Hereditary nonpolyposis colorectal cancer/Lynch syndrome is characterized by a dominantly inherited predisposition to early onset colorectal carcinoma and certain extracolonic tumors. The syndrome is caused by germline mutations in DNA mismatch repair (MMR) genes, predominantly MLH1 and MSH2.1 More than 1,500 different variants in MMR genes have been reported, approximately half of which may be pathogenic.2, 3 In target tissues developing cancer, germline mutation is typically accompanied by a second “hit”, which may be somatic deletion (loss of heterozygosity), point mutation or promoter methylation of the remaining wild-type allele.4 Identification of inherited predisposition is important because it enables targeted clinical surveillance, which significantly reduces cancer morbidity and mortality in Lynch syndrome families.5 Conventional screening by exon-specific methods, such as sequencing, fails to detect a pathogenic change in 14–51% of families meeting the Amsterdam criteria,6, 5 and the reported mutation-negative fractions may vary between 30 and 92% in families not meeting these criteria.7–9 Among consecutive colorectal cancer patients fulfilling the revised Bethesda guidelines,10 and showing evidence of MMR deficiency in tumor tissue, ∼90% may remain mutation-negative.11
New technical advances, such as conversion technology,12 semiquantitative multiplex PCR13, 12 and multiplex ligation-dependent probe amplification (MLPA),14 in addition to the more traditional Southern blot hybridizations,15 have led to the identification of large genomic rearrangements, primarily deletions, in a variable proportion of point-mutation negative cases and families. Moreover, epigenetic mechanisms may account for cancer susceptibility in occasional cases, as suggested by the recent discovery of germline epimutations in MLH116 and MSH2.17
We addressed large genomic rearrangements and constitutive promoter hypermethylation as explanations for families/cases lacking germline mutations in MMR genes by conventional screening but displaying loss of MLH1, MSH2 or MSH6 expression in tumor tissue. We also evaluated if different types of MMR gene changes were associated with distinct clinicopathological features within the present study series and relative to families in which MMR gene point mutations had been identified previously.
Our material consisted of a research cohort and a clinic-based cohort (Fig. 1). Each index patient represented a separate family except for one (2 index individuals existed for family F181 from the clinic-based cohort). The research cohort included 11 putative or verified Lynch syndrome families from the nation-wide Hereditary Colorectal Cancer Registry of Finland.18 Among the original 81 kindreds ascertained through family history suggestive of Lynch syndrome, these 11 families showed no detectable germline changes by exon-specific screening methods19; yet, their disorder was found to be linked to the major MMR genes based on aberrant MMR protein expression and microsatellite instability (MSI) in tumor tissue.20 Besides point mutations, Mutation 1, a prevalent founder mutation in the Finnish population that consists of a 3.5-kb genomic deletion of MLH1 Exon 16 and flanking introns had been excluded by a specific test.21 Among the families, 45% fulfilled the Amsterdam Criteria I or II6, 22 and 55% at least 1 criterion included in the revised Bethesda guidelines.10
The clinic-based cohort (n = 34) consisted of patients referred to clinical genetic units of 5 hospital districts in Finland for suspected Lynch syndrome. Suspicion of Lynch syndrome was based on clinical features (21% met the Amsterdam criteria, 65% the revised Bethesda criteria and 15% none of these criteria), combined with the loss of MLH1, MSH2 or MSH6 protein expression (and MSI) in tumor tissue. All patients had undergone full exon sequencing of the implicated MMR genes and specific testing for Mutation 1 in a diagnostic laboratory at the Helsinki University Central Hospital and showed no mutations. All mutation-negative index individuals identified in the diagnostic laboratory during September, 2004–September, 2007, who consented to participation in the present investigation were included.
Information on index patients (n = 51) from verified Lynch syndrome families from the Hereditary Colorectal Cancer Registry of Finland, all previously diagnosed with point mutations in MLH1 (n = 41) or MSH2 (n = 10), was used for clinical comparisons. The families met the Amsterdam or revised Bethesda guidelines (Table II below).
