Regulation of MLH1 mRNA and protein expression by promoter methylation in primary colorectal cancer: a descriptive and prognostic cancer marker study

Abstract

Background

In colorectal cancer MLH1 deficiency causes microsatellite instability, which is relevant for the patient’s prognosis and treatment, and its putative heredity. Dysfunction of MLH1 is caused by sporadic gene promoter hypermethylation or by hereditary mutations as seen in Lynch Syndrome. The aim of this study was to determine in detail how DNA methylation regulates MLH1 expression and impacts clinical management.

Methods

Colorectal cancer samples were collected from 210 patients. The laboratory methods used to study these samples included methylation specific multiplex ligation-dependent probe amplification (MS-MLPA), real-time quantitative PCR (qPCR), and immunohistochemistry (IHC).

Results

We found that the MLH1 mRNA and protein expression levels were highly related. MS-MLPA was successful in tumors from 195 patients. In these tumors, hypermethylation was observed in promoter regions A (n = 57), B (n = 30), C (n = 28), and D (n = 47), and in intron 1 (n = 25). The promoter region C and intron 1 methylation levels were found to be excellently suited for discriminating between low and high gene expression levels, whereas those of promoter regions A, B and D were less specific. Hypermethylation in any region (n = 77) served as an independent prognostic factor (hazard ratio 0.56, 95 % confidence interval 0.36–0.89, p = 0.01).

Conclusions

MLH1 inactivation through hypermethylation was found to be related to improved survival. Hypermethylation in promoter region C and intron 1 served as the most specific markers for this inactivation.

Introduction

The mismatch repair (MMR) gene MLH1 plays a key role in the development of colorectal cancer. Proper analysis of this gene may serve to address major clinical questions, including (i) what is the prognosis of the disease, (ii) how and when should it be treated and (iii) is it hereditary and is monitoring indicated for the patients and its relatives?

MLH1 deficiency gives rise to alterations in the length of repetitive DNA sequences, termed microsatellites, throughout the genome [1, 2]. This phenomenon, microsatellite instability (MSI), has been correlated with a better survival [3]. MSI is thought to be linked to a reduced response to chemotherapy, both pre-clinically [4–8] and clinically [9–11], and may thus be used as a marker for deselecting stage II colorectal cancer patients for adjuvant chemotherapy [12]. The most prevalent heritable colorectal cancer syndrome, Lynch syndrome, is caused by germline mutations in MMR genes, in about half of the cases MLH1 [13]. In contrast to these germline mutations, sporadic MSI is usually caused by acquired promoter hypermethylation of the MLH1 gene [14–16]. Consequently, methylation analysis can be of help in selecting patients for genetic counselling and screening [17, 18]. For practical screening purposes, the rare occurrence of hypermethylation in hereditary cases [19] is of minor importance.

Gene regulation by methylation is commonly observed in normal and cancer cells. Both mRNA and protein expression levels may be reduced when the promoter regions are hypermethylated [20]. As of yet, however, data on the regulation of MLH1 by promoter hypermethylation are contradictory, since most of them have been obtained from either cell culture studies or rather small clinical samples.

The aim of the present study was to perform a comprehensive analysis of the mechanisms underlying the regulation of MLH1 gene expression by promoter hypermethylation in colorectal cancer patients eligible for surgery. Our hypothesis was that different methylation patterns will have a different impact on MLH1 gene expression and, hence, on the clinical interpretation. A specific issue to be addressed was the effect of methylation within different regions of the gene on both mRNA and protein expression, with the aim to better guide clinicians through the evaluation of methylation reports. Furthermore, we wanted to investigate the impact of MLH1 methylation and expression on patient survival.

Material and methods

Patients and tissue samples

In this study, 210 patients undergoing surgery for colorectal adenocarcinomas were included. All these patients were included consecutively and unselected and only after informed consent was obtained, according to the protocol approved by the regional ethics committee. Tissues from primary tumors and lymph node metastases, if available, were formalin fixed and paraffin embedded. Also, tissues from the primary tumors were stored fresh frozen together with matched blood samples.

