Hereditary nonpolyposis colorectal cancer (HNPCC) is an autosomal dominantly inherited predisposition to develop early onset epithelial cancers, especially colorectal cancer and endometrial cancer, as a result of germline mutations occurring in DNA mismatch repair (MMR) genes.1–6 There is, however, considerable variation in disease expression (age of onset and tumor site) in this disorder which cannot be entirely explained by the type and position of the mutation in the respective genes.7 Several reports have shown that genetic polymorphisms may be contributing factors to disease in HNPCC.8–11
Two single nucleotide polymorphisms (SNPs) in genes involved in the cell cycle, Aurora-A and Cyclin D1, have been associated with the age of onset of colorectal cancer (CRC) in HNPCC patients.12, 13 HNPCC patients homozygous for the wild-type allele (TT) in the T91A SNP (F31I) in Aurora-A developed CRC ∼7 years earlier than patients carrying the variant allele.12 This association was increased after stratifying for the G870A SNP at codon 242 in Cyclin D1, which had previously been associated with an earlier age of disease diagnosis by an average of 11 years compared to patients homozygous for the wild-type allele.13
Aurora-A is involved in the normal cell cycle but is overexpressed in a variety of malignancies.14–22 Aurora-A positively regulates the G2-to-M phase of the cell cycle, and the activation of Aurora-A in late G2 is inhibited by DNA damage.15 DNA repair and cell-cycle control might be linked together as it has been suggested that MMR genes are necessary to activate the G2-M checkpoint in the presence of certain types of DNA damage.23
Cyclin D1 has an important role in the G1-to-S phase in the cell cycle.24 The G870A SNP in Cyclin D1 enhances alternate splicing of the gene, and the protein encoded by the alternate transcript may have a longer half life.25 Several studies have suggested that there is a relationship between the G870A SNP in Cyclin D1 and the age of disease onset in HNPCC. The relationship between Cyclin D1 and disease expression in HNPCC appears to be complex as one study shows an association between the polymorphism and age of disease onset,13 whereas a second study from Finland failed to show a similar relationship, but did observe a decreased age of disease onset for homozygote wild-type and mutant alleles compared to heterozygotes,26 suggestive of heterosis. The major difference between the 2 studies was the predominance of hMSH2 mutation carriers in one population compared to hMLH1 carriers in the other.
One of the major shortcomings with all of the previous reports on the relationship between the age of disease onset and respective genotype in Aurora-A and Cyclin D1 was the number of individuals used in each study, all of which were relatively small.12, 13, 26 To understand the biological basis of the relationship between disease phenotype and polymorphisms in Aurora-A and Cyclin D1, additional studies on larger groups of HNPCC patients are required.
In this report, we have genotyped 312 HNPCC individuals for the G>A SNP at codon 242 in Cyclin D1 and the 91T>A SNP (F31I) in Aurora-A, all with confirmed causative germline mutations in either hMSH2 or hMLH1.
Material and methods
All the participants selected for this study had previously been diagnosed with HNPCC, and the selection criteria were based on the molecular diagnosis of HNPCC. The 312 participants harbor germline mutations in either hMLH1 or hMSH2. There were 177 cases with germline hMLH1 mutations and 135 with hMSH2 mutations, of which there were 255 (82%) nonsense insertion, deletion or splice mutations (leading to a truncated protein) and 57 (18%) missense mutations described as pathogenic in the International Society for Gastrointestinal Hereditary Tumours (InSiGHT) mutation database. Fifty three 53 (17%) of the missense mutations were in hMLH1 mutation carriers, while 4 (1%) were in hMSH2 mutation carriers. Local ethics committees in Poland and Australia approved the study.
Of the 312 HNPCC participants, 220 have been described previously.27 Ninety-two new samples from Poland have since been added to the study. The new samples include 33 CRC probands, 6 uterine cancer probands and 53 relatives (57%). Briefly, the sample population consists of 86 participants from Australia and 226 participants from Poland, 22 (26%) of the Australian and 131 (58%) of the Polish participants were relatives of probands. Of the 312 participants, 157 had been diagnosed with colorectal cancer (CRC). The distribution of second cancers was similar between hMLH1 and hMSH2 mutation carriers. The 6 probands diagnosed with endometrial cancer were considered as controls since we were specifically focusing on colorectal cancer and not on any other malignancy common in HNPCC.
The median age of participants diagnosed with CRC and unaffected MMR gene mutation carriers were 43 years and 39.5 years, respectively. The median age did not differ between hMLH1 and hMSH2 mutation carriers; 43 years for both groups of participants diagnosed with CRC and 40.5 years for hMLH1 and 39 years for hMSH2 carriers for unaffected MMR gene mutation carriers.
