Aberrant DNA methylation profile in pleural fluid for differential diagnosis of malignant pleural mesothelioma

Abstract

Malignant pleural mesothelioma (MPM) usually develops pleural fluid. We investigated the value of DNA methylation in the pleural fluid for differentiating MPM from lung cancer (LC). Pleural fluid was collected from 39 patients with MPM, 46 with LC, 25 with benign asbestos pleurisy (BAP) and 30 with other causes. The methylation of O ⁶ ‐methylguanine‐DNA methyltransferase (MGMT), p16 ^INK4a, ras association domain family 1A (RASSF1A), death‐associated protein kinase (DAPK), and retinoic acid receptor β (RARβ) was examined using quantitative real‐time PCR. DNA methylation of RASSF1A, p16 ^INK4a, RARβ, MGMT and DAPK was detected in 12 (30.8%), 3 (7.7%), 11 (28.2%), 0 (0.0%) and five patients (12.8%) with MPM, and in 22 (47.8%), 14 (30.4%), 24 (52.2%), 1 (2.2%) and six patients (13.0%) with LC, respectively. The mean methylation ratios of RASSF1A, p16 ^INK4a and RARβ were 0.37 (range 0.0–2.84), 0.11 (0.0–2.67) and 0.44 (0.0–3.32) in MPM, and 0.87 (0.0–3.14), 1.16 (0.0–5.35) and 1.69 (0.0–6.49) in LC, respectively. The methylation ratios for the three genes were significantly higher in LC than in MPM (RASSF1A, P = 0.039; p16 ^INK4a, P = 0.005; and RARβ, P = 0.002). Patients with methylation in at least one gene were 3.51 (95% confidence interval, 1.09–11.34) times more likely to have LC. Hypermethylation seemed no greater with MPM than with BAP. Extended exposure to asbestos (≧30 years) was correlated with an increased methylation frequency (P = 0.020). Hypermethylation of tumor suppressor genes in pleural fluid DNA has the potential to be a valuable marker for differentiating MPM from LC. (Cancer Sci 2012; 103: 510–514)

Malignant pleural mesothelioma (MPM) is an aggressive neoplasm associated with asbestos exposure (AE). The diagnosis of MPM is challenging. Patients with MPM usually develop pleural fluid as the first presentation,^{( 1 )} but the cytological examination of the fluid yields a diagnosis in only 26% of cases.^{( 2 )} For diagnostic confirmation, invasive work‐up is required, such as thoracoscopic exploration.

In fact, pleural fluid is demonstrated under various conditions. However, an exact diagnosis based on pleural fluid is difficult. In particular, differentiation between malignant and nonmalignant pleural fluid is a critical clinical problem. A cytological evaluation could detect tumor cells in no more than 60% of malignant cases,^{( 3 )} and blindly performed pleural needle biopsy adds little detection sensitivity. Several investigators have sought to improve the differential diagnosis of malignant pleural fluid by measuring tumor markers. Shi^{( 4 )} reports the usefulness of measuring pleural carcinoembryonic antigen to diagnose malignant pleural fluid. Similar findings are reported regarding cytokeratin 19 fragment 21–1, and carbohydrate antigen (CA) 125, CA15‐3 and CA19‐9.^{( 5 )} We previously reported that the concentration of receptor‐binding cancer antigen expressed on SiSo cells was higher in malignant pleural fluid than in nonmalignant pleural fluid.^{( 6 )} However, the usefulness of these markers has not yet been fully established in clinical practice.

In addition, if malignant cells are detected in pleural fluid, differentiation between MPM and other malignancies, such as lung cancer (LC), are sometimes problematic. The differentiation requires immunohistochemical analysis using multiple markers.^{( 7 )} Recently, soluble mesothelin‐related peptides,^{( 8} , ^{9 )} osteopontin^{( 10 )} and N‐ERC/mesothelin^{( 11 )} have been demonstrated to be promising markers for differentiating MPM from LC. However, there is no single biomarker established for this differentiation. Furthermore, if malignant cells are not detected, there is still a high probability of mesothelioma, especially for patients with a history of occupational AE. In such cases, differentiation from other asbestos‐related diseases, such as benign asbestos pleurisy (BAP), is also a critical issue. A reliable clinical marker to support rapid and accurate diagnosis of pleural fluid is greatly needed.

