The first two authors contributed equally to this work.
Haemochromatosis HFE gene polymorphisms as potential modifiers of hereditary nonpolyposis colorectal cancer risk and onset age
Article first published online: 3 FEB 2009
Copyright © 2009 UICC
International Journal of Cancer
Volume 125, Issue 1, pages 78–83, 1 July 2009
How to Cite
Shi, Z., Johnstone, D., Talseth-Palmer, B. A., Evans, T.-J., Spigelman, A. D., Groombridge, C., Milward, E. A., Olynyk, J. K., Suchy, J., Kurzawski, G., Lubinski, J. and Scott, R. J. (2009), Haemochromatosis HFE gene polymorphisms as potential modifiers of hereditary nonpolyposis colorectal cancer risk and onset age. Int. J. Cancer, 125: 78–83. doi: 10.1002/ijc.24304
- Issue published online: 27 APR 2009
- Article first published online: 3 FEB 2009
- Accepted manuscript online: 3 FEB 2009 12:00AM EST
- Manuscript Accepted: 22 DEC 2008
- Manuscript Received: 28 JUL 2008
- Hunter Medical Research Institute
- The University of Newcastle and the Pomeranian Medical University Szczecin
- colorectal cancer;
Hereditary nonpolyposis colorectal cancer (HNPCC) is characterized by germline mutations in DNA mismatch repair genes; however, variation in disease expression suggests that there are potential modifying factors. Polymorphisms of the HFE gene, which cause the iron overload disorder hereditary haemochromatosis, have been proposed as potential risk factors for the development of colorectal cancer (CRC). To understand the relationship between HNPCC disease phenotype and polymorphisms of the HFE gene, a total of 362 individuals from Australia and Poland with confirmed causative MMR gene mutations were genotyped for the HFE C282Y and H63D polymorphisms. A significantly increased risk of developing CRC was observed for H63D homozygotes when compared with combined wild-type homozygotes and heterozygotes (hazard ratio = 2.93, p = 0.007). Evidence for earlier CRC onset was also observed in H63D homozygotes with a median age of onset 6 years earlier than wild type or heterozygous participants (44 vs. 50 years of age). This effect was significant by all tests used (log-rank test p = 0.026, Wilcoxon p = 0.044, Tarone-Ware p = 0.035). No association was identified for heterozygosity of either polymorphism and limitations on power-prevented investigation of C282Y homozygosity or compound C282Y/H63D heterozygosity. In the Australian sample only, women had a significantly reduced risk of developing CRC when compared with men (hazard ratio = 0.58, p = 0.012) independent of HFE genotype for either single nucleotide polymorphisms. In conclusion, homozygosity for the HFE H63D polymorphism seems to be a genetic modifier of disease expression in HNPCC. Understanding the mechanisms by which HFE interrelates with colorectal malignancies could lead to reduction of disease risk in HNPCC. © 2009 UICC
Hereditary nonpolyposis colorectal cancer (HNPCC) occurs as a result of germline mutations to DNA mismatch repair (MMR) genes, including hMSH2 and hMLH1.1–4 The disease is inherited in an autosomal dominant manner and predisposes individuals to the development of early-onset epithelial cancers, accounting for approximately 5% of colorectal cancer (CRC) cases.5 It has been demonstrated that the high level of variation in disease expression (age of onset and tumour site) cannot be solely attributed to differences in mutations of MMR genes.6 There is now considerable evidence to suggest that disease expression is influenced by polymorphisms in other genes.7–11
The role of high body iron levels in modifying the risk of CRC has been investigated by several groups but is still unclear.12–16 The genetic iron overload disorder hereditary haemochromatosis is characterized by high iron indices and progressive parenchymal iron overload and occurs due to a problem in restricting iron uptake (reviewed in Refs.17–19). Although clear associations have been established between haemochromatosis and liver disease, studies investigating the correlation between haemochromatosis and other pathologies have yielded conflicting results.20–23
The primary cause of classical hereditary haemochromatosis has been ascribed to single nucleotide polymorphisms (SNPs) in the HFE gene, in particular the 845G > A SNP which results in the substitution of a tyrosine residue for a cysteine at position 282 (C282Y) and is present in 10–15% of individuals of northern European descent. The more common but less penetrant 163C > G SNP (H63D) is present in 15–30% of individuals.17, 22–28 A recent longitudinal study has demonstrated that up to 30% of men and 1% of women homozygous for the C282Y polymorphism develop iron overload that subsequently manifests as a disease phenotype.29 The risk of developing CRC increased 3-fold in C282Y homozygotes when compared with matched controls without the mutation.30
A number of other epidemiological studies have also investigated the impact of HFE genotype on CRC risk, with mixed results.31–34 Most studies exploring the link between HFE genotype and the risk of developing CRC have approached the problem by selecting subjects diagnosed with CRC and comparing the frequency of HFE polymorphisms with matched controls. In this study, we have instead used a group of patients with confirmed HNPCC-causative mutations in either hMSH2 or hMLH1 to assess the impact of HFE genotype on HNPCC disease expression. Specifically, patients with HNPCC with confirmed causative germline mutations in either hMSH2 or hMLH1 have been genotyped for both the G > A SNP at position 845 and the C > G SNP at position 163 of the HFE gene.
