DLEC1 and MLH1 promoter methylation are associated with poor prognosis in non-small cell lung carcinoma.

The significance of chromosome 3p gene alterations in lung cancer is poorly understood. This study set out to investigate promoter methylation in the deleted in lung and oesophageal cancer 1 (DLEC1), MLH1 and other 3p genes in 239 non-small cell lung carcinomas (NSCLC). DLEC1 was methylated in 38.7%, MLH1 in 35.7%, RARbeta in 51.7%, RASSF1A in 32.4% and BLU in 35.3% of tumours. Any two of the gene alterations were associated with each other except RARbeta. DLEC1 methylation was an independent marker of poor survival in the whole cohort (P=0.025) and in squamous cell carcinoma (P=0.041). MLH1 methylation was also prognostic, particularly in large cell cancer (P=0.006). Concordant methylation of DLEC1/MLH1 was the strongest independent indicator of poor prognosis in the whole cohort (P=0.009). However, microsatellite instability and loss of MLH1 expression was rare, suggesting that MLH1 promoter methylation does not usually lead to gene silencing in lung cancer. This is the first study describing the prognostic value of DLEC1 and MLH1 methylation in NSCLC. The concordant methylation is possibly a consequence of a long-range epigenetic effect in this region of chromosome 3p, which has recently been described in other cancers.

Lung cancer is one of the most common causes of cancer death. The overall 5-year survival rate for surgical resection of stage I non-small cell lung cancer (NSCLC) can achieve 60–75% while survival rates in stage II–IV patients remain poor. Unfortunately only a small subset responds to currently available treatments. Thus, it is important to identify and characterise new molecular markers and gene targets to improve the accuracy of prognosis and develop more targeted treatment strategies to improve the clinical management of lung cancer.

Allelic loss of chromosome 3p is one of the most frequent and earliest documented events in lung cancer, with deletions at 3p24–26, 3p21.3, 3p21.1–21.2, 3p14.2 and 3p12–13, suggesting the presence of multiple tumour suppressor genes on 3p (Hung ; Wistuba ; Zabarovsky ). Recent work has revealed the involvement of frequent epigenetic alterations in the inactivation of many 3p candidate genes, including BLU, FHIT, RASSF1A, RARβ and SEMA3B (Dammann ; Virmani ; Zochbauer-Muller ; Zabarovsky ; Ito ). Detection of methylated genes in serum and sputum DNA from lung cancer patients has also raised the possibility of using DNA methylation as an early detection marker (Esteller ; Palmisano ; Belinsky ; Usadel ).

Methylation of the MLH1 gene in 3p22.3 and its correlation with a mismatch repair defect and high microsatellite instability (MSI-H) is well characterised in sporadic colorectal cancer, where this phenotype is associated with better patient survival (Sinicrope ). In NSCLC MLH1 methylation has been described with frequencies ranging from 7 to 59% (Yanagawa ; Safar ) but in the absence of MSI-H (Benachenhou ; Okuda ). LOH within the MLH1 gene has also been detected in 55% (Benachenhou ) and reduced MLH1 expression in 59% of lung cancers (Xinarianos ). These intriguing findings have been followed by a recent report that MSH2, but not MLH1, methylation is a marker of poor prognosis in a Taiwanese cohort of nonsmoking female NSCLC patients (Hsu ). It remains to be determined if a mismatch repair gene defect has a role in lung carcinogenesis and why it is not associated with typical MSI-H.

The deleted in lung and oesophageal cancer 1 (DLEC1) gene is located about 1 Mb centromeric from MLH1 (Figure 1A). The 3p21.3 region was identified as one of the common deleted regions in lung cancer. Four candidate genes in this region were analysed but no evidence of their involvement in cancer development was found (Ishikawa ). Further analysis led to the identification of the DLC1 gene (Daigo ), which was later renamed DLEC1. Loss of DLEC1 expression has been observed in lung, oesophageal, renal, ovarian and nasopharyngeal carcinoma cell lines and primary tumours and functional analyses strongly suggest that DLEC1 is a tumour suppressor gene (Daigo ; Kwong , 2007). Promoter hypermethylation has been shown to be responsible for silencing of DLEC1 in ovarian cancer and in nasopharyngeal carcinoma (Kwong , 2007) but there has been no comprehensive methylation analysis reported for lung cancer.