Aberrant hypermethylation in CpG‐rich promoter regions of tumor suppressor genes interferes with gene transcription and can contribute to the development and progression of various cancers by abolishing tumor suppressor gene function.^{( 12 )} These types of hypermethylation have been examined in recent studies of LC^{( 13} , ¹⁴ , ^{15 )} and MPM.^{( 16} , ¹⁷ , ¹⁸ , ^{19 )} We recently demonstrated that testing for aberrant promoter methylation of ras association domain family 1A (RASSF1A), p16 ^INK4a, retinoic acid receptor β (RARβ) and O ⁶ ‐methylguanine‐DNA methyltransferase (MGMT) in the pleural fluid DNA might identify malignant pleural fluid.^{( 20 )}

In the present study, we examined whether promoter hypermethylation in pleural fluid DNA could be used for differential diagnosis of pulmonary diseases. The principal aim of this study was to examine the usefulness of the methylation profile for differentiating MPM from other diseases, especially LC or BAP. We also discuss the correlation between methylation status and AE.

Materials and Methods

Study population.  The subjects included 140 patients from Okayama Rosai Hospital (n = 114) and National Hospital Organization Yamaguchi‐Ube Medical Center (n = 26) between 1996 and 2008. Of the 140 patients, 46 were diagnosed with LC, 39 were diagnosed with MPM, 25 were diagnosed with BAP and 30 were diagnosed with other pulmonary diseases, such as tuberculous pleurisy, empyema or simple pleurisy. The characteristics of the study population are summarized in Table 1. All patients provided written informed consent under the approval of the appropriate institutional review boards. Clinical information was obtained from medical records. Histological subtypes of LC and MPM were based on World Health Organization (WHO) classification.^{( 21 )} The clinical stage of the disease was assessed using the International Staging System for LC^{( 22 )} and the International Mesothelioma Interest Group criteria for MPM.^{( 23 )} The diagnosis of BAP was based on the exclusion of other specific causes in a patient with AE,^{( 24 )} in which malignant diseases were ruled out based on thoracoscopic examination. Patients with MPM and BAP were assessed for history of AE by obtaining occupational history with an in‐person questionnaire or interview. None of the patients with LC had a history of AE.

Table 1.

 Patient characteristics

	MPM	LC	BAP	Others
N	39	46	25	30
Age (years)
Median (range)	67 (47–89)	71 (45–92)	70 (58–87)	77 (30–91)
Gender
Male/female	36/3	38/8	25/0	24/6
Smoking status
S/NS/unknown	22/10/7	38/6/2	18/5/2	24/6/0
Asbestos exposure period (years)
Median (range)	37 (3–50)	–	32 (3–50)	–
Histology
epi/bi/sar/unknown	26/6/5/2	–	–	–
ad/sq/sm/nd	–	36/2/2/6	–	–
Stage
I/II/III/IV/unknown	7/9/11/9/3	–	–	–
III/IV/relapse	–	14/20/12	–	–

ad, adenocarcinoma; BAP, benign asbestos pleurisy; bi, biphasic; epi, epithelioid; LC, lung cancer; MPM, malignant pleural mesothelioma; nd, not determined; NS, non‐smoker; S, smoker; sar, sarcomatoid; sm, small cell carcinoma; sq, squamous cell carcinoma.