Material and methods
The study was conducted in accordance with the Declaration of Helsinki. Approval for the study was obtained from Hunter Area Research Ethics Committee (Australia), University of Newcastle Human Research Ethics Committee (Australia) and the Ethics Committee of the Pomeranian Academy of Medicine (Poland). Written, informed consent was obtained for all participants.
A total of 362 participants (181 from each of Australia and Poland) were genotyped for at least 1 of the 2 main HFE gene polymorphisms, C282Y and H63D. The Polish sample comprised 86 probands/families and 95 relatives (52%) belonging to 42 families collected from within and around the region of Szczecin, Pomerania. The Australian sample contained 76 probands/families and 105 relatives (58%) belonging to 42 families. The Australian samples were all clinic-based recruits from the state of New South Wales, collected between the years 1997 and 2005. Demographic information of both Australian and Polish participants is shown in Table I. There was no significant difference in the median age of disease onset of CRC between the Australian (41 years of age) and Polish samples (43 years of age).
|Colorectal cancer (n = 86)||No colorectal cancer (n = 85)||Colorectal cancer (n = 76)||No colorectal cancer (n = 105)|
|Age of onset1 (yr)|
|Mean (SD)||41 (12)||40 (14)||43 (10)||38 (16)|
|MMR gene mutation|
|Unable to genotype||2||2||1||0|
|Unable to genotype||8||2||6||5|
All participants had previously been diagnosed with HNPCC based on the molecular diagnosis of a pathogenic germline mutation in either hMLH1 or hMSH2; 104 and 77 Australian participants and 96 and 85 Polish participants harboured causative mutations in the hMLH1 or hMSH2 gene, respectively. Of these, 9 Australian and 25 Polish participants carried a missense mutation in the hMLH1 gene described as pathogenic in the International Society for Gastrointestinal Hereditary Tumours (InSiGHT) mutation database and Mismatch Repair Genes Variant Database. There was no statistically significant difference in the median age of CRC onset between missense mutation carriers and truncating mutation carriers (43 years of age in both groups in the Polish sample, 43 and 41 years of age in the Australian sample). There was also no significant difference in the median age of CRC onset between hMLH1 and hMSH2 mutation carriers (42 vs. 44 years of age in the Polish samples, 41 vs. 39 years of age in the Australian samples).
A total of 86 Australian and 76 Polish participants had been diagnosed with CRC. The median age of participants without CRC was 5 years less than individuals with CRC [41 years (range 17–72) for affected participants vs. 37 years (13–71) for unaffected participants in the Australian sample, 43 (18–72) for affected participants vs. 38 (6–81) for unaffected participants in the Polish sample].
Several participants were diagnosed with extracolonic tumours, which are listed in Table II. Of the 86 Australian and 76 Polish participants with CRC, there were only 9 patients diagnosed with CRC as their second cancer, 2 in the Polish group and 7 in the Australian group.
|CRC+ group||CRC− group||CRC+ group||CRC− group|
|Breast and ovarian||1|
|Uterine and breast and kidney||1|
|Uterine and breast||1|
|Lymphoma and bladder||1|
Allelic discrimination was performed on an ABI PRISM 7900 Sequence Detection System (PE Applied Biosystems, Foster City, CA), using the primers (Invitrogen) and probes (Applied Biosystems) listed in Table III. The assay was performed under universal conditions as previously described.35 A random sample representing 10% of the Australian and 10% of the Polish population was regenotyped to ensure the accuracy of the assay.
