RASSF1A, APC, ESR1, ABCB1 and HOXC9, but not p16INK4A, DAPK1, PTEN and MT1G genes were frequently methylated in the stage I non-small cell lung cancer in China

Abstract

Purpose

To identify the DNA methylation biomarkers for the detection of the stage I non-small cell lung cancer (NSCLC).

Materials and methods

The methylated state of p16INK4A, ESR1, HOX9, RASSF1A, DAPK1, PTEN, ABCB1, MGMT, APC and MT1G genes that have been reported frequently methylated in lung cancer was determined using methylation-specific PCR in four lung cancer cell lines, 124 cancer tissues of the stage I NSCLC and 26 non-cancerous disease tissues.

Result

The RASSF1A (53/124, 42.74%), APC (49/123, 39.52%), ESR1 (37/124, 29.84%), ABCB1 (31/124, 24.19%, MT1G (25/124, 20.16%) and HOXC9 (17/124, 13.71%) genes were more frequently methylated in the lung tissue from the stage I NSCLC than the non-cancerous lesion patients (2/26, 7.69%, P < 0.01; 2/26, 7.69%, P < 0.01; 2/26, 7.69%, P < 0.05; 1/26, 3.85% P < 0.01; 0/26 0%, P value: <0.01; 0/26, 0%, P < 0.05, respectively). p16INK4A was methylated in 28/124 (22.56%) of cancer tissues and 2/26 (7.69%) of non-cancerous tissues (P value >0.05). No significant association between the methylated state of the genes and the smoking, age or the pathologic types (squamous carcinoma, adenoma and the mixed types) was found. However, p16INK4A methylation was more frequently detected in the male (23/80, 28.75%) than the female (5/44, 11.36%, P > 0.05) patients. MGMT was barely methylated: 1/67, 1.49%), while DAPK1 and PTEN were not at all methylated in the cancer groups.

Conclusions

Methylation analysis in tissue of RASSF1A, APC, ESR1, ABCB1 and HOXC9 genes confirmed 79.8% of the existing diagnosis for the stage I NSCLC at specificity: 73.1%. The insufficiency of predicting disease onset in China, using the previously recommended targets (MGMT, DAPK1 and PTEN) in the United States reflects a potential disease disparity between these two populations. Alternatively, methylated state of this set of genes may be more specific to the late rather than the early stage of NSCLC.

Introduction

Lung cancer is the leading cause of cancer death in men and women in the United States and Western Europe. Over 160,000 Americans die of this disease every year. The 5-year survival rate is 15%—significantly lower than that of other major cancers (Jemal et al. 2007). Owing to the rapid industrialization and increase in the smoking consumption in society, lung cancer presents as the number 1 cancer type of threats in China (Song et al. 2008) (Globoscan 2002, http://www-dep.iarc.fr/). The non-small cell lung cancer (NSCLC) is the most common type, accounting for 85–90% of the total cases (Molina et al. 2008) and its current 5-year survival rate varies within the range from 2 to 47% for different stages and different pathological classes (squamous carcinoma versus adenocarcinoma) (Risch and Plass 2008). Its early detection is crucial to the prolonged survival of the disease. Imaging and cytology-based screening strategies have been employed for its early detection for many years (Anglim et al. 2008), but little progress, judged from whether and how much in the reduction of mortality, has been made. A better understanding of both genetic and epigenetic mechanisms underpinned with the disease initiation and progression should aid in the development of the better diagnosis, clinical management and outcome prediction of this devastating malignancy. DNA hypermethylation is recognized as an important mechanism for tumor suppressor gene inactivation that underlies the tumor-specific biological profile, as well as the promising biomarkers for the early detection of various types of cancer, including NSCLC.

