Article Text

Download PDFPDF

Original article
Tumour MLH1 promoter region methylation testing is an effective prescreen for Lynch Syndrome (HNPCC)
  1. K Newton1,
  2. N M Jorgensen2,
  3. A J Wallace2,
  4. D D Buchanan3,4,5,
  5. F Lalloo6,
  6. R F T McMahon7,8,
  7. J Hill1,
  8. D G Evans6
  1. 1Department of General Surgery, Manchester Royal Infirmary, Central Manchester University Hospitals NHS Trust, Manchester, UK
  2. 2Genomic Diagnostics Laboratory, Manchester Centre for Genomic Medicine, Central Manchester University Hospitals NHS Foundation Trust, Saint Mary's Hospital, Manchester, UK
  3. 3Cancer and Population Studies Group, Queensland Institute of Medical Research, Herston, Queensland, Australia
  4. 4Oncogenomics Group, Genetic Epidemiology Laboratory, Department of Pathology, The University of Melbourne, Parkville, Victoria, Australia
  5. 5Centre for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, The University of Melbourne, Parkville, Victoria, Australia
  6. 6Manchester Centre for Genomic Medicine, Central Manchester University Hospitals NHS Foundation Trust, Saint Mary's Hospital, Manchester, UK
  7. 7Department of Histopathology, Manchester Royal Infirmary, Central Manchester University Hospitals NHS Foundation Trust, Manchester, UK
  8. 8Manchester Medical School, University of Manchester, Manchester, UK
  1. Correspondence to K Newton, Genomic Diagnostics Laboratory, Manchester Centre for Genomic Medicine, Central Manchester University Hospitals NHS Foundation Trust, Saint Mary's Hospital, Oxford Road, Manchester M13 9WL UK; Katynewton2012{at}doctors.org.uk

Abstract

Background and aims Lynch syndrome (LS) patients have DNA mismatch repair deficiency and up to 80% lifetime risk of colorectal cancer (CRC). Screening of mutation carriers reduces CRC incidence and mortality. Selection for constitutional mutation testing relies on family history (Amsterdam and Bethesda Guidelines) and tumour-derived biomarkers. Initial biomarker analysis uses mismatch repair protein immunohistochemistry and microsatellite instability. Abnormalities in either identify mismatch repair deficiency but do not differentiate sporadic epigenetic defects, due to MLH1 promoter region methylation (13% of CRCs) from LS (4% of CRCs). A diagnostic biomarker capable of making this distinction would be valuable. This study compared two biomarkers in tumours with mismatch repair deficiency; quantification of methylation of the MLH1 promoter region using a novel assay and BRAF c.1799T>A, p.(Val600Glu) mutation status in the identification of constitutional mutations.

Methods Tumour DNA was extracted (formalin fixed, paraffin embedded, FFPE tissue) and pyrosequencing used to test for MLH1 promoter methylation and presence of the BRAF c.1799T>A, p.(Val600Glu) mutation 71 CRCs from individuals with pathogenic MLH1 mutations and 73 CRCs with sporadic MLH1 loss. Specificity and sensitivity was compared.

Findingss Unmethylated MLH1 promoter: sensitivity 94.4% (95% CI 86.2% to 98.4%), specificity 87.7% (95% CI 77.9% to 94.2%), Wild-type BRAF (codon 600): sensitivity 65.8% (95% CI 53.7% to 76.5%), specificity 98.6% (95% CI 92.4% to 100.0%) for the identification of those with pathogenic MLH1 mutations.

Conclusions Quantitative MLH1 promoter region methylation using pyrosequencing is superior to BRAF codon 600 mutation status in identifying constitutional mutations in mismatch repair deficient tumours.

