Genetic variation in genes for the xenobiotic-metabolizing enzymes CYP1A1, EPHX1, GSTM1, GSTT1 and GSTP1 and susceptibility to colorectal cancer in Lynch syndrome

Abstract

Individuals with Lynch syndrome are predisposed to cancer due to an inherited DNA mismatch repair gene mutation. However, there is significant variability observed in disease expression, likely due to the influence of other environmental, lifestyle, or genetic factors. Polymorphisms in genes encoding xenobiotic-metabolizing enzymes may modify cancer risk by influencing the metabolism and clearance of potential carcinogens from the body. In this retrospective analysis, we examined key candidate gene polymorphisms in CYP1A1, EPHX1, GSTT1, GSTM1, and GSTP1 as modifiers of age at onset of colorectal cancer among 257 individuals with Lynch syndrome. We found that subjects heterozygous for CYP1A1 I462V (c.1384A>G) developed colorectal cancer 4 years earlier than those with the homozygous wild-type genotype (median ages 39 and 43 years, respectively; log-rank test P = 0.018). Furthermore, being heterozygous for the CYP1A1 polymorphisms, I462V and Msp1 (g.6235T>C), was associated with an increased risk for developing colorectal cancer [adjusted hazard ratio for AG relative to AA = 1.78, 95% CI = 1.16–2.74, P = 0.008; and hazard ratio for TC relative to TT = 1.53, 95% CI = 1.06–2.22, P = 0.02]. Since homozygous variants for both CYP1A1 polymorphisms were rare, risk estimates were imprecise. None of the other gene polymorphisms examined were associated with an earlier onset age for colorectal cancer. Our results suggest that the I462V and Msp1 polymorphisms in CYP1A1 may be an additional susceptibility factor for disease expression in Lynch syndrome since they modify the age of colorectal cancer onset by up to 4 years.

Introduction

Lynch syndrome, commonly known as hereditary non-polyposis colorectal cancer (HNPCC), is caused by an inherited pathogenic germline mutation in one of the DNA mismatch repair (MMR) genes, MLH1, MSH2, MSH6, or PMS2 (1). The mutations have an autosomal dominant pattern of inheritance and result in deficient mismatch repair, predisposing individuals with these mutations to early onset of cancers of the colon, the endometrium, and, less frequently, the stomach, ovaries, small intestine, biliary and uroepithelial tracts, skin, and brain (2).

Wide variation in disease expression both within and between families, particularly in age at onset (3), has been observed among individuals with Lynch syndrome, suggesting that other genetic and environmental factors may modify the effect of the inherited single-gene mutations. Evidence for the role of other genetic factors may also lie in the fact that 25–30% of colorectal cancer (CRC) cases that occur annually are considered familial (4) . Of these cases, about 5–14% are attributable to known inherited deleterious mutations (5). The remaining familial component of incident CRCs is likely due to the effect of variation in common low-penetrance genes or “modifier” genes (4). Therefore, it is of interest to examine low-penetrance genes as potential modifiers of risk for cancer onset in individuals with Lynch syndrome.

Cancer risk resulting from human exposure to exogenous chemicals (xenobiotics), like polycyclic aromatic hydrocarbons (PAH), that are ubiquitous environmental, dietary, and tobacco carcinogens may vary according to the ability to clear the xenobiotics from the body. Polymorphisms in the genes that encode enzymes involved in the metabolism of PAHs (that is, xenobiotic-metabolizing enzymes) such as the cytochrome P450 group (CYPs), the microsomal epoxide hydrolase group (mEH or EPHX) and the glutathione-S-transferase group (GSTs), result in varying activity levels of these enzymes, which can then influence xenobiotic clearance. The metabolism of PAHs involves both activation (phase I) and detoxification (phase II) reactions by these enzymes. During activation, reactive intermediates are formed that can bind to DNA and result in adducts that cause mutations if not repaired, thereby initiating carcinogenesis (6). It is likely that the expression and activity levels of the xenobiotic-metabolizing enzymes determine the relative level of activation and detoxification of carcinogens. These levels are important because increased levels of activation, decreased detoxification, or both, may increase cancer risk. Therefore, we hypothesize that the variation in enzyme activity due to polymorphisms in metabolic genes may explain some of the differences seen in disease risk for CRC in individuals with Lynch syndrome. We identified several candidate metabolic genes that display variation in activity levels of their expressed enzymes and selected common polymorphisms in these genes (i.e., with a minor allele frequency > 5%), to test our hypothesis.