Altered DNA methylation is a common hallmark for human cancer (Esteller 2008; Zhu and Yao 2009), reflecting the epigenetic disturbance that is likely to contribute to the changed patterns of the gene expression and, therefore, the unique phenotypes of cancer cells. The hypermethylated state of the promoter CGI of the genes, encoding the proteins that negatively regulate the cell proliferation and maintain the genome stability, such as tumor suppressor and DNA-repairing mediators, correlates well with the transcriptional silencing state of their expression in cancer cells. Therefore, this set of the targets has been intensively tested for their utility in cancer diagnosis and prognosis. Some of them have been reported specific to NSCLC from the studies in the patients outside of the mainland of China. The environmental factors that vary drastically with countries exert a great impact to both carcinogenesis and disease characteristics, which is believed to effect chiefly on the epigenetic makeup of cell. Aiming to identify the valuable DNA methylation biomarkers for the early detection of this type of cancer, we have methylation profiled, in the lung tissue of cohort of the stage I non-small cell lung cancer (124 diseases) and a cohort of the non-cancerous lesion patients (24 cases), ten methylated genes “informative: for NSCLC the United States: APC (Yanagawa et al. 2003), DAPK1 (Liu et al. 2007), p16INK4A (Nakata et al. 2006), PTEN (Marsit et al. 2005), RASSF1A (Nakata et al. 2006), ABCB1 (Anglim et al. 2008), MGMT (Liu et al. 2006), ESR1 (Tsou et al. 2005), HOXC9 (Anglim et al. 2008) and MT1G (Anglim et al. 2008). Although a significant association of the methylated RASSF1A (53/124, 42.74%), APC (49/123, 39.52%), ESR1 (37/124, 29.84%), ABCB1 (31/124, 24.19%, MT1G [25/124, 20.16%) and HOXC9 (17/124, 13.71%)] with the stage I disease was found, the remaining were either found little (p16INK4A) or no association with the stage I NSCLC (MGMT, DAPK1 and PTEN).

Materials and methods

Sample collection and genome DNA isolation

Lung cancer cell lines A549 (carcinoma, ATCC No: CCL-185TM),NCI-H460 (large cell cancer, ATCC No: HTB-177TM), NCI-H446 (small cell lung cancer, ATCC No: HTB-171TM) and SPC-A-1 (lung adenoma, Cell Bank in Shanghai No:TCHu 53) were cultured in l-DMEM medium containing 10% FBS at 37°C in a 95% air, 5% CO₂ humidified incubator to the log phase of proliferation stage before cell collection. BJ cell line (human fibroblast from skin, ATCC No: CRL-2522TM) was used as the control for the normal human cells.

With the informed consents obtained from all eligible patients and approval of the ethics committee, 124 tissue samples from the Stage I NSCLC and 26 from the non-cancerous respiratory diseases were collected from Shanghai Chest Hospital, China (Table 1 for the patient profiles, including the cell types, ages, sex and smoking etc). The tumor-node-metastasis (TNM) staging/classification of the patient cohorts were performed according to the WHO classification (Sobin and Wittekind 2002; Travis et al. 2004). The genomic DNA was prepared from both cell lines and the frozen clinical tissues using conventional proteinase K/organic extraction methods as previous described (Yu et al. 2007, 2002)

Table 1.

Clinical profile of the lung cancer patients and controls

	Non-small cell lung cancer cases (n = 124)	Non-cancerous lung lesions (n = 26)
Gender
Female	44	7
Male	80	19
31–40	3	0
41–50	12	5
51–60	42	10
61–70	43	6
70–	24	5
Range	32–79	42–75
Median	61.13	59.48
Stage category I	124
Squamous	38
Adenocarcinoma	70
Adenosquamous	11
Others	5
Pulmonary tuberculosis		6
Bronchiectasis		4
Pulmonary abscess		5
Organizing pneumonia		2
Pulmonary sclerosing hemangioma		3
Pulmonary giant lymph node hyperplasia		1
Pulmonary hamartoma		2
Pulmonary sequestration		1
Pulmonary inflammatory pseudotumor		2

^aAccording to the World Health Organization Guidelines for the conduct of tobacco-smoking surveys among health professionals (WHO 1984)

Bisulfite treatment of DNA and methylation-specific PCR (MSP)

The primer pairs for MSP analysis were directly taken from the literatures for APC (Esteller et al. 2000), CDKN2 (Rosas et al. 2001), DAPK1 (Zochbauer-Muller et al. 2001), PTEN (Yu et al. 2002), ESR1 (Wang et al. 2008) and RASSF1A (Lo et al. 2001) or designed for ABCB1, MGMT, HOXC9 and MT1G with the assistance of two relevant software (http://www.ebi.ac.uk/emboss/cpgplot/index.html) and (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) (Table 2). The bisulfate conversion and PCR analyses were done as described previously (Yu et al. 2007; Yu et al. 2002), followed by sequencing validation of the representative PCR products that have been T cloned.