Methylation specific MLPA

Methylation levels were determined in five different regions within the MLH1 locus using methylation specific multiplex ligation-dependent probe amplification (MS-MLPA) (ME011, MRC-Holland, Amsterdam, The Netherlands) [18, 21]. Previously, the MLH1 promoter region has been divided into four regions, A through D, according to Deng et al. [22]. The assay covers each of these regions and, in addition, a methylation-prone site within intron 1. The four steps in the MS-MLPA procedure, i.e., denaturation and hybridization, ligation and digestion with methylation sensitive endonuclease, PCR amplification, and fragment and data analysis, were conducted according to the manufacturer’s instructions (outlined in Fig. 1). If any level of methylation was observed, the sample was considered positive. Since it has previously been suggested to consider a case methylation positive if the mean methylation in regions C and D was above 15 % [23], this value was also included for comparison.

Fig. 1.

Methylation specific multiplex ligation-dependent probe amplification (MS-MLPA). Denaturation and hybridization followed by ligation and digestion with methylation sensitive endonuclease were performed. Capillary electrophoresis after PCR amplification gave rise to panels 1 and 2. (1) Peak ratio of sample/reference sample indicates which sequences show aberrant copy numbers. (2) A high peak ratio of sample/reference sample indicates which sequences are methylated. A semi-quantitative estimate of the percentage of methylated DNA is provided. In this sample, between 50 % and 70 % of MLH1 is methylated (probably due to mix with non-tumor DNA). The results for regions A-D and intron 1 (I) are marked

Real-time quantitative PCR

Two sequential 5 and 10 μm sections from primary tumor tissues and available lymph node metastases were stained with hematoxylin-eosin and Nuclear Fast Red, respectively. The first section was used for morphology evaluation and the second, neighbouring, section was used for laser microdissection (PALM Microlaser Technologies AG, Bernried, Germany) or manual microdissection of neoplastic cells to ensure at least 90 % tumor cells in the sample. RNA was isolated according to a proprietary procedure of Response Genetics Inc. (US patent no. 6,248,535) and converted to cDNA as previously described [24, 25]. The relative MLH1 gene expression level was measured using β-actin as internal reference in a fluorescence-based real-time detection system (ABI PRISM 7900, Applied Biosystems, Foster City, CA, USA). A comparative method was used according to the manufacturer’s instructions [26], only modified to take three calibrators into account. The primers and probes used were AGGAGTCGACCCTCTCAGG (MLH1 forward), GTCCACTTCCAGGAGTTTGG (MLH1 reverse), Probe #36 (Universal ProbeLibrary, Roche, IN, USA) (MLH1 probe), TGAGCGCGGCTACAGCTT (β-actin forward), TCCTTAATGTCACGGACGATTT (β-actin reverse) and 6FAM-ACCACCACGGCCGAGCGG (β-actin probe). Amplicon sizes were 67 and 60 base pairs for MLH1 and β-actin, respectively. The primer and probe sets for both MLH1 and β-actin were validated for up to 39 and 28 PCR cycles, respectively. The amplifications for both target and reference mRNAs were proven to be equally efficient within these intervals. If the threshold of MLH1 expression detection surpassed 39 cycles, the level of expression of the sample was set to those corresponding to 39 cycles. If the threshold of β-actin expression surpassed 28 cycles, the quality of the sample was considered to be insufficient.