Figure 1

(A) Schematic drawing of the short arm of chromosome 3 and the relative location of the RARβ, MLH1, DLEC1, RASSF1A and BLU genes. (B) DLEC1 (NM_005106) and GAPDH expression using RT–PCR (two upper panels) and methylation status using MSP (two bottom panels) in lung cancer cell lines and in normal human lung tissue. (C) Restoration of DLEC1 expression and concomitant demethylation of the CpG island in H1299 cells using the 5-aza treatment.

In this study, we investigated if promoter hypermethylation of DLEC1 is found in lung cancer and whether it has any prognostic significance. We determined the relationship of DLEC1 methylation with patient clinicopathologic variables and other 3p molecular markers, in particular MLH1, RARβ, RASSF1 and BLU methylation.

Patients and methods

Lung cancer patients

We reviewed the NSCLC surgery database maintained by the one cardiothoracic surgeon (BMC) for the period of 1994–2000. Patients who had received induction chemotherapy or for whom sufficient tissue was not available, were excluded. The final cohort had 155 (64.9%) men and 84 women (35.1%) with a median age at diagnosis of 68 years (range, 41–87 years) and a median survival time of 36.9 months (range, 1–113 months). Data on survival was obtained from the Cancer Registry of NSW, by routine follow-up visits or contact with the patient's general practitioner. Overall survival was measured from the date of surgery to the date of death or the date of last follow-up, censored patients being those who were alive at the time of last follow-up.

This study cohort consisted of 92 (38.7%) adenocarcinomas (ADC), 54 (22.7%) large cell carcinomas (LCC), and 92 (38.7%) squamous cell carcinomas (SCC). These tumours were classified according to the American Joint Committee on Cancer (AJCC) tumour-node metastasis classification (Grondin and Liptay, 2002) and consisted of 153 (64.0%) stage I and 86 (36.0%) stage II tumours (Table 2). The study was approved by the Ethics Review Committee of the Royal Prince Alfred Hospital (approval no. X02-0216).

DNA extraction and bisulphite treatment

Hematoxylin and Eosin-stained sections from paraffin-embedded tissue blocks were reviewed by an anatomical pathologist (WAC) for tumour and matching normal tissue specimens. Six to twelve serial 4 μm sections of each block were used for DNA extraction, depending on the size of the tissue. DNA extraction was carried out using the Puregen Genomic DNA purification kit (Gentra Systems, MN, USA). Sodium bisulphite conversion was performed as previously described (Millar ).

Expression of DLEC1 in lung cancer cell lines

Five lung cancer cell lines, A427, A549, NCI-H292, NCI-H1299 and NCI-H358, were used. Total RNA and DNA were extracted from cell pellets using RNeasy® Mini Kit and DNeasy® Tissue Kit (Qiagen GmbH Inc., Germany), respectively. Normal human adult lung RNA samples were purchased from Stratagene (Stratagene, CA, USA). One microgram of RNA from each sample was used in a reverse transcription reaction using GeneAmp RNA PCR kit (Applied Biosystems, CA, USA). Expression of DLEC1 was assessed by RT–PCR (DLEC1-F: 5′-TTCCTCCCTCGCCTACTC-3′; DLEC1-R: 5′-AAACTCATCCAGCCGCTG-3′). The primer pair was designed across exons 1 and 2 of the main DLEC1 transcript NM_005106. GAPDH was used as control.

To investigate if methylation regulates expression of DLEC1, cancer cells were treated with 5-aza-2′-deoxycytidine, a DNA methyltransferase inhibitor. Freshly seeded cells were grown overnight in normal medium, which was then replaced with medium containing 1 μM of 5-aza (Sigma-Aldrich Corporation, MO, USA). Cells were allowed to grow for 72 h, with 5-aza-containing medium changed every 24 h, and harvested for DNA and RNA extraction. A cell viability of >70% was retained after 72 h of treatment.