Table 2.

Primer list for MSP

Gene symbol	GenBank no.	Sense 5′–3′	Antisense 5′–3′	Amplicon lication relative to transcription start	Size (bp)	References
RASSF1A M	XM_040961	GTGTTAACGCGTTGCGTATC	AACCCCGCGAACTAAAAACGA	−82  to +176	95	(Lo et al. 2001)
RASSF1A U		TTTGGTTGGAGTGTGTTAATGTG	CAAACCCCACAAACTAAAAACAA	+70 to +178	109
APC M	NT_034772	TATTGCGGAGTGCGGGTC	TCGACGAACTCCCGACGA	−163 to −66	98	(Esteller et al. 2000)
APC U		GTGTTTTATTGTGGAGTGTGGGTT	CCAATCAACAAACTCCCAACAA	169 to −62	108
ABCB1 M	NT_007933	TAGATGATTGTTTTCGGTTCGG	CCCCAATAATTCAACTAATACGC	+112,279  to +112,130	150
ABCB1 U		ATAGATGATTGTTTTTGGTTTGG	CCCCAATAATTCAACTAATACACAT	+ 112,280  to +112,130	151
MT1G M	NT_010498	TGCGGTGTGCGTTTAGTT	AAACCCAACAACCAACGA	−177 to −1	176
MT1G U		TGTGGTGTGTGTTTAGTTGTG	CCAACAACCAACAACTATTTTTA	−177 to −5	172
ESR1 M	NT_025741	ACGAGTTTAACGTCGCGGTC	ACCCCCCAAACCGTTAAAAC	+539 to +649	110	(Wang et al. 2008)
ESR1 U		TGGGTGAGGTGTATTTGGAT	AAACACCACAACCTCAAACC	+531 to +655	124
HOXC9 M	NC_000012	ACGCGTTCGTCGGTAGTA	GCGCTCCCTACACGATTC	+582 to +824	242
HOXC9 U		GGTTTGTTTGGGGAGTTGT	AAATCCTCACCCCCTCAC	+524 to +761	237
p16INK4A M	NM_000077	TTATTAGAGGGTGGGGCGGATCGC	ACCCCGAACCGCGACCGTAA	−80 to 69	149	(Rosas et al. 2001)
p16INK4A U		TTATTAGAGGGTGGGGTGGATTGT	CAACCCCAAACCACAACCATAA	−80 to 71	151
MGMT M	NT_008818	AGCGTCGTTGTTTTGTGC	CGCTTTCAAAACCACTCG	−439 to −254	186
MGMT U		TTGGTAGTGTTGTTGTTTTGTGT	CATCCTACAACCCCCACA	−457 to −249	209
DAPK1 M	NT_023935	TCGGTAATTCGTAGCGGTAG	TACTCACCCGAACGCCTA	+57 to +234	178	(Zochbauer-Muller et al. 2001)
DAPK1 U		GGGATTTGGTAATTTGTAGTGG	CCTAACTACTCACCCAAACACCT	+52 to +240	189
PTEN M	NT_030059	GTTTGGGGATTTTTTTTTCGC	AACCCTTCCTACGCCGCG	−1,441 to −1,264	176	(Yu et al. 2002)
PTEN U		TATTAGTTTGGGGATTTTTTTTTTGT	CCCAACCCTTCCTACACCACA	−1,446 to −1,261	186

Statistics

The association between gene methylation and clinical pathological variables was statistically tested. The hypermethylation frequency with a 95% confidence interval (95% CI; α = 0.05) was calculated using SPSS software. The hypermethylated incidence in non-small cell lung cancer versus the respiratory system lesion patients was completed by a 2X2 Fisher’s exact test and tested by the R statistical package. The receiver operating characteristics of both specificity and sensitivity of the sets was constructed as previously described (Yu et al. 2007).