Sample collection and DNA extraction.  Pleural fluid was collected in syringes and supernatant was isolated by centrifugation at 400g for 10 min, and then stored at −80°C until use. DNA was extracted from the supernatant using a QIAamp DNA Blood Midi Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Tumor DNA was extracted from formalin‐fixed, paraffin‐embedded tissues that were obtained by surgical resection or autopsy from patients with MPM, using the QIAamp DNA Mini Kit (Qiagen) according to the manufacturer’s instructions. The researchers were unaware of the diagnosis of each patient.

Bisulfite treatment.  Sodium bisulfite conversion of unmethylated cytosine residues to uracil in samples of pleural fluid DNA and tissue sample DNA obtained from patients was performed using the DNA Methylation Bisulfite Assay Kit, CpGenome (Millipore, Billerica, MA, USA) according to the manufacturer’s instructions. All bisulfite‐modified DNA was resuspended in TE buffer (10 mmol/L Tris, 0.1 mmol/L EDTA [pH 7.5]) and used immediately or stored at −20°C until use. DNA from SBC‐3,^{( 25 )} a small‐cell lung cancer cell line with promoter methylation of all tested genes, was used as a positive control.

Quantitative real‐time PCR.  Quantitative real‐time PCR was performed with locus‐specific primers and dual labeled fluorogenic probes. Methylation of RASSF1A, p16 ^INK4a, RARβ, MGMT and DAPK was examined using β‐actin as the internal control for DNA quantification. The DNA sequences of primers and probes for these genes were based on published data.^{( 26} , ^{27 )} PCR was set up in a reaction volume of 25 μL containing 1× Taqman universal PCR master mix (Applied Biosystems, Foster City, CA), 500 nmol/L of each primer, 150 nmol/L of probe and 3 μL bisulfite‐treated DNA samples. After the initial denaturing steps at 50°C for 2 min and at 95°C for 10 min, 50 cycles at 94°C for 15 s and at 60–64°C for 1 min followed. The methylation ratio was defined as the ratio of the fluorescence emission intensity values for methylated RASSF1A, p16 ^INK4a, RARβ, MGMT and DAPK to those of ß‐actin. Positive methylation was defined with >0.001 of the methylation ratio.

Statistical analysis.  We determined the methylation ratios of pleural fluid samples for each disease. These values were analyzed using Kruskal–Wallis one‐way analysis of variance, followed by the Mann–Whitney U‐test. Areas under the receiver operating characteristics (ROC) curve were calculated using standard techniques. An unconditional logistic regression model was applied to estimate the odds ratios and 95% confidence intervals (CI). Crude and multivariate models were examined.

Statistical significance was defined as P < 0.05. All statistical analyses were conducted using SPSS version 10 (SPSS, Chicago, IL, USA).

Results

Methylation frequency of five genes.  The frequency of DNA methylation of tested genes in each patient group is shown in Table 2. The frequency of methylation of each gene was higher in malignant diseases (MPM and LC) than under nonmalignant conditions (BAP and others), except for MGMT (RASSF1A 40.0% vs 10.9%, P < 0.001; p16 ^INK4a 20.0% vs 3.6%, P = 0.005; RARβ 41.2% vs 10.9%, P < 0.001; DAPK 12.9% vs 3.6%, P = 0.053). These results confirmed our previous findings that methylation analysis is useful in distinguishing malignant and nonmalignant pleural effusions.^{( 20 )}

Table 2.

 Frequency of methylation

	Number of patients (%)
	MPM	LC	BAP	Others
N	39	46	25	30
RASSF1A	12 (30.8)	22 (47.8)	6 (24.0)	0 (0.0)
p16 INK4a	3 (7.7)	14 (30.4)	2 (8.0)	0 (0.0)
RARβ	11 (28.2)	24 (52.2)	4 (16.0)	2 (6.7)
MGMT	0 (0.0)	1 (2.2)	0 (0.0)	1 (3.3)
DAPK	5 (12.8)	6 (13.0)	0 (0.0)	2 (6.7)

BAP, benign asbestos pleurisy; LC, lung cancer; MPM, malignant pleural mesothelioma.

