Early onset MSI-H colon cancer with MLH1 promoter methylation, is there a genetic predisposition?

High frequency of microsatellite instability (MSI-H) is the hallmark of tumors with a mismatch DNA repair (MMR) deficiency. This deficiency leads to an accumulation of somatic mutations, especially in repetitive coding or non-coding DNA sequences (microsatellites) in the genome. MSI-H in colon cancer is found in the context of Lynch syndrome, previously known as hereditary non-polyposis colorectal cancer (HNPCC), in which germline mutations in one of four mismatch repair genes (primarily in MLH1 and MSH2 [1] and to a lesser extent in MSH6[2], PMS2 [3] or deletions in EPCAM/TACSTD1 (leading to MSH2 methylation) [4, 5] are found. Approximately 15% of cases are due to somatic biallelic or hemiallelic DNA methylation of the CpG-rich MLH1 promoter sequence, which is associated with gene silencing [6]. Colon cancers with sporadic MSI-H are observed more frequently in females and are often located proximal to the splenic flexure [7].

A clear association between increased age and occurrence of sporadic MSI-H colon cancer was described in 2002 by Young et al. [8]. The combination of age at diagnosis and three pathological features (tumor heterogeneity, peritumoral lymphocytes and tumor-infiltrating lymphocytes) allowed positive identification of 94.5% of MSI-H cancers as either Lynch syndrome or sporadic [8]. As normal aging colon mucosa shows global hypomethylation and specific hypermethylation of tumor associated genes, this epigenetic accumulation can explain the association between sporadic MSI-H colon cancer and older age [9‐11]. The rare sporadic cases diagnosed at a relatively young age can provide insight into the etiology of sporadic MSI-H. As young patients with sporadic MSI-H colon cancer are subjected to MLH1 methylation without this age-associated epigenetic accumulation, a defect of DNA methylation maintenance or direct targeting of MLH1 for methylation could be expected.

MLH1 methylation has been one of the hallmarks of the CpG island methylator phenotype (CIMP) since the phenotype was first described in 1999 [12]. The high levels of methylation found in CIMP-high colon tumors suggest that a causative genetic or epigenetic defect influences the spread and initiation of methylation [13]. Somatic BRAF mutations, MLH1 methylation and sporadic MSI-H are associated with CIMP-positive (CIMP-high and CIMP-low combined) colon tumors in which the bulk of aberrant methylation can be found [13, 14]. Aberrant methylation in CIMP-high tumors is thought to arise through an increase in de novo methylation. KRAS mutations have also been associated with elevated levels of aberrant DNA methylation, although discrepancies between marker panels and techniques showed variable levels of methylation. In general, KRAS mutations are associated with CIMP-low (also annotated CIMP2) colon tumors, in which increased levels of aberrant methylation can be detected to some extent, but at lower levels than in the CIMP-high tumors [13, 14].

The underlying causes leading to MLH1 promoter hypermethylation and subsequently to sporadic MSI-H colon cancer are still largely unknown. A relatively new concept in the field of genetics is germline epimutation. Although rare, multiple studies have described inherited and de novo germline methylation of MLH1 in patients with Lynch-like colon cancer [15‐21]. Cases with confirmed or probable MLH1 epimutations are documented to have the same range of tumors as described in Lynch syndrome patients, predominantly early-onset MSI colorectal cancer and endometrial cancer. Although possible, inheritance of the MLH1 epimutation is described as very weak, as the MLH1 epimutation is unstable in the germline [17, 18, 20, 21]. Paradoxically, patients suspected of having a genetic disorder based on a strong family history may be less likely to carry an epimutation [18]. Germline epimutations are thus highly suspected in young patients presenting with an MSI tumor without a clear family history. Increased risk of MSI-H tumors [22] and tumor-specific MLH1 methylation [23] might also be associated with a single-nucleotide polymorphism (SNP) -93 bp from the MLH1 transcription start site (rs1800734). This MLH1 G>A polymorphism is associated with increased age of onset and CIMP and BRAF mutations in individuals with MSI-H tumors [24].

Another possible factor contributing to aberrant DNA methylation is inactivation of GADD45A [25], although this finding was later disputed [26, 27]. GADD45A (growth arrest and DNA-damage inducible protein 45 alpha) is a nuclear protein involved in maintenance of genomic stability, DNA repair and cell growth suppression [28, 29]. A recent publication has found GADD45A to be a key regulator of active DNA demethylation in Xenopus oocytes and cell lines through a DNA repair-induced mechanism [25]. Specific short interfering RNA (siRNA)-mediated knockdown of GADD45A and GADD45B in the colon cancer cell line RKO induced hypermethylation of MLH1, THBS1 and p16, three genes known to be involved in carcinogenesis of different types of tumors by DNA methylation [25].

