MLH1 Promoter Methylation Frequency in Colorectal Cancer Patients and Related Clinicopathological and Molecular Features

Purpose To describe the frequency of MLH1 promoter methylation in colorectal cancer (CRC); to explore the associations between MLH1 promoter methylation and clinicopathological and molecular factors using a systematic review and meta-analysis. Methods A literature search of the PubMed and Embase databases was conducted to identify relevant articles published up to September 7, 2012 that described the frequency of MLH1 promoter methylation or its associations with clinicopathological and molecular factors in CRC. The pooled frequency, odds ratio (OR) and 95% confidence intervals (95% CI) were calculated. Results The pooled frequency of MLH1 promoter methylation in unselected CRC was 20.3% (95% CI: 16.8–24.1%). They were 18.7% (95% CI: 14.7–23.6%) and 16.4% (95% CI: 11.9–22.0%) in sporadic and Lynch syndrome (LS) CRC, respectively. Significant associations were observed between MLH1 promoter methylation and gender (pooled OR = 1.641, 95% CI: 1.215–2.215; P = 0.001), tumor location (pooled OR = 3.804, 95% CI: 2.715–5.329; P<0.001), tumor differentiation (pooled OR = 2.131, 95% CI: 1.464–3.102; P<0.001), MSI (OR: 27.096, 95% CI: 13.717–53.526; P<0.001). Significant associations were also observed between MLH1 promoter methylation and MLH1 protein expression, BRAF mutation (OR = 14.919 (95% CI: 6.427–34.631; P<0.001) and 9.419 (95% CI: 2.613–33.953; P = 0.001), respectively). Conclusion The frequency of MLH1 promoter methylation in unselected CRC was 20.3%. They were 18.7% in sporadic CRC and 16.4% in LS CRC, respectively. MLH1 promoter methylation may be significantly associated with gender, tumor location, tumor differentiation, MSI, MLH1 protein expression, and BRAF mutation.


Introduction
Colorectal cancer (CRC) is one of the most common malignancies, representing the third most common cancer in men and the second in women worldwide [1]. One of the genetic pathways in the development of CRC is the microsatellite instability (MSI) [2].
Microsatellites are repeated DNA sequences which occur approximately every 50-100 Kb base pairs throughout the human genome [3,4]. Multiple studies have indicated that about 90% of the Lynch Syndrome (LS) [5,6] and 10% to 15% of sporadic CRC can be detected of MSI [3,4]. MSI in LS and sporadic CRC occurs through two different mechanisms. In LS, MSI is mainly caused by germline mutation of mismatch repair genes [7]. MSI in sporadic CRC is commonly due to methylation induced silencing of the MLH1 gene [8].
DNA methylation refers to the presence of a methyl group on a cytosine residue [9]. DNA methylation of tumor suppressor genes leading to transcriptional inactivation has been identified as an important mechanism in human carcinogenesis [10,11]. MLH1 gene, as a number of suppressor genes, is prone to be silenced by promoter methylation in CRC [8,12,13].
Since the first report of MLH1 promoter methylation in sporadic colon tumors [8], the prevalence of MLH1 promoter methylation have been widely studied not only in sporadic but also in LS CRC. However, the results are inconsistent. The frequency of MLH1 promoter methylation in sporadic CRC varied from 0.0% [14] to 66.9% [15]. It varied from 0.0% [16] to 21.4% [17] in LS CRC.
BRAF and KRAS are important members of RAS/RAF/MAPK signaling pathway, which regulates cell growth, proliferation, differentiation, and apoptosis in malignant and nonmalignant cells [18]. BRAF mutation has been shown to be associated with MLH1 promoter methylation [19,20]. Whereas, MLH1 promoter methylation was few detected in KRAS mutant CRC [21]. The associations between MLH1 promoter methylation and BRAF and KRAS mutation in CRC have been widely studied with inconsistent results [15,21,22,23]. The associations between MLH1 promoter methylation and other clinicopathological and molecular characteristics of CRC such as tumor location, tumor staging, tumor differentiation, family history, MSI, and MLH1 protein expression were also widely studied. However, the results are inconsistent. Therefore, we conducted a systematic review and meta-analysis to accurately estimate the frequency of MLH1 promoter methylation in LS and sporadic CRC, and the associations between MLH1 promoter methylation and clinicopathological/molecular characteristics of CRC.

Search Strategy and Selection Criteria
We conducted a systematic literature search using PubMed and Embase from January 1, 1997 to September 7, 2012 to identify all the relevant English-language articles. The following keywords were used: ''methylation'' and ''MLH1'' and ''promoter'' and ''colorectal cancer'' and/or ''carcinoma'' or ''tumor'' or ''neoplasm''. We also hand-searched the reference lists of the retrieved articles and reviews for additional articles.
The inclusion/exclusion criteria were as follows: (1) papers on MLH1 promoter methylation in unselected CRC were included. In contrast, papers that selected subgroups were excluded (such as selected based on age, tumor staging and ulcerative colitisassociated CRC); (2) sporadic CRC and/or LS related CRC remained as specific selected groups, often stratified by MSI status and/or MLH1 expression loss; (3) data regarding the DNA methylation of tumor tissue of CRC were included in the pooled analysis, whereas data regarding the DNA methylation of normal colonic mucosa [24,25,26], serum [27,28], and peripheral blood leukocyte [29,30,31] of CRC were excluded; (4) studies that investigated multiple CRCs were excluded [32,33,34]; (5) case reports were excluded; (6) repetitive reports were unified by using the latest or the largest edition; (7) paper with insufficient or duplicated data were excluded.

Data Extraction
Two authors (X. and X.P) independently conducted literature searches to identify all possible papers that met the inclusion criteria. Disagreements were settled by consensus or a third review (Y.B.N) for adjudication. The following information were extracted from every eligible study: authors, publication year, continent, country, patient source, sample size, methylation detecting method, positive frequency, gender, family history, tumor location (proximal and distal), tumor staging, and promoter regions.

Classification of Family History
Patients had no family history of cancer regardless of the onset age were categorized as sporadic CRC. LS was diagnosed if a patient with family history met either Amsterdam criteria (I or II) [35,36] or Bethesda criteria (original or revised) [37,38] or confirmed with germline mutation in a DNA mismatch repair gene [39,40]. The unselected CRC tumors were defined as patients from nature population or hospital-based. The unselected CRC tumors included sporadic and LS CRC, which were defined as total CRC.

Tumor Staging and Differentiation
Tumor staging was categorized as I, II, III and IV stages based on the TNM classification (The Union for International Cancer Control [UICC]) [41]. Stage I: Cancer has begun to spread, but is still in the inner lining. Stage II: Cancer has spread to other organs near the colon or rectum. It has not reached lymph nodes. Stage III: Cancer has spread to lymph nodes, but has not been carried to distant parts of the body. Stage IV: Cancer has been carried through the lymph system to distant parts of the body. Differentiation was graded on a scale of poor, moderate or well differentiation.

Promoter Regions
Promoter regions tested were noted as the A, B, C and D regions proposed by Deng et al. [42], where primer sequences were given. Promoter regions were checked against the sequence 21000 to 21, relative to the start codon of MLH1.

Molecular Classification
MSI is typically assessed by analyzing five microsatellite markers (BAT25, BAT26, D2S123, D5S346, and D17S250) suggested by the National Cancer Institute [43]. One study expanded this panel to ten markers, which made the diagnosis of MSI CRC easier [38]. In this meta-analysis, three categories of MSI status were defined according to the following criteria: two or more loci out of five loci with instability (or $30-40% of loci if a larger panel of markers was used) was defined as MSI-H; one locus with instability (or ,30-40% of loci in larger panels) was defined as lower-level microsatellite instability (MSI-L); and no loci with instability (or no apparent instability in larger panels) was defined as microsatellite stable (MSS). For papers without detailed information about MSI-H and MSI-L, only two levels of microsatellite instability status could be categorized: MSI-positive (MSI) and MSI-negative (MSS). BRAF and KRAS status were classified into mutant and wild type. MLH1 protein expression status was defined as positive or negative.

Statistical Analysis
The pooled frequency of MLH1 promoter methylation and 95% confidence intervals (95% CI) were estimated. The frequency of MLH1 promoter methylation was compared in different tumor characteristics. Heterogeneity among studies was evaluated with Cochran's Q test [44] and the I 2 statistic [45,46]. When heterogeneity was not an issue (I 2 values ,50%), a fixed effect model was used to calculate parameters. If there was substantial heterogeneity (I 2 values $50%), a random-effects model was used to pool data and attempt to identify potential sources of heterogeneity based on subgroup analyses. The pooled OR was estimated for the association between MLH1 promoter methylation and clinicopathological, molecular features. P values tailed less than 0.05 were considered statistically significant.
Publication bias was evaluated with funnel plot, Begg's rank correlation [47], and Egger's regression [48]. If publication bias existed, the trim and fill method was used to adjust the pooled frequency, pooled OR and 95% CI [49]. Data were calculated with Comprehensive Meta-Analysis V2.
Results 752 relevant articles were identified for initial review according to the inclusion and exclusion criteria. After screening, in-formation for 10528 individuals from 96 studies was reviewed and included in the meta-analyses. Figure1 showed the detailed selection process of articles.

Family History
The frequency of MLH1 promoter methylation was 18

Publication Bias
For the frequency of MLH1 promoter methylation in MSI CRC and MSI-H CRC, the funnel plot seemed asymmetry ( Figure S2 A and B). Funnel plot for the association between MLH1 promoter methylation and tumor location (proximal vs. distal) also seemed asymmetry ( Figure S3). Begg's rank correlation and Egger's regression methods further supported the significant publication bias. With the trim and fill method, the adjusted frequency of MLH1 promoter methylation decreased from 55.8% to 36.7% in MSI CRC and from 62.6% to 53.5% in MSI-H CRC. The pooled OR for the association between MLH1 promoter methylation and tumor location decreased from 3.804 (95% CI: 2.715-5.329) to 3.172 (95% CI: 2.323-4.331).

Discussion
Our meta-analysis suggested that the frequency of MLH1 promoter methylation in total CRC was 20.3%. They were 18.7% in sporadic CRC and 16.4% in LS CRC, respectively; significant associations were observed between MLH1 promoter methylation and gender, tumor location, tumor differentiation, MSI, MLH1 protein expression, and BRAF mutation.
The pooled MLH1 promoter methylation frequencies were 16.4% and 20.5% in 4 population-based studies and 16 hospitalbased studies (One study [53] included 1061 population-based and 172 hospital-based CRC). In total CRC, the frequency of the MLH1 promoter methylation between hospital-based and population-based studies was not significantly different (P = 0.279) (Table1). A, B, C and D regions in the MLH1 promoter were commonly tested for methylation. However, only one study tested the MLH1 promoter methylation in ''A'' region [69], three studies tested the MLH1 promoter methylation in ''C'' region [56,57,70], other 15 studies did not provide the specific A, B, C or D regions in total CRC. The MLH1 promoter methylation frequency in ''A'' region (66.9%) was significantly higher than in ''C'' region in CRC (26.4%; P = 0.001) (Table S1). It may be due to the variation of methylation status in different regions of the MLH1 promoter.
The MLH1 promoter methylation was detected in total of 12 studies with 968 MSI-H CRC with a frequency of 62.6%. After the adjustment by trim and fill method, the pooled frequency decreased to 53.5%. The pooled MLH1 promoter methylation in 249 sporadic MSI-H CRC was 73.6%, significantly higher than in 95 LS MSI-H CRC (15.3%). The following may explain our results: in sporadic CRC, MSI-H was mainly caused by MLH1 promoter methylation [13,71]; whereas, in LS CRC, MSI-H was mainly caused by MMR inactivation because of germline mutation [72].
In sporadic CRC, our meta-analysis indicated that the MLH1 promoter methylation frequency in 308 CRC with MLH1 protein expression (9.8%), which was lower than in 156 CRC without MLH1 protein expression (69.8%, P,0.001). In CRC with loss of MLH1 protein expression, the MLH1 promoter methylation was significantly higher in sporadic CRC (69.8%) than in LS CRC (37.8%, P = 0.026). In sporadic MSI-H CRC with loss of MLH1 protein, the MLH1 promoter methylation frequency was 86.3%. MLH1 promoter methylation could explain more fraction of MLH1 gene silencing in sporadic CRC than that in LS CRC. In this systematic review and meta-analysis, we can see that the highest frequency of MLH1 promoter methylation was in MSI-H CRC with loss of MLH1 protein, the following in sporadic CRC without MLH1 protein expression, and the lowest in the sporadic CRC with MLH1 protein expression.
The frequency of MLH1 promoter methylation in BRAF mutated total CRC was 53.2%, significantly higher than in BRAF wild type total CRC 13.7% (P = 0.001). In contrast, the MLH1 promoter methylation frequency in KRAS mutated total CRC  [73]. In addition, the methylation of MGMT was associated with KRAS mutant CRC but not of BRAF mutant CRC could also support the results of our meta-analysis [20].
Our study suggested that the frequency of MLH1 promoter methylation was higher in female, proximal tumor location, and poor differentiation. Study reported that MSI CRC had a very distinct clinicopathological phenotype, which was commonly mucinous, poorly differentiated, presenting at earlier Dukes' stage, and in the proximal side of the colon [74]. MSI CRC was also commonly female and older at diagnosis. Moreover, in sporadic CRC, MSI was mainly caused by MLH1 promoter methylation. Therefore, MSI CRC and MLH1 promoter methylation CRC may have similar clinicopathological phenotype. However, the underlying mechanisms need to be investigated.
Heterogeneity persisted in our meta-analysis. The followings may explain the sources of heterogeneity. Firstly, MLH1 promoter methylation was tested in different promoter regions. One study tested in ''A'' region; three studies tested in ''C'' region; other 15 studies did not supply specific regions tested. Additionally, various ages of the study subjects may also explain the heterogeneity. Genes of individual are progressively methylated with aging due to chromosomal instability [75]. However, only three of 19 studies provided the information of age of study subjects [16,54,55].
Although this meta-analysis provides some robust results, limitations also existed like all the meta-analysis. Firstly, dietary factors, smoking and drinking alcohol may affect the MLH1 promoter methylation. Lack of these original data of the studies reviewed limited our further evaluation of their effect on MLH1 promoter methylation [76,77,78]. Secondly, lacking of the original data limited our further evaluation of the interactions between the clinicopathological and molecular variables in CRC. Thirdly, the prevalence of MLH1 promoter methylation in CRC may increase with aging. However, the majority of studies did not provide this information, which limited our further evaluate their effect on MLH1 promoter methylation.
In summary, this systematic review and meta-analysis yield some conclusions: the MLH1 methylation in total CRC was 20.3%; they were 18.7% in sporadic CRC and 16.4% in LS CRC, respectively; MLH1 promoter methylation may be significantly