  • Genetics
  • Cancer: colon

Statistics from Altmetric.com

Introduction

Lynch syndrome (LS) is responsible for 3%–4% of all colorectal cancer (CRC) and is the most common cause of hereditary CRC.1 ,2 It is caused by mutations in one of the DNA mismatch repair (MMR) genes MLH1, MSH2, MSH6 and PMS2. Mutations result in MSI-H (microsatellite instability high) cancers. Identification of families with LS is necessary to initiate screening and to reduce CRC mortality.3–5

Diagnosis of LS is complicated by the expense and time-consuming nature of constitutional mutation analysis. Family history criteria and tumour-derived biomarkers are used to prescreen to select patients for germline testing. The Amsterdam II criteria were designed to select research families for linkage analysis. They are currently used, somewhat inappropriately, for clinical purposes to select individuals at high risk of having a MMR gene mutation. Patients who meet these criteria have at least 60% chance of a mutation.6 These criteria are inherently specific but consequently have low sensitivity. Much work has been done over the last decade to improve the identification of non-Amsterdam Lynch families. The revised Bethesda guidelines described in 20047 are sensitive but have low specificity. They have been criticised for being overly complicated and are little used in clinical practice.8 Tumour MSI and MMR protein immunohistochemistry (MMR IHC) are currently used in conjunction with the revised Bethesda guidelines (or other medium risk criteria). The sensitivity of MSI is 89% for MLH1 and MSH2, but less than 80% for MSH6 and PMS2, with a specificity of 90% for all genes.9 MSI testing is impractical for population-based screening due to the need for a molecular genetics laboratory. MMR IHC may be preferable in patients meeting Bethesda guidelines because of the low sensitivity of MSI for detecting MSH6 and PMS2 gene mutation carriers. MMR IHC has a sensitivity of 100% and a specificity 91.5% for the detection of MLH1 carriers, a sensitivity 87.5% and specificity of 88.5% for the detection of MSH2 carriers.10 Prescreening of all newly diagnosed CRCs (population-based) is used in some specialist centres in the USA and Europe (none in the UK) in order to identify families not meeting clinical criteria for LS. A multicentre study of over 10 000 newly diagnosed CRC probands, found that MMR tumour testing was the most effective strategy for the identification of mutation carriers (sensitivity 100%, specificity, 93.0%, diagnostic yield 2.2% compared to use of the Bethesda guidelines; sensitivity 87.8% specificity, 97.5%, diagnostic yield, 2.0% p<0.001).11

There are two independent molecular pathways which lead to MSI-H (MMR deficient) CRC.12 MSI-H cancers occur in LS and also as a result of epigenetic silencing of the MLH1 gene through hypermethylation of its promoter. This occurs in around 13% of sporadic CRCs.12 These cancers are also associated with the BRAF c.1799T>A, p.Val600Glu mutation13 and are not familial. LS cancers are characterised by MSI-H, a normal (unmethylated) MLH1 gene promoter region, and wild-type (wt) BRAF (ie, c.1799T, p.Val600). MMR IHC is able to effectively identify patients for MSH2, MSH6 testing. Sporadic defects in these genes are rare, so protein loss is highly indicative of a constitutional abnormality. However, MSI and MMR IHC are not specific enough to identify MLH1 constitutional mutation carriers because of this large group of sporadic cancers with MLH1 deficiency. A method of differentiating between these groups of cancers is required.

BRAF mutation testing has been suggested. The methodology is well established and is currently in use in some centres. However, BRAF testing has low specificity. MLH1 promoter region methylation testing is attractive as a better prescreen test. Methylation is thought to be the first step in pathogenesis of cancers with sporadic loss of MLH1, and is thought to be rare in Lynch cancers.14 Lack of methylation should, therefore, be more specific for the identification of constitutional mutation carriers. Constitutional MLH1 methylation has been reported as a rare cause of mutation-negative LS (four cases reported).15–19 This may confound the use of methylation as a prescreen, but the incidence of this is likely to be extremely low. Tumour MLH1 promoter region methylation has not previously been tested in a large group of patients. While a number of methods for MLH1 methylation analysis have been developed, most are technically difficult (particularly in formalin fixed, paraffin embedded (FFPE) tissue) and expensive.