Genotyping was performed on an ABI PRISM® 7500 Real-Time PCR System (PE Applied Biosystems, Foster City, CA), using primers and probes from Assay-by-Demand (Applied Biosystems) for Aurora-A (rs2273535, assay ID: C__25623289_10) and Cyclin D1 (rs603965, assay ID: C_744725_1). The assay was performed under universal conditions previously described.27
Pearson's χ2 was used to evaluate the deviation from the expected Hardy-Weinberg equilibrium and for comparison of the distributions of the genotypes, while odds ratio (OR) and 95% confidence intervals (CI) were calculated using a 2 × 2 table. Kaplan-Meier survival analysis was used to plot the proportion of the population that was cancer free versus the subject's age of diagnosis of colorectal cancer (CRC). By comparing the Kaplan-Meier survival curves plotted by genotype, we tested the association between age of diagnosis of CRC and genotypes. The Wilcoxon's test, which emphasizes observations from early diagnosis, log-rank test, which gives more weight to later ages and Tarone-Ware test, which is an intermediate of the 2 other tests, were used to examine the homogeneity of the survival curves. If the results were insignificant, the p value of the log-rank test was reported. The significance levels of all tests were set at p < 0.05. All statistical analyses were performed with Intercooled Stata 8.0 (Stata Corp., College Station, TX) and GraphPad Instat version 3.06 (GraphPad Software, San Diego, CA).
The genotypes and allele frequency distribution for Aurora-A (rs2273535) and Cyclin D1 (rs603965) are presented in Table I.
Table I. Study Demographics of the Aurora-A and Cyclin D1 SNPS in HNPCC Patients According to Disease Expression (Affected with CRC (CRC+) and Unaffected MMR Gene Mutation Carriers (CRC−))
Ten samples (2 from the nonaffected hMLH1 group, 4 from the nonaffected hMSH2 group and 2 each from the affected hMSH2 and hMLH1 groups) consistently failed to amplify for the Aurora-A SNP and were consequently not included in the statistical analysis.
The frequencies of both SNPs were in Hardy-Weinberg equilibrium. The allele frequencies observed in this study were not different from those reported for Caucasian subjects (http://snp500cancer.nci.nih.gov), p = 0.90 for Aurora-A and p = 0.95 for Cyclin D1. The genotype frequencies were not significantly different between Australian and Polish samples (data not shown). When comparing genotype frequencies for the Aurora-A and Cyclin D1 SNPs between unaffected MMR gene mutation carriers (CRC–) and the colorectal cancer MMR gene mutation carriers (CRC+), no significant differences were observed. Similarly, no significant differences were observed between male or female participants, affected or not.
A significant difference was observed for the Cyclin D1 SNP frequency between hMLH1 and hMSH2 mutation carriers when comparing individuals with the wild-type genotype (GG) to individuals with any variant allele (GA and AA combined), p = 0.03.
Kaplan-Meier survival analysis
No significant difference between the age of diagnosis of CRC and genotype was observed for either of the 2 SNPs in this study when assessing the subject group, whether comparing the 3 genotypes or combining heterozygotes and variant genotype (see Fig. 1). The median age of diagnosis of CRC, or the age at which 50% of the population is cancer-free, was not significantly different between the genotypes for Aurora-A or Cyclin D1. Similarly, when dividing the subject group by gender, no difference was observed between age of diagnosis of CRC and genotype between females and males.
For Cyclin D1, there is a significant difference between age of diagnosis of CRC and genotype in the hMSH2 mutation carriers, but not hMLH1 mutation carriers (see Fig. 2). The hMSH2 mutation carriers harboring the wild-type (GG) genotype developed CRC on average 30 years later than individuals with the variant genotype (AA) and 24 years later than individuals with the heterozygote genotype (GA), p = 0.03 (log-rank test), p = 0.03 (Wilcoxon test) and p = 0.02 (Tarone-Ware test). When combining the heterozygote and variant genotype (GA + AA), there was a difference of 26 years, p = 0.01 (log-rank test), p = 0.04 (Wilcoxon test) and p = 0.02 (Tarone-Ware test); see Table II.
Table II. Median Age of Diagnosis of CRC (Age at which 50% of the Population is Cancer Free) in HNPCC Participants for Aurora-A and Cyclin D1; Subject Group and hMSH2 Mutation Carriers
The subject group includes 298 samples (14 samples without information about age were excluded).
hMSH2 carriers include 130 samples (5 samples without information about age were excluded).