In contrast to MSI-H colon cancer, chromosomal instability (CIN) is enhanced and more pronounced in tumors with a low frequency of microsatellite instability (MSS or MSI-L tumors). This relationship can also be deduced from the observation that MSS/MSI-L tumors often are aneuploid, whereas MSI-H tumors mostly are peri-diploid. Lynch syndrome-associated MSI-H colon cancer hardly shows chromosomal copy number alterations, and the few alterations are mainly restricted to copy neutral LOH (cnLOH) at the mutated locus, especially in MLH1 mutated cases [30]. However, sporadic MSI-H colon cancer and MSI-H from patients with unclassified variants in MMR genes seem to show an enhanced (although subtle) number of chromosomal aberrations [30‐33].

We studied 46 MSI-H colon tumors showing loss of MLH1 expression and its heterodimer PMS2 and methylation of the MLH1 promoter. Pathogenic germline MMR mutation were excluded. We have primarily focused on comparing relatively young patients with patients of older ages to identify a possible cause for MLH1 methylation in young individuals with colon cancer. Tumors were characterized for somatic BRAF, KRAS, GADD45A and the MLH1 -93G>A polymorphism (rs1800734), as these genetic factors could play a causative role in MLH1 promoter methylation. Whenever material was available, germline MLH1 methylation status was studied and DNA sequencing for germline GADD45A mutations was performed. In order to analyze whether the younger patients exhibit an intrinsic higher methylation tendency in their genome, the methylation status of eight CIMP markers was determined in the tumors. In a selected subset of young patients, whole genome SNP array analysis was performed on formalin-fixed, paraffin-embedded (FFPE) tumor tissue and compared with previously published data to search for recurrent chromosomal aberrations involved in MLH1 methylation.

We characterized 13 MSI-H colon cancer cases with MLH1 promoter methylation from patients with an age of onset below 50 years and compared these data with those obtained from 33 MSI-H cases of patients over 50 years of age. Based on the Bethesda guidelines, which recommend MSI testing for all colorectal cancers in patients diagnosed before 50 years of age, we used the cutoff age of 50 for our comparisons [36]. The mean age of the study cohort was 61 years (SD is 31 years). The majority of tumors originated from the proximal colon (n = 36, 78% of total), while a low percentage (n = 5, 11%) of MSI-H tumors originated distal from the splenic flexure. All tumors showed loss of expression of nuclear MLH1 and its heterodimer PMS2, confirming the deleterious effect of MLH1 promoter methylation. Both MSH2 and MSH6 stained positive and pathogenic germline mutations in any of the four mismatch repair genes were identified in none of the patients.

Since CpG island hypermethylation (including MLH1) in colon mucosa is considered to be age-related [9], the finding of hypermethylation of MLH1 at a younger age is unexpected. Since 2002, several manuscripts pointed to the existence of MLH1germline methylation [15‐21]. More recently, MSH2 methylation due to an inherited deletion in the 3'end of EPCAM/TACSTD1 was also discovered [4]. Methylation of MLH1 can also be found in addition to a germline MMR mutation, as described by Rahner et al. [45]. We studied 13 MMR germline mutation-negative patients with MSI-H colon cancer (mostly right-sided) at ages of onset under 50 years. These data were compared with those obtained from a control group of 33 patients with an age of onset above 50. The presence of (somatic) promoter methylation of MLH1 in the tumors made Lynch syndrome unlikely. We identified two female patients with ages of onset of 33 and 60 years harboring germline MLH1 methylation. Relatively young patients without a strong family history who present a MSI-H tumor with loss of MLH1 and PMS2 protein expression are suggested as candidates for MLH1 germline epimutation screening [17, 21]. We identified one patient with germline MLH1 methylation in seven tested cases who were less than 50 years of age, giving a frequency of ~14%. Although the low number of tested samples in this study makes this percentage not representative, this number is not significantly higher than the frequency range of 0.6-13% described in studies screening for germline MLH1 methylation in Lynch syndrome-suspected patients [5]. The discovery of germline MLH1 methylation in a patient aged 60 years at diagnosis is surprising, as the patients with germline MLH1 methylation described prior to this study (n = 25) have a mean age of diagnosis of 37 years with a range of 17-46 [5].