DNA was extracted from lymphocytes and lymphoblasts as described23 and from formalin-fixed paraffin-embedded specimens of tumor and matching normal tissues according to Isola et al.24 The appropriate institutional review boards of the Helsinki University Central Hospital and Jyväskylä Central Hospital approved this study.
Immunohistochemical staining for MMR proteins was performed as described20 using the following mouse primary antibodies; anti-MLH1 (clone G168-15, Pharmingen, San Diego, CA), anti-MSH2 (clone FE-11, Calbiochem/Oncogene Research, San Diego, CA) and anti-MSH6 (clone 44, Transduction Laboratories, San Diego, CA).
Microsatellite instability analysis
For the research cohort, MSI status was determined using the Bethesda panel (BAT25, BAT26, D5S346, D2S123 and D17S250).25 For the clinic-based cohort, BAT25 and BAT26 were used because the sensitivity and specificity of these 2 mononucleotide repeat markers, or BAT26 alone, to diagnose high-degree instability has been found to be comparable to all 5 Bethesda markers together.26 Tumors were categorized into microsatellite-unstable (MSI, practically equivalent to MSI-high) or microsatellite-stable (MSS); we did not aim to distinguish MSI-low from MSS because MSI-low shows a generally poor correlation with MMR protein deficiency.27 Two or more markers among the entire Bethesda panel (research cohort) or either of the 2 mononucleotide repeats (clinic-based cohort) had to be unstable for the diagnosis of MSI. Tumors were considered MSS if fewer markers or none were unstable.
Single nucleotide primer extension analysis
For allele-specific expression analysis, as well as for loss of heterozygosity analysis, single nucleotide primer extensions utilizing an A/G polymorphism at nucleotide 655 in Exon 8 of MLH1 were performed as described in Renkonen et al.20 and Ollikainen et al.4
Multiplex ligation-dependent probe amplification
MLPA kits from MRC-Holland (Amsterdam, the Netherlands) were used to screen DNA samples for large genomic rearrangements in MLH1 and MSH2 (SALSA MLPA kit P003) and in MSH6 (SALSA MLPA kit P008) following the manufacturer's instructions (http://www.mrc-holland.com28). Normal DNA specimens derived from lymphocytes from healthy controls were included in every assay. For each MLPA reaction, 100 ng of DNA was used. Fluorescently labeled PCR products from MLPA reactions were analyzed on an ABI 3730 sequencer using the Genemapper software, version 3.0 (Applied Biosystems, Foster City, CA). A reduction of 40% or more in the peak area when compared with normal controls suggested a single-copy deletion.
Long-range genomic PCR
To determine the breakpoints for the MLH1 deletion observed in F36 by MLPA, long-range PCR was carried out using the Expand Long Template PCR System (Roche Diagnostics, Mannheim, Germany). The region between 373 bp upstream of Exon 1 and c.256 in Exon 3 was amplified with primers 5′-GATCACCT CAGCAGAGGCAC-3′ (forward) and 5′-GCAGTTTACTAG TAGTGAACC-3′ (reverse), using conditions recommended by the manufacturer. Upon separation on a 0.8% agarose gel, fragments specific for the mutant allele (1.1 kb) and wild-type allele (7.5 kb) appeared. The product corresponding to the mutant allele was cut out, purified and sequenced.