Guidelines for constitutional mutation testing for cancer susceptibility genes suggest a threshold of 10% risk.20 Using Bayes theorem, specificity and sensitivity of any prescreen test can be applied to individuals with differing risk determined by their family history of cancer. Individuals who fulfil Amsterdam II criteria have a pretest probability of harbouring a mutation of 60%.6 Individuals who fulfil the revised Bethesda guidelines and have loss of MLH1 in their tumour have a pretest probability of at least 10.5%.21–23 Patients from the general population who have MLH1 loss (tumour) have a pretest probability of 4.0%.23–26 We have previously shown that MSI testing alone is not an appropriate prescreening tool in Amsterdam criteria (I and II) positive families. Even if their tumour is microsatellite stable, the risk of having a mutation remains greater than 10%.27 Given that a recent Health Technology Assessment study has recommended that all CRCs in patients aged 60 years of age or younger should be prescreened for tumour mismatch repair deficiency (MMRd),28 strategies need to be developed to deal with the large number of MMRd tumours most of which will be the result of MLH1 promoter methylation in the tumour and not be caused by constitutional mutations.

The aim of this study was to: (1) develop a simple, cheap, reproducible method for quantitative MLH1 promoter region methylation analysis in FFPE tissue, (2) compare this with BRAF c.1799T>A p.Val600Glu mutation testing in patients whose CRCs demonstrate loss of MLH1 protein expression and (3) assess additional benefit of adding a methylation assay to BRAF testing in order to select patients for constitutional MLH1 mutation testing.

Methods

Ethical approval was obtained from South Manchester (UK) Research Ethics Committee.

Participants

To compare tumour MLH1 promoter methylation testing and BRAF c.1799T>A p.Val600Glu somatic mutation testing for selecting patients for constitutional MLH1 mutation analysis, two groups of patients with MLH1-deficient tumours were identified; those patients with pathogenic constitutional MLH1 mutations and patients with MLH1 promoter hypermethylated CRCs with MLH1 loss.

CRCs from patients with known pathogenic constitutional MLH1 mutations were identified from the Familial Colorectal Cancer Registry (Central Manchester University Hospitals NHS Foundation Trust, UK n=22). Additional cases were provided by The Jeremy Jass Memorial Pathology Tissue Bank of the Australasian Colorectal Cancer Family Registry (ACCFR: U01 CA097735).29 MMR IHC to identify CRCs with loss of MLH1 protein expression was performed (by the ACCFR).30 Screening for constitutional mutations in MLH1, MSH2, MSH6 and PMS2 was performed for all probands recruited from high-risk clinics, and for population-based probands who had a CRC with evidence of MSI or loss of MMR protein expression by IHC. Mutation testing was performed as previously described,31 ,32 and 49 ACCFR CRC cases with loss of MLH1 protein expression from patients with known pathogenic MLH1 mutations (n=49) were included.

Semiquantitative MMR IHC, as previously described,10 was conducted on 86 consecutive right-sided CRCs (sporadic MSI-H tumours occur more frequently in the right colon12) from patients aged over 50 years who did not fulfil Amsterdam or Bethesda criteria identified at Manchester Royal Infirmary. Patients known to have LS, familial adenomatous polyposis (FAP) or inflammatory bowel disease were excluded. Those with MLH1 loss were considered to be sporadic MLH1 loss cancers (n=33). MMR deficiency was confirmed by MSI analysis (MSI Analysis V.1.2 (Promega, USA)). Additional MLH1 loss cases were provided by ACCFR.29 CRC cases with MLH1 loss were classified as sporadic (n=40) based on the presence of the BRAF mutation and/or methylation of the MLH1 gene promoter, and did not harbour a pathogenic mutation in the MLH1 gene. Detection of BRAF mutation was determined on CRC tissue DNA using an allele-specific PCR assay as previously described.33 Methylation of the MLH1 gene promoter region was assayed using MethyLight qPCR on sodium bisulphite converted tissue DNA where samples with a percent of methylated reference greater than or equal to 10 were classified as positive for MLH1 methylation.13 ,14

Cancer specimens

Manchester samples: H&E slides were reviewed by an experienced Consultant Gastrointestinal Pathologist (RFTM), and areas which contained at least 70% cancer cells were selected; 10 µm thick slices were taken from the corresponding FFPE block for DNA extraction. ACCFR samples: approximately six 4 µm thick FFPE slices mounted on microscope slides were obtained per cancer. Each slice was reported to contain at least 70% tumour tissue. The tissue was manually removed from the slides and placed into 1.5 mL Eppendorf tubes for DNA extraction.