The CYP group of enzymes is involved in the oxidation of many xenobiotics. CYP1A1 encodes the principal enzyme that metabolizes PAHs. In the phase I reaction, PAHs are metabolically activated, which may generate highly reactive mutagenic metabolites. Two common single nucleotide polymorphisms (SNPs) in CYP1A1, a non-coding Msp1 polymorphism in the 3′ untranslated region (nucleotide T to C) and an exon 7 polymorphism Ile462Val, (nucleotide A to G), which are in linkage disequilibrium (LD) (7), have been commonly examined for cancer risk. The functional significance of the Msp1 polymorphism is uncertain but the I462V polymorphism results in higher enzyme activity compared to the homozygous wild-type genotype and the enzyme is expressed in the colon (8). De Jong et al. (9) found little evidence that polymorphic variants of CYP1A1 influence the risk for sporadic CRC; however, Slattery et al. (10) have suggested that these variant alleles increase the risk for sporadic CRC among smokers. Similarly, Talseth et al. (11) found an association of the CYP1A1 Msp1 variant with an increased CRC risk in Lynch syndrome.

EPHX1 plays an important role in both the activation and detoxification of PAHs. Two EPHX1 polymorphisms, Tyr113His and His139Arg, result in reduced and increased enzyme activity, respectively. Gsur et al. (12) demonstrated that genetically reduced mEH activity may be protective against lung cancer, but these same EPHX1 polymorphisms have not been implicated in risk for sporadic CRC (13;14). We were unable to find any prior studies evaluating their role in the risk for CRC in individuals with Lynch syndrome.

GSTs are a superfamily of proteins that perform the phase II detoxification reactions of PAHs and other xenobiotics. GSTs catalyze the conjugation of reduced glutathione to a variety of potentially carcinogenic electrophilic and hydrophobic compounds. This detoxification reaction inactivates the compounds and renders them water-soluble so they can be readily excreted through urine or bile (15). GSTs include glutathione S-transferase theta1, mu1, and pi1 (GSTT1, GSTM1, and GSTP1). Null alleles exist as a common polymorphism for the GSTT1 and GSTM1 genes. Total or partial deletion of these genes results in no enzyme being produced (16;17). An overall increased risk for sporadic CRC has been described for the GSTT1 null allele, in a report by de Jong et al. based on 11 studies (9) but their report was inconclusive for the association, or lack thereof, between CRC and GSTM1 or GSTP1. In a recent study examining Lynch syndrome individuals, Felix et al. (18) reported that males with the GSTM1 and GSTT1 null genotype were at a 3-fold increased risk for an earlier age at onset of CRC compared to those with no deletion of the genes. Another study reported a 6-year shift in the median age at onset among 150 MLH1 mutation carriers with an earlier onset age associated with the null alleles of both GSTM1 and GSTT1 (19). However, two earlier studies found no influence on age at CRC onset for either GSTM1 (20;21) or GSTT1 polymorphisms (21). In our study, we report on some of the participants previously examined by Jones et al. (20), but our study includes substantially more subjects.

Although many of the genes encoding xenobiotic-metabolizing enzymes have been examined for their influence on the age at onset for CRC in Lynch syndrome, the results have not been consistent. This inconsistency might be due to the limited sample size or heterogeneous populations used in previous studies. A major strength of this study is the larger, predominantly white study population, consisting of a cohort of known carriers of pathogenic MMR gene mutations from many different families. Therefore, we are reporting on a population that is a unique resource for examining possible variation in disease expression associated with metabolic genes in this high-risk group. Specifically, we assessed the association between common polymorphisms in genes involved in xenobiotic-metabolism—CYP1A1 (rs4646903:g.6235T>C and rs1048903:c.1384A>G p.I462V), EPHX1 (rs1051740:c.339T>C p.Y113H and rs2234922:c.418A>G p.H139R), GSTM1 (deletion), GSTT1 (deletion), and GSTP1 (rs1695:c.330A>G p.I105V and rs1138272:c.343C>T p.A114V)—and age at onset of CRC to determine if genetic variation in any of these candidate metabolic genes modifies the age at onset of CRC in individuals with Lynch syndrome.

