Development of a multiplex methylation specific PCR suitable for (early) detection of non-small cell lung cancer

Abstract

Lung cancer is a worldwide health problem and a leading cause of cancer-related deaths. Silencing of potential tumor suppressor genes (TSGs) by aberrant promoter methylation is an early event in the initiation and development of cancer. Thus, methylated cancer type-specific TSGs in DNA can serve as useful biomarkers for early cancer detection. We have now developed a “Multiplex Methylation Specific PCR” (MMSP) assay for analysis of the methylation status of multiple potential TSGs by a single PCR reaction. This method will be useful for early diagnosis and treatment outcome studies of non-small cell lung cancer (NSCLC). Genome-wide CpG methylation and expression microarrays were performed on lung cancer tissues and matched distant non-cancerous tissues from three NSCLC patients from China. Thirty-eight potential TSGs were selected and analyzed by methylation PCR on bisulfite treated DNA. On the basis of sensitivity and specificity, six marker genes, HOXA9, TBX5, PITX2, CALCA, RASSF1A, and DLEC1, were selected to establish the MMSP assay. This assay was then used to analyze lung cancer tissues and matched distant non-cancerous tissues from 70 patients with NSCLC, as well as 24 patients with benign pulmonary lesion as controls. The sensitivity of the assay was 99% (69/70). HOXA9 and TBX5 were the 2 most sensitive marker genes: 87% (61/70) and 84% (59/70), respectively. RASSF1A and DLEC1 showed the highest specificity at 99% (69/70). Using the criterion of identifying at least any two methylated marker genes, 61/70 cancer samples were positive, corresponding to a sensitivity of 87% and a specificity of 94%. Early stage I or II NSCLC could even be detected with a 100% specificity and 86% sensitivity. In conclusion, MMSP has the potential to be developed into a population-based screening tool and can be useful for early diagnosis of NSCLC. It might also be suitable for monitoring treatment outcome and recurrence.

Introduction

Lung cancer is a leading cause of cancer-related deaths worldwide. Its incidence and mortality is steadily on a rise.¹ It accounts for nearly 1 600 000 cases (13% of all cancer cases) and 1 400 000 (18%) cancer deaths globally per year.² It is the most common type of cancer detected in the male population (350 000 cases, 21.7% of all cancers) and is the second most common type of cancer in the female population (170 000 cases, 14.3%) after breast cancer in China.³ An increase of 1.6% in lung cancer incidence per year was reported from 1988 to 2005.⁴ The incidence rate for lung cancer in China was reported to be 53.6 per 100 000 in 2009.⁵ The mortality rate for lung cancer in China was 45.6 per 100 000 in 2009, accounting for 25.2% of cancer deaths in 2009.⁵ It has shown a dramatic increase of 4.7 times over the past three decades.⁶

Lung cancer is divided into non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC).⁷ NSCLC accounts for about 80% of the global lung cancer cases and is the leading cause of worldwide cancer-related deaths.⁸ Histologically, NSCLC is classified into adenocarcinoma (AC), squamous cell carcinoma (SCC), and large cell carcinoma.⁷ NSCLC exhibits a slower growth and spread as compared with SCLC;⁹ however, most NSCLC patients are diagnosed at an advanced stage of the disease.¹⁰ The major reasons for late diagnosis are the late appearance of symptoms and a lack of reliable biomarkers for its early detection.⁹

It has been proven that activation of oncogenes and inactivation of tumor suppressor genes (TSGs) are the main causes of lung cancer initiation and development. However, epigenetic changes may provide an alternative yet important mechanism of TSG silencing by means of DNA methylation, histone modification and posttranscriptional gene regulation by non-coding RNA,^11-13 and thus play important roles in tumorigenesis.^14-16 It has even been suggested that an epigenetic process can initiate cancer before any mutations occur.¹⁷ This makes epigenetic changes potentially interesting biomarkers for early diagnosis in cancer.^17-20 Current advances on epigenetic biomarkers in lung cancer have been reviewed in a recently publication.¹³ Among all epigenetic modifications, DNA methylation is the best characterized and it has also been reported to correlate with clinical characteristics of lung cancer.¹³ Aberrant methylation of cancer related genes has been suggested to be the most common mechanism of their inactivation,²¹ and is commonly found in lung cancer.²² Hypermethylation of critical genes involved in important cellular functions such as cell cycle regulation, differentiation, adhesion, and apoptosis was reported to be downregulated in lung cancer,^13,23 and this might contribute to the development of chemotherapy resistance.²⁴ DNA methylation of genes such as RASSF1A,²⁵ RARβ,²⁵ P16,²⁵ and SHOX2^26,27 has been evaluated for early detection of lung cancer.⁹ Promoter hypermethylation of NR2E1, OSR1, and OTX1 was also reported to be a good marker for early detection of both NSCLC subtypes.²⁸ Hypermethylation on P16 and MGMT was detectable three years before any clinical symptoms of SCC subtype NSCLC.²⁹ The development of reliable methods using DNA methylation marker genes for early detection of NSCLC will thus become a valuable tool for early screening for NSCLC and, ultimately, for estimation of prognosis and therapeutic targeting.