MSI, IHC and MS-PCR assays

The tissues included in this study overlap partly with those of a previous study [17]. The MSI status was determined by comparing the allele lengths of five microsatellite markers (D2S123, BAT25, BAT26, D5S346, and D17S250) in both tumor and normal (blood) DNA. Immunohistochemical (IHC) reactions were performed on 4 μm sections using an anti-MLH1 antibody (Becton Dickinson Clone G168-15, 1:250). Antigen retrieval was done by microwaving in TEG-buffer at pH 9.0, and visualization of the reaction was performed using EnVision + (DAKO, Glostrup, Denmark) with diaminobenzidine followed by counterstaining with hematoxylin. Similar reactions were carried out for the other three relevant MMR proteins PMS2, MSH2 and MSH6. For methylation specific PCR (MS-PCR) the sodium bisulifide conversion method was used [27]. The primers used were according to those reported by Grady et al. [28] and the assessment of the methylation status in promoter region C according to the method reported by Deng et al. [22].

Statistical analyses

Statistical analyses were performed using the software Stata 12 (StataCorp LP, Texas, USA) package. Fisher’s exact test, χ² test, and Wilcoxon test were used where appropriate. Non-parametric correlations were analyzed using Spearman’s correlation test. Receiver operating characteristics (ROC) were used to describe the correlation between the binary classifier ‘methylated or not’ and the continuous variable ‘gene expression’. Survival analyses were performed using the Kaplan-Meier method. Survival time was defined as the time from surgery to death. The last check for death dates was median 6.5 years after the last inclusion. Univariate and multivariate Cox regression models for overall survival were performed comprising the standard variables age, sex, stage and localization, and the experimental markers MLH1 gene expression and methylation in any region. P-values of <0.05 were considered statistically significant.

Results

MLH1 mRNA expression in primary tumors

After qPCR of β-actin as a control, total RNA isolated from 208 primary colorectal adenocarcinomas was considered to be of sufficient quality, thereby excluding only two cases from further analyses. Using subsequent real-time qPCR, we found that the MLH1 gene exhibited an overall low expression level. In 42 tumors the expression level was found to be outside the validated interval and, consequently, estimated as described in the Material and methods section. The relative MLH1 expression levels measured are listed in Table 1, and they were correlated to clinicopathological and molecular characteristics as outlined in the Material and methods section and reference [17]. The median age of the 208 patients was 71 years (range 41 to 91 years) and age as an ordinal variable was not correlated to gene expression (p = 0.37, Wilcoxon). Neither were sex, localization and stage. In 61 cases a pair-wise comparison was possible between gene expression in lymph nodes and corresponding primary tumors. There was a moderate, but significant, correlation (Spearman’s rho = 0.365, p = 0.004).

Table 1.

Relative MLH1 mRNA expression in relation to clinicopathological characteristics, MLH1 protein expression (IHC), microsatellite instability (MSI) or microsatellite stability (MSS), methylation specific PCR for region C (MS-PCR region C), BRAF V600E mutation, and methylation levels in promoter regions A to D and intron 1. Data is also given for a combined measure where methylation is considered positive if the mean percentage of methylated DNA is at least 15 % in regions C and D (CD15)

Variable		n	Median	Lower quartile	Upper quartile	P-value
Total		208	0.85	0.59	1.30
Age
	Years	208	70.96	61.93	78.95	0.37
Sex
	Female	100	0.86	0.43	1.41
	Male	108	0.84	0.66	1.20	0.66
Localization
	Right colon	58	0.68	0.30	1.01
	Left colon	67	0.91	0.64	1.42
	Rectum	83	0.91	0.70	1.38	0.19
Stage
	1	17	0.83	0.70	1.12
	2	82	0.82	0.47	1.20
	3	73	0.81	0.62	1.26
	4	36	1.02	0.68	1.59	0.83
IHC
	Negative	28	0.30	0.20	0.42
	Positive	180	0.95	0.70	1.42	<0.0001
MSI
	MSI	29	0.30	0.20	0.45
	MSS	179	0.96	0.70	1.42	<0.0001
MS-PCR region C
	Methylated	28	0.30	0.20	0.46
	Unmethylated	180	0.95	0.70	1.42	<0.0001
BRAF V600E
	Mutation	30	0.35	0.20	0.62
	Wild type	178	0.91	0.69	1.41	<0.0001
Region A
	Methylated	55	0.65	0.30	1.52
	Unmethylated	138	0.87	0.68	1.30	0.07
Region B
	Methylated	29	0.30	0.20	0.60
	Unmethylated	164	0.93	0.69	1.43	<0.0001
Region C
	Methylated	26	0.30	0.20	0.45
	Unmethylated	167	0.96	0.70	1.44	<0.0001
Region D
	Methylated	45	0.54	0.30	0.81
	Unmethylated	148	0.96	0.69	1.45	<0.0001
Intron 1
	Methylated	24	0.30	0.20	0.40
	Unmethylated	169	0.95	0.69	1.43	<0.0001
CD15
	Methylated	26	0.30	0.20	0.45
	Unmethylated	167	0.96	0.70	1.44	<0.0001