|C282Y||Primer-sense||5′-TCA CAT ACC CCA GAT CAC AAT GA-3′|
|Primer-antisense||5′-GGC TGG ATA ACC TTG GCT GTA C-3′|
|Probe wild type||6Fam-TGC TCC ACC TGG CAC GTA TAT CTC TG|
|Probe mutation||Vic-TGC TCC ACC TGG TAC GTA TAT CTC TGC TC|
|H63D||Primer-sense||5′-TCT ACT GGA AAC CCA TGG AGT TC-3′|
|Primer-antisense||5′-GAA GCT TTG GGC TAC GTG GAT-3′|
|Probe wild type||6Fam-CGA CTC TCA TGA TCA TAG AAC ACG AAC AGC|
|Probe mutation||Vic-CGA CTC TCA TCA TCA TAG AAC ACG AAC AGC|
Pearson's Chi-square test was used to evaluate the deviation from the expected Hardy–Weinberg equilibrium and for comparison of the distributions of the genotypes. Published data on large, randomized Australian and Polish samples were used to determine if the genotypic and allelic frequencies of the HNPCC groups were consistent with those predicted for the corresponding population.22, 25, 36 Odds ratios and 95% confidence intervals (CI) were calculated using a 2 × 2 table and applying Fisher's exact test (two sided). Kaplan-Meier survival analysis was used to plot the proportion of the population that was cancer free vs. the subject's age of diagnosis of CRC. In the patients who had developed CRC, age of CRC onset was defined as the patient's age at diagnosis. The age of unaffected HNPCC carriers at the most recent follow-up was treated as censored in this analysis. By comparing the Kaplan–Meier survival curves plotted by genotype, we tested the association between age of diagnosis of CRC and genotype.
Cox proportional hazard regression models were used to analyze the effect of the SNPs with adjustment for potential confounding variables in the study (gender, hMLH1/hMSH2 mutation group, mutation type and genotype) to decrease the bias. To correct for any correlation in time to onset of CRC that may exist among family members, the robust variance correction approach was applied.
The significance levels of all tests were set at p < 0.05. All statistical analyses were performed with Intercooled Stata 8.2 (Stata Corp., College Station, TX) and GraphPad Instat version 3.06 (GraphPad Software, San Diego, CA).
Genotype frequency distribution and odds ratios
In both the Australian and Polish HNPCC groups, the genotype distributions of both the C282Y and H63D SNPs were in Hardy–Weinberg equilibrium (data not shown) and neither the genotype nor the allele frequency deviated significantly from those reported in the literature for large, randomized, Australian or Polish population samples of similar ethnicity.22, 36 However, there was a significant difference in the C282Y genotype frequencies of the Australian and Polish samples (p < 0.001), and as a result the 2 samples were considered separately in investigations of this SNP. In the Australian sample, genotype data for the C282Y and H63D SNPs were not available for 2 participants and 12 participants, respectively, while in the Polish sample genotype data were not available for the C282Y and H63D SNPs for 1 participant and 11 participants, respectively.
Odds ratios were calculated to compare the frequency of participants with either 1 or 2 copies of the mutant allele with wild type. Odds ratios and genotype frequencies of the 2 SNPs were analyzed in comparisons of CRC presentation, gender and MMR gene mutation type (Table IV). There were no significant differences in genotype frequencies for either SNPs in any comparison.
|Australian (n = 179)||147||32||0|
|Controls (n = 3011)||2571||424||16||0.2391|
|CRC+ (n = 85)||69||16||0|
|CRC− (n = 84)||68||16||0||0.9991|
|OR = 1.022||95% CI = 0.47–2.19||1.0003|
|CRC+ (n = 85)||69||16||0|
|No cancer (n = 71)||58||13||0||0.9351|
|OR = 0.972||95% CI = 0.43–2.18||1.0003|
|Female (n = 109)||87||22||0|
|Male (n = 70)||60||10||0||0.3151|
|OR = 0.662||95% CI = 0.29–1.49||0.4243|
|hMLH1 (n = 102)||86||16||0|
|hMSH2 (n = 77)||61||16||0||0.3791|
|OR = 1.412||95% CI = 0.65–3.04||0.4943|
|Polish (n = 180)||173||7||0|
|Controls (n = 1517)||1398||117||2||0.1551|
|CRC + (n = 75)||74||1||0|
|CRC− (n = 100)||99||6||0||0.1341|
|OR = 4.492||95% CI = 0.53–38.08||0.2423|
|CRC+ (n = 75)||74||1||0|
|No cancer (n = 83)||77||6||0||0.0721|
|OR = 5.772||95% CI = 0.68–49.08||0.1203|