MLH1 methylation analysis

MLH1 promoter methylation was quantified using a novel pyrosequencing assay developed to UK Good Laboratory Practice standard. This assay relies on sodium bisulphite conversion. The EpiTect plus FFPE Bisulphite kit (Qiagen, UK) was used as per manufacturer's instructions, with an additional overnight tissue lysis step, to extract and bisulphite modify genomic tumour DNA from approximately 10 µm thick sections of FFPE tissue. An area of the MLH1 promoter from −248 to −178, known to be functionally significant34 was amplified (see figure 1). Each sample was tested in triplicate. A CpGenome Universal Methylated DNA control (Millipore, cat no. S7821) was used.

Figure 1

MLH1 promoter region and BRAF PCR conditions.

The amplicons were sequenced using the Pyrosequencer (PSQ 96MA). Sequencing primer: GAATTAATAGGAAGAG. Pyrograms were analysed independently by two blinded scientists. Greater than 10% methylation at each cytosine,35 in at least two of the three triplicates was considered significant. Figure 2 shows a typical pyrogram of methylated MLH1 tumour DNA. Genotype: CmGGACAGCmGATTTTTAACmGCmG (methylated).

Figure 2

Pyrogram of methylated MLH1 promoter region of tumour DNA.

BRAF c.1799T>A p.Val600Glu mutation analysis

The region of BRAF codon 600 (exon 15) was sequenced using the Pyrosequencer (PSQ 96MA) and associated software (Qiagen, UK). Tumour DNA was extracted from the FFPE tissue using the Qiagen EZ1 robot in conjuction with the Tissue Extraction kit (Qiagen, UK) as per the manufacturer's guidelines. Codon 600 was amplified using a non-nested 25 µL PCR. Each sample was tested in triplicate. The amplicons were sequenced using the Pyrosequencer (PSQ 96MA) and associated software. Sequencing primer 5′-TGATTTTGGTCTAGCTACA-3′. Pyrograms were genotyped by two independent blinded scientists.

Statistical methods

The performance characteristics of unmethylated MLH1 promoter region and wt BRAF c.1799T>A p.Val600Glu and for the identification of cancers from the individuals with a constitutional MLH1 mutation were analysed using Diagnostic test 2×2 tables (StatsDirect, Cheshire UK). Each test was analysed separately and in conjunction by applying the ‘either positive’ and ‘both positive’ rules. A Bayesian calculation was used to calculate the post-test risk of harbouring a constitutional MLH1 mutation in groups of patients with differing a priori risk. This was calculated from all relevant published studies. In this setting, the post-test risk of an individual harbouring an MLH1 mutation can be calculated taking into account the pretest (a priori) risk which is determined by their family history, and also their test result.

Results

Seventy-one CRCs from pathogenic constitutional MLH1 mutation carriers, and 73 sporadic cancers with MLH1 protein loss were analysed. Somatic MLH1 promoter region methylation was found in 4/71 (5.6%) tumours from constitutional MLH1 mutation carriers and 64/73 (87.7%) sporadic cancers with MLH1 loss. Somatic BRAF mutation was found in 1/71 (1.4%) tumours from constitutional MLH1 mutation carriers and 48/73 (65.8%) sporadic cancers with MLH1 loss (see table 1); 33/73 sporadic MLH1 loss cancers were known to be wt for constitutional MLH1 gene mutations. Of these 28/33 (84.8%) had somatic MLH1 methylation and 18/33 (54.5%) had somatic BRAF mutation. Of the 40 that had not been tested for constitutional mutations, 36/40 (90.0%) had somatic MLH1 methylation and 30/40 (75.0%) had somatic BRAF mutation.