Results

MSP profiling in lung cancer cell lines and the normal human cell line

Established cancer cell lines are expected to share the most biological features and the underlying molecular mechanisms, including the DNA methylation with the cancer in situ. Therefore, methylation profiling by MSP was carried out in four lung cancer cell lines: A549 (lung carcinoma), NCI-H460 (large cell carcinoma), NCI- H446 (non-small cell carcinoma) and SPC-A-1 (lung adenoma), the identity of the PCR products was confirmed by sequencing. BJ human fibroblast cell line, as the control for the normal diploid cells, was found in a full unmethylated state of each of all the ten genes analyzed (Fig. 1 and data not shown). As shown in Table 3 and Fig. 1, APC was fully hypermethylated in both A549 and SPC-A-1 and heterozygously methylated in both NCI-H460 and NCI- H446 cell lines. RASSF1A was fully hypermethylated in both A549 and NCI-H446 cell lines, but unmethylated in both NCI-H460 and SPC-A-1 cell lines. ESR1 was heterozygously methylated in both A549 and SPC-A-1, but unmethylated in both NCI-H460 and NCI- H446 cell lines. p16INK4A was heterozygously methylated in both NCI-H460 and SPC-A-1 cancer cell lines. PTEN was heterozygously methylated in SPC-A-1, but was unmethylated in A549, NCI-H460 and NCI- H446 cancer cell lines. ABCB1 was heterozygously methylated in both A549 and SPC-A-1 cancer cell lines, but unmethylated in both NCI-H460 and NCI-H446 cancer cell lines. MT1G was heterozygously hypermethylated in both NCI-H460 and SPC-A-1, but unmethylated in both A549 and NCI- H446 cancer cell lines. HOXC9 was heterozygously methylated in NCI-H446, while unmethylated in A549, NCI- H460 and SPC-A-1 cancer cell lines. Both DAPK1 and MGMT were unmethylated in all cancer cell lines.

Fig. 1.

The MSP profile and sequencing verification. The MSP profile and sequencing verification of both methylated and unmethylated targeted regions of the ten informative genes in each of four lung cancer cell lines: A549, NCL-H460, NCL-H446 and SPC-A-1. Both the electrophoretic patterns of the representative MSP data and the sequencing verification are shown. SssI is the positive control with the DNA of the normal liver tissue in vitro methylation modified. BJ is the bisulfate DNA from BJ, a normal fibroblast cell line. H ₂ O the no template control. The target identity of each panel is indicated. The wild-type sequence is aligned with the sequences of the T-vector cloned with the representative PCR product

Table 3.

Methylation profile from lung cancer cell lines

1 in the white background, the unmethylated in both alleles, 2 in gray background one allele methylated and the other unmethylated, and 3 in the dark background methylated in both alleles

MSP profiling in NSCLC and non-cancerous lesion control groups

We MSP profiled each promoter CGI in the cancer tissues from a subset of the stage I NSCLC patient cohort (67 cases) as well as the lung tissues of a subset of the non-cancerous disease cohort (13 cases) (Table 1). The following, then, were profiled in the remaining cases of both NSCLC and control cohorts: RASSF1A, APC, ABCB1, MT1G, ESR1 and HOXC9.

Death-associated protein kinase 1 (DAPK1) functions as a positive mediator of apoptosis and associated with Alzheimer’s disease. DAPK1 gene promoter methylation frequency in these tumors was 32.8% (40/122) and did not differ according to the patients’ smoking status, tumor histology or tumor stage (Liu et al. 2007). DAPK1 was unmethylated in all four lung cancer cell lines (Fig. 1; Table 3) as well as in both the cancerous and non-cancerous tissues (Fig. 2; Table 4).The tumor suppressor gene, PTEN encodes a lipid phosphatase that negatively regulates the phosphatidylinositol 3-kinase/AKT cell survival pathway, underlying the cell growth and invasion. Methylation of PTEN was previously found in 26% (39/151 cases) of NSCLC (Marsit et al. 2005), but was unmethylated in the cancer tissues (Table 4; Fig. 2).

Fig. 2.