Materials and methods

Study population

The study population, which consisted of a cohort of individuals with Lynch syndrome has been described previously (22;23). Briefly, most Lynch syndrome probands were recruited from The University of Texas M. D. Anderson Cancer Center’s gastrointestinal and other clinics from September 1994 to July 2007. Through the probands (n=119), other first-degree relatives (n=74) and more distantly related family members (n=64) carrying the inherited mutation were recruited to the study. Study participants were from 130 families. The size of the families varied between 1 and 10 members of which 80 (31%) were singletons, 150 (58%) were between 2-7 members per family and the remaining were single families with 8, 9 and 10 members each. Criteria for mutation testing for Lynch syndrome were young age at onset of CRC (< 45 years), presence of a strong family history (met Amsterdam or relaxed Amsterdam criteria (24)), and suggestive tumor characteristics (defined as presence of microsatellite instability and loss of staining for an MMR protein). A total of 260 confirmed carriers of a MMR gene mutation were recruited. We excluded 3 subjects with mutations in MSH6 because MSH6 is associated with a variant form of Lynch syndrome with later age at CRC onset (25). We also did not include any individuals who tested positive for variants of unknown significance in the MMR genes since the pathogenicity of these mutations is not known. Among the remaining 257 study participants, there were 81 different MMR mutations which included deletions (25.7%), insertions (4.3%), nonsense (20.6%), splice site (27.2%) and missense mutations (22.2%). We confirmed that all of these mutations were pathogenic, particularly the missense mutations. They were either reported to be pathogenic by a Clinical Laboratory Improvement Act certified laboratory or confirmed to be pathogenic from the International Collborative Group-HNPCC InSight database (http://www.insight-group.org/) or from the published literature (26-29). All participants provided informed consent, and this retrospective study was approved by the Institutional Review Board.

For most participants (65%), information about demographic characteristics and medical history was obtained from a self-administered health, habits, and history questionnaire. Where questionnaire data were not available (i.e., when individuals were deceased or lost to follow-up), information was abstracted from the subjects’ medical records. CRC date of diagnosis was confirmed for the cases by review of pathology reports. Dates of diagnosis for adenoma or other cancers were similarly confirmed from medical records, pathology reports, or both.

Genotyping

All participants provided 10-ml samples of blood from which DNA was extracted using the AUTOPURE LS Automated DNA Purification Instrument (Gentra Systems Inc., Minneapolis, MN) according to the manufacturer’s instructions. Genotyping of SNPs, with the exception of GSTM1 and GSTT1 deletions, was done using the polymerase chain reaction (PCR) and pyrosequencing methods described by Chen et al. (22). The primers used in these procedures are listed in supplementary Table 1 available at the journal website online. The PCR reaction mixture was initially incubated at 95°C for 6 min, followed by 45 cycles at 95°C for 15 s, 59°C for 30 s for both CYP1A1 SNPs, 59°C for 30 s for both EPHX1 SNPs, 59°C for 30 s for the GSTP1 A114V SNP, 61.5°C for 30 s for the GSTP1 I105V SNP and 56°C for 30 s for GSTM1/GSTT1, followed by 72°C for 15 s, and then an extension of 72°C for 5 min. A multiplex PCR method was used for the GSTM1 and GSTT1 deletions, and the genotypes were ascertained on a 1.2% agarose gel by examining the gels for bands of the appropriate sizes, as described by Abdel-Rahman et al. (30). Positive and negative controls were included and 5% of the samples were run in duplicate for each genotyping assay with 100% concordance.