MLH1 methylation in primary tumors

MS-MLPA was successfully carried out in tumors from 195 patients. By using this method, in most cases (n = 118) no methylation was observed. Hypermethylation at any level was, however, observed in 77 cases: in 57 cases (29.2 %) in region A, in 30 cases (15.4 %) in region B, in 28 cases (14.4 %) in region C, in 47 cases (24.1 %) in region D, and in 25 cases (12.8 %) in intron 1. In 30 cases (15.4 %) the mean methylation of region C and D was at least 15 %. Methylation results in region C as estimated by both MS-MLPA and by MS-PCR were compared for validation purposes [23]. By doing so, an excellent correlation was observed (Spearman’s rho = 0.958, p < 0.0001).

MLH1protein expression, MSI and BRAF V600E status

Using immunohistochemistry (IHC), 28 tumors were found to be negative for MLH1 protein expression. Subsequent DNA analysis revealed MSI in all of them. One additional case showed MSI, but was positive for MLH1 protein expression. This latter tumor, however, lacked expression of one of the other mismatch repair proteins, MSH2. In DNA extracted from 30 tumors the BRAF V600E mutation was detected (Table 1). All three markers were found to be significantly correlated to MLH1 mRNA expression (p < 0.0001 in all cases).

Correlation between MLH1 expression and methylation

As shown in Table 1, highly significant correlations were observed between MLH1 mRNA expression and methylation in promoter regions B through D and intron 1. Methylation of region A was only trend-wise correlated to MLH1 mRNA expression (p = 0.07). To further illustrate the relationship between methylation and MLH1 mRNA expression, ROC analyses were performed (Fig. 2). By doing so, methylation levels of regions C and intron 1 were found to be excellently suited for discriminating between low and high expression, with areas under the curve (AUC) of 0.92 and 0.93, respectively. The protein expression levels measured by IHC and the combined methylation scores of regions C and D were also highly correlated to MLH1 mRNA expression (AUC = 0.93 and 0.92, respectively). Methylation of promoter regions A, B and D were to a lesser degree related to MLH1 mRNA expression levels.

Fig. 2.

Receiver operating characteristics (ROC) showing the ability of the binary classifier ‘methylation in a specific region yes/no’ or ‘protein expression by immunohistochemistry (IHC)’ to discriminate between high and low MLH1gene expression. When the classifier separates high and low gene expression perfectly, the curve is at the upper left corner and the area under the curve (AUC) is 1. When there is no correlation between methylation and gene expression, the curve follows a 45° line from lower left corner to upper right corner and the AUC = 0.5. AUC and 95 % confidence intervals (CI) are given in the lower right panel

Any methylation of any of the five regions was found to be significantly correlated with IHC status and this was also the case with the combined CD15 score (mean percentage of methylation in regions C and D > 15 %; p < 0.0001). Two-by-two comparisons are shown in Table 2, with the highest χ² for intron 1, CD15 and region C. The lowest χ² were observed for regions A, D and B. Three tumors lacked protein expression without concurrent DNA methylation. In two cases no tissue was available for further analyses. The third case was analyzed for germline mutations in MLH1 on the suspicion of Lynch syndrome, but none were found [17].

Table 2.

Association between methylation levels in different regions and protein expression. Two-by-two comparisons showing the correlations between protein expression (IHC) and methylation in different regions of the MLH1 promoter and intron 1. Data are presented for 195 cases (see text)

		Methylated			Pearson χ2
		Yes**	No
Region A
IHC	Negative	25	3*	28
	Positive	32	135	167
		57	138	195	57.0
Region B
IHC	Negative	25	3*	28
	Positive	5	162	167
		30	165	195	137.2
Region C
IHC	Negative	25	3*	28
	Positive	3	164	167
		28	167	195	149.3
Region D
IHC	Negative	26	2*	28
	Positive	21	146	167
		47	148	195	84.5
Intron 1
IHC	Negative	25	3*	28
	Positive	0	167	167
		25	170	195	171.0
CD15
IHC	Negative	25	3*	28
	Positive	2	165	167
		27	168	195	156.0

CD15 = positive for methylation if mean percentage of methylation in regions C and D is at least 15 %. Fisher’s exact test is in every case <0.0001. Numbers in italic are summed up. *Negative IHC without promoter hypermethylation is an indicator for a potential germline mutation

In several tumors more than one site was methylated, thereby giving rise to different methylation patterns. In Table 3 these patterns are listed together with the mRNA expression data. It appears that the cases with methylation of all five regions (ABCDI; n = 24) exhibited the lowest mRNA expression levels. Tumors with isolated methylation of region D, A or AB exhibited similar or even higher expression levels than the unmethylated tumors.

Table 3.

MLH1 methylation patterns and concomitant mRNA expression levels. The relative MLH1 mRNA expression level is given for the different patterns of methylation observed. Regions A to D are located in the gene promoter and region I in intron 1. Data are presented for 193 cases (see text)

Pattern	n	Median	MLH1 gene expression
			Lower quartile	Upper quartile	Minimum	Maximum
ABCDI	24	0.30	0.20	0.40	0.09	0.80
AD	1	0.50	0.50	0.50	0.50	0.50
ABC	2	0.53	0.47	0.60	0.47	0.60
D	20	0.85	0.74	1.28	0.35	3.58
No methylation	118	0.88	0.66	1.30	0.10	7.44
AB	3	1.31	1.04	3.10	1.04	3.10
A	25	1.52	1.19	1.73	0.37	3.42

MLH1 methylation status acts as survival predictor

In univariate survival analyses, any methylation of the MLH1 promoter (depicted in Fig. 3) was found to act as a significant predictor of survival (p = 0.01), whereas the MLH1 mRNA expression level was not (p = 0.13). Standard variables significant for survival were stage (p < 0.001) and age (p = 0.03). Sex (p = 0.11) and localization (p = 0.17) were not found to act as significant predictors.

Fig. 3.

Kaplan-Meier curves showing overall survival of patients with any methylation of the MLH1 promoter region compared to patients without methylation (p = 0.01, logrank)

In our multivariate analysis we included the four standard variables and methylation or mRNA expression. In this analysis only stage (hazard ratio (HR) 2.64, 95 % confidence interval (CI) 2.02–3.44, p < 0.001), age (HR = 1.04, 95 % CI 1.02–1.06, p < 0.001), and methylation (HR = 0.56, 95 % CI 0.36–0.89, p = 0.01) were found to act as significant predictors of overall survival.

Discussion

By dividing the MLH1 promoter region into regions A through D relative to their distance to the transcription start site, the relevance of specific methylation sites was established by Deng et al. [29]. By doing so, lack of MLH1 mRNA expression in colorectal carcinoma cell lines could be correlated to hypermethylation in region C and to a lesser extent in regions B and D. Subsequently, Deng et al. expanded their study through the inclusion of primary colorectal tumors and, concordantly, found a correlation between hypermethylation of region C and loss of MLH1 protein expression [22]. In an unselected larger cohort of primary colorectal cancer samples we have now resolved a strong correlation between hypermethylation of specific CpG islands within the MLH1 locus and the reduction in mRNA and protein expression levels. Our main findings (i) confirm that region C acts as a primary site for regulation of MLH1 gene expression through methylation, (ii) that assessment of methylation by MS-MLPA in regions D, B and especially A should not be used as an indicator of inactivated MLH1expression and (iii) that methylation of intron 1 may serve as a new and superior marker of decreased MLH1 mRNA and protein expression.

Promoter methylation represents one mechanism to (de)regulate mRNA and protein expression levels and, ultimately, cellular phenotypes. Other mechanisms include point mutations, gross chromosomal rearrangements, gene amplifications and microRNA-based mechanisms [30]. The observed high correlation between MLH1 methylation levels and both mRNA and protein expression levels indicates that methylation acts as a prime mechanism for its regulation in colorectal cancer. In only three cases we observed absence of MLH1 protein expression without concurrent gene methylation. One explanation for these latter observations may be the presence of inherited point mutations [31], but this possibility remains to be tested.

In the past, several laboratory methods and interpretation tools have been used for the assessment of MLH1 methylation [32], and it has been found that 60 % of the hitherto published studies were based on the analysis of non-specific regions within the MLH1 promoter [33]. Whereas specific regions relevant for MLH1 expression were reported to include C and D, our current data indicate that methylation measured by MS-MLPA in region C and intron 1 are most specific. The unspecific nature of region A was recently underscored by showing that this region was methylated in 42 % of MLH1 mutation carriers and in 17 % of sporadic microsatellite stable cases [34].

MLH1 mRNA expression in clinical samples has also been reported by others. Müller et al. [35] analyzed 27 sporadic cases and 26 hereditary cases and found a strong correlation between mRNA and protein expression. Ioana et al. [36] analyzed 29 sporadic cases and found increased levels of MLH1 mRNA expression in polyps and tumors compared to matched normal tissues. Ide et al. [37] compared MLH1 mRNA expression with methylation in the distal promoter region A in 94 primary tumors. They did not include the other, more specific, proximal regions of the promoter and the mRNA expression levels were not found to correlate with protein expression levels.

Hypermethylation of the MLH1 promoter was found to be a positive predictor of survival both in univariate and multivariate analyses. In the search for new prognostic markers, methylation may be a relevant candidate, since it is prevalent (40 % of colorectal tumors) and holds prognostic information beyond age, sex, stage and localization. The prognostic value of the MS-MLPA based assessment of MLH1 as observed here should be confirmed in an independent patient cohort and coupled to different carcinogenic pathways [38].

There are some limitations of this study. Firstly, in some cases the clinical material was sparse and it was not possible to obtain new samples if the quality was low. This also limited the possibility of conducting validation analyses using other methods. Secondly, the region specific methylation was analyzed by MS-MLPA. This method depends on the presence of specific restriction sites for the endonuclease and, in addition, relies on the performance of the probes. A low level of unspecific methylation may result from the use of sub-optimal probes. Thirdly, while RNA was extracted after microdissection, DNA was not. MS-MLPA requires larger amounts of nucleic acid to be feasible, and the best results were obtained in fresh frozen tumor samples compared to formalin fixed samples. This makes direct comparisons of methylation and expression levels difficult and, therefore, we have chosen not to analyze methylation as a continuous variable.

In conclusion, we found that MLH1 promoter methylation assessment by MS-MLPA in primary colorectal adenocarcinomas is feasible and, in addition, that the MLH1 methylation level serves as an independent prognostic marker. In daily practice, most laboratories analyze one region of the MLH1 promoter and, based on our current data, this should be either region C or intron 1. Next to IHC [39], MLH1 methylation analysis [40] should be readily available in order to enable optimal clinical management of individual patients with respect to prognosis, treatment choice and putative heredity.