Methylation-specific PCR

Deleted in lung and oesophageal cancer 1 methylation status was assessed by a fluorescence based real-time detection quantitative methylation-specific PCR (MSP) with primers DLEC-m1, DLEC-m2 (Table 1) and a TaqMan® probe 5′-6FAM-TAATCAAACTTACGCTCACTTCGTCGCCG-BHQ1-3′ (Biosearch Technology, CA, USA) (Weisenberger ). A reference gene MYOD1 was employed to normalise the DNA input of each sample as previously described (Eads ; Kohonen-Corish ). Quantitative real-time PCR was performed for DLEC1 and MYOD1 in parallel using the RealMasterMix Probe ROX (Eppendorf, Hamburg, Germany) in the ABI7900HT Sequence Detection System (Applied Biosystems, CA, USA). Deleted in lung and oesophageal cancer 1 methylation was scored as present when the value of (DLEC1/MYOD1 × 100%)⩾5 or absent if the value is <5. All samples were run in duplicate.

Table 1

PCR primers used in the promoter methylation analysis of RARβ, MLH1, DLEC1, RASSF1A and BLU

Gene	Forward primer	Reverse primer
RARβ-m a	5′-TCGAGAACGCGAGCGATTC-3′	5′-GACCAATCCAACCGAAACGA-3′
MLH1-m	5′-AGCGATTTTTAACGCGTAAGC-3′	5′-CTCAATACCTCGTACTCACG-3′
DLEC1-m	5′-TTTCGTTGCGTATTTAAGATATTTC-3′	5′-CGTAACGCTCATTCTCGCTACC-3′
DLEC1-u b	5′-TAGTTTTGTAGTTTGGTTTTGTT-3′	5′-ACAAAATATCTTAAATACACAACA-3′
RASSF1A-m	5′-TTAGCGAAGTACGGGTTTAATC-3′	5′-CTACCGTATAAAATTACACGCG-3′
BLU-m	5′-CGTGGGTTATAGTTCGAGAAAGC-3′	5′-AACGAATTAACCGCGCCTACGC-3′
MYOD1 c	5′-CCAACTCCAAATCCCCTCTCTAT-3′	5′-TGATTAATTTAGATTGGGTTTAGAGAAGGA-3′

Primers specific for methylated, bisulphite converted DNA.

Primers specific for unmethylated, bisulphite converted DNA.

Control primers not including CpG sites specific for bisulphite converted DNA.

Methylation-specific PCR of other chromosome 3p genes RARβ, MLH1, RASSF1A and BLU was carried out (Table 1) together with MYOD1 amplification, as previously described (Eads ; Kohonen-Corish ). PCR steps included 30 s for denaturing, annealing and extension (40 cycles), initial denaturation and final elongation for 10 min, and annealing temperatures of 55°C (MLH1), 57°C (MYOD1), 63°C (BLU), and 58°C (RARβ, RASSF1A).

Immunohistochemistry and MSI analysis

MLH1 expression on tissue microarrays was analysed as part of a previous study (Cooper ). Matched normal bronchial mucosa or peripheral lung parenchyma specimens were used as control tissue for each patient. MLH1 expression was scored semiquantitatively by multiplying the percentage of cells showing nuclear expression and the intensity of staining using a 3-tier grading system (1=weak, 2=moderate and 3=strong staining). Reduced MLH1 expression was taken for a score less than 100, of the maximum score of 300. MSI was analysed as previously described (Kohonen-Corish , 2006), except that only two markers BAT25 and BAT26 were evaluated, which are sufficient for detecting high MSI (Suraweera ).

Statistical and survival analysis

Correlation between DLEC1 methylation and clinicopathologic parameters was determined using the χ2 test while survival analysis was performed using the Kaplan–Meier log-rank and Cox Proportional Hazards Model in the StatView package, and P<0.05 was regarded as statistically significant. Only those variables that were significant predictors of survival outcome in univariate analysis were incorporated into multivariate analyses.

Results

High correlation between promoter methylation and loss of expression of DLEC1

Expression of DLEC1 was assessed by RT–PCR in five lung cancer cell lines. While DLEC1 was expressed in normal lung tissue, no expression was detected in the A427, A549 and H1299 lung cancer cell lines (Figure 1B). We assessed DLEC1 methylation using methylation-specific PCR (MSP). Only the methylated allele was detected in the three cell lines where DLEC1 was not expressed, while both the unmethylated and methylated alleles were detected in cell lines expressing DLEC1 (Figure 1B). Methylation was rare in normal lung tissue (2.5%, 200 specimens analysed). To determine whether methylation directly regulates the silencing of DLEC1, the cell line H1299 was treated with 5-aza, a DNA methyltransferase inhibitor. After 3 days of 5-aza treatment, DLEC1 expression was restored and demethylation observed (Figure 1C).