Statistical analysis

We estimated the genotype frequencies for each of the candidate metabolic gene polymorphisms and tested them for Hardy-Weinberg equilibrium. We used the Kaplan-Meier product limit method to assess the probability that subjects with a particular genotype would remain free of CRC. Thus, we defined CRC as the failure event; data for all other subjects were censored on the date of last contact for those remaining free of CRC, on the date of diagnosis of adenoma or cancer other than CRC, or on the date of death due to other causes. Total analysis time at risk for all participants consisted of the period from the date of birth to the time of the first CRC event for cases, or the date of censoring for the rest. The disease-free survival curves by genotype were compared using the log-rank (LR) test. The median age at onset was defined as the age at which 50% of the participants remained cancer-free. We performed Cox proportional hazards regression analysis to estimate hazard ratios (HR) and 95% confidence intervals (95% CI) for risk of CRC at any age, comparing the polymorphic to the wild-type genotype. Gender, ethnicity (white or other), and MMR gene mutated (MLH1 or MSH2) were included in the Cox model as potential confounding factors. We obtained unadjusted and adjusted HRs for the main effect of each of the candidate metabolic genes. We further performed stratified analysis (using Kaplan-Meier plots and Cox regression) for the SNPs that had a significant main effect by levels of the other variables, such as gender, ethnicity, gene mutated, to examine whether the effect of the SNP varied within these variables. We also generated interaction terms for SNPs significantly associated with CRC risk with each of the other gene polymorphisms (dichotomized as homozygous wild-type = 0; heterozygous and homozygous variant = 1; or non-deleted = 0 and deleted = 1 for GSTM1 and GSTT1) and tested for multiplicative interaction in the Cox model by including each of the main effect terms and the term for interaction. We used the Huber-White robust variance correction as applied in STATA 8.0 (StataCorp LP, College Station, TX) to correct for any correlations in time to onset of CRC among family members (31;32). The robust variance estimator adjusts for within-cluster correlation (data not independent within groups, but independent across groups) appropriately correcting for within-family correlation of the age of onset. We tested whether the hazard ratios were proportional using the method described by Grambsch et al. (33).

Results

Of the 257 participants, 120 (46.7%) developed CRC as the first cancer. We found no differences in the CRC-free survival time or the median age at onset of CRC by gender, ethnicity, or MMR gene mutated (LR test P > 0.05 for each of the variables; Table 1). All of the SNPs analyzed were in Hardy-Weinberg equilibrium (exact P > 0.05; Table 2). The allele frequencies were close to those described in the literature (we compared allele frequencies in our patients with those reported for white populations since our study participants were predominantly non-Hispanic whites). We compared the genotypic frequencies for each of the gene SNPs between subjects with and without CRC, using Pearson’s χ² tests and found no difference in the frequencies of EPHX1 and GST genotypes (data not shown). For CYP1A1, the genotype frequencies for both the Msp1 (χ²₍₂₎ 6.23; P = 0.04) and I462V (χ²₍₂₎ 7.63; P = 0.02) SNPs were significantly different between the patients with and without CRC (however, the endpoint of our study was the time to onset of CRC by genotype).

Table I.

Participant characteristics, frequency of CRC events, and comparison of disease-free survival

	n (%)	No. of CRC events (failures)
	n = 257	Total = 120	Log-rank test*P
Gender
Male	112	60	0.19
Female	145	60
Ethnicity
White	215 (83.7)	100	0.9
Black	25 (9.7)	13
Hispanic	16 (6.2)	7
Asian	1 (0.4)	0
Non-White (Black, Hispanic, and Asian combined)	42 (16.3)	20	0.83
MMR gene mutated
MLH1	108 (42)	55	0.24
MSH2	149 (58)	65

Age† (years)		Mean (SD) Min/Max	Student’s t-test P
CRC +	120 (46.7)	41.9 (10.4) 20/70	0.34
CRC −	137 (53.3)	43.3 (12.4) 18/84

^*Test for equality of the Kaplan-Meier estimates of CRC-free survival by participant characteristics.

^†Age at diagnosis for the CRC-affected and age at censoring for the unaffected (censored at date of last contact or date of diagnosis of adenoma/other cancer/death).

Table 2.

Genotype frequencies, comparison of disease-free survival by genotype, and Cox proportional hazards analysis for the age-associated risk of CRC by genotype.