Table 1

BRAF c.1799T>A p.Val600Glu mutation and MLH1 promoter region methylation in MLH1 mutation carriers and sporadic MLH1 loss cancers

The mean percentage of methylation (of the four CpGs examined in triplicate for each sample) of the methylated tumours was 75.8%, and the median 81.5% (range 19.3%–100%). There was no difference in the level of methylation between the MLH1 methylated tumours from the sporadic group and those tumours that demonstrated MLH1 methylation from constitutional MLH1 mutation carriers.

We have demonstrated that a normal (unmethylated) MLH1 promoter region has a higher specificity (87.67% (95% CI 77.88% to 94.2%)) than wt BRAF (65.75% (95% CI 53.72% to 76.47%)) and similar sensitivity (normal MLH1 promoter region 94.37% (95% CI 86.2% to 98.44%), wt BRAF (98.59% (92.4% to 99.96%)) for the identification of MLH1 mutation carriers.

When used in combination, a wt BRAF OR normal MLH1 promoter region result has the highest sensitivity for the identification of MLH1 mutation carriers (100% (95% CI 94.94% to 100%)), but has the lowest specificity (63.01% (50.91% to 74.03%)); wt BRAF AND normal MLH1 has the lowest sensitivity (92.96% (95% CI 84.33% to 97.67%)) but the highest specificity (90.41% (81.24% to 96.06%)) (tables 2 and 3).

Table 2

Diagnostic Test (2 × 2 tables) analysis for the identification of MLH1 mutation carriers

Table 3

Pretest and post-test probabilities of being an MLH1 mutation carrier for three a priori risk groups

Applying MLH1 methylation and/or BRAF mutation analysis to tumour DNA from patients who have a dMMR cancer and fulfil Amsterdam II criteria does not significantly improve diagnostic prediction of MLH1 mutation. Although the identification of wt BRAF and/or normal MLH1 promoter region strongly suggests that a constitutional MLH1 mutation is present, mutant BRAF and/or MLH1 hypermethylation does not bring the (post-test) risk of a mutation to below 10%.

Applying MLH1 methylation and/or BRAF mutation analysis to tumour DNA from patients who fulfil the revised Bethesda guidelines and have loss of MLH1 is informative. An unmethylated MLH1 promoter region suggests a risk of having a constitutional MLH1 mutation of 31.4% while the identification of wild-type BRAF suggests a post-test risk of 19.0%. The finding of MLH1 methylation gives a risk of 1.2%. The finding of mutant BRAF gives a risk of 1.6%.

Applying MLH1 promoter region methylation analysis to tumour DNA from patients from the general population who have a tumour with MLH1 loss is informative. The finding of normal MLH1 promoter region suggests a risk of a constitutional MLH1 mutation of 14.0% compared with the finding of methylated MLH1 promoter region which suggests a risk of 0.4%. However, applying BRAF mutation analysis to tumour DNA from patients from the general population who have MLH1 loss is uninformative. Wt BRAF only indicates a risk of 7.7%, so constitutional testing would not necessarily be indicated. Applying both BRAF and MLH1 analysis is only informative if both are wt/normal.

The ACCFR had performed somatic MLH1 promoter methylation and BRAF analysis on a proportion of their sporadic MLH1 loss CRCs. MLH1 methylation result was consistent in 21/22 cases, and BRAF result was consistent in 39/39 cases with the pyrosequencing assays performed in this study.

Discussion

This study is the first large-scale assessment of specificity and sensitivity of pretesting MLH1 CRC loss for the presence of a constitutional MLH1 mutation. We have demonstrated the use of pyrosequencing-based testing of CRC tumour DNA for MLH1 promoter methylation in a large cohort of CRCs. A novel MLH1 promoter region methylation assay has been developed to good laboratory practice (GLP) standards and its clinical use demonstrated in the assessment of patients meeting Bethesda guidelines and in population-based prescreening for LS. A single assay is time and cost effective, and may encourage the introduction of prescreening into routine clinical practice. Unmethylated MLH1 has a sensitivity of 94.4% and a specificity of 87.7% for the identification of MLH1 mutation carriers from a group of cancers with MLH1 loss MMRd. As such methylation is more effective than BRAF testing.