|Female (n = 111)||106||5||0|
|Male (n = 69)||67||2||0||0.5881|
|OR = 0.632||95% CI = 0.12–3.36||0.7093|
|hMLH1 (n = 96)||91||5||0|
|hMSH2 (n = 84)||82||2||0||0.3281|
|OR = 0.442||95% CI = 0.08–2.35||0.4513|
|Australian (n = 169)||127||40||2|
|Controls (n = 2603)||1758||711||62||0.2321|
|CRC + (n = 78)||59||18||1|
|CRC− (n = 83)||61||21||1||0.9471|
|OR = 1.122||95% CI = 0.55–2.28||0.8573|
|CRC+ (n = 78)||59||18||1|
|No cancer (n = 71)||52||18||1||0.9451|
|OR = 1.142||95% CI = 0.54–2.37||0.8513|
|Female (n = 105)||75||28||2|
|Male (n = 64)||52||12||0||0.2491|
|OR = 0.582||95% CI = 0.27–1.23||0.1993|
|hMLH1 (n = 98)||74||22||2|
|hMSH2 (n = 71)||53||18||0||0.4541|
|OR = 1.052||95% CI = 0.52–2.12||1.0003|
|Polish (n = 170)||127||37||6|
|Controls (n = 1505)||1087||380||38||0.4821|
|CRC + (n = 70)||51||15||4|
|CRC− (n = 100)||76||22||2||0.4341|
|OR = 0.852||95% CI = 0.42–1.71||0.7213|
|CRC+ (n = 70)||51||15||4|
|No cancer (n = 79)||57||20||2||0.5561|
|OR = 1.042||95% CI = 0.50–2.13||1.0003|
|Female (n = 106)||76||25||5|
|Male (n = 64)||51||12||1||0.3881|
|OR = 0.652||95% CI = 0.31–1.36||0.2783|
|hMLH1 (n = 91)||68||20||3|
|hMSH2 (n = 79)||59||17||3||0.9831|
|OR = 1.002||95% CI = 0.50–2.01||1.0003|
Our HNPCC samples contained only 3 C282Y/H63D compound heterozygotes and no C282Y homozygotes. In the absence of a C282Y mutation, any phenotypic impact of the H63D SNP on iron indices is generally only observed in homozygotes. We assessed the risk of developing CRC associated with homozygosity for the H63D SNP compared with heterozygous or wild-type genotypes using Cox proportional hazard regression models. After correcting for the inclusion of family members, a hazard ratio of 2.93 (95% CI = 1.34–6.38, p = 0.007) was calculated for the combined Australian and Polish samples. This result suggests an increased risk of developing CRC in individuals with the homozygous H63D genotype.
No association between CRC and heterozygosity for either SNPs was identified. There was a decreased risk of CRC for women when compared with men in the Australian sample only (hazard ratio = 0.58, 95% CI = 0.38–0.89, p = 0.012). This was independent of HFE genotype.
Kaplan–Meier survival analysis
The possibility that HFE genotype might influence the age of disease onset was assessed using Kaplan-Meier survival analysis. Since H63D homozygosity was found to affect CRC risk, we investigated the effect on disease onset of this genotype compared with combined wild type and heterozygotes. Kaplan–Meier analysis was performed for the combined Australian and Polish samples (Fig 1). H63D homozygosity was significantly associated with earlier age of CRC onset compared with combined wild type and heterozygotes (log-rank test p = 0.026, Wilcoxon test p = 0.044 and the Tarone-Ware test p = 0.035). Participants homozygous for the H63D polymorphism developed CRC on average 6 years earlier than participants who were heterozygous or wild type for the SNP (median age, 44 vs. 50 years).
In an attempt to explain the decreased risk of CRC in women when compared with men in the Australian sample, we also investigated differences in age of CRC onset by gender in this sample. Male participants in the Australian sample developed CRC on average 7 years earlier than female participants (median age, 44 vs. 51 years). Gender was significantly associated with age of CRC onset using the log-rank test, which is most heavily weighted toward later ages (log-rank test, p = 0.027). Trends toward significance were observed with the Wilcoxon test (p = 0.116) and the Tarone-Ware test (p = 0.058), which are weighted toward younger ages and intermediate ages, respectively.
The results of the current study indicated that homozygosity of the HFE H63D mutation may act as a modifier of the effects of HNPCC-causative mutations in MMR genes, increasing the risk of developing CRC. In addition, there was evidence for earlier CRC onset age in H63D homozygotes. Our results suggest that the median age of disease onset could be as much as 6 years earlier in H63D homozygotes (Fig 1), who represent around 2.5% of the Australian and Polish general populations.