	HWE Exact P	n (%)	CRC events	Log-rank test P	Univariate HR (95% CI)	P	Adjusted*HR (95% CI)	P
		Total n = 257	120
CYP1A1 I462V
AA	0.15	222 (86.4)	99	0.018	Ref	-	Ref	-
AG		32 (12.4)	21		1.81 (1.19-2.75)	0.005	1.78 (1.16 - 2.74)	0.008
GG†		3 (1.2)	0		-	-	-	-
Any G (AG+GG)		35 (13.6)	21	0.036	1.63 (1.08 2.44)	0.02	1.62(1.06-2.45)	0.02
CYP1A1 Msp1
TT	0.62	187 (72.8)	83	0.06	Ref	-	Ref	-
TC		66 (25.7)	37		1.41 (1.01-1.97)	0.04	1.53(1.06 - 2.22)	0.02
CC†		4 (1.5)	0		-	-	-	-
Any C(TC+CC)		70 (27.2)	37	0.17	1.31 (0.93-1.84)	0.13	1.44 (0.99 - 2.09)	0.06
EPHX1 Y113H
TT	0.65	130 (50.6)	66	0.46	Ref	-	Ref	-
TC		108 (42.0)	46		0.9 (0.63-1.27)	0.55	0.88 (0.62-1.26)	0.49
CC		19 (7.4)	8		0.64 (0.27-1.51)	0.31	0.62 (0.25-1.51)	0.29
Any C (TC+CC)		127 (49.4)	54	0.37	0.85 (0.61-1.19)	0.34	0.83 (0.59-1.17)	0.29
EPHX1 H139R
AA	0.58	155 (60.3)	77	0.34	Ref	-	Ref	-
AG		92 (35.8)	37		0.77 (0.55-1.08)	0.13	0.74 (0.52-1.04)	0.09
GG		10 (3.9)	6		1.18 (0.40-3.49)	0.76	1.17 (0.4-3.4)	0.77
Any G (AG+GG)		102 (39.7)	43	0.26	0.81 (0.58-1.14)	0.22	0.78 (0.55-1.1)	0.16
GSTP1 I105V
AA	1	120 (46.7)	55	0.75	Ref	-	Ref	-
AG		112 (43.6)	54		1.13 (0.81-1.57)	0.48	1.11 (0.79-1.54)	0.54
GG		25 (9.7)	11		0.93 (0.48-1.79)	0.83	0.85 (0.44-1.65)	0.64
Any G (AG+GG)		137 (53.3)	65	0.64	1.09 (0.79-1.5)	0.78	1.06 (0.76-1.46)	0.73
GSTP1 A114V
CC	1	227 (88.3)	105	0.49	Ref	-	Ref	-
CT		30 (11.7)	15		0.83 (0.5-1.34)	0.44	0.75 (0.44-1.28)	0.3
GSTM1
+ (Gene present)	-	144 (56.0)	68	0.81	Ref	-	Ref	-
− (Null)		113 (44.0)	52		0.96 (0.66-1.38)	0.81	0.98 (0.67-1.43)	0.92
GSTT1
+ (Gene present)	-	198 (77.0)	95	0.79	Ref	-	Ref	-
− (Null)		59 (23.0)	25		1.06 (0.75-1.49)	0.73	1.05 (0.74-1.49)	0.78

^*Adjusted for gender, ethnicity (white or non-white), and gene mutated (MLH1 or MSH2 ).

^†Hazard ratios omitted as estimates are unreliable with small numbers. HWE = Hardy-Weinberg equilibrium.

Kaplan-Meier survival curves, comparing CRC-free survival by genotype, were significantly different for the CYP1A1 I462V SNP (comparing AA, AG, and GG, LR test P = 0.018; for AG and GG genotypes combined, LR test P = 0.036) and marginally so for the CYP1A1 T>C SNP (comparing genotypes TT, TC, and CC, LR test P = 0.059) (Figure 1). CRC-free survival did not differ by genotype for any of the other polymorphisms or gene deletions (P > 0.05). The median age at onset for CRC was 39 years for patients with the CYP1A1 I462V AG genotype compared to 43 years for those with the wild-type AA genotype. Using Cox proportional hazards regression analysis, we also found an increased risk of early onset CRC associated with the AG genotype compared to the AA genotype (HR = 1.81, 95% CI = 1.19–2.75, P = 0.005). Because of the rarity of the CYP1A1 homozygous variant genotypes for both Msp1 and I462V, we did not analyze them separately (risk estimates were unstable and have been omitted) but did combine them with the heterozygous genotypes to calculate the risk estimates (Table 2). On stratified analysis for the effects of various CYP1A1 genotypes by gender and gene mutated on risk of early disease, we found that female gender and presence of the MSH2 mutation were significantly associated with the risk of early onset CRC (Table 3). Although the direction of association was the same among males and females, and among MLH1 and MSH2 mutation carriers, the magnitude of effect was smaller among males and MLH1 mutation carriers and the results did not achieve statistical significance, perhaps due to smaller numbers in those categories. Furthermore, since our study population was predominantly white, we analyzed the non-Hispanic whites versus others and found that a significant association persisted with an increased risk related to the AG genotype compared to the AA (HR = 1.73, 95% CI = 1.08–2.77, P = 0.021) among the non-Hispanic whites (Table 3). Even though gender, ethnicity, or gene mutated did not have significant effects at the α = 0.05 level in the Cox model, we retained them as covariates as per convention for their role as potential confounders of the modifier gene-CRC association. The crude and adjusted hazard ratios are listed in Table 2. We tested for and found no evidence of multiplicative interaction by gender, ethnicity, or gene mutated in the Cox model (Wald χ² test P > 0.05). All the estimates were obtained by applying robust correction using the cluster function in STATA.