The occurrence of the methylated genes in non-small cell lung cancer and non-cancerous disease control. The methylation frequency (%; Y axis) of each gene (X axis) in NSCLC patients (column 2, Table 4) and non-cancerous disease control (column 3, Table 4) was plotted. The hypermethylation frequency with a 95% confidence interval (α = 0.05) was calculated and indicated. P value for each gene is presented as vertical line in the plot. The positions of P values <0.01 and <0.05 in the plot are shown

Table 4.

MSP profiles of stage I non-small cell lung cancer and controls

Gene symbol	Non-small cell lung cancer/frequency (%) (n = 124)	Non- cancerous disease(s)/frequency (%) (n = 26)	Fisher exact test, right tail P value
RASSF1A	53/42.74	2/7.69	<0.01
APC	49/39.52	2/7.69	<0.01
ABCB1	16/24.19	1/3.85	<0.01
MT1G	25/20.16	0	<0.01
ESR1	37/29.84	2/7.69	<0.05
HOXC9	17/13.71	0	<0.05
P16INK4A	28/22.58	2/7.69	>0.05

	(n = 67)	(n = 14)
MGMT	1/1.49	0	>0.05
DAPK1	0	0	1
PTEN	0	0	1

MGMT encodes a DNA repair protein O6-methylguanine-DNA methyltransferase that removes alkyl adducts from the O6 position of guanine. Its inactivation by the promote CGI methylation was first reported in colo-rectal cancer, mechanistically underlying the increase in G–A mutation (Esteller et al. 1999). It was also methylated as high as 30.3% (27/126 cases) in NSCLC (Liu et al. 2006). However, it was only methylated in 1 of 67 cases (1.49%) of the cancer tissues (Table 4).

It has been previously established that transcriptional silencing of the tumor suppressor gene, RASSF1A was frequently mediated by the promoter CGI methylation in the human cancer (Lo et al. 2001), including NSCLC (Marsit et al. 2005) where 71 of 146 cases were reported hypermethylated. In this study, it was methylated at a compatible frequency: in the previous study in the patient cohort in the United States, in 53 of 124 cases (42.74) of cancer tissues and only 2 of 26 cases the non-cancerous tissues (7.69%), displaying a significant cancer state specificity (P < 0.01). Similarly, the promoter CpG island (CGI) hypermethylation/transcription silencing of APC was first reported in colo-rectal cancer (Esteller et al. 2000) and then in lung cancer. For instance, the increased level of DNA methylation at the APC promoter CGI in the non-small cancer tissues was found in 37% (28/75) of the cases tested (Yanagawa et al. 2003). In this study, the methylated APC was detected in 49 of 124 cases (39.52%) of cancer tissues and 2 of 26 cases (7.69%) of the non-cancerous controls, showing a significant cancer state association (P < 0.01). ESR1 (short for estrogen receptor alpha) encodes the main mediator of estrogen effect in breast epithelia and has also been shown to be activated by epidermal growth factor (EGF), methylation of which was found in 43% (3/7 cases) of lung adenocarcinoma tissue (Tsou et al. 2005). In this study, ESR1 methylation was also specific to the stage I NSCLC (P value <0.05) is, in 37 of 124 cases (29.84%), and in 2 of 26 cases (7.69%) of the non-cancerous tissues. ABCB1, short for ATP-binding cassette, subfamily B, member 1, encodes a protein that actively pumps the small molecule out from the cell. Its hypermethylated state has been frequently reported in various types of human cancer, including the squamous cell lung cancer where it was methylated in 100% cases in a small cohort (Anglim et al. 2008). We found that it was methylated in 31 of 124 cancer cases (24.19%) and in 1 of 26 cases (3.69%) of the non-cancerous tissues, showing a tight cancer state association (P < 0.001).