Promoter methylation of DLEC1, MLH1, RARβ, RASSF1A and BLU in lung cancer

We employed MSP to assess the promoter methylation status of the five 3p candidate genes in 239 NSCLCs. Methylation was detected in 123 patients (51.5%) for RARβ, 86 (36.0%) for MLH1, 93 (38.9%) for DLEC1, 78 (32.6%) for RASSF1A and 85 (35.6%) for BLU (Table 2). Next we investigated the relationship between methylation of each set of two out of five genes. Significant correlation was observed between DLEC1 and MLH1 (P=0.0002), DLEC1 and RASSF1A (P=0.0003), and RASSF1A and BLU methylation (P=0.017). MLH1 methylation was also associated with RASSF1A (P=0.0006) and BLU (P=0.0005) (Table 3). Methylation of at least one of the five genes was detected in 204 of 239 (85.4%) patients; methylation of at least two genes in 139 (58.2%); three genes in 77 (32.2%); four genes in 36 (15.1%); and methylation of all five genes was detected in only nine (3.8%) patients.

Table 2

Association of clinicopathologic variables with poor overall survival and with RARβ, MLH1, DLEC1, RASSF1A and BLU promoter methylation in the cohort of 239 NSCLC patients.

	Kaplan– Meier log-rank	RARβ methylation			hMLH1 methylation			DLEC1 methylation			RASSF1A methylation			BLU methylation
		No. of patients			No. of patients			No. of patients			No. of patients			No. of patients
Clinicopathologic parameters	P-value	+a	−b	P-value	+	−	P-value	+	−	P-value	+	−	P-value	+	−	P-value
Age	0.5754
<68 (n=117)		63	54	0.4706	39	78	0.4032	49	68	0.3567	41	76	0.4371	34	83	0.0397
>68 (n=122)		60	62		47	75		44	78		37	85		51	71
Gender	0.1680
Male (n=155)		76	79	0.3068	57	98	0.7293	61	94	0.8488	49	106	0.6468	53	102	0.5475
Female (n=84)		47	37		29	55		32	52		29	55		32	52
Tumour differentiation	0.2864
Poor (n=103)		48	55	0.3388	44	59	0.3106	44	59	0.7712	35	68	0.3738	34	69	0.2847
Moderate (n=113)		65	48		35	78		41	72		37	76		41	72
Well (n=20)		9	11		6	14		7	13		4	16		10	10
Unknown (n=3)		1	2		1	2		1	2		2	1		0	3
Histologic type	0.5207
ADC (n=92)		56	36	0.0317	21	71	0.0047	29	63	0.0696	25	67	0.0008	42	50	0.0278
LCC (n=54)		20	34		22	32		19	35		29	25		17	37
SCC (n=92)		47	45		42	50		44	48		23	69		25	67
Primary tumour stage	0.1014
T1 (n=64)		31	33	0.8517	16	48	0.0995	17	47	0.0555	19	45	0.8406	22	42	0.7303
T2 (n=154)		81	73		62	92		66	88		52	102		57	97
T3 (n=21)		11	10		8	13		10	11		7	14		6	15
Regional lymph node status	<0.0001
N0 (n=174)		91	83	0.6728	54	120	0.0091	53	121	<0.0001	52	122	0.1378	68	106	0.0632
N1 (n=65)		32	33		32	33		40	25		26	39		17	48
AJCC staging	<0.0001
Stage I (n=153)		80	73	0.7341	46	107	0.0110	43	110	<0.0001	45	108	0.1562	62	91	0.0327
Stage II (n=86)		43	43		40	46		50	36		33	53		23	63
Total		123	116		86	153		93	146		78	161		85	154

+Indicates number of patients with a methylated tumour.

−Indicates number of patients with no methylation detected.

