Objective Recently, six colorectal cancer (CRC) susceptibility loci have been identified, and two single-nucleotide polymorphisms (SNPs)—rs16892766 (8q23.3) and rs3802842 (11q23.1)—from two of these regions have been found to be significantly associated with an increased CRC risk in patients with Lynch syndrome. The objective of this study was to genotype nine SNPs within these six loci to confirm previous findings and investigate whether they act as modifiers of disease risk in patients with Lynch syndrome.
Design The patient cohort consisted of 684 mutation-positive patients with Lynch syndrome from 298 Australian and Polish families. Nine SNPs were genotyped: rs16892766 (8q23.3), rs7014346 and rs6983267 (8q24.21), rs10795668 (10p14), rs3802842 (11q23.1), rs10318 and rs4779584 (15q13.3), and rs4939827 and rs4464148 (18q21.1). The data were analysed to investigate possible associations between the presence of variant alleles and the risk of developing disease.
Results An association between SNP rs3802842 on chromosome 11q23.1 and rs16892766 on chromosome 8q23.3 and the risk of developing CRC and age of diagnosis was found in MLH1 mutation carriers. Female MLH1 mutation carriers harbouring the homozygous variant genotype for SNP rs3802842 have the highest risk of developing CRC. When the number of risk alleles for the two SNPs combined was analysed, a difference of 24 years was detected between individuals carrying three risk alleles and those carrying no risk alleles.
Conclusion The authors were able to replicate the association between the CRC susceptibility loci on chromosomes 8q23.3 and 11q23 and the risk of developing CRC in patients with Lynch syndrome, but the association could only be detected in MLH1 mutation carriers in this study.
- CRC susceptibility loci
- lynch syndrome
- molecular genetics
Statistics from Altmetric.com
Lynch syndrome (hereditary non-polyposis colorectal cancer (HNPCC)) is an autosomal dominantly inherited cancer predisposition which is associated with mutations in genes involved in DNA mismatch repair (MMR).1–4 It is characterised by the diagnosis of colorectal cancer (CRC) at a relatively young age, neoplastic lesions that display DNA microsatellite instability, and an increased incidence of extracolonic malignancies.5 6 The most recent disease penetrance estimates in Lynch syndrome suggest that, by age 70 years, 45% of men and 33% of woman will develop CRC and 15% of women will develop endometrial cancer due to mutations in the DNA MMR pathway.7 Disease expression in patients with Lynch syndrome is highly variable, as observed in unrelated and related patients harbouring the same mutations.8
Recently, genome-wide association studies and a linkage study have identified six CRC susceptibility loci.9–12 The loci identified were associated with risk of disease and as such are likely to represent low-penetrance risk alleles. One report in 2009 identified two loci, 8q23.3 and 11q23.1, to be associated with CRC risk in patients with Lynch syndrome,13 thereby providing some evidence that these two loci may act as modifier genes in this condition.
In this study, we genotyped nine single-nucleotide polymorphisms (SNPs) from the six CRC susceptibility loci in 684 patients harbouring germline mutations in DNA MMR genes to confirm which of these CRC susceptibility loci behave as modifiers in patients with a diagnosis of Lynch syndrome.
Patients with Lynch syndrome harbouring a germline mutation in one of the DNA MMR genes, MLH1, MSH2 or MSH6, were selected for this study. All participants gave written informed consent before the study for their DNA to be used for further institutional ethics-approved research into their condition (review board in both Australia and Poland). Samples were collected and processed at Hunter Area Pathology Service, NSW, Australia and at the Pomerian Academy of Medicine, Poland. Each participant (n=684) had previously contributed blood from which DNA was extracted using the salt precipitation method.14 All the samples were genotyped, and information about the nine SNPs can be seen in online supplementary table 1. A description of the genotyping method, information about the sample population, and statistical analysis can also be seen in online supplementary file 1.
The significance levels of all tests were set at p<0.05. Applying Bonferroni correction for multiple testing gives a corrected significant threshold of: p<0.05/9=0.0055. Any results with a p value lower than 0.0055 were considered significant.
The Australian and Polish samples were analysed both individually and as a combined group. The genotype frequency between the Australian and Polish populations were significantly different for two SNPs: rs4464148 (18q21), p=0.008; and rs7014346 (8q24.21), p=0.0001. However, both were in Hardy–Weinberg equilibrium (HWE) in their respective populations. The genotype frequencies of the studied SNPs are presented in table 1.
The genotype frequencies of the nine SNPs observed in the MSH6 mutation carriers are not presented separately in table 1 because of the low numbers of MSH6 mutation carriers in this study (n=51). One SNP, rs4464148 (18q21), was significantly different in the MSH6 group from that observed in the MSH2 group (p=0.0016).
Results of genotype frequency analysis and Kaplan–Meier survival analysis can be seen in online supplementary file 1.