MT1G encodes a key member of the metallothionein protein, loss of which expression has been regarded instrumental to cancer formation and progression (Ferrario et al. 2008). The hypermethylated state of this gene was found in 95% cases of squamous cell lung cancer (Anglim et al. 2008). However, its methylated state was detected in 20.16% (25/124 cases and in none of 26 non-cancerous cases, exhibiting a clear specificity to the stage I disease (P < 0.01). HOXC9, encoding C9 member of the homeobox superfamily, was found hypermethylated in the squamous lung cancer frequently: 66%(Anglim et al. 2008). Its methylated state was found in 17 of 124 cancer cases (13.71%) and in none of 26 non-cancerous tissue control showing a significant cancer state association.

p16INK4A is a classic tumor suppressor, frequently methylated and lost in function in most types of cancer studied, including lung cancer. For instance, its methylated state was reported in 49 of 224 cases (21.9%) of NSCLC (Nakata et al. 2006). It was hypermethylated in 28 of 124 cases (22.58%) in cancer tissues. But, its methylated state was also found in 2 of 26 cases (7.69%) of the non-cancerous tissues and its cancer state association fails to survive by the stringent statistic test (Table 4). A plausible explanation for this observation may suggest that the methylation of p16INK4A may be specific to the premalignant state of the cancer, as proposed from the correlation of its methylation in the sputum DNA with the late onset of lung cancer (Palmisano et al. 2000).

Discussion

DNA methylation is a promising biomarker for cancer detection (Esteller 2008; Zhu and Yao 2009). DNA methylation profile underlines the gene expression pattern for the cancerous behaviors at the present time and reflects the past history of the interaction between the epigenomic interface of cells and the environment factors (Zhu and Yao 2009). Its value as the biomarker for cancer detection, especially for the early stage of the disease has been highly suggested. To determine the value of the set of the methylated genes that are “informative” for lung cancer in the United States (Table 5), both the stage I NSCLC (124 cases) and the non-cancerous lesion patient groups in China were recruited for MSP analysis.

Table 5.

Receiver operating characteristics of the informative sets for non-small cell lung cancer detection

No.	Gene					TP/FN (124)	FP/TN (26)	Sensitivity (%)	Specificity (%)
								TP/(TP + FN)	TN/(TN + FP)
1	RASSF1A					53/71	2/24	42.7	92.3
2	RASSF1A	APC				79/45	4/22	63.7	84.6
3	RASSF1A	APC	ESR1			93/31	6/20	75.0	76.9
4	RASSF1A	APC	ESR1	HOXC9		97/27	6/20	78.2	76.9
5	RASSF1A	APC	ESR1	HOXC9	ABCB1	99/25	7/19	79.8	73.1

Both sensitivity (%), TP/(TP + FN) (column 5, Table 5) and specificity (%), TN/(TN + FP) (column 6, Table 5) of each gene sets were calculated and plotted

TP true positive, FN false negative, FP false positive, TN true negative, NB inclusion of MT1G does not improve the detectability and therefore eliminated

RASSF1A, APC, ESR1, ABCB1, MT1G and HOXC9 were found methylated at a frequency no less than 13.71% (17 of 124 cases) of the stage I disease and no more than 7.69% (2 of 26 cases) of the non-cancerous tissues (Table 3), exhibiting a significant association with the stage I NSCLC. A comparable frequency of the methylated forms of both RASSF1A and APC were detected, while the methylated state other four informative genes (ESR1, ABCB1, MT1G and HOXC9) was found at a significantly lower incidence in this study than that in the previous reports (Table 4). This may indicate that the methylation of the later four genes may more specific to the late stage of disease. No significant association of the methylated state of any single target or six targets in combination with the age, gender, cell types and smoking habit (Table 1) has been identified. The p16INK4A methylation seems more tightly associated with the male (23/80, 28.75%) than female (5/44, 11.36%, P < 0.05), the implication of this observation to the clinical management of this disease remains to be assessed.

Methylation profiling of RASSF1A, APC, ESR1, HOXC9 and ABCB1, confirms 79.8% (99/124 cases) of the existing diagnosis of the stage I NSCLC at a specificity of 73.1%