The SALSA methylation-specific (MS)-MLPA ME011 kit (MRC Holland, Amsterdam, the Netherlands) was used to detect methylation in MMR gene promoter regions using probes that contain a digestion site (or occasionally 2 digestion sites) for the methylation-sensitive HhaI enzyme. All reactions were carried out and results analyzed according to the manufacturer's instructions (http://www.mrc-holland.com). The kit includes 5 probe pairs for MLH1 (with the respective HhaI sites located at −638, −402, −251/−245, −8 and +220 relative to the initiating ATG; GenBank accession number U26559), 3 probe pairs for MSH2 (HhaI positions at −264, −200 and +133/+148 relative to ATG; AB006445) and 3 probe pairs for MSH6 (HhaI positions at −293, −100/−143 and −64 relative to ATG; U73732). Normal DNA specimens derived from lymphocytes from healthy controls were included in every assay. For each MLPA reaction, 150 ng of DNA was used. Dosage ratio was calculated as the peak area of a given probe divided by the sum of the peak areas of all control probes in HhaI-digested sample relative to undigested sample.29 A dosage ratio of 0.15 or higher, corresponding to 15% of methylated DNA, was interpreted to indicate promoter methylation.30
Bisulphite modification and sequencing
DNA (1 μg) was modified using the CpGenome DNA Modification Kit (Chemicon, Temecula, CA). For methylation analysis of the MLH1 promoter region, a fragment from −370 to −49 relative to the initiating ATG was amplified from bisulphite-modified DNA using methylation-unbiased primers MLH1 DEG 5′ and MLH1 DEG 3′ from Suter et al.16 In F36 with a deletion of Exons 1–2 and flanking areas, these primers amplified the wild type allele only because the MLH1 DEG 3′ region was lost. For analysis of the deletion-containing allele, the MLH1 DEG 5′ primer was combined with a methylation-unbiased reverse primer (5′-ACAATTTACTAATAATAAACCTTTCACATAC-3′) from Intron 2 beyond the deletion. The PCR products obtained with methylation-unbiased primers were sequenced to determine the methylation status.
To examine allele-specificity of methylation in N2, bisulphite-converted DNA amplified with primers MLH1 DEG 5′ and MLH1 DEG 3′ was first cloned into a pCR2.1-TOPO vector using the TOPO TA Cloning System (Invitrogen, Carlsbad, CA), followed by sequencing of DNA extracted from clones.
For methylation analysis of the MSH2 promoter, a region from −651 to −42 relative to initiating ATG was evaluated in 2 overlapping fragments, using methylation-unbiased primers NP1-F and NP1-R from Chan et al.,17 for distal area, as well as primers 5′-GGGTTTTAAGTTTTGTAGTTGAGTAAAT-3′ (forward) and 5′-CCCACACCCACTAAACTATTTCC-3′ (reverse) for proximal area.
Methylation-specific PCR (MSP) was performed using HotStarTaq DNA polymerase (Qiagen, Valencia, CA) in a final reaction volume of 25 μl containing 24 ng of bisulphite modified DNA. The manufacturer's standard protocol was applied to design cycling conditions for HotStarTaq polymerase. Primers from Kim et al.31 were used for methylation analysis of the MLH1 promoter, where primers for the methylated reaction (M) covered a region from −269 to −206 and those for the unmethylated reaction (U) a region from −288 to −192 relative to the initiating ATG. The PCR reaction consisted of 36 cycles, and the annealing temperature was 55°C and 57°C for the M and U reactions, respectively. PCR products were visualized on a 2% agarose gel with UV transillumination. The RKO cell line with verified MLH1 promoter methylation and lymphocyte-derived DNA from a healthy control served as positive and negative controls, respectively.
The t-test for independent samples or Fisher's exact test (2-tailed) was used to evaluate the statistical significance of differences between the groups.
This study included 45 index patients suspected of having Lynch syndrome, who were ascertained as described in material and methods and in Figure 1. All individuals were mutation-negative by genomic sequencing and specific testing for a prevalent founder mutation (Mutation 1), and lacked MLH1, MSH2 or MSH6 protein in tumor tissue. Table I shows the essential clinical and molecular data for each index patient. In the entire series, the average age at onset of the first Lynch-syndrome associated cancer (which was colorectal cancer in 82%) was 49 years. Twenty-seven percent met the Amsterdam criteria and 62% the revised Bethesda criteria. The most important (although not statistically significant) differences between the research and clinic-based cohorts were later age at onset, lower frequency of families fulfilling the Amsterdam criteria, and higher relative share of tumors lacking MLH1 in the clinic-based cohort.