There are several techniques for the detection of DNA methylation and each technology has its advantages and disadvantages (for reviews see refs. 30–32). A suited method to screen large number of samples for methylation of specific genes is methylation specific PCR (MSP) as it is rapid and also cost-effective.³⁰ MSP is very sensitive and allows detection of a methylated gene among 1000 unmethylated copies of genomic DNA.³³ To address the limitations of various methylation detection assays, we previously developed a PCR-based assay—Multiplex methylation specific PCR (MMSP)—to simultaneously analyze the methylation status of multiple markers in a single reaction.³⁴ Now we have modified the MMSP to be suited for lung cancer using other marker genes. We show that it works on NSCLC biopsy samples.

We used methylation and expression microarrays that provide a global assessment of epigenetic and genetic alterations and their effects on expression. On the basis of this analysis and published data 38 TSGs were selected and screened for methylation status by MSP on DNA from lung cancer and matched distant non-cancerous and benign pulmonary lesion tissues from Tianjin Lung Cancer Institute, Tianjin Medical University General Hospital, P. R. China. According to the sensitivity and specificity of each gene among the screened markers we identified six TSGs— HOXA9, TBX5, PITX2, CALCA, RASSF1A, and DLEC1—, which were included in a PCR based “Multiplex methylation specific PCR (MMSP)”-assay for the validation of NSCLC. The method was capable of generating simultaneous multiplex information by a single PCR reaction, used to establish criteria for diagnosis.

Results

Identifying candidate genes using expression and genome-wide CpG methylation microarray

The average expression (DiffScore) and methylation (DeltaBeta) levels of 22 markers in three lung cancer and matched distant non-cancerous tissue pairs (no. 3, 23, and 27) are shown in Table 1 and their expression data in Figure 1. Based on this, the most informative 11 of the 22 markers were selected for the further study. These 11 markers were ALDH1A2, AQP1, CDO1, EPOR, GPR124, HLF, HOXA5, HSPA12B, TAL1, TBX5, and ZNF177.

Table 1. The average expression (DiffScore) and methylation (DeltaBeta) values of 22 commonly methylated and downregulated markers in three NSCLC and matched non-cancerous tissue samples.

UCSC_REFGENE_NAME	Average DiffScorea	Average DeltaBetab
ABCA3	153	0.3
ALDH1A2	332	0.4
AMT	83	0.2
AQP1	204	0.3
C7	271	0.4
CDO1	320	0.5
COX7A1	195	0.2
EPAS1	238	0.3
EPOR	52	0.2
GAS7	125	0.2
GPR124	276	0.4
HLF	338	0.4
HOXA5	226	0.3
HSPA12B	211	0.2
KCNA5	158	0.2
PLEKHB1	118	0.2
RGS5	78	0.1
SLC15A2	105	0.2
TAL1	181	0.3
TBX5	268	0.4
TRHDE	154	0.3
ZNF177	345	0.5

^a DiffScore obtained from the comparison of the expression of three NSCLC tissue samples using distant non-cancerous controls as a reference for each sample. ^bAverage methylation (DeltaBeta) obtained from the comparison of methylation of three NSCLC tissue samples using matched distant non-cancerous controls as a reference for each sample. Shaded: Tested by MSP. Non-shaded: Not tested by MSP.

Figure 1. Expression data for 22 commonly methylated genes that were methylated and showed lower expression in three NSCLC compared with their matched non-cancerous tissue samples (Fig. 3). These genes were also reported by Lokk et al.⁹ In addition, five other genes have also been included (ADCY4, RASSF1A, FLJ21511, SNRPN, and RYR2), which were identified only by Lokk et al.⁹ but not by our own array data.

Validating the hypermethylation status of candidate genes in tumor using methylation specific PCR (MSP)

Together with the additional 27 methylation regulated TSGs identified in the literature, 38 (11+27) markers were pre-screened on a small group of DNA samples from lung cancer and matched distant non-cancerous tissue by MSP.

From all these 38 markers with 92 pairs of primers, the six markers with the highest sensitivity and specificity were HOXA9, TBX5, PITX2, CALCA, RASSF1A, and DLEC1 (Table 2).

Table 2. Sensitivity and specificity of all 38 MSP markers.