Tumor tissues were obtained from 18 of 39 cases with MPM and sufficient DNA was extracted in 15 of these 18 cases. The methylation profile of tumor tissues was consistent with those of pleural fluid in 13 of 15 cases (86.7%) (Fig. 1), indicating that the methylated DNA detected in pleural fluid was released from the tumor cells.

Figure 1.

 Summary of methylation profile for three genes in 15 corresponding tissue and pleural fluid samples of malignant pleural mesothelioma. Black boxes: methylated samples. White boxes: unmethylated samples.

Methylation status of malignant pleural mesothelioma and lung cancer.  Next, we focused on the methylation profile of malignant pleural fluid and directly compared the methylation status of MPM to that for LC. The distributions of the methylation ratios of RASSF1A, p16 ^INK4a and RARβ are shown in Figure 2. The methylation ratios of these three genes were significantly higher in LC than in MPM. In an ROC analysis, the best value of the area under the ROC curve to distinguish LC from MPM was obtained as 0.752 (95% CI 0.661–0.843), with a cut‐off level of 0.05 for the methylation ratio. With the cut‐off level of the ratios in at least one of these three genes, the sensitivity to distinguish LC from MPM was 67.4%. The specificity and positive predictive values were 79.5% and 79.5%, respectively. The methylation ratios of these three genes in the adenocarcinoma subtype of LC were also significantly higher than in MPM (RASSF1A P = 0.049; p16 ^INK4a P = 0.003; and RARβ P = 0.009).

Figure 2.

 Box and whisker plots of methylation ratios for three genes. (a) RASSF1A, (b) p16INK4a and (c) RARβ. The bottom and top of the box represent the 25th and 75th percentile (lower and upper quartile), respectively. The band near the middle of the box is the 50th percentile (median). LC, lung cancer; MPM, malignant pleural mesothelioma.

Given that LC exhibited a higher level of aberrant methylation than MPM, we analyzed the odds ratios for LC. Table 3 shows the results of a crude logistic regression analysis of the correlation between the methylation profile and the odds of LC. Patients with at least one methylated gene were 2.86 (95% CI, 1.09–7.53) times more likely to have LC than were patients with MPM (Table 3, model 1). To consider the imbalance in the baseline characteristics, we conducted a similar analysis adjusting for age, gender and smoking status, which was considered to be associated with the risk of LC. After adjusting for these factors, patients with methylation in at least one gene were 3.51 (95% CI, 1.09–11.34) times more likely to have LC (Table 3, model 2). These results indicate that methylation of these tumor suppressor genes is associated more closely with LC than MPM.

Table 3.

 Methylation status and risk of lung cancer compared to malignant pleural mesothelioma

	MPM	LC	Model 1			Model 2
			Odds ratio	95% CI	P	Odds ratio	95% CI	P
0	16	9	1			1
≥1	23	37	2.86	1.086–7.532	0.03	3.513	1.089–11.338	0.036
Age (≥70 to <70 years)						3.213	1.070–9.650	0.037
Gender (male to female)						0.314	0.054–1.829	0.198
Smoking status (smoker to never smoked)						4.089	0.958–17.445	0.057

CI, confidence interval; LC, Lung cancer; MPM, malignant pleural mesothelioma.

Methylation status and asbestos exposure.  We next examined the correlation between methylation status and AE. For this purpose, we analyzed the methylation in the pleural fluid DNA of asbestos‐related diseases, MPM and BAP. As a result, there was no difference concerning the methylation ratios between MPM and BAP. The methylation ratios of RASSF1A, p16 ^INK4a and RARβ in BAP were lower than those in LC (RASSF1A P = 0.024; p16 ^INK4a P = 0.025; and RARβ P = 0.002), and higher than in other pleural diseases, including tuberculous pleurisy, empyema and simple pleurisy (RASSF1A P = 0.005; p16 ^INK4a P = 0.118; and RARβ P = 0.248).