While these findings will require substantiation in other populations, they support a possible relationship between iron dysregulation and CRC risk. Although mechanisms cannot be established by a genetic epidemiological study of this nature, it seems likely that iron is involved, in view of the roles of the HFE gene in iron metabolism, the previously reported effects of H63D homozygosity on iron status37 and existing evidence that iron status can modify CRC. Because iron levels in haemochromatosis patients can usually be maintained at normal levels through phlebotomy and regulating factors such as diet, this might have the potential to substantially reduce CRC risks or delay onset by several years in people with HNPCC-associated MMR gene mutations.
However, the possibility of other mechanisms not directly reflecting abnormal body iron status cannot be ruled out. Homozygosity of the H63D polymorphism increases the risk of the neurodegenerative brain disease amyotrophic lateral sclerosis in the absence of apparent effects of C282Y polymorphism,38–40 suggesting that in some tissues the H63D mutation might have pathological consequences that are not directly related to whole body iron status. It will be important to validate the findings on H63D and also to investigate the effects of C282Y homozygosity in larger HNPCC samples, preferably in conjunction with information on patient's iron status, to determine the mechanisms involved and the role of iron.
While this is the first time that the H63D polymorphism has been specifically associated with HNPCC, there is some previous evidence for association of both the H63D and the C282Y polymorphism with CRC in general.33, 34 Power has limited past studies, because the homozygous and compound heterozygous mutations that have been associated with the greatest increases in iron loading and potentially the highest disease risks are relatively rare. For this reason, some studies analyzed all HFE mutation genotypes as a single group, which may dilute observed effects. Although past epidemiological studies of HFE genotype and CRC risk have had mixed results, an American study of 475 CRC case patients and 833 control subjects found an odds ratio of 1.4 for participants with any HFE mutation after adjustment for a range of factors including age, gender and total iron intake.34 The increased risk predominantly occurred in the quartile with greatest dietary iron intakes. In addition, a recent study of a large Australian sample has found that homozygosity for the C282Y SNP is associated with a 3-fold increase in the risk of developing CRC in men.30 This suggests that the effects of HFE on CRC may not be limited only to MMR gene mutation carriers, although such effects may be stronger when both types of mutation are present simultaneously.
Heterozygosity for either the H63D or C282Y SNP does not seem to have any modifying effect in either the Australian or Polish samples, although it is possible that small effects may be detectable with very large samples. Although heterozygosity for C282Y or H63D has been reported to have a range of effects in other diseases (reviewed in Ref.41), these genotypes are not usually associated with significant changes in iron parameters.26, 42–44 For these reasons, for our final analyses it was considered more appropriate to compare mutant homozygotes with combined heterozygotes and wild-type homozygotes, as is usually done in most studies of HFE gene SNPs. However, although this was effective in revealing the potential modifying effect of H63D homozygosity on HNPCC development, we were not able to do this for the C282Y SNP, because of its relative rarity and the lack of C282Y homozygotes in the samples. Stronger modifying effects may occur in C282Y homozygotes or C282Y/H63D compound heterozygotes, because it is well established that iron indices are increased most in individuals with these genotypes (reviewed in Refs.17, 18, 45).
Gender affects both the onset age and site of first tumour manifestation in HNPCC. In women, the age of onset of CRC is delayed 5 to 10 years when compared with men.46 This trend was also observed in the Australian sample investigated in our study; however, this was not evident in the Polish sample. This may reflect environmental influences or other underlying genetic differences between the Australian and Polish populations. Gender is also a factor in the manifestation of iron loading as a result of HFE genotype, affecting men much earlier in life than women.47 In a larger sample, it is possible that HFE genotype may show a contribution to the earlier onset of CRC in men when compared with women.
In summary, these results suggest that the homozygous H63D genotype may act as a modifier of HNPCC-causative mutations in MMR genes to increase the risk of developing CRC. Whether this is due to iron dysregulation or some other effect involving the HFE gene remains to be determined but in any event, understanding the mechanisms involved could lead to new treatment options that substantially reduce CRC risks in people with MMR gene mutations. Further studies are required in larger samples to confirm this association and to determine if homozygous C282Y or compound C282Y/H63D heterozygous HFE genotypes also modify the risk of manifesting HNPCC.
The authors thank the participants for contributing to this study.
- 30Homozygosity for the C282Y mutation in the HFE gene is associated with increased risk of colorectal and breast cancer in Australian population. Am J Hematol 2007; 82: 575., , , , , , , , , , , , et al.
- 37Initial screening transferrin saturation values, serum ferritin concentrations, and HFE genotypes in Native Americans and whites in the hemochromatosis and iron overload screening study. Clin Genet 2006; 69: 48–57., , , , , , , , , , .