MMR protein IHC should be used as the first-line LS prescreening test in patients meeting Bethesda guidelines and cancers with typical LS features histologically. If MSH2, MSH6 or PMS2 proteins are absent, this indicates high risk of LS and the individual should be tested for constitutional mutations in the relevant gene. If MLH1 protein is absent, tumour DNA should be subjected to MLH1 promoter methylation testing. If methylation is absent, this indicates high risk (>10%) of LS, and constitutional MLH1 mutation analysis should be conducted.

It was previously unknown whether BRAF mutation or MLH1 promoter region methylation or both, is best able to distinguish between sporadic MLH1 loss of CRCs and cancers from patients with constitutional MLH1 mutations. Previous studies have examined small numbers of patients, and thus, there is no guide for clinical practice.36–38

Sporadic MMR-deficient cancers with loss of MLH1 (or MSI-high) are associated with MLH1 gene silencing through the epigenetic effect of promoter region methylation. In this study, MLH1 promoter region methylation was found in 64/73 (87.7%) sporadic MMR-deficient CRCs; 33/73 (45.0%) of the sporadic MMR-deficient CRCs are known to be negative for constitutional MLH1 mutations; 28/33 (84.8%) were found to have MLH1 promoter methylation. Methylation was consistent across all four cytosine residues in the functional area of the promoter region as described by Deng et al.34 Pyrosequencing allows accurate quantification of methylation; 9/73 (12.3%) sporadic MMR deficient CRCs were found to have normal MLH1 promoter region. Two were found to have BRAF mutation. Of the seven that were wt BRAF and normal MLH1 promoter region, four had been tested for constitutional MLH1 mutations by the ACCFR and were found to be negative. The remaining three had been classified as sporadic MMR-deficient due to the patient's age (over 50 years) and lack of family history. The aetiology of these cancers without promoter region hypermethylation is unclear, but possible factors include loss of protein expression, somatic mutation of MLH1 and loss of heterozygosity.39 It is feasible that the MLH1 promoter displayed mosaic or heterogeneous patterns of methylation for the CpGs dinucleotides captured in the pyrosequencing amplicon but enough of the surrounding CpGs dinucleotides were methylated to result in loss of MLH1 protein expression. Alternatively, the three untested patients may be carriers of constitutional MLH1 mutations.

It has been thought that MLH1 promoter methylation is found exclusively in sporadic MMR-deficient CRCs.24 ,34 ,40–42 The current study is the largest dataset of MLH1 mutation carriers tested for MLH1 promoter region methylation. MLH1 promoter methylation was found in 4/71 (5.6%) mutation carriers (see table 4). Of these, 1/22 (4.5%) was a Manchester patient (from an Amsterdam family), and 3/49 (6.1%) were from the Australasian Colon Cancer Family Registry (two from Amsterdam families) A first-degree relative of the Manchester patient (with the same constitutional MLH1 mutation) has an unmethylated somatic MLH1.

Table 4

MLH1 promoter region methylation in MLH1 mutation heterozygotes

This low MLH1 promoter methylation frequency in mutation carriers is supported by a recent literature review and meta-analysis43 which found eight positively methylated tumours in MLH1 mutation carriers taken from 12 studies (5.56%). It has been suggested that sporadic inactivation of the second normal MLH1 allele by hypermethylation may be the ‘second hit’ event in mutation carriers.44 While somatic MLH1 promoter region methylation is an infrequent event in constitutional MLH1 mutation carriers, this data demonstrates that it is not rare and supports the hypothesis that it may function as the second hit event. These findings also suggest that the discovery of MLH1 hypermethylation does not exclude the diagnosis of LS. While the sensitivity of MLH1 hypermethylation testing is adequate for low/moderate-risk individuals, it is not for high-risk patients (3/4 MLH1 mutation carriers who had MLH1 promoter methylation were from Amsterdam families).