48 years (n = 171)
47 years (n = 75)
72 years (n = 24)
47 years (n = 103)
49 years (n = 151)
48 years (n = 71)
50 years (n = 14)
47 years (n = 72)
42 years (n = 35)
Since the identification of the genetic basis of HNPCC many studies have been undertaken to identify modifier genes that may explain, at least in part, the variation in disease expression in this cancer syndrome. Functional studies of the 2 SNPs investigated in this study indicate that the variant allele of Cyclin D1 (A) increases alternate splicing25 of the gene, which has been associated with nonsmall cell lung cancer,25 squamous cell carcinoma of the head and neck28 and the prompt for this current study, the age of diagnosis of CRC in HNPCC.13 The variant allele of Aurora-A (A) has been associated with an increased risk of ovarian cancer,29 esophageal squamous cell carcinoma,30 breast cancer31 and a variety of other cancer types, including CRC32 as well as a later age of onset in HNPCC.12
The focus of this study has been on 2 genes involved in cell-cycle control, as they have been shown to be associated with the age of diagnosis of CRC in HNPCC patients.12, 13, 26 Our results did not differ significantly with respect to the genotype frequencies in the original publications, but we were unable to show any significant association overall with the age of disease onset and genotype. In addition, when the study population was divided on the basis of hMLH1 or hMSH2 mutation carriers, no association with the T91A SNP in Aurora-A and age of disease onset was observed. The most likely explanation for previous results showing a positive relationship between the Aurora-A SNP and age of disease onset is the high likelihood of a Type 1 error occurring by virtue of the limited population size used in the previous study (125 patients were studied, originating from 60 families12).
With respect to cyclin D1 and disease risk, a more complicated relationship with disease risk and the G870A SNP has been observed. Our studies indicate that there is a significant association between this SNP in cyclin D1 and hMSH2 mutation carriers and the age of CRC diagnosis. The first report of an association with an earlier onset of disease was based on a very limited number of mutation positive cases (n = 86), which was dominated by hMSH2 mutation carriers (57 hMSH2 carriers vs. 28 hMLH1 mutation carriers).13 A second group investigated 146 HNPCC patients, 141 patients harbored the hMLH1 mutations and the remaining 5 harbored hMSH2 mutations,26 and did not find the same relationship. Furthermore, a larger study from Germany33 failed to find any association between the G870A polymorphism and the age of disease onset in HNPCC. In our study, we have investigated 135 and 177 hMSH2 and hMLH1 carriers, respectively. When all HNPCC cases are analyzed together, similar to Kruger et al.,33 we failed to identify any association with the G870A SNP and the age of diagnosis of CRC. If, however, the HNPCC cases are subdivided into hMLH1 or hMSH2 mutation carriers we observed a significant difference in the age of colorectal cancer onset in the hMSH2 carriers associated with the G870A SNP. In the study by Kruger et al., the HNPCC population was not subdivided into hMLH1 or hMSH2 carriers,33 and we are therefore unable to compare directly this observation. Our observation that there is a relationship between the G870A SNP and the age of diagnosis of CRC suggests that there is an interaction between hMSH2 and cyclin D1. However, there is little information in the literature to support a direct relationship, but it does not exclude this possibility.
Potential limitations of the current study include population stratification, which could be a confounder, but this should not be affecting our results as we are searching for modifying polymorphisms affecting disease expression in HNPCC patients (defined entity), but cannot be excluded. Environmental factors that are different in the 2 countries could potentially affect the results. We do not believe this is the case, as it has been shown that for most of the common disease-associated polymorphisms, ethnicity is likely to be a poor predictor of an individual's genotype.34 The subject groups in this study have similar genotype frequencies (58, 37 and 5% for Aurora-A and 24, 51 and 24% for Cyclin D1) to Caucasian controls (58, 35 and 6% for Aurora-A and 26, 48 and 26% for Cyclin D1) and are not significantly different from the frequencies reported in the study by Chen et al.12 The results could be due to multiple statistical testing; however, the association of the Cyclin D1 SNP with age of onset of CRC in hMSH2 mutation carriers is intriguing.
Although this study is one of the larger cohorts of HNPCC patients examined, more HNPCC populations need to be studied to confirm these results. Looking for modifying polymorphisms influencing disease expression has proven to be a difficult task, as controversial results seem to be the rule rather then the exception. Nevertheless, it is becoming evident that genetic modifiers of disease risk in HNPCC do exist and have the potential to improve predictive genetic counseling in this disorder.