Table I. Clinicopathological Features and Molecular Findings of the Index Patients Organized According to Immunohistochemical Abnormality
C, clinic-based cohort; R, research cohort; AC, Amsterdam criteria 1 and 2; B, Bethesda guidelines 1–5; N, normal result; ND, not done.
Contiguous exons; 2, co-segregation with cancer phenotype; 3, reverse-transcription-PCR analysis of RNA; 4, allele-specific expression analysis by single nucleotide primer extension; 5, long-range genomic PCR.
Skin 25 years & 26 years, rectum 42 years & 44 years, endometrium 45 years, colon 50 years
MSH2 del ex 9–10
1, 2, 3
Endometrium & ovary 48 years, colon & rectum 55 years
Colon 50 years
B: 1, 3
Colon 41 years
B: 1, 3
Colon 56 years
Large genomic rearrangements
Twelve out of 45 patients (27%) were found to have large genomic rearrangements, including 4 patients (4/11, 36%) from the research cohort and 8 patients (8/34, 24%) from the clinic-based cohort (Fig. 1, Table I). All rearrangements were deletions comprising 1 or several exons or the entire gene. The deletion of MLH1 Exons 1–2 in F36 was fully characterized on the nucleotide level (see below). In other cases, deletions were confirmed by other methods (Table I). Deletions were present in 3/25 (12%), 9/16 (56%) and 0/4 (0%) among index patients lacking MLH1, MSH2 or MSH6 expression, respectively. The difference in deletion frequency between MLH1 and MSH2 was statistically significant (p = 0.0043).
Two patients with MLH1 protein loss in tumor tissue (1 from each cohort) showed a germline epimutation in MLH1 (Fig. 1). Of the 5 tested CpG-containing regions, 3 were methylated in F36 (Fig. 2a) and all in N2 (Fig. 2b) (the probes are numbered from 5′ to 3′ terminus so that 1 is located the furthest away in the promoter region, see material and methods and Figures 3a and 3b, below). The epimutations identified by MS-MLPA were independently confirmed by 2 other methods, MSP (Fig. 2c) and bisulphite sequencing of the MLH1 promoter region (data not shown). Among the index patients with MSH2 protein loss in tumor tissue, only 1 (F88; Table I) displayed promoter methylation in a region (HhaI site -200 relative to the initiating ATG) that has not been associated with lost expression of MSH2 in the literature, whereas 2 other HhaI-containing regions (−264 and +133/+148) were intact. Subsequent bisulphite sequencing of the MSH2 promoter region around HhaI sites monitored by MS-MLPA showed no evidence of methylation at the surrounding CpG sites. There were no cases of MSH6 promoter methylation. The overall frequency of epimutations was therefore 2/45 (4%) in our series.
Patient F36 was diagnosed with colorectal cancer at age 22 and 45 and with endometrial cancer at 40 years (Table I). MS-MLPA performed on blood, normal colonic mucosa and colorectal cancer revealed MLH1 methylation with Probes 1–3 corresponding to HhaI sites between −638 and −245 in the promoter region, whereas Probes 4 and 5 (−8 and +220) yielded no methylation signal (Fig. 2a). For the methylated sites, the methylation dosage ratio was around 0.5 in blood and normal mucosa, compatible with monoallelic methylation, but close to 1 in tumor tissue, implying somatic loss of the unmethylated allele. The latter suggestion was confirmed by a primer extension analysis utilizing a common polymorphism at c. 655 of MLH1, which revealed loss of the wild-type allele (G) in tumor DNA (with an A to G ratio of 3.1 in tumor relative to normal DNA; in a balanced situation the value is 1). We have previously demonstrated a complete loss of mRNA expression from allele A in this patient, indicating that A is specific to the mutant allele.20
Long-range PCR on genomic DNA from blood of the same patient subsequently showed that the regions not showing methylation by MS-MLPA were located within a large genomic deletion occurring in the same patient (Figs. 3a and 3b). In this deletion, a C (AGGCAC) in the promoter region, 68 bp upstream of Exon 1, was joined to a C (CAGGAC) in Intron 2, 709 bp upstream of Exon 3, resulting in the deletion of Exons 1 and 2. The deletion was apparently mediated by a shared sequence (CAC) at the site of the breakpoint. Bisulphite sequencing of fragments specific to the deletion-containing and normal alleles (see Material and methods) confirmed that the methylation observed in the region between −638 and −245 by MS-MLPA indeed originated from the mutant allele (data not shown). In conclusion, patient F36 had both a large genomic deletion and a germline epimutation, which overlapped in the same allele, and the observed loss of expression from the mutant allele20 could result from either change.
Patient N2 was diagnosed with colon cancer at 35 years of age and with endometrial cancer at 45 years. All 5 tested CpG-containing regions were methylated in blood, normal colonic mucosa and colorectal tumor, and the dosage level (around 0.5) suggested monoallelic methylation (Fig. 2b). Tumor tissue did not reveal any significant loss of heterozygosity (data not shown), which was compatible with the observation of similar methylation levels in neoplastic and non-neoplastic tissues. To formally prove that methylation affected one allele and left the other allele intact blood DNA from patient N2 was subjected to bisulphite conversion, after which the promoter region was amplified with methylation-unbiased primers,16 cloned and sequenced. Although the patient was homozygous for a promoter polymorphism (A/G at −93), which could therefore not be used as an allele-specific marker, sequencing of the individual clones (n = 17) showed that either all CpG sites (6 clones) or none (11 clones) were methylated, proving our hypothesis. The patient was heterozygous for the c.655 polymorphism in MLH1 Exon 8, which made it possible to investigate the mRNA expression pattern of the 2 MLH1 alleles in blood; only 1 allele was expressed (Fig. 4), in accordance with the interpretation that the other allele was silenced by methylation.
The present index patients with abnormal MLH1 or MSH2 expression were divided into groups with vs. without large genomic deletions and compared to index individuals from our nation-wide registry known to carry MLH1 or MSH2 point mutations (Table II). In both MLH1- and MSH2-associated groups, deletion patients showed clearly stronger family histories than those without deletions. When compared to patients with point mutations, deletion patients did not differ in any essential respect. An interesting feature of MSH2 point mutation carriers was the predominance of distal over proximal colorectal tumors, and the same was true for patients with large genomic deletions of MSH2 (p = 0.021 for difference between MLH1 and MSH2, with point mutation and large deletion groups combined).
Table II. Clinical Comparison of the Present Patients With and Without Deletion in MLH1 OR MSH2, and Relative to Patients with Point Mutations
Ratio of colonic to extracolonic tumors in index patients
Ratio of proximal to distal colorectal cancer in index patients
The MLH1-associated subgroup without genomic deletion in Table II includes 1 patient with epimutation (N2) as the sole demonstrable germline defect. Some features of this patient were typical of Lynch syndrome, such as 2 different cancers of the Lynch syndrome spectrum, both diagnosed at an early age (Table I). The family history, however, was atypical because a parent had died of peripheral nerve sheath tumor, but no Lynch syndrome-associated cancers were present in any first or second-degree relative. Presentation as apparently sporadic cases may characterize germline epimutation carriers in general (see discussion below).
Our study was based on families/cases in which clinical and immunohistochemical findings gave a reason to suspect Lynch syndrome despite the absence of MMR gene point mutations. The present frequency of large genomic deletions, 27%, which is already in the upper range of deletion rates reported for comparable series in the literature14, 32–34 does not include a prevalent founder mutation (Mutation 121) that had been excluded at the outset. Among the original 81 families from which the present research cohort of 11 families originated, 29 are known to harbor Mutation 1; additionally, 1 family has a genomic deletion of MLH1 Exons 3–519 that was independently diagnosed in another member of the same family as part of the present clinic-based cohort. Combined with the 4 new deletions presently identified in the research cohort (Fig. 1), the overall deletion frequency rises to 34/81 (42%), which may be the highest frequency of large genomic rearrangements reported in any cohort to date. In other populations, too, certain genomic deletions may be enriched because of founder effect, e.g., MLH1 Exon 11 deletion in China35 and MSH2 Exon 1–6 deletion in North America.9
Repetitive elements, such as Alu repeats, promote recombinations in MMR genes as shown for Mutation 121 and other cases.9, 36, 37 Alu repeats are present in almost every intron of MLH1 and MSH236 and offer a likely explanation for several of the present observations, such as the scattering of genomic deletions throughout the genes, the existence of some potential hot spots for rearrangements (e.g., MLH1 Introns 3 and 6 and MSH2 Introns 7 and 8), and the more frequent involvement of MSH2 than MLH1. F36 that we describe represents a novel case of 2 overlapping events, deletion and epimutation. The chronological and mechanistic relationship between these events is unknown. Because DNA methylation is connected to transcription, it is possible that the loss of essential transcriptional elements because of the deletion triggered secondary hypermethylation. On the other hand, if epimutation was the primary event, it might have induced a conformational change that facilitated recombination and consequent deletion.
Screening MLH1, MSH2 and MSH6 promoters for constitutive hypermethylation, that is epimutations, resulted in the detection of MLH1 epimutation in 2 patients from our material (4%), and consequences on allele-specific mRNA expression were verified by primer extension assays. A limited number of studies evaluating the incidence of constitutive MLH1 promoter hypermethylation in comparable series of suspected Lynch syndrome patients are available and have arrived at frequencies between 0.6% and 13%.16, 38–41 We could not find any incidence figures for MSH2 or MSH6 epimutations in the published literature, but at least 1 family with MSH2 epimutation is known to exist.17 Three studies17, 40, 42 provide evidence of heritability or transgenerational epigenetic inheritance of germline epimutation. Hitchins et al.40 report the reversion of the affected allele to the normal active state in the offspring. MLH1 epimutation cases detected by Morak et al.42 are compatible with several different patterns (heritability, de novo methylation, mosaic or incomplete methylation, no evidence of heritability). The remaining investigations indicate the absence of epimutation in other family members.16, 39, 41, 43 In our study, none of 5 first- or second- degree relatives of patient N2 had the epimutation (data not shown). This complex behavior of epimutations is compatible with the observation that these are only seldom, if at all, associated with classical Lynch syndrome families.16, 17, 38–41
No genetic or epigenetic germline defect was detectable in almost half of our cases with lost MSH2 protein and in a majority of MLH1 or MSH6-associated cases, leaving their genetic aetiology unknown. Certain complex structural alterations (e.g., inversions)44 as well as regulatory changes in non-coding regions45 remain a theoretical possibility. Isolated loss of PMS2 protein in tumor tissue may point to germline mutations in PMS2, which may account for 5% of Lynch syndrome cases.46 In our study, PMS2 was unlikely because all cases with absent PMS2 also lacked MLH1, suggesting secondary instability of PMS2.47 We mostly did not have tumor samples available from several affected family members to be able to evaluate concordance for immunohistochemical staining patterns within families. Given that most of our cases met the revised Bethesda guidelines at best it is feasible that MMR gene silencing in tumor tissue occasionally resulted from entirely somatic mechanisms. We addressed the latter issue in our research cohort, from which we had the necessary tumor samples available, and found that at least for MLH1, somatic hypermethylation is a plausible mechanism (data not shown).
In conclusion, we demonstrate that large genomic deletions (mainly MSH2) may explain a significant proportion and germline epimutations (in MLH1) a smaller fraction of point mutation-negative families/cases with MMR protein loss in tumor tissue. In contrast to genomic deletions, which are associated with strong family histories for Lynch syndrome, epimutations occur in patients with multiple early-onset tumors without any significant family history. The identification of such defects is important for the diagnosis, counseling and management of the patients and their families.
The authors thank Mrs. Kirsi Pylvänäinen, Mrs. Tuula Lehtinen, Mrs. Katja Kuosa and Ms. Outi Kajula for sample collection, and Ms. Saila Saarinen for laboratory analyses. They also thank Dr. Michael Woods, Dr. Elise Renkonen and Dr. Lauri Aaltonen for helping in some initial mutation analyses.