Fig. 1.

Kaplan-Meier plots comparing the age at onset of CRC according to CYP1A1 I462V genotypes (A) AA, AG, and GG (B) AA and AG+GG; and CYP1A1 Msp1 T>C genotypes (C) TT, TC, and CC (D) TT and TC+CC.

Table 3.

Stratified analysis for CYP1A1 I462V and Msp1 by gender, gene mutated, and ethnicity.

	HR (95% CI)	P	HR (95% CI)	P
	Male		Female
CYP1A1 I462V
AA	Ref		Ref
AG	1.65 (0.88 - 3.10)	0.12	1.99 (1.12 - 3.53)	0.019
AG+GG	1.65 (0.88 - 3.10)		1.61 (0.91 - 2.86)	0.09
CYP1A1 Msp1
TT	Ref		Ref
TC	1.34 (0.75 - 2.4)	0.33	1.58 (0.99 - 2.5)	0.052
TC+CC	1.34 (0.74 - 2.41)	0.33	1.39 (0.87 – 2.24)	0.17
	MLH1		MSH2
CYP1A1 I462V
AA	Ref		Ref
AG	1.3 (0.82 - 2.03)	0.24	2.24 (1.22 - 4.12)	0.009
AG+GG	1.15 ( 0.79 - 1.81)	0.52	2.06 (1.14 - 3.72)	0.017
CYP1A1 Msp1
TT	Ref		Ref
TC	1.3 (0.82 - 2.07)	0.27	1.74 (1.12 - 2.69)	0.013
TC+CC	1.19 (0.74 - 1.91)	0.46	1.62 (1.04 - 2.54)	0.034
	White		Other
CYP1A1 I462V
AA	Ref		Ref
AG	1.73 (1.09 - 2.77)	0.021	2.19 (0.82 - 5.88)	0.12
AG+GG	1.53 (0.97 - 2.43)	0.066	2.11 (0.8 - 5.63)	0.13
CYP1A1 Msp1
TT	Ref		Ref
TC	1.5 (0.99 - 2.26)	0.052	1.04 (0.52 - 2.08)	0.91
TC+CC	1.41 (0.93 - 2.12)	0.1	0.93 (0.44 - 1.96)	0.85

In analyzing the CYP1A1 Msp1 SNP, which was in significant LD with the CYP1A1 I462V SNP (D’ = 0.97; R² = 0.44; P < 0.0001), we found that the TC genotype was associated with an increased hazard for earlier onset of CRC compared to the TT genotype (adjusted HR = 1.62, 95% CI = 1.06–2.45, P = 0.02) and that this was more evident among MSH2 mutation carriers, as in the case of the CYP1A1 I462V. The proportional hazards assumption was tested using the Schoenfeld residuals and was not violated for either of the CYP1A1 SNPs (P > 0.05 for the global test).

None of the other SNPs analyzed was associated with an earlier age at onset or a difference in risk by genotype. However, we found evidence for multiplicative interaction between CYP1A1 I462V and EPHX1 Y113H (Wald χ² P = 0.036; likelihood ratio test P = 0.044) with a greater than multiplicative hazard ratio for the combined effect of having a variant allele of both these SNPs (HR = 3.09, 95% CI = 1.58–6.04, P = 0.001). Further, on stratified analysis (Figure 2) we found that in the presence of any polymorphic allele of the EPHX1 Y113H SNP (genotypes TC and CC), having any polymorphic allele of CYP1A1 I462V (genotypes AG and GG) was associated with a significantly earlier median age of CRC onset at 37 years compared to 42 years (LR test P = 0.002) for having the CYP1A1 I462V homozygous wild-type allele (AA genotype) . We did not detect interaction of CYP1A1 I462V SNP with any of the other polymorphisms (P for interaction term > 0.05).