Marker	Sensitivity %a	Specificity %b
ADCY4	33 (2/6)	100 (6/6)
ALDH1A2	0 (0/3)	100 (3/3)
APC	27 (3/11)	82 (9/11)
AQP1	100 (3/3)	0 (0/3)
ASC	75 (3/4)	0 (0/4)
BNC1	43 (6/14)	93 (13/14)
BVES	60 (3/5)	60 (3/5)
CALCA	67 (2/3)	100 (3/3)
CDH13	20 (1/5)	100 (5/5)
CDO1	33 (1/3)	100 (3/3)
CNTNAP2	33 (1/3)	67 (2/3)
CYB5R2	0 (0/2)	100 (2/2)
DAPK	25 (1/4)	100 (4/4)
DLEC1	67 (2/3)	100 (3/3)
E-CADH	25 (1/4)	100 (5/5)
EPOR	100 (3/3)	0 (0/3)
GDNF	33 (1/3)	100 (3/3)
GPR124	100 (3/3)	0 (0/3)
HLF	33 (1/3)	67 (2/3)
HOXA5	100 (3/3)	0 (0/3)
HOXA9	80 (4/5)	100 (5/5)
HSPA12B	33 (1/3)	100 (3/3)
ITGA9	33 (1/3)	100 (3/3)
KLK10	29 (2/7)	100 (7/7)
LOX	0 (0/3)	100 (3/3)
MYOD1	33 (1/3)	100 (3/3)
OPCML	33 (1/3)	100 (3/3)
P16	17 (1/6)	100 (6/6)
PAX6	100 (3/3)	67 (2/3)
PITX2	75 (3/4)	75 (3/4)
RARβ	25 (1/4)	100 (4/4)
RASSF1A	67 (2/3)	100 (2/2)
TAL1	33 (1/3)	100 (3/3)
TBX5	86 (12/14)	86 (12/14)
TSLC1	29 (2/7)	71 (5/7)
WIF1	33 (1/3)	100 (3/3)
WNT7A	0 (0/3)	100 (3/3)
ZNF177	33 (1/3)	100 (3/3)

^a Sensitivity: number of positive cases in NSCLC patients / total number of NSCLC cases tested. ^bSpecificity: (total number of tested matched distant non-cancerous controls – number of positive cases in tested matched distant non-cancerous controls) / total number of tested matched distant non-cancerous controls. Shaded: Marker genes selected for MMSP assay.

Detection of combination of marker genes by multiplex methylation specific PCR (MMSP)

Subsequently, these six markers were employed in an MMSP assay testing 70 lung cancer with matched non-cancerous controls and 24 additional normal lung tissue samples from patients with benign pulmonary lesion (Fig. 2). The level of methylation in cancer DNA was clearly different from matched control DNAs. The sensitivity and specificity of all individual markers included in the MMSP is given in Table 3. HOXA9 showed the highest sensitivity (87% [61/70]), whereas DLEC1 was lowest, with a sensitivity of 37% (26/70). RASSF1A and DLEC1 provided the highest specificity as 99% (69/70), whereas the TBX5 showed the lowest specificity (86% [60/70]). DNAs from 24 normal lung tissue samples from patients with benign pulmonary lesion were negative using these six methylation markers.

Figure 2. MMSP assay applied to DNAs from NSCLC (T) matched non-cancerous control (MC) and normal (N) tissue samples. A mixture of cell-line DNA from A549 and H1299 was used as a positive control (see Method and Materials section). P1: 27 ng A549 and H1299 DNA in 1:1 ratio; P2: 13 ng A549 and H1299 DNA in 1:1 ratio to confirm the semi-quantitative nature of MMSP assay. Water was used as a blank control.

Table 3. Methylation status of individual marker genes in NSCLC with matched distant non-cancerous controls by MMSP.

Marker	Sensitivity % (methylated/total)	Specificity % (unmethylated/total)
HOXA9	87 (61/70)	97 (68/70)
TBX5	84 (59/70)	86 (60/70)
PITX2	77 (54/70)	97 (68/70)
CALCA	59 (41/70)	93 (65/70)
RASSF1A	47 (33/70)	99 (69/70)
DLEC1	37 (26/70)	99 (69/70)

Concurrent methylation of the six MMSP markers in NSCLC tissue and matched distant non-cancerous control lung tissue are summarized in Table S1. Among the 70 NSCLC tissue samples, one case (<1%) was free of methylation for all six methylation markers. The remaining 69 NSCLC tissue samples were methylated for at least any one marker. Out of the 70 matched distant non-cancerous controls, 79% (55/70) were free of methylation for all six markers while the remaining 21% (15/70) samples were methylated for at least any one marker. None of the matched distant non-cancerous controls was found to be concurrently methylated in more than three markers.

The sensitivity and specificity of the panel of these six methylation markers in NSCLC patients in combinations of markers is presented in Table 4. Among the 70 NSCLC tissue samples, 61 samples were methylated for two up to six markers in different patterns. Sensitivity (Sn) and specificity (Sp) were as follows: at least any 2 markers: Sn: 87% (61/70), Sp: 94% (66/70); at least any 3 markers: Sn: 76% (53/70), Sp: 97% (68/70); at least any 4 markers: Sn: 60% (42/70), Sp: 100% (70/70); at least any 5 markers: Sn: 49% (34/70), Sp: 100% (70/70) and all six markers: Sn: 21% (15/70), Sp: 100% (70/70). Thus, the sensitivity and specificity revealed by at least any 2 of the marker genes reached levels that are fit for detection and diagnosis.

Table 4. Cumulative values of sensitivity and specificity using combinations of 1 – 5 marker genes for detection of NSCLC.

No. of Methylated markers	0	1				2				3					4		5			6
HOXA9	-	-	+	-	-	+	+	-	-	+	+	-	+	+	+	+	+	+	+	+
TBX5	-	-	-	-	+	-	+	+	+	+	-	+	+	+	+	+	-	+	+	+
PITX2	-	-	-	+	-	+	-	+	-	-	+	+	+	-	+	+	+	+	+	+
CALCA	-	+	-	-	-	-	-	-	+	+	-	-	-	-	+	-	+	+	+	+
RASSF1A	-	-	-	-	-	-	-	-	-	-	-	-	-	+	-	+	+	-	+	+
DLEC1	-	-	-	-	-	-	-	-	-	-	+	+	-	-	-	-	+	+	-	+
NSCLC tissue samples (n: 70)	1	1	3	1	3	3	3	2	0	3	0	1	5	2	3	5	2	8	9	15
Matched Controls (n: 70)	55	3	0	1	7	0	0	0	2	0	1	0	0	1	0	0	0	0	0	0
At least any 5 markers																	*Sn: 49 (34/70) #Sp: 100 (70/70)
At least any 4 markers															*Sn: 60 (42/70) #Sp: 100 (70/70)
At least any 3 markers										*Sn: 76 (53/70) #Sp: 97 (68/70)
At least any 2 markers b						*Sn: 87 (61/70) #Sp: 94 (66/70)
At least any 1 marker a		*Sn: 99 (69/70) #Sp: 79 (55/70)

^a As long as there is any one marker gene methylated, that sample is considered as positive. ^bOnly if there are 2 marker genes or more than 2 marker genes are methylated in the sample, that sample is considered as positive. ^*Sn: Sensitivity: % (methylated / total cancer samples). ^#Sp: Specificity: % (unmethylated / total matched control). Shaded squares with + sign: Methylated. Non-shaded squares with - sign: Unmethylated.

Thirteen different combinations of the six methylated markers were found in all NSCLC tissue samples. The sensitivity and specificity of all these 13 combinations are shown in Table S2. The combination of HOXA9, TBX5, and PITX2 (with and without RASSF1A) showed the highest sensitivity using at least any two markers—86% (61/70)—with 97% (68/70) specificity. An increase in the number of minimal markers from two to three decreased the sensitivity for all the combinations.

A comparison of the clinical parameters of the patients in accordance with the MMSP data are shown in Table 5. Methylation of RASSF1A was significantly different between treated and untreated patients (Chi² test, P = 0.026). HOXA9 showed a significant difference (Chi² test, P = 0.0495) between smokers and non-smokers. DLEC1 also showed some difference in smokers vs. non-smokers, but it was not statistically significant (Chi² test, P = 0.088). We also found some difference in the methylation status of the DLEC1 marker gene between subtypes of NSCLC but, upon statistical analysis, the difference turned out to be non-significant. There were no significant differences between the cases and matched non-cancerous controls in terms of gender, age, type and stage of the disease.

Table 5. Comparison of clinical characteristics of NSCLC patients with outcome for methylation marker genes, individual or in combinations.

Clinical Parameter		Sensitivity of MMSP Marker Genes % (+/total)						At least any two Marker Genes
		HOXA9	TBX5	PITX2	CALCA	RASSF1A	DLEC1	Sensitivity % (+/total)	Specificity % (-/total)
	Total	87 (61/70)	84 (59/70)	77 (54/70)	59 (41/70)	47 (33/70)	37 (26/70)	87 (61/70)	94 (66/70)
Gender	Male	84 (37/44)	89 (39/44)	73 (32/44)	59 (26/44)	45 (20/44)	43 (19/44)	89 (39/44)	93 (41/44)
	Female	92 (24/26)	77 (20/26)	85 (22/26)	58 (15/26)	50 (13/26)	27 (7/26)	85 (22/26)	96 (25/26)
Age	<62.5	86 (31/36)	86 (31/36)	81 (29/36)	58 (21/36)	44 (16/36)	44 (16/36)	86 (31/36)	94 (34/36)
	>62.5	88 (30/34)	82 (28/34)	74 (25/34)	59 (20/34)	50 (17/34)	29 (10/34)	88 (30/34)	94 (32/34)
Smoking	Yes	84 (32/38)	87 (33/38)	76 (29/38)	63 (24/38)	45 (17/38)	45 (17/38)	89 (34/38)	97 (37/38)
	No	100 (22/22)	86 (19/22)	77 (17/22)	55 (12/22)	50 (11/22)	23 (5/22)	91 (20/22)	91 (20/22)
Treatment	Yes	78 (7/9)	89 (8/9)	67 (6/9)	56 (5/9)	11 (1/9)	33 (3/9)	89 (8/9)	100 (9/9)
	No	88 (50/57)	82 (47/57)	77 (44/57)	60 (34/57)	51 (29/57)	37 (21/57)	86 (49/57)	93 (53/57)
Type	AC	86 (30/35)	80 (28/35)	74 (26/35)	54 (19/35)	51 (18/35)	29 (10/35)	86 (30/35)	94 (33/35)
	SCC	89 (31/35)	89 (31/35)	80 (28/35)	63 (22/35)	43 (15/35)	46 (16/35)	89 (31/35)	94 (33/35)
M Stage	M0	79 (50/63)	75 (47/63)	71 (45/63)	59 (37/63)	43 (27/63)	38 (24/63)	86 (54/63)	97 (61/63)
	M1	100 (7/7)	100 (7/7)	86 (6/7)	43 (3/7)	71 (5/7)	29 (2/7)	100 (7/7)	71 (5/7)
Stage	I+II	89 (25/28)	86 (24/28)	71 (20/28)	57 (16/28)	39 (11/28)	32 (9/28)	86 (24/28)	100 (28/28)
	III+IV	86 (36/42)	83 (35/42)	81 (34/42)	60 (25/42)	52 (22/42)	40 (17/42)	88 (37/42)	90 (38/42)

Shaded squares indicate a P value <0.05.

The criterion of at least any two markers could not discriminate between the different clinical parameters. Importantly, however, it identified 86% (24/28) of the patients at early stages I or II with better five year survival than late stages, and with a 100% specificity. Thus, this criterion may be applicable for future screening and early tumor discovery in high-risk populations.

Discussion

It has been shown that a panel of three to four well-selected methylation markers can identify an abnormality in 70–90% cancers and also discriminate between different subtypes.^35-37 In addition, it can also be suggestive of the aggressiveness and outcome of the tumor.³⁸ Based on this, we carefully selected six cancer-related markers and one quality control marker, β-ACTIN, to be used in our MMSP assay. The primers for the six potential TSGs, namely HOXA9, TBX5, PITX2, CALCA, RASSF1A, and DLEC1, were designed to detect aberrant methylation in DNA samples from NSCLC patients. All 24 normal control samples from patients with benign pulmonary lesion were unmethylated at all MMSP marker genes.

The sensitivity of individual markers included in the MMSP panel ranged from 37% to 87%, with a specificity ranging from 86% to 99% (Table 3). HOXA9 marker showed the highest sensitivity (87%), whereas the RASSF1A and DLEC1 displayed the highest specificity (99%).

The sensitivity and specificity of methylation markers in the MMSP assay was as follows: at least any one marker: Sn 99% and Sp 79%; at least any two markers: Sn 87% and Sp 94%; and at least any three markers: Sn 76% and Sp 97%. In matched distant non-cancerous control samples, 21% (15/70) were methylated for at least any one marker. As epigenetic changes occur early during the multistep process of tumorigenesis,^17-20,28,29 it is quite possible that methylation of some genes had already started in distant, apparently non-cancerous tissue, which in fact was pre-cancerous,

Detecting aberrant methylation of a single-gene biomarker is not enough to diagnose a cancer.³⁵ This is because tumorigenesis is a multistep process that involves various other factors as well. It was obvious from Table 4 that an increase in sensitivity was always accompanied by a decrease in specificity. Therefore, in order to avoid the chances of any false positive cases, the most balanced and diagnostically applicable threshold could be given by determining the presence of at least any two methylation markers, where the sensitivity and specificity were as 87% and 94% respectively. With the criterion of at least any two methylation markers we could detect 86% (24/28) samples at the early stage I or II of the disease, with 100% (28/28) specificity. The lower overall sensitivity of the MMSP assay at this threshold value was due to the lower sensitivity of the marker DLEC1 marker. Although we screened 38 TSGs, no other suitable marker could be found that could replace DLEC1 to improve the sensitivity. Although DLEC1 sensitivity of 37% was not comparable to that of the other five markers (Table 3), DLEC1 showed the highest specificity (99%). This makes it a good biomarker candidate for early diagnosis of NSCLC.

To optimize the sensitivity and specificity of the MMSP, 13 different combinations of markers were compared (Table S2). The highest sensitivity was 86%, when the methylation of at least two markers was taken into account whereas, with this criterion, the highest specificity was 100%. When this result was compared with the methylation of at least three markers, the specificity ranged between 97% - 99%, while the highest sensitivity was 76%. This analysis suggests that at least two markers criterion would be the optimal choice for the MMSP assay. According to this data (Table S2) there were two combinations of MMSP markers showing the highest sensitivity and specificity with the criterion of at least any two markers. One such combination was HOXA9, TBX5, and PITX2, while the other combination was of HOXA9, TBX5, PITX2, and RASSF1A. In both these combinations, the sensitivity and specificity were 86% (61/70) and 97% (68/70), respectively. This raises the possibility to reduce the MMSP markers to four.

The markers included in the MMSP have been described to have a role in early tumorigenesis. HOXA9 marker is a member Homeobox genes superfamily that encode transcription factors, which regulate embryonic morphogenesis in animals.³⁹ These proteins have an essential role in developmental gene regulation and thus are potential targets during tumorigenesis.⁴⁰ HOXA9 was found to be methylated in lung cancer Korean patients by using pyrosequencing technique and was suggested to be a potential marker for early diagnosis⁴¹ and to play a role in the pathogenesis of NSCLC patients.⁴² Induction of HOXA9 expression inhibits the invasive property of lung cancer cell lines with low expression of HOXA9.⁴² TBX5 is a member of T-box family of genes that encodes transcription factors regulating the developmental processes, cell–cell signaling, cardiac muscle cell proliferation, induction of apoptosis and cell migration.^43-46 Its ectopic expression in osteosarcoma (U2OS) and lung cancer (H1299) cells results in apoptosis and inhibition of cell proliferation.⁴³ TBX5 has been found to be methylated in colon cancer and could serve as a potential biomarker for prognosis. It has been found to suppress tumor cell proliferation and metastasis in colon cancer cell lines.⁴⁷ PITX2 is involved in embryonic and fetal development.^48,49 DNA methylation of PITX2 has been found to be an independent prognostic biomarker for the progression of the disease in NSCLC patients.⁵⁰ It was found that the higher methylation of the PITX2 gene correlates with lower risk of progression-free survival of the disease, making this marker helpful in optimizing individualized therapy.⁵⁰ CALCA is also considered to have a very important role in embryonic and fetal growth.⁵¹ It has been found to have different roles in vascular tone,⁵² cell growth and differentiation,⁵³ migration,⁵⁴ and adhesion.⁵⁵ CALCA was also found to be methylated in NSCLC and a positive correlation between CALCA methylation and invasion or adhesion was observed in NSCLC.⁵⁶ CALCA methylation is also correlated with poor prognosis of lung cancer.⁵⁷ RASSF1A is a member of RASSF1 family of proteins that possess tumor suppressive properties.⁵⁸ Its expression is also lost or downregulated in lung cancer cell lines.⁵⁹ Downregulation of RASSF1A expression by promoter hypermethylation has been reported in at least 37 tumor types.⁶⁰ RASSF1A was found to be methylated in 32% of stage IA lung adenocarcinomas and correlated with pleural invasion and poor differentiation.⁶¹ It affects tumor suppression through the control of microtubule polymerization with a potential role in genome stability.⁶² RASSF1A is related to cell differentiation, proliferation and survival but its use as an early detection or prognosis of NSCLC has been reported to be inefficient.⁶³ The biological function of DLEC1 is still not clear but it was found to be methylated in lung cancer and to be a marker of poor prognosis, independently of stage of the disease.⁶⁴ DLEC1 methylation was found in 39% of lung cancer samples and was correlated with clinical stage, lymph nodes metastasis and poor prognosis.^65,66

Considering at least any two markers as a positive criterion, we achieved a diagnostic rate of 87% (61/70) with a specificity of 94% (66/70). More importantly with this criterion we could detect 86% (24/28) of the stage I and II samples with a specificity of 100% (28/28). MMSP did not detect methylation in any of the markers in any of the 24 normal lung tissue samples from patients with benign pulmonary lesion whereas its detection rate for any one marker for cancer sample was 99% (69/70). Although this is a very satisfactory outcome for the MMSP, one might still consider combining it with detection of “diagnostic” mutations, particularly when applied to other forms of cancer.

The present data was generated on tissue samples. The next step will be to extend this study to DNA in serum, sputum, or bronchoalveolar lavage fluid from lung cancer patients and compare the results to those from controls and tissue samples DNA. This would be necessary in order to establish this sensitive and cost effective technique as an early detection tool. Further studies are required to test the feasibility of MMSP assay as a population-based screening tool in high-risk population as well as a way of monitoring tumor recurrence. The methylation status of multiple potential TSGs can be evaluated in a single reaction using only routine lab equipment. As pattern and timing of methylation status in certain genes are associated with defined biological behaviors,⁶⁷ we suggest that detection of TSG methylation in body fluids by MMSP-assay method might not only provide diagnostic information, but also has the potential of predicting specific behavior of individual tumors.

MMSP assay has the potential to be used to monitor metastatic potential of NSCLC, but this can only be addressed by follow-up studies of a cohort. MMSP should be tested as a tool to monitor clinical follow-up of treatment (outcome research) and metastasis risk in NSCLC. Prognostic markers, such as DLEC1^64,65 and PITX2⁵⁰ in the MMSP assay will be particularly useful for this purpose.

Materials and Methods

Cell lines and clinical samples

Human cell line A549 and H1299 of NSCLC were cultured in RPMI medium containing 10% fetal calf serum (FCS) at 37 °C with 5% CO₂. Seventy NSCLC tissues and matched distant non-cancerous tissues from pathology-verified patients and 24 normal lung tissue control samples from patients with benign pulmonary lesion were obtained from Tianjin Lung Cancer Institute of Tianjin Medical University General Hospital, PR China (ethical approval: no 03-203 Stockholm, Sweden and ethical committee, Tianjin Medical University, China). The matched distant non-cancerous tissue samples were taken from the distant part of the lung—from a location away from the primary tumor—that was confirmed to be non-cancerous by the pathologist. These samples were stored at −80 °C until the DNA was extracted as stated below.

The demographic and clinical characteristics of the recruited subjects are summarized in Table 6.

Table 6. Demographic and clinical characteristics of NSCLC and normal controls.

Clinical characteristics		NSCLC patients (n:70)	Non-matched Controls (n:24)
Gender	Male	44 (63%)	18 (75%)
	Female	26 (37%)	6 (25%)
Age (in years)	Average	63	59
	Median	63	58
	Range	40–70	26–80
Smoking	Yes	38 (63%)	10 (42%)
	No	22 (37%)	13 (54%)
	No data	10 (14%)	1 (4%)
Treatment	Yes	9 (13%)
	No	57 (81%)
	No data	4 (6%)
Type	AC	35 (50%)
	SCC	35 (50%)
Metastasis	M0	63 (90%)
	M1	7 (10%)
Stage	I	25 (36%)
	II	3 (4%)
	III	35 (50%)
	IV	7 (10%)

DNA extraction and conversion by bisulfite modification

DNAs were extracted from cell lines, lung cancer, and matched distant non-cancerous tissues and normal control samples and purified by conventional phenol/chloroform and ethanol extraction method. The procedure for bisulfite modification of DNAs was slightly modified from the protocol of Olek et al.⁶⁸ and has been reported in Zhang et al.³⁴ and 400 ng of genomic DNA was used as a starting material. A549 and H1299 cell line DNA were mixed in 1:1 ratio to be used as a positive control for MMSP.

Expression microarray

Expression profiling was performed on three lung cancer and matched distant non-cancerous tissues from NSCLC patients. These three patients were selected randomly among patients without a history of radio/chemotherapy. All these three samples were from SCC patients with a smoking history of 20–40 cigarettes per day for 10–40 years, i.e., a typical high-risk group. The expression profiling was performed using the Affymetrix microarray platform and standard procedures. Total RNA was extracted with the Trizol reagent (Invitrogen) from primary lung cancer tissues and the distant non-cancerous tissues from the lung of these three NSCLC patients. After extraction the RNA was purified using the Oligotex mRNA Midi kit (Qiagen), and prepared for hybridization to the Human Genome U133 Plus 2.0 microarray, with over 54 000 probe sets (purchased from the Affymetrix) after the manufacturer’s instruction. CEL files were generated by Affymetrix GeneChip Command Console (AGCC). The data fulfilled the requirements of being MIAME compliant.

The pre-processing step was done with the R version 2.15.0 and a number of bioconductor packages. CEL files were corrected for background and normalized using the RMA function from the affy package. Present/absent calls were made by mas5calls function from the same package. The normalized arrays passed quality control provided by the function QCReport from the affyQCReport package.

After pre-processing, genes were considered to be differentially expressed based on signal ratio of tumor samples vs. matched non-cancerous tissue as baseline. Genes with signal ratio ≤1 were considered downregulated and were selected for further analysis.

Genome-wide CpG methylation microarray

Methylation array was performed on lung cancer and matched distant non-cancerous tissues from the same three NSCLC patients as above. The experiments were performed using Illumina platform and standard procedures. The same samples prepared as described before were hybridized to Infinium® HumanMethylation450 BeadChip following manufacturer’s protocols. IDAT files were acquired and normalized with GenomeStudio from Illumina. DiffScore and DeltaBeta were calculated. Genes with DiffScore ≥ 13 and DeltaBeta ≥ 0.1 were selected for further analysis.

By combining expression profiling and methylation microarray results, genes with lower expression values and hypermethylation in cancer tissue were selected from each sample pair. Figure 3 represents the overlapping genes found in both the data sets as a Venn plot. Totally 1392 common genes were acquired from the three matched samples.

Figure 3. Area proportional Venn diagram showing the number of genes that were hypermethylated and expressed at a low level in three NSCLC biopsies (no. 5, 23 and 27). The matched distant non-cancerous tissue was used as a reference for each sample.

Functional analysis was done using DAVID Bioinformatics tools.^69,70 Top clusters of genes were related within the actin and cytoskeleton-ontology (enrichment score 10.43), angiogenesis (enrichment score 7.37), and RAS/Rho related activities (enrichment score 6.42). For detailed functional analysis results, see supplementary file 1.

Methylation specific PCR (MSP)

Out of the 1392 above stated marker genes, 22 were also reported by Lokk et al.⁹ among 643 genes listed in their supplementary file. On the basis of their average DiffScore (expression) and DeltaBeta (methylation) we selected 11 out of 22 genes identified in both sets for further study with MSP. These 11 marker genes were ALDH1A2, AQP1, CDO1, EPOR, GPR124, HLF, HOXA5, HSPA12B, TAL1, TBX5, and ZNF177. In addition, 27 methylation regulated marker genes in NSCLC were identified from other publications that showed significant sensitivity and specificity. We also took into account if they were reported to be involved in tumorigenesis related pathways such as cell cycle regulation, differentiation, adhesion, and apoptosis etc. For this purpose we selected those markers that were tested on lung cancer in different areas e.g., Australia, Japan, USA, Korea, Europe, and China in order to find common markers for NSCLC patients from different geographical, ethnic, environmental, and cultural backgrounds. These 27 marker genes included ADCY4,⁷¹ APC,⁷² ASC,⁷³ BNC1,⁷⁴ BVES,⁷⁵ CALCA,^56,57 CDH13,⁷⁵ CNTNAP2,³⁸ CYB5R2,⁷⁶ DAPK,⁷⁵ DLEC1,⁶⁶ E-CADH,⁷⁵ GDNF,⁷⁷ HOXA9,^41,42,78,79 ITGA9,⁸⁰ KLK10,^81,82 LOX,⁷⁴ MYOD1,⁴² OPCML,⁷⁷ P16,^25,74 PAX6,²⁸ PITX2,⁷⁷ RARβ,^25,66,75 RASSF1A,^25,66,74,75 TSLC1,⁸³ WIF1,¹⁹ and WNT7A.⁸⁰ Thus, all together, 38 (11+27) marker genes were pre-screened on a small group of DNA samples from lung cancer and matched distant non-cancerous tissue by MSP. The identification and selection of the marker genes is also shown as flowchart in Figure 4.

Figure 4. Flowchart showing the selection of marker genes for validation by MSP and subsequent development of the MMSP assay.

We designed one to six pairs of MSP primers for all these 38 markers using MethPrimer⁸⁴ online software. In total we tried 92 pairs of MSP primers for all 38 TSGs. The MSP primers covered different CpG sites in the CpG rich regions of the promoters or CpG island shore regions of theses TSGs. Primers for all potential TSGs were specific for methylated bisulfite converted sequence. Then we screened the DNA samples using MSP protocol from some of the lung cancer tissues and matched distant non-cancerous tissues for all these 38 markers. In order to save the precious human material, we initially used only three matched pairs to analyze methylation status of marker genes with MSP assay. If a marker was methylated in two out of these three cancer samples and was not methylated in any control samples then we applied the marker to the larger sample size. As a result, HOXA9, TBX5, PITX2, CALCA, RASSF1A, and DLEC1 were found to be differentially methylated in NSCLC tissue DNA compared with matched distant non-cancerous tissue DNA samples and potentially valuable markers for development of the MMSP assay. The primer sequences used for these six markers are given in Table 7.

Table 7. Sequence of the primers selected for MMSP.

Marker		Primer Sequence (5’ – 3’)	Size	Reference
β-ACTIN	F	TTTTTAGGGA GGAGTTGGAA GTAGT	212bp	*
	R	AAAATATACC CTCCCCCATA CC
CALCA	MF	CGGAATTTTT TCGATTTATA GC	107bp	*
	MR	AAAACCCTAT AAAAACGACG AC
DLEC1	MF	GATTAAGCGA TGACGGGATT C	193bp	66
	MR	ACCCGACTAA TAACGAAATT AACG
HOXA9	MF	GGTTAATGGG GGCGCGGGCG TC	77bp	*
	MR	TCATATAACA ACTTAATAAC ACCG
TBX5	MF	GGGACGCGTA AAATTTAGAA TC	130bp	*
	MR	AACACAAAAC CGAAAAACGT C
PITX2	MF	CGTTATTAGT TGAAGGTAAG GTCG	172bp	*
	MR	AACACCGAAA AATACAATCC G
RASSF1A	MF	GTGTTAACGC GTTGCGTATC	93bp	85
	MR	AACCCCGCGA ACTAAAAACG A

^* Designed in our lab using MethPrimer online software.⁸⁴

For each PCR reaction, 4 µl (27 ng) of bisulfite-modified DNA was added in a final volume of 25 µl of PCR mixture containing 1.8 × PCR buffer, 5 mM MgCl₂, 0.3 nM deoxynucleotide triphosphates, primers (0.1 µM each per reaction) and one unit of Taq Platinum (Invitrogen). Water was used as blank control.

MSP amplifications were performed at 95 °C for 3 min, followed by 4 cycles at 94 °C for 1 min, 60 °C for 30 s, and 65 °C for 45 s, which was then followed by 28 amplification cycles at 94 °C for 1 min, 56 °C for 1 min, and 65 °C for 45 s. It was followed by final elongation step at 65 °C for 4 min. MSP products were analyzed by 2.5% agarose gel electrophoresis stained with ethidium bromide.

Multiplex methylation specific PCR (MMSP)

Based on the sensitivity and specificity of these 38 markers tested by MSP, six were selected for MMSP: HOXA9, TBX5, PITX2, CALCA, RASSF1A, and DLEC1. A 1:1 mixture of DNA from NSCLC cell line A549 and H1299 was used as control, in order to get a positive signal for all our marker genes. The CALCA signal was weak in A549 and the HOXA9 marker gene was unmethylated in H1299. The MMSP assay also included the house keeping gene β-ACTIN as quality control of input DNAs. The primers for β-ACTIN were designed to amplify bisulfite-converted genomic DNA without distinguishing between methylated and unmethylated CpGs.

For each MMSP PCR reaction, 4 µl (27 ng) of bisulfite-modified DNA was used in a final volume of 25 µg of PCR mixture containing 1.8 × PCR buffer, 5 mM MgCl₂, 0.3 nM deoxynucleotide triphosphates, primers (β-ACTIN: 10 nM, HOXA9: 120 nM, TBX5: 20 nM, PITX2: 20 nM, CALCA: 40 nM, RASSF1A: 280 nM and DLEC1: 40 nM) and 2.5 unit of Taq Platinum (Invitrogen). Water was used as blank control.

The conditions for MMSP amplification and gel electrophoresis were same as stated for MSP.

Glossary

Abbreviations:

AC

adenocarcinoma

AGCC

Affymetrix gene chip command console

FCS

fetal calf serum

HCl

hydrochloric acid

MIAME

minimum information about a microarray experiment

MMSP

Multiplex Methylation Specific PCR

MSP

methylation specific PCR

NSCLC

non-small cell lung cancer

RMA

robust multi-array average

RPMI media

Roswell Park Memorial Institute media

SCC

squamous cell carcinoma

SCLC

small cell lung cancer

Sn

sensitivity

Sp

specificity

TE

Tris EDTA

TSG

tumor suppressor gene

10.4161/epi.29499