Table 3

P-values for pairwise correlation of promoter methylation in the 3p genes

	MLH1	DLEC1	RASSF1A	BLU
	P-value	P-value	P-value	P-value
RARβ	0.0691	0.1032	0.5543	0.7343
MLH1	—	0.0002	0.0006	0.0005
DLEC1	—	—	0.0003	0.0550
RASSF1A	—	—	—	0.0173

MLH1 expression in lung cancer tissue and MSI

Expression of MLH1 was previously determined using immunohistochemistry on tissue microarrays in 105 of the 239 patients (Cooper ). MSI was analysed in the whole cohort of 239 patients. Reduced MLH1 expression was detected in seven of the 105 cancers including an apparent loss of MLH1 expression in two cancers, but none of the matching DNA specimens prepared from a larger area of the tumour showed any MSI using markers BAT25 and BAT26. Also, none of the seven cancers with reduced MLH1 expression showed MLH1 promoter methylation. In the rest of the cohort MSI-H was detected in a stage 1B ADC (one marker) and a stage 2A LCC (both markers), of which only the latter was methylated in MLH1. There was no significant correlation between reduced MLH1 expression and survival (P=0.421).

Methylation of DLEC1 and MLH1 are associated with poor patient survival

A statistically significant association between methylation and histologic type was observed, where MLH1 methylation had a higher frequency in SCC (45.6%) and LCC (40.7%) compared with ADC (22.8%); RASSF1A methylation was associated with LCC (53.7%); BLU and RARβ methylation with ADC (45.7% and 60.9%). Furthermore, MLH1 and DLEC1 methylation were associated with the presence of regional lymph-node metastases and AJCC stage II. No association was observed between methylation of the five genes and age of diagnosis, gender or tumour differentiation status, except that BLU methylation was more common in older patients (Table 2).

Methylation of DLEC1 (P=0.0005), MLH1, (P=0.004), and RASSF1A (P=0.024) as well as regional lymph node status (P<0.0001) and AJCC stage (P<0.0001) were associated with poorer overall survival (Figure 2 and Table 2). RARβ and BLU methylation were not prognostic in the whole NSCLC cohort using the Kaplan–Meier log-rank analysis (P=0.313 and 0.474). Regional lymph node metastases and AJCC stage are two of the known prognostic factors for NSCLC and these two parameters are dependent predictors of survival in our cohort. Therefore, a bivariate analysis with the molecular marker predictor (DLEC1, MLH1 or RASSF1A methylation) and AJCC stage was set up. Methylation of either DLEC1 or MLH1 but not RASSF1A was a prognostic indicator independent of AJCC stage in the entire patient cohort (Table 4). Deleted in lung and oesophageal cancer 1 methylation was also a prognostic factor independent of AJCC stage in the SCC subgroup of patients (HR, 1.754; 95% CI, 1.023–3.007; P=0.041) and MLH1 methylation in LCC (HR, 2.926; 95% CI, 1.358–6.308; P=0.006).

Figure 2

Kaplan–Meier log-rank analysis of overall survival of NSCLC patients stratified by promoter methylation of (A) DLEC1; (B) MLH1, (C) RASSF1A, (D) RARβ, (E) BLU, (F) DLEC1/MLH1, (G) MLH1/RASSF1A and (H) DLEC1/RASSF1A (*P<0.05; **P<0.005; ***P<0.0001).

Table 4

Association of gene promoter methylation with survival in a univariate and a bivariate analysis, which takes into account AJCC stage of cancer