Fig. 2.

Kaplan-Meier plots for CYP1A1 I462V SNP comparing genotype AA with AG+GG, stratified by EPHX1 Y113H

Discussion

This study investigated eight polymorphisms in five candidate genes involved in the metabolism of xenobiotics to determine if they had any effect on the age at onset of CRC as an indicator of phenotypic variation in individuals with Lynch syndrome. The modifying effect of these genes was examined among individuals with a common background for increased susceptibility to CRC due to an inherited deleterious mutation resulting in deficient mismatch repair. Our most prominent finding was an observed shift in the median age at onset of CRC to 4 years earlier among CYP1A1 I462V heterozygotes compared to those with the homozygous wild-type genotype. In addition, subjects with the heterozygous genotype had an ~80% higher risk for CRC each year than subjects with the homozygous wild-type genotype. A similar trend in risk was observed for the CYP1A1 Msp1 variant, although the risk estimates were lower. Since the CYP group of enzymes is predominantly involved in activation reactions and the CYP1A1 I462V polymorphism results in higher enzyme activity, it is reasonable to assume that the increased CYP1A1 activity may be associated with an increase in the level of activated metabolites that have the potential to cause DNA damage and initiate carcinogenesis.

Comparing our results to evidence in the literature, we found that in a recent study, Talseth et al. (11) examined four of the same polymorphisms we did—CYP1A1 Msp1, GSTP1 I105V, and GSTM1 and GSTT1 deletions—in a mixed Australian (n = 86) and Polish (n = 134) population of MMR gene mutation carriers. The researchers did not find a statistically significant difference between the age at diagnosis of CRC by genotype for any of the genes tested, although they did detect a nonsignificant difference of up to 8 years in the median age at onset of CRC for three of the SNPs (authors did not specify which SNPs). Talseth et al. also found that only the CYP1A1 Msp1 mutant genotype was present significantly more in the CRC-affected subjects than in those without CRC (P = 0.03). This finding was confirmed by our results which showed a significant excess of Msp1 heterozygotes among the CRC affected (P = 0.04). Similarly, for the I462V SNP there were significantly more subjects with the heterozygous genotype among the CRC cases (P = 0.02) than among the cancer free. However, Talseth et al. did not examine the CYP1A1 I462V, which is the SNP that we also found to be most strongly associated with an earlier onset of CRC.

The GSTM1 and GSTT1 deletion polymorphisms have been extensively examined in individuals with Lynch syndrome (18-21). In a Finnish population of 150 MLH1 mutation carriers, the null alleles of GSTM1 and GSTT1 each shifted the median age of onset of CRC to 6 years earlier; the authors, however, did not report whether these differences were statistically significant (19). Another recent study examining a homogeneous cohort of 129 South African individuals carrying a single predisposing mutation in MLH1 reported that men who were carriers of both the GSTM1 and GSTT1 null alleles had a 3-fold increase in risk of developing CRC compared to men who had neither null allele (18). There are likely to be inherent genetic differences and differences in environment and lifestyle factors between the two study populations that could explain why we did not find similar results. Nevertheless, our null results for GSTM1 and GSTT1 do corroborate the findings of two previous studies on MMR gene mutation carriers, neither one of which found an association between GSTM1 and GSTT1 deletions and early onset of CRC (18;19). Of these, the study by Jones et al. was on a smaller subset of our population (i.e. 104 MMR gene mutation carriers from 59 families) and a similar analysis detected no association of GSTM1 with age at onset of colorectal cancer. (20). Though GSTM1 deletion is common, being present in almost 50% of the general population, the enzyme is not expressed in the colonic mucosal cells (34); hence it may play a very limited role in the metabolism of xenobiotics in the colon. This could perhaps explain why most studies have not found an association of GSTM1 with risk for sporadic or hereditary cases of CRC. The putative influence of GSTT1, however, is biologically plausible because this gene is abundantly expressed in the colon (34) and variation in GSTT1 enzyme levels could influence xenobiotic detoxification in the colon. Although our study did not find a modification of risk for CRC by GSTT1 in individuals with Lynch syndrome, compiled evidence appears to indicate that those with the gene deletion, that is, those lacking the GSTT1 enzyme, are likely to be at an increased risk for sporadic CRC (9;35;36). Our study could have detected a hazards ratio of ~1.6 for this polymorphism with 80% power at a significance level of 5%, but a larger sample size would be needed to detect less penetrant effects of this (or other) variants on risk.