Assaying a single gene confirmed no more than 43.74% (RASSF1A) of the stage I cancer (Tables 4, 5). However, the methylation analysis of multiple genes that were hypermethylated in the cancer at a statistically significantly higher frequency than the non-cancerous controls (4 genes with P < 0.01: RASSF1A, APC, ABCB1 and MT1G and 2 gene: ESR1 and HOXC9, with P < 0.05) (Fig. 2; Table 4) should confirm more cases by the existing diagnosis. However, four genes in this set were also methylated at a frequency no more than 2 of 26 cases (7.69%) of the non-cancerous control group, inclusion of which as the informative sets of the methylated targets for cancer detection would reduce the specificity for the cancer detection. The receiver operating characteristic (ROC) of both specificity and sensitivity was constructed for the sets consisting of two up to five genes (RASSF1A, APC, ESR1, ABCB1 and HOXC9) (Fig. 3; Table 5). The sensitivity was 42.7.3% (53/124) for RASSF1A alone, 63.7% (79/124) for RASSF1A and APC, 75% (93/124) for RASSF1A, APC, ESR1 and HOXC9, 78.2% (97/124) as well as 79.2%(99/124) for RASSF1A, APC, ESR1, HOXC9 and ABCB1. It was noticed that a further recruitment of MT1G would not improve the detectability of the set. As expected, the specificity by methylation profiling to detect the stage 1 disease was reduced from 92.3, 84.6, 76.9, 76/9 and 73.1% for the sets ranging from one to five targets, respectively. The clinical utility of the suggested set(s) of the targets for the stage 1 disease that are recommended from this study remains to be evaluated in large cohorts of NSCLC in a near future.

Fig. 3.

Receiver operating characteristics of the informative sets for non-small cell lung cancer detection. Both the sensitivity (%), TP/(TP + FN) (column 5, Table 5) and specificity (%),TN/(TN + FP) (column 6, Table 5) of each gene sets were calculated and plotted.TP true positive, FN false negative, FP false positive, TN true negative. Inclusion of MT1G does not improve the detectability and therefore eliminated

The possible disease disparity between the NSCLC in China and its US counterparts

The environmental factors that vary drastically with countries exert a great impact to both carcinogenesis and the clinical profile of the disease, which is believed to effect chiefly on the epigenetic makeup of cell. Three of ten genes that have been reported frequently methylated in NSCLC patients in the United States were either unmethylated [DAPK1 (Liu et al. 2007) and PTEN (Marsit et al. 2005)], or barely methylated [MGMT (Liu et al. 2006)] (Table 6), implicating the disease disparity in the NSCLC populations between the mainland of China and the United States. Alternatively, the methylated state of these three genes may not be associated with the early stage NSCLC.

Table 6.

Methylation frequency in the literature and in this study

Gene symbol	Cases/frequency (%)	Specimen	In this study (%)	References
RASSF1A	71/146 (49%)	Non-small cell lung cancer	42.74	(Marsit et al. 2006)
APC	28/75 (37%)	Non-small cell lung cancer	39.52	(Yanagawa et al. 2003)
ABCB1	100%	Squamous cell lung cancer	24.19	(Anglim et al. 2008)
MT1G	95%	Squamous cell lung cancer	20.16	(Anglim et al. 2008)
ESR1	3/7 (43%)	Lung adenocarcinoma	29.84%	(Tsou et al. 2005)
HOXC9	66%	Squamous cell lung cancer	13.71%	(Anglim et al. 2008)
p16INK4A	49/224 (21.9%)	Non-small cell lung cancer	22.58%	(Nakata et al. 2006)
MGMT	37/122 (30.3%)	Non-small cell lung cancer	1.49%	(Liu et al. 2006)
DAPK1	40/122 (32.8%)	Non-small cell lung cancer	0.00%	(Liu et al. 2007)
PTEN	39/151 (26%)	Non-small cell lung cancer	0.00%	Marsit et al. (2005a, b)

This report has provided a good set of the candidate targets for methylation analysis to detect the stage I NSCLC in China. It is desirable to further validate this set of informative targets in larger patient cohorts as well as assess their methylation status in the plasma DNA of the same clinical setting.

Abbreviations

APC

Adenomatous polyposis coli

DAPK1

Death-associated protein kinase 1

p16INK4A

Cyclin-dependent kinase inhibitor 2A

PTEN

Phosphatase and tensin homolog

RASSF1A

RAS association family 1A

ABCB1

ATP-binding cassette, subfamily B, member 1

MGMT

O(6)-methylguanine-DNA-methyltransferase

ESR1

Estrogen receptor alpha

HOXC9

Homeobox C9

MT1G

Metallothionein 1G