	Cox Proportional Regression Model
	Univariate	Bivariate	Univariate	Bivariate	Univariate	Bivariate	Univariate	Bivariate
Clinical or molecular marker	All samples (No.=239)		ADC (No.=92)		LCC (No.=54)		SCC (No.=92)
AJCC stage
HR	2.795	2.580–2.728	3.126	2.767–3.042	2.930	2.438–3.256	2.594	2.411–2.597
95% CI	2.009–3.889		1.808–5.404		1.417–6.056		1.506–4.467
P	<0.0001	<0.0001	<0.0001	<0.0007	0.004	<0.021	0.0006	<0.002
MLH1/DLEC1
HR	2.075	1.668	2.347	1.415	2.146	2.284	1.731	1.525
95% CI	1.428–3.015	1.138–2.447	1.210–4.555	0.683–2.932	0.939–4.906	0.992–5.257	0.984–3.043	0.863–2.696
P	0.0001	0.009	0.012	0.350	0.070	0.052	0.057	0.146
MLH1/RASSF1A
HR	1.951	1.688	1.702	1.245	3.011	3.223	1.390	1.206
95% CI	1.307–2.913	1.127–2.529	0.832–3.483	0.595–2.607	1.395–6.501	1.482–7.008	0.677–2.854	0.586–2.482
P	0.001	0.011	0.146	0.561	0.005	0.003	0.370	0.611
DLEC1/RASSF1A
HR	1.505	1.137	1.763	1.102	1.292	0.925	1.338	1.213
95% CI	1.015–2.233	0.759–1.705	0.927–3.352	0.548–2.214	0.574–2.909	0.395–2.163	0.686–2.607	0.623–2.362
P	0.042	0.533	0.084	0.785	0.536	0.857	0.393	0.571
MLH1
HR	1.621	1.421	2.034	1.512	2.591	2.926	1.007	0.972
95% CI	1.164–2.257	1.016–1.988	1.171–3.535	0.842–2.714	1.220–5.499	1.358–6.308	0.597–1.697	0.576–1.640
P	0.004	0.040	0.012	0.166	0.013	0.006	0.980	0.915
DLEC1
HR	1.783	1.471	1.681	1.337	1.454	1.103	1.968	1.754
95% CI	1.283–2.479	1.050–2.062	0.983–2.872	0.766–2.333	0.698–3.029	0.514–2.366	1.155–3.353	1.023–3.007
P	0.0006	0.025	0.058	0.306	0.317	0.802	0.013	0.041
RASSF1A
HR	1.474	1.259	1.397	1.174	2.453	1.902	1.138	1.021
95% CI	1.049–2.071	0.892–1.776	0.796–2.451	0.661–2.084	1.133–5.310	0.856–4.230	0.620–2.089	0.556–1.876
P	0.025	0.190	0.244	0.584	0.023	0.115	0.676	0.947

Abbreviations: AJCC=American Joint Committee on Cancer; HR=Hazard Ratio; CI=confidence interval.

We then investigated if concordant methylation of two genes affect patient prognosis (Figure 2; Table 4). Concordant MLH1/DLEC1 methylation was associated with poorer overall survival in both univariate (HR, 2.075; 95% CI, 1.428–3.015; P=0.0001) and bivariate (HR, 1.668; 95% CI, 1.138–2.447; P=0.009) analyses. Also, MLH1 methylation was prognostic in combination with RASSF1A methylation independent of AJCC stage in all patients (HR, 1.688, 95% CI, 1.127–2.529; P=0.011) and particularly in the LCC cohort (HR, 3.223; 95% CI, 1.482–7.008; P=0.003).

Discussion

Deleted in lung and oesophageal cancer 1 is a candidate tumour suppressor gene in multiple cancers. Although the function of DLEC1 is unclear, it suppresses tumour growth or reduces invasiveness of cancer cells (Daigo ; Kwong , 2007). In this study, we demonstrate for the first time that the DLEC1 promoter is methylated in lung cancer. The demethylating agent 5-aza reversed loss of mRNA expression in lung cancer cell lines. Frequent DLEC1 methylation (34.2%) was observed in NSCLC and was most common in SCC (47.8%). DLEC1 methylation was cancer-specific, as it was only rarely detected in matching normal lung tissue, and was strongly associated with stage II tumours and the spread of cancer to regional lymph nodes (P<0.0001). DLEC1 methylation was also associated with shorter overall survival in the whole cohort and in the SCC group of patients, and this remained statistically significant upon bivariate analysis with AJCC stage (Table 4). As there is no antibody available for DLEC1, we could not determine what proportion of methylated tumours would show loss or reduced DLEC1 protein expression. However, it has been previously demonstrated that DLEC1 RNA expression was lost in eight of 30 primary lung cancers and that this was not due to gene mutations (Daigo ).

The MLH1 gene is located within 1 Mb of DLEC1 in a locus that shows 55% LOH in NSCLC (Benachenhou ). Therefore, there has been some interest in determining the biological significance of reduced MLH1 gene expression and promoter methylation in lung cancer. As gene alterations can cause either increased sensitivity or resistance of tumours to chemotherapy treatment, we excluded those patients who had received induction chemotherapy prior to surgery to avoid a possible bias in the molecular analyses. MLH1 methylation was found in 36% of the cancers but did not result in the loss of gene expression in the 105 cancers analysed with immunohistochemistry. Only 6.7% of the cancers showed reduced MLH1 expression with stringent criteria (<100 of the maximum score of 300) and none of these specimens were methylated. We also found that MLH1 methylation was patchy and/or monoallelic in region C of the MLH1 promoter by using combined bisulphite-restriction analysis (COBRA, Hitchins ) (data not shown). This is consistent with the finding that MSI is extremely rare in NSCLC.