BRAF gene mutations are found in 5%–15% of all CRCs.45 ,46 They are more frequent in cancers from Jass's subtypes 1 (MSI-H, chromosome stable, CpG Island methylator phenotype (CIMP) high, methylated MLH1 promoter region; 13% of all CRCs) and 2 (MSI-low or stable, chromosome stable, CIMP-high, partial MLH1 methylation; 8% of all CRCs).12 Both are thought to originate in serrated lesions. BRAF mutation is thought to be an unequivocal marker of the serrated neoplasia pathway. The discovery of a BRAF mutation is thought to rule out LS.47 In the current study, BRAF mutation was found in 48/73 (66%) sporadic MLH1 loss CRCs. This is consistent with previous studies.42 ,48 A BRAF mutation was detected in 1/71 (1.4%) CRCs from MLH1 mutation carriers. BRAF mutations have previously been reported as a rare finding in patients with LS,49 ,50 and are thought to represent a mixed lineage of cancer predisposition. Walsh et al have reported two families with evidence of LS and probable additional constitutional factors causing serrated neoplasia.47 Senter et al investigated 99 probands with Lynch spectrum cancers that demonstrated loss of PMS2 on IHC. Constitutional PMS2 mutations were detected in 62%, and three (one exon 10 deletion, two c.736_741del6ins11) of these were found to have tumour BRAF mutation.31 It is likely that BRAF mutation is a rare finding in LS, and that its occurrence represents the influence of other molecular pathways, as suggested by Walsh.47

It has been suggested that in MMR-deficient cancers, BRAF mutation is a surrogate marker for MLH1 promoter methylation. However, there is now evidence that BRAF mutation occurs in only 50%–75% of sporadic MLH1 loss cancers. In a series of 270 CRCs, Wang et al found BRAF mutations in 42/123 (34%) MMR-deficient cases. BRAF was closely associated with MLH1 methylation (30/36 (83.3%) MLH1 hypermethylated cases also had a BRAF mutation).48 In a large population-based study, Woods et al examined 68 MSI-H CRCs for BRAF and MLH1 methylation in order to prescreen for constitutional mutation testing. BRAF mutation was closely but not exclusively associated with MLH1 methylation; 31/40 (78%) of the hypermethylated tumours had BRAF mutations. In the current study, 46/73 (63.0%) sporadic MLH1 loss cancers had both BRAF mutation and MLH1 methylation, but 18/73 (24.7%) had only MLH1 methylation. This is consistent with previous studies.42 ,48

A recent heath technology assessment (HTA) report has established that it would be cost effective for the NHS to introduce systematic testing for LS of all CRCs up to age 70 years.28 This report addressed the issue of excluding sporadic MLH1 cases from requiring unnecessary referral to clinical genetics services. The cost-effectiveness analysis allows for the increased costs of performing additional tests and the inherent reduction in sensitivity when more tests are performed serially in an attempt to increase specificity. Our data suggest that only 65% of cases without constitutional mutations will be identified by using BRAF alone. This is increased to 90% by adding a MLH1 methylation test. There are around 16 800 new cases of CRC up to age 70 years each year in the UK.51 Around 2200 (13%) of these will be sporadic MLH1 cases. The increased specificity of additional MLH1 methylation testing (90%) rather than BRAF alone (65%) would reduce the number of cases requiring genetic counselling and testing from around 780 to 220, a reduction of 550 cases each year. In our laboratory, MLH1 mutation testing costs around £483, MLH1 methylation testing costs around £138, and BRAF testing around £69. On average in the UK, a new person's appointment with a genetic counsellor or physician costs around £500 and a follow-up appointment around £350. Adding MLH1 methylation testing into systematic testing would cost around £300 000 per year. The cost saving each year would be around £700 000 (£450 000 for counselling and £250 000 for constitutional MLH1 analysis).