In contrast to the group with an age of onset above 50 years, only some (4/11) of the MLH1 methylated MSI-H tumors from patients below 50 years showed high levels of CIMP marker methylation (CIMP-high). For the patient group with an age of onset below 50 years the CIMP-high status completely overlapped with BRAF mutations. As both BRAF and KRAS mutations have been observed in the earliest identified colonic neoplasms, and recent papers have provided evidence that induction of the ras oncogenic pathway will result in DNA hypermethylation, a causative effect of BRAF/KRAS mutations is likely [24, 46‐50]. Instead of widespread CpG island methylation in non-BRAF mutated tumors in the early onset patient group, methylation seems to be largely restricted to the MLH1 locus. Although the existence of locus-restricted methylation may be a reflection of the Gaussian curve of methylation patterns in relation to age, this finding may suggest a distinct, non-BRAF associated mechanism of MLH1 methylation. However, all tumors here were selected upon MLH1 promoter methylation which may explain the fact that MLH1 is methylated more frequent than all other CIMP genes. As the methylation mechanism is (at least partly) age related, and progressive, a similar selection of tumors methylated on one of the other CIMP markers would have also shown more frequent methylation on these than other CIMP markers including MLH1 and occurring in tumors not reaching the CIMP-high classification yet. A progressive methylation and CIMP appearance according to age similar as that shown in Additional file 3: Table S2 favors the argument that MLH1 methylation in these young patients is a reflection of the Gaussian curve of methylation patterns in relation to age.

An alternative hypothesis concerning the association between BRAF mutation and DNA methylation is that promoter methylation and silencing of specific target genes such as IGFBP7 by promoter methylation could favor the selection of activating BRAF mutations, since the oncogenic effect of activated BRAF would be enhanced in the absence of IGFBP7's inhibitory function [51]. Since promoter hypermethylation is partly age related the occurrence of IGFBP7 hypermethylation and BRAF mutation would also explain the diminished occurrence in the young sporadic MSI-H patient group [50]. This role of BRAF in aberrant methylation initiation will have to be elucidated in the future. The locus-specific, non-BRAF associated mechanism of MLH1 methylation suggested in our study should be addressed in a larger group of early onset sporadic colon cancer patients with MLH1 methylation to provide additional insights.

In patients of older ages, there is an association between somatic MLH1 methylation and the MLH1 -93G>A polymorphism [22, 24, 49]. Indeed, when we compared our group of patients above 50 years of age with the published control groups of Raptis et al., and Samowitz et al., we observed an enrichment of the A allele. We explored the possibility that the A allele was more prevalent in the sporadic MSI-H at early ages. However a similar distribution in both age groups was found, no significant enrichment could be found for the cases under 50 years. The hypothesis of Samowitz et al., which suggests an increased likeliness of MLH1 methylation in the presence of a CIMP/BRAF mutation background and a MLH1 -93 G>A polymorphism, excludes young onset patients because of low levels of BRAF mutations [24]. Although Samowitz did find a significant difference in A allele distribution between MSI-H colon cancer age groups, our cohort of sporadic MSI-H colon cancer patients with MLH1 methylation excluded patients with a germline MMR gene mutation, which might explain the difference found between our studies.

Knockdown and overexpression experiments of GADD45A in Xenopus laevis led to the suggestion that deregulation of GADD45A's role in active DNA demethylation could give rise to aberrant methylation [25]. The absence of pathogenic somatic and germline mutations in human GADD45A observed in our study and data published during this study [26, 27] suggest that a role for GADD45A mutations in aberrant hypermethylation in human colon tumors is unlikely.

In a subset of tumors (including five with an age of onset under 50 years), whole genome SNP array analysis of FFPE tumor tissue was used to assess possible causative loci for MLH1 methylation. Our copy number and cnLOH analysis identified patterns in agreement with literature describing limited chromosomal instability in sporadic MSI-H colon tumors with MLH1 methylation. The extent of copy number abnormalities (CNA) identified here is in agreement with that found by Trautman et al. and by van Puijenbroek et al. [30, 33]. In patients under 50 years, no specific genomic pattern was identified, although two cases showed overlapping alterations at chromosome 4q. The smallest region of overlap (region 4q35.1-4q35.2) encompasses the cancer associated genes TLR3, CDKN2AIP, ING2, CASP3 and SORBS2, none of which are thought to cause aberrant DNA methylation. The four regions of cnLOH that showed infrequent overlap in the 15 tumors tested are not known as such. The cnLOH of 3p21.31-26.3, found in a 44 and a 62 year old, encompasses the 3p22.2 region where MLH1 is located. Such cnLOH is not typical for sporadic MSI-H colon carcinomas, but is more readily found in tumors containing pathogenic MLH1 mutations [30]. We can not rule out that the identified cnLOH regions may harbor loci involved in MLH1 methylation. However, the odds are against such a suggestion.

Although our study did not identify a cause for MLH1 methylation in sporadic MSI-H colon cancer with an age of onset below 50 years, we observed methylation to be almost restricted to the MLH1 locus in patients without a BRAF mutation. We show that this early onset group consists of two sub-groups: those which are CIMP-high and contain a BRAF mutation (resembling sporadic MSI-H in the older age group to a great extend) and those with wild-type BRAF and limited methylation in addition to MLH1 methylation.

Genomic analysis did not provide recurrent aberrations leading to identification of a possible cause of MLH1 methylation in the cases under 50 years. Lastly, we excluded a role for somatic and germline GADD45A mutations in the tumorigenesis of early onset sporadic MSI-H colon cancer.