It is intriguing therefore, that MLH1 methylation showed strong prognostic significance, which is reported here for the first time. It was a marker of poor survival in the whole cohort, and particularly in the LCC subgroup, with both univariate and bivariate analyses. This is in contrast to colorectal adenocarcinoma where MLH1 methylation causes the MSI-H phenotype, which has improved prognosis. There was a high correlation between DLEC1 and MLH1 methylation (P=0.0002). As for DLEC1, MLH1 methylation was associated with stage II tumours and spread to regional lymph nodes. Concordant methylation of MLH1 and DLEC1 was also a marker of poor prognosis independent of stage in the whole cohort (Table 4).

The close correlation between MLH1 and DLEC1 methylation may be a consequence or a byproduct of a long-range epigenetic effect in this region of chromosome 3p. The first such chromosomal region reported was 2q14.2, which shows modification of chromatin structure such as histone H3-K9 methylation in colon cancer cells. This results in clusters of both methylated and unmethylated genes being coordinately suppressed (Frigola ). It has recently been shown that DLEC1 and MLH1 are also subject to long-range epigenetic regulation in colon cancer (Hitchins ). Multiple genes in this region can be simultaneously silenced through promoter hypermethylation and histone methylation in MSI-positive colorectal cancers. This effect appears to extend centromeric from the MLH1 gene and does not always reach DLEC1 in all specimens. In bladder cancer there is also evidence of such long-range epigenetic regulation around the DLEC1 gene, but here the predominant mechanism is gene silencing through histone methylation rather than CpG methylation, and MLH1 was not analysed (Stransky ). The two genes, which were analysed in both studies, DLEC1 and its neighbour PLCD1, are silenced through DNA methylation and H3-K9 dimethylation in colorectal cancer whereas in bladder cancer they are silenced through histone H3-K9 trimethylation. This suggests that there are tissue-specific differences in this regulation. Therefore, if such long-range epigenetic regulation of chromosome 3p is also operating in lung cancer, it is possible that some genes in the region may be affected less than others. As a consequence the overall methylation in this region could serve as a marker of poorer prognosis but only some genes show complete loss of function.

The other three genes analysed in this study RASSF1A, BLU and RARβ are known to be methylated in lung cancer and all have shown functional characteristics of tumour suppressor genes (Toulouse ; Shivakumar ; Agathanggelou ). RARβ is located 12 Mb telomeric from MLH1, and RASSF1A and BLU about 12 Mb centromeric from DLEC1. RASSF1 methylation was also highly correlated with DLEC1 (P=0.0003) and MLH1 methylation (P=0.0006), whereas RARβ was methylated independent of the other genes. The correlation between RASSF1A and BLU methylation observed here (P=0.017) has also been described previously (Agathanggelou ). However, none of these three markers were as strongly prognostic as DLEC1 and MLH1 methylation in this cohort. In a previous study RASSF1A methylation correlated with poor survival (Tomizawa ), but this has not been confirmed in all cohorts (Toyooka ; Choi ). Here, RASSF1A methylation was a prognostic marker in univariate analyses but not independent of stage, as was also observed previously (Choi ). It was interesting that concordant methylation of MLH1 with RASSF1 was an independent marker of poor prognosis. This suggests that a possible long-range epigenetic effect may extend centromeric from the DLEC1 locus but not telomeric from the MLH1 locus.

Taken together, our study has described two new prognostic markers, methylation of DLEC1 and MLH1 on chromosome 3p. Methylation of these two genes is clearly associated with each other and with methylation of RASSF1 and BLU, which are ∼12 Mb centromeric from DLEC1. MLH1 methylation itself does not lead to gene silencing in lung cancer and the biological significance of DLEC1 methylation also needs further study. In any case, concordant methylation of MLH1 with DLEC1 or RASSF1A is a valuable prognostic indicator in lung cancer. Future studies should reveal whether DLEC1, another gene or perhaps multiple genes in this region are functionally the most important in lung carcinogenesis.