We then analyzed the associations between the methylation status of RASSF1A, p16 ^INK4a and RARβ and clinical variables in MPM and BAP. An association was exhibited between methylation status and age, gender, smoking status and the period of AE in multivariable analysis (Table 4). Among the independent variables used in the multivariate model, the period of AE was considered to be the factor most associated with methylation status. These results suggest that methylation of these genes would be associated with AE.

Table 4.

 Multivariate analysis of clinical variables and methylation status in patients with asbestos related diseases (MPM and BAP)

	Odds ratio	95% CI	P‐value
Age (≥70 to <70 years)	0.224	0.053–0.942	0.041
Sex (male to female)	0.038	0.002–0.856	0.040
Smoking status (smoker to never smoked)	4.745	0.884–25.473	0.069
Asbestos exposure period (≥30 to <30 years)	6.393	1.332–30.688	0.020

BAP, benign asbestos pleurisy; CI, confidence interval; MPM, malignant pleural mesothelioma.

Discussion

The principal aim of this study was to examine the value of the methylation profile of pleural fluid DNA in MPM in order to differentiate it from LC. For this purpose, we examined promoter hypermethylation of tumor suppressor genes in pleural fluid DNA from patients with various types of pulmonary diseases, including MPM, LC and BAP, using quantitative real‐time PCR. There are few reports in which methylation profiles have been examined in pleural fluid DNA. In previous reports concerning LC tissue, the prevalence of methylation of RASSF1A, p16 ^INK4a, RARβ, DAPK and MGMT was 30–50, 25–52, 10–38, 26–35 and 30–50%, respectively.^{( 28} , ²⁹ , ³⁰ , ³¹ , ^{32 )} In the current study, the proportion of methylated DNA in pleural fluid for each gene in LC was 47.8% for RASSF1A, 30.4% for p16 ^INK4a, 52.2% for RARβ, 2.2% for MGMT and 13.0% for DAPK. We have no clear explanation for the low prevalence of methylation of DAPK and MGMT. There might be some interference with detection of methylation of the genes caused by contamination by mesothelial cells in the pleural fluid. This should be clarified in the future. In addition, the methylation profile of tumor tissues of MPM in the current study was consistent with those of pleural fluid in most cases. These results indicate that methylation of these genes can be detected in pleural fluid DNA at similar proportions as tumor tissue.

The current study demonstrates the usefulness of a methylation profile in pleural fluid DNA for the differential diagnosis between MPM and LC. MPM and LC have been found to have different molecular characteristics. Toyooka^{( 33 )} reports significantly increased methylation of p16 ^INK4a (P < 0.01), RARβ (P = 0.02) and MGMT (P < 0.01) in lung adenocarcinoma as compared with MPM. There are other reports that gene deletion is responsible for the inactivation of p16 ^INK4a in MPM, leading to a different pattern for the methylation profile in MPM and LC.^{( 16} , ^{34 )} In the present study, the prevalence of hypermethylation of the three genes (RASSF1A, p16 ^INK4a and RARβ) was significantly higher in LC than in MPM. These results indicate that methylation analyses would be an option to support the differentiation of MPM from LC. We found that the sensitivity, specificity and positive predictive values for the methylation ratio of >0.05 in one or more genes for the diagnosis of LC were 67.4%, 79.5% and 79.5%, respectively. To improve the specificity and sensitivity, there should be a search for genes specifically methylated in either LC or MPM. The use of methylation‐array technology could lead to identification of such genes. There is growing interest in the use of gene expression profiling for the differential diagnosis of MPM and LC. Further studies are warranted to search for the best combination of genes examined and to apply the method to pleural fluid DNA.

Another important subject in the current study was the association between the methylation status and AE. To examine this association, we analyzed the methylation profile in the pleural fluid DNA from patients with MPM and BAP. BAP was first described by Eisenstadt^{( 35 )} in 1964 as fibrotic pleuritis and might occur early or late after AE. It is noteworthy that the prevalence of aberrant hypermethylation in MPM and BAP was almