Background
Colorectal cancer is a heterogeneous disease that arises from a number of distinct molecular pathways [
1]. The majority arise from conventional colorectal adenomas in which the initiating event is usually inactivation of the
APC tumor suppressor gene [
2,
3]. An important subgroup of colorectal cancers bear a mutation in the
BRAF oncogene [
4] and these cancers arise from serrated polyps initiated by the
BRAF mutation [
5]. There is a very strong association between
BRAF mutation in colorectal cancer and aberrant DNA methylation of CpG islands which is associated with gene silencing when it occurs in promoter areas [
6]. This has been described as the CpG Island Methylator Phenotype (CIMP) [
7]. One of the important genes sometimes silenced by methylation is
MLH1 which encodes a mismatch repair protein. Loss of MLH1 expression results in mismatch repair deficiency and the rapid accumulation of mutations manifested as microsatellite instability (MSI) [
8]. MSI cancers have a good prognosis but not all colorectal cancers with
BRAF mutation and CIMP silence
MLH1 and those that remain microsatellite stable (MSS) have a particularly poor prognosis [
9].
There are two types of serrated polyp from which
BRAF mutant cancers arise. The most common is the sessile serrated adenoma which occurs predominantly in the proximal colon and in older women [
1]. They are characterized by abnormal crypt architecture but do not have cytological dysplasia. They typically have both
BRAF mutation and evolving CIMP but not MLH1 silencing or MSI. Development of cytological dysplasia in a sessile serrated adenoma (SSAD) is associated with rapid progression to invasive malignancy, it is at this stage that methylation-induced silencing of MLH1, and development of MSI may occur. These lesions ‘caught in the act’ of progressing to malignancy are rarely observed in the clinic, and account for approximately 1% of all sessile serrated adenomas. We have recently curated a series of dysplastic sessile serrated adenomas and shown that 75% of SSAD progress methylate MLH1, are MSI, and thus progress to
BRAF mutant MSI cancers. For unknown reasons, 25% do not silence MLH1 and become
BRAF mutant MSS cancers [
10]. The second type of serrated polyp with malignant potential is the traditional serrated adenoma (TSA) which is an uncommon polyp occurring in the distal colon with an equal gender distribution [
11].
BRAF mutation is present in 67% and the majority of these polyps show CIMP. They have a high malignant potential but even during malignant conversion silencing of MLH1 is extremely rare [
11]. Thus TSAs are a source of
BRAF mutant MSS cancers.
Whether the promoter of
MLH1 becomes sufficiently methylated to silence the gene in the setting of CIMP may not be a random, stochastic process. Several studies have associated a series of single nucleotide polymorphisms in the
MLH1 promoter with the occurrence of methylation-induced silencing in large series of cancers [
12,
13]
. The study by Mirakuya and colleagues found a significant association between
MLH1 methylation and the A allele of the rs1800734 single nucleotide polymorphism in a consecutive, unselected series of colorectal cancers, stratifying cancers into negative, partial or full methylation using bisulphite sequencing. Rs1800734 (or
MLH1–93) is a polymorphism 93 base pairs from the
MLH1 translation start site. Subsequent studies have indicated a shift in protein binding as a result of this G > A polymorphism [
13]. Further, a recent study by Liu et al. showed that the A allele was able to regulate an upstream gene, DCLK3, in a trans-acting manner [
14]. They were unable to demonstrate a relationship between the polymorphism and methylation in vivo, but only MSS cell lines were studied [
14]. The effect of the polymorphism on methylation may only occur in a particular cellular context.
We hypothesized that the A allele of MLH1–93 is an important factor influencing methylation-induced silencing of MLH1 in the permissive environment of a BRAF mutant SSAD but not in the context of TSA.
Methods
Sample selection
Samples were obtained from Envoi Specialist Pathology (Envoi) Brisbane, Australia, over a six-year period and are part of two previously published series [
10,
11]. Envoi Specialist Pathology is a community based specialist gastroenterology practice. These series include polyps and cancers removed both endoscopically and surgically. Tissue from Envoi was embedded in formalin fixed paraffin embedded (FFPE) blocks, with DNA extracted using chelex, as previously reported [
15]. Cancers were obtained in a fresh state from patients undergoing surgery at the Royal Brisbane and Women’s Hospital, Brisbane, Australia, and from FFPE blocks at Envoi. Fresh samples were extracted using salt precipitation [
16] and FFPE samples were extracted using chelex. For the control cohort, blood samples were taken from consenting patients who presented to gastroenterology clinics in Brisbane for investigation of symptoms and in whom subsequent colonoscopy showed no polyps or cancer.
Pathological assessment
Each sample was review by independently by two expert pathologists. Criteria for the diagnosis of a traditional serrated adenoma can be found in Bettington et al., 2015 [
11]. Criteria for the diagnosis of a dysplastic sessile serrated adenoma can be found in Bettington et al., 2017 [
10].”
BRAF and CIMP analysis
The BRAF V600E mutation was assessed in each sample using allelic discrimination as previously reported [
11]. We assessed CIMP status using a methylation specific PCR with a marker panel consisting of
NEUROG1, SOCS1, CACNAIG, IGF2 and RUNX3 as reported by Weisenberger and colleagues [
6]. To avoid the potential confounding of MLH1 loss secondary to Lynch Syndrome, only polyps and cancers bearing the
BRAF
V600E
mutation were included. BRAF mutation has previously been shown to be an excellent marker of somatic MLH1 loss due to promoter hypermethylatioon [
17].
MLH1 methylation and immunohistochemical analysis
For SSAD,TSA and cancer cohorts,
MLH1 methylation was determined by bisulfite conversion, followed by methylation specific qPCR as previously reported [
10]. MLH1 protein expression was assessed by immunohistochemistry using previously reported methods [
11], staining patterns were analyzed by an experienced gastrointestinal pathologist (MB).
SNP genotyping analysis
MLH1–93 genotypes were determine by high resolution melt analysis using 2.4 mM MgCl2, 0.24 mM dNTP, 0.24uM forward primer (5-‘TGACTGGCATTCAAGCTGTC-3’), 0.24uM reverse primer (5’-TTCAGCCAATCACCTCAGTG-3′), 0.24uM SYTO9, 1X DNA polymerase GoBuffer (Promega, Wisconsin USA), 1 unit GoTaq DNA Polymerase (Promega, Wisconsin USA) and 1 ng template DNA. The PCR thermal conditions were 95 °C for 120 s; 40 cycles of: 94 °C for 30s, 60 °C for 30s, 72 °C for 45 s followed by 95 °C for 300 s, 50 °C for 120 s and high resolution melt from 75 °C to 87 °C ramping by 0.2 °C / step) and consequent high resolution melt profile analysis. High resolution melt profile was confirmed using Sanger sequencing (Forward primer: 5’ TCTGCTCCTATTGGCTGGAT3’, Reverse primer: 5’ CCCTCCGTACCAGTTCTCAA3’).
Statistical analysis
Statistical analysis was carried out in GraphPad Prism 7. For categorical variables, a χ2 test was used for contingencies >2 × 2, with Fishers Exact test used for 2 × 2 contingencies. For percentage of methylated reference comparisons, a Mann-Whittey-U test was used. The null hypothesis was rejected at p < 0.05.
Ethical approval
The study was approved by the QIMR Berghofer Medical Research Institute Human Research Ethics Committee and the Royal Brisbane and Women’s Hospital Ethics Committee. All participants gave informed written consent prior to participation in this study.
Discussion
Sessile serrated adenomas progress to malignancy following the development of focal dysplasia [
10]. Approximately 75% of dysplastic SSA develop hypermethylation at
MLH1, lose mismatch repair function and develop the MSI phenotype, whilst the rest remain mismatch repair proficient [
10]. Factors involved in this bifurcation are currently unknown. The present study provides evidence that this is influenced by an inherited predisposition to
MLH1 hypermethylation via a series of germline regulatory single nucleotide polymorphisms. Our data indicates a significant increase in the A-allele at
MLH1–93 in
BRAF mutant, mismatch repair deficient, dysplastic sessile serrated adenomas and colorectal cancers. Further, we demonstrate a dose-dependent increase in promoter localized CpG island hypermethylation in the presence of A-alleles in the cellular context of dysplastic sessile serrated adenoma.
Intriguingly, we observed similar allele and genotype frequencies in our traditional serrated adenoma cohort as are present in our local control cohort. Traditional serrated adenomas are nearly universally microsatellite stable lesions [
18]. Our data indicates that while the
MLH1–93 A allele predisposes sessile serrated adenomas to
MLH1 hypermethylation and mismatch repair deficiency, this is not the case for traditional serrated adenomas. Instead, we propose that traditional serrated adenomas arise through distinct molecular pathways that will not, regardless of regulatory genetic changes, methylate the
MLH1 promoter. This is despite the presence of the
BRAF
V600E
mutation and CIMP. It is possible that there are quantitative and qualitative differences in CIMP and interestingly less TSAs and
BRAF mutant, mismatch repair proficient cancers met the definition of a high level of CIMP. We postulate that
BRAF mutant MSS colorectal cancers with the AA-genotype arise in traditional serrated adenomas.
The mechanism by which the A-allele promotes, or the G-allele prevents, methylation is unclear. Perera and colleagues [
19] used EMSA assays to demonstrated the modulation of the binding of nuclear proteins to the region by the MLH1–93 G > A SNP. We and other groups [
12,
13] have used bioinformatics approaches to estimate the effects of the polymorphism on transcription factor binding, identifying numerous candidate protein binding events, including the destruction of TFAP4, Pbx1b and Myf-5 binding sites and creation of AP-3, HNF-3b and GCR binding sites in the presence of the A-allele. Savio and colleagues [
13] used ChIP assays to demonstrate the diminished binding of TFAP4 in cell lines of AA-genotype confirming the accuracy of at least one of our predictions. Interestingly, TFAP4 is under-expressed in CIMP-positive cancers. TFAP4 may share similar affinity for specific sequences as the protein complexes involved in maintenance of CIMP, and hence could be repressed in order to promote the CIMP phenotype.
The loss of mismatch repair function and development of MSI within sessile serrated adenomas with dysplasia is highly clinically relevant as these lesions evolve rapidly into invasive cancer, often in less than 12 months [
10].
BRAF mutant MSI colorectal cancers have an excellent 5 year survival of 84.6%, while microsatellite stable
BRAF mutant colorectal cancers have a significantly reduced 5-year survival of 40.5% [
9]. There is no evidence that the
MLH1–93 polymorphism makes an individual more likely to develop sessile serrated adenomas but if they do, the present study suggests the outcome is likely to be better if they carry the A allele, especially if they are homozygous AA because if a cancer develops it is likely to be MSI. However, other factors must also be important as a number of SSADs with loss of MLH1 expression possessed the GG genotype. These lesions may have polymorphisms in other regions of the genome modulating methylation at the locus, or possess other risk factors for
MLH1 promoter hypermethylation. Understanding other genetic and environmental risk factors that predispose a sessile serrated adenoma to MLH1 retention will aid in evaluating patients who are at risk of developing these particularly aggressive cancers, and may inform surveillance guidelines.
Conclusion
In conclusion, inheritance of the A allele is associated with a dose dependent increase in methylation at the
MLH1 promoter in dysplastic sessile serrated adenomas. The homozygous A genotype appears to
strongly predict the development of mismatch repair deficiency at the transition to dysplasia in this context. However, the A allele is insufficient to generate
MLH1 methylation and loss of protein expression in other cellular contexts, such as traditional serrated adenoma in the present study and in PBMCs as reported by Miyakura et al. [
12].
We propose that the MLH1 polymorphism is an important risk factor for development of MLH1 methylation but only in certain cellular environments such as sessile serrated adenomas and BRAF mutant colorectal cancers arising from sessile serrated adenomas. Collectively, these findings inform our understanding of the mechanism by which MLH1 methylation can occur in the setting of serrated colorectal neoplasia. Understanding the implications of germline polymorphisms in the epigenetic modulation of gene expression may inform screening guidelines and risk stratification for patients with sessile serrated adenomas.