Kaplan–Meier survival analysis
The average and median age of diagnosis of CRC (as 1st primary tumour) in the combined subject group is 43 years (range 17–72). There is no significant difference between the age of diagnosis of CRC or extracolonic HNPCC and the sample population (Australian and Polish) (see online supplementary figure 1A and 1B, respectively).
When analysing the combined sample population, we found a significant difference between age of diagnosis of CRC and the genotypes for SNP rs3802842 (11q23.1) in MLH1 mutation carriers (figure 1A). This result was also consistent with the analysis of the Australian and Polish populations separately, with all three tests being significant in the Australian population and a trend observed in the Polish population (figure 1B and 1C). Cox regression analysis shows that, when family relationships are taken into account, the homozygous variant genotype (CC) in MLH1 mutation carriers is associated with an increased risk of CRC (HR=2.67, 95% CI=1.35 to 5.26 and p=0.0050). Further analysis revealed an increased risk of CRC in women harbouring MLH1 mutations (HR=3.19, 95% CI=1.46 to 7.01 and p=0.004).
In the Australian population, the genotypes for SNP rs16892766 (8q23.3) also have an impact on the age of diagnosis of CRC in MLH1 mutation carriers (log-rank p=0.0023, Wilcoxon p=0.0013, and tware p=0.0011), with individuals who harbour the heterozygous genotype (AC) developing CRC 12 years earlier than those who harbour the homozygous wild-type genotype (AA).
Kaplan–Meier—total number of risk alleles
The highest number of risk alleles in any individual was 12 out of a possible 18. No significant results were detected in the subject group (or if divided by gender or gene) when an individual's total number of risk alleles and the effect on the development of CRC was analysed, either by itself or when divided into risk groups (data not shown). The data were also analysed with two risk alleles as the reference (instead of zero risk alleles as the reference) considering that only very few people had zero and one risk allele, but no significant difference was observed.
The two SNPs giving significant results, rs3802842 (11q23.1) and rs16892766 (8q23.3), were combined to check whether clustering of low-risk alleles (zero, one, three and four risk alleles) had an effect on the risk of developing disease or age at onset of CRC. A highly significant difference between the number of risk alleles and the age of diagnosis of CRC can be seen in the MLH1 mutation carriers (figure 2). MLH1 mutation carriers who harbour three risk alleles also have a significantly increased risk of developing CRC (HR=5.52, p=0.003, 95% CI=1.80 to 16.91) compared with individuals who harbour no risk alleles, whereas having two risk alleles resulted in a trend towards higher risk (HR=1.9, p=0.067 and 95% CI=0.96 to 3.77).
In 2009 Wijnen et al13 showed that two of six recently identified CRC susceptibility loci,9–12 rs16892766 (8q23.3) and rs3802842 (11q23.1), are significantly associated with CRC risk in Lynch syndrome. The risk was significantly associated with the number of risk alleles in a dose-dependent manner, where the effects were stronger in female carriers than male carriers.13
Individuals carrying the AA or AC genotype for SNP 3802842 (11q23.1) in the present study develop CRC on average 10 years later than expected compared with unselected people diagnosed as having Lynch syndrome, suggesting a protective effect against the development of CRC for these genotypes. In addition, the CC genotype also confers a significantly increased risk of CRC compared with the AA genotype. The results are not significant if analysed with extracolonic cancer as the end point for the survival analysis, confirming the role of the loci in CRC development. Little is known about the function of SNP rs3802842, but it is located in a gene-rich region of chromosome 11q23 where four open reading frames (LOC120376, FLJ45803, C11orf53 and POU2AF1) and a potential polymorphic binding site target for microRNA that is in high linkage disequilibrium have been reported.11
There is also a highly significant difference between the genotypes of SNP rs16892766 (8q23.3) and age of diagnosis of CRC in Australian MLH1 mutation carriers (12 years difference). SNP rs16892766 (8q23.3) maps to a gene called EIF3H. EIF3H is a translation factor involved in the control of translation initiation, which is an important step in the regulation of gene expression.15 Deregulation causes abnormal gene expression leading to altered cell growth and cancer.16 17
The Dutch study13 and the present study have the same number of participants, with similar power but our patient cohort consists of 298 families compared with 127 families in the Dutch study, so we cannot rule out the effect of other genetic factors that may influence disease risk in the Dutch population. SNP rs3802842 was associated with an increased risk of CRC in both studies: in all females in the Dutch study and in all MLH1 mutation carriers in the present study, but MLH1 mutation-positive females from the present study had the highest risk of developing CRC.
