may be more likely to develop their neoplasia through this pathway and are therefore amenable to a prevention strategy that targets this particular mechanism.

Although the study has a number of strengths, this retrospective, secondary analysis, conducted on a subset of the trial participants, may be prone to false discovery. Second, dichotomization of the bile acid variables (using the median value as the cut point, to yield equally sized groups) may have resulted in underestimating the magnitude of the association between the genotypes and bile acid levels. Furthermore, the uniqueness of the sample set, which includes only polyp formers, limits the generalizability of our results. Despite these limitations, the overall findings show consistency with biological mechanisms (CYP7A1 variation as a determinant of bile acid levels and CYP7A1 gene regulation as a putative UDCA target) and, more importantly, replicate prior associations between genetic variation in CYP7A1 and risk of CRA development (27, 28).

Our results suggest that targeting the bile acid pathway with UDCA was efficacious for CRA prevention in a subset of participants with high activity CYP7A1 alleles and, possibly, in individuals with lower baseline bile acid levels. These results lend support for the bile acid hypothesis of carcinogenesis in the colorectum, a highly modifiable risk factor. Importantly, our findings show the significant role that genetic variation plays in modifying intervention responsiveness, which, in the parent trial, undermined our ability to adequately test UDCA's efficacy in targeting bile acid-induced carcinogenesis. This issue extends across the chemoprevention field, with similar examples reported for SEP15 and selenium (36), ODC1 and difluoromethylornithine (37), FMO3 and sulindac (38), and ODC and aspirin (39). We believe that consideration and integration of knowledge on biological mechanisms and their intermediates, along with genetic information from GWAS, are needed to improve trial design and are critical for the success of future chemoprevention and treatment efforts. Our results also suggest the need to either tailor the selection of agents by patient genetic/phenotypic background or develop and screen pathway-specific prevention agents with broader efficacy in the human population by integrating biomarker endpoints in early-phase prevention trials.

No potential conflicts of interest were disclosed.

This work was supported by Colon Cancer Prevention Program Project grant (National Cancer Institute/NIH PO1 CA4108), Cancer Center Core Service grant (CA023074), and NIH RO1 CA 135300.

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.