One of the limitations of our study was a lack of complete data on different sources of exposure to PAHs, which include smoking and consumption of well-done meat. It is likely that the influence of the genetic polymorphisms is dependent on the level of xenobiotic exposure (i.e., the risk associated with the metabolic genes could be further modified by the level of the substrates upon which they act). As demonstrated by Slattery et al. (10), the impact of smoking on sporadic CRC risk is modified by the CYP1A1 genotype. Further, we had limited power to test the joint effects of the various genes or gene-gene interaction in our Cox model. Although we found that the adverse effect of the CYP1A1 I462V variant allele was further increased among individuals with the EPHX1 Y113H variant allele, showing a 5-year earlier shift in median age at onset and a 3-fold increased risk by year compared to having the CYP1A1 homozygous wild-type allele, these results are based on small numbers and would require validation in larger studies. It is pertinent to note that the interaction effect seen between EPHX1 and CYP1A1 is biologically plausible since both these enzymes are sequentially involved in the biotransformation of PAHs. In the initial phase I reaction, CYP1A1 converts benzo(a)pyrene, to the active benzo(a)pyrene 7,8 epoxide. This is hydrated by EPHX1 to a transhydrodiol derivative benzo(a)pyrene 7,8 diol that is less toxic (37), but the diol derivative is a primary substrate for CYP enzymes that further oxidize it to the highly reactive benzo(a)pyrene 7,8 dihydrodiol 9,10 epoxide (BPDE) – a potent mutagen capable of reacting with DNA to form adducts. Thus these and other genes may interact to play a more complex role in carcinogenesis. Future studies on larger numbers of mutation carriers using statistical tools such as Random Forest and CART analysis may be useful in determining the presence of gene-gene interactions and allow risk predictions to be made based on a combination of genotypes.

It is important to note that although we found an earlier age at CRC onset associated with two SNPs in the CYP1A1 gene that are in LD, it is possible that neither one of these is the causative SNP. The increased risk may be due to genetic variation at another locus that happens to be in further LD with the CYP1A1 SNPs examined in our study. Within CYP1A1, a reduced risk for sporadic CRC has been described for two other CYP1A1 SNPs, T461N and −1738A>C (13). The T461N SNP has a very low minor allele frequency (< 5%) and the functional relevance of the −1738A>C SNP is still unknown, but it would be of interest to examine these two CYP1A1 SNPs in future studies on individuals with Lynch syndrome.

Although we had limited power to examine the association of the SNPs with age at onset by different ethnic subgroups, our results were consistent whether we analyzed the overall population or restricted the analysis to non-Hispanic whites alone. Further, although we examined genetic variation at eight different loci, we do not believe that our results could be due to type I error as a consequence of multiple testing since the polymorphisms examined were in candidate metabolic genes that were selected a priori based on biological or epidemiological evidence for their role in affecting CRC risk. Besides, the univariate P value for CYP1A1 I462V (P=0.005) meets the stringent threshold P- value of <0.00625 (0.05/n=8 tests) on applying the Bonferroni correction for multiple testing.

There is a possibility that some degree of selection bias may have influenced the results of our study because hospital identified probands are likely to be younger than Lynch syndrome cases identified from the general population. This may be mitigated to some extent in our study by the inclusion of family members with Lynch syndrome, both CRC-affected and non-affected, to the analysis which perhaps restores the age at onset distribution closer to that in the population. Additionally, since the participants were not enrolled on the basis of their genotype, selection bias would not falsely indicate an association even though there may be an earlier shift in the age at onset.

Finally, the results of our study indicate that among MMR gene mutation carriers, the age at onset of CRC may be modified by the CYP1A1 I462V and Msp1 polymorphisms; the I462V polymorphism increasing risk for an earlier onset age by 4 years on average. Therefore, combining the knowledge of an individual’s CYP1A1 genotype along with other genetic markers and environmental factors that influence CRC risk in MMR gene mutation carriers may improve risk estimates and help identify genetically susceptible high-risk subgroups with Lynch syndrome that could benefit most from intensive screening and chemopreventive or surgical interventions.