We have been able to replicate the association seen in the Dutch study, but only in MLH1 mutation carriers; this may be due to an over-representation of MLH1 mutation carriers in the Dutch study compared with ours, but unfortunately this information was not available. Both studies could detect an association with increasing number of risk alleles, calculated by utilising the genotypes for the two SNPs (rs3802842 (11q23.1) and rs16892766 (8q23.3)) that were found to be significant in both studies. Having more risk alleles increases the risk of developing CRC in the Dutch subject group and in MLH1 mutation carriers in our study. Our findings suggest that SNP rs3802842 (11q23.1) is perhaps a better candidate as a modifying locus for disease expression in MLH1 mutation-positive patients with Lynch syndrome (HR=2.67 in MLH1 mutation carriers (rs3802842), HR=3.19 in female MLH1 mutation carriers (rs3802842) and HR=5.52 in MLH1 mutation carriers when the two SNPs (rs16892766 and rs3802842) are combined) as opposed to conferring disease susceptibility in sporadic CRC cases.
The genotype frequencies of two SNPs were significantly different between the two study populations in the present study. Both of the SNPs were in HWE in their respective populations. The minor and major allele frequency of the study cohort was also compared with the HapMap European control population (n=120) available at the National Centre for Biotechnology Information (NCBI) for the nine SNPs, and no significant difference in the allele frequencies was observed. Potential limitations of the study include population stratification, which could not be formally tested because of the low number of SNPs studied. This should not be a major bias, but it cannot be ruled out. Differences in environmental factors could potentially influence the result in addition to any genetic variance between the two populations. However, it has been shown that, for most of the common disease-associated polymorphisms, ethnicity is likely to be a poor predictor of genotype.18 As shown in online supplementary figure 1, there is no difference between the age of diagnosis of CRC or extracolonic HNPCC between the two populations. Alternatively, the difference between the two study populations may be due to a type 1 error occurring in the Australian data or a type 2 error in the Polish population.
The evidence presented in this report suggests that there are covert differences between the biological activities of MSH2 and MLH1. This is supported by evidence that there are different functional domains on MLH1 and MSH2 that could ostensibly result in subtle differences between the two DNA MMR proteins that are influenced by independent genetic factors.19 The search for modifying polymorphisms that influence disease expression in patients with Lynch syndrome has proven to be a difficult task, as controversial results seems to be the rule rather than the exception.20–24 Nevertheless, we believe that the elucidation of modifiers in Lynch syndrome is important as they have the potential to improve the specificity of genetic counselling. The association of SNPs rs3802842 (11q23.1) and rs16892766 (8q23.3) in MLH1 mutation carriers is very appealing and requires further investigation in a larger independent cohort or by performing meta-analysis and/or functional studies to elucidate how these SNPs influence disease risk.
In conclusion, we were able to replicate the association between the CRC susceptibility loci on chromosomes 8q23.3 and 11q23 and the risk of developing CRC in patients with Lynch syndrome, but the association could only be detected in MLH1 mutation carriers in the present study. Our study could not replicate the clustering of low-risk variants from the six CRC susceptibility loci combined as a risk factor for the development of CRC, but the number of risk alleles when the two significant SNPs (rs3802842 (11q23.1) and rs16892766 (8q23.3)) were combined influenced age of diagnosis of cancer and the risk of developing disease in MLH1 mutation carriers.
What is already known about this subject
Six colorectal cancer (CRC) susceptibility loci have recently been identified.
The six loci identified were associated with risk of disease and as such are likely to represent low-penetrance risk alleles for CRC.
Two single-nucleotide polymorphisms (SNPs)—rs16892766 (8q23.3) and rs3802842 (11q23.1)—from two of these regions have been found to be significantly associated with an increased CRC risk in patients with Lynch syndrome.
What this study adds
An association between SNPs, rs3802842 (11q23.1) and rs16892766 (8q23.3), and the risk of developing CRC in patients with Lynch syndrome and age of diagnosis was found in MLH1 mutation carriers only. In addition, carrying the variant allele of rs3802842 (11q23.1) confers an increased risk of CRC, with the highest risk seen in females.
When the number of risk alleles for the two SNPs combined is analysed, a difference of 24 years in the age of diagnosis of CRC can be detected between individuals carrying three risk alleles and those with no risk alleles, and a 5.52 times increased risk of developing CRC for the individuals carrying three risk alleles.
Female MLH1 mutation carriers harbouring the homozygous variant genotype for SNP rs3802842 have the highest risk of developing CRC.
How might it impact on clinical practice in the foreseeable future
The evidence presented in this report suggests that there are covert differences between the biological activities of MSH2 and MLH1. If the findings of this study are confirmed, MLH1 mutation carriers should be offered additional genetic testing to target individuals for early screening for CRC, which will lower the overall morbidity and mortality associated with Lynch syndrome.
We thank the participants for contributing to this study, and Patrick McElduff for help with statistical analysis.
Funding This study was supported by grants from the Hunter Medical Research Institute and the Gladys M Brawn Memorial Fund through the University of Newcastle.
Competing interests None.
Ethics approval This study was conducted with the approval of the 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).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.