There are some limitations to the current study. A proportion of the sporadic samples did not undergo constitutional testing for MLH1. Even full sequencing and a dosage test of MLH1 may miss mutations such as deep intronic splicing mutations, and sensitivity may therefore be reduced. Clendenning et al have reported the discovery of an intronic MSH2 mutation, 478 bp upstream from exon 2 causing LS.52 As such, some of the ‘sporadic’ MLH1 loss CRCs may have had an undetected constitutional mutation. However, the rate of non-methylated MLH1 and wt BRAF in the 40 (4/40, 10%) sporadic tumours with mutation testing was not different to that in the 33 (3/33, 9.1%) untested sporadic cases. Additionally, the untested sporadic patients may harbour constitutional methylation.

Constitutional MLH1 promoter region methylation has been described as a rare (33 reported cases) finding in CRC.16–18 53–57 It is thought that this epimutation is usually erased in the gametes, but inheritance has been demonstrated in four cases.15–17 ,19 A recent study from the German hereditary non-polyposis colorectal cancer (HNPCC) consortium investigated 32 mutation-negative suspected Lynch cases with MSI-H and MLH1 loss CRC. They report one case of heritable partial MLH1 promoter methylation, which is induced by a large genomic duplication including the complete MLH1 gene and the promoter.15 This suggests that even in mutation-negative Lynch cases, the finding of constitutional methylation is low.

Constitutional MLH1 methylation is likely a rare cause for CRC tumour DNA MLH1 promoter region methylation, and a rare cause for LS, although the true incidence is unknown. In 10%–15% of suspected Lynch cases, no disease-causing mechanism can be detected. In these cases, it may be prudent to test for constitutional MLH1 promoter methylation.

Schofield et al reported a population-based screening programme using IHC, MSI and BRAF testing in CRCs in patients aged below 60 years. In the cohort of 270, 70 were MSI-H. 82 had loss of MMR protein expression. BRAF testing was conducted on 76 tumours. Twenty-five mutant BRAF tumours were excluded from further testing; 45 ‘Red Flag’ cases were identified (MSI-H and loss of MSH2 or MSH6, OR MSH-H and loss of MLH1/PMS2 and wtBRAF); 31 were tested for constitutional mutations; 15 mutation carriers (7 MLH1, 2 MSH2, 3 PMS2 and 3 MSH6) were identified. The incidence of constitutional mutation in their ‘Red Flag’ cases is 48%. Our study has demonstrated that IHC followed by MLH1 methylation testing is likely to have a higher ‘hit’ rate due to the higher specificity of MLH1 methylation compared with BRAF (88 vs 66% in our study). Using IHC as the initial test avoids additional expense of MSI and allows the appropriate gene to be targeted for constitutional testing.

Identification of families with LS is vital to enable reduction in morbidity and mortality with screening. The use of population-based prescreening has been hampered by a lack of evidence for the specificity and sensitivity of MLH1 promoter region methylation analysis for the detection of mutation carriers. It is hoped that this current study provides that evidence.

Conclusion

MLH1 promoter region methylation analysis is simple, reproducible and cheap. It can be used in conjunction with MMR IHC in the prescreening of low and moderate-risk patients for LS mutation testing. This will be a vital addition to BRAF testing when population assessment of MMR deficiency is introduced. Amsterdam criteria CRCs should be tested for constitutional mutation regardless of CRC prescreening status.

Acknowledgments

Keeling C and Griffiths-Davies L (Department of Histopathology, Manchester Royal Infirmary, Central Manchester University Hospitals NHS Foundation Trust) for conducting immunohistochemistry. Newton K, Jorgensen NM, Lalloo, F, Hill J, and Evans DG are supported, by Manchester NIHR BRC. Biospecimens were provided by The Jeremy Jass Memorial Pathology Bank, Australasian Colorectal Cancer Family Registry (U01 CA097735).

References

Footnotes

  • Contributors KN, AJW, JH, FL, DGE contributed to the study concept, study design, acquisition of data, analysis and interpretation of data, drafting of the manuscript and critical revision of the manuscript for important intellectual content. Jorgensen NM contributed to acquisition of data, analysis and interpretation of data, drafting of the manuscript and critical revision of the manuscript for important intellectual content. DDB contributed to acquisition of data, analysis and interpretation of data and critical revision of the manuscript for important intellectual content.

  • Competing interests None.

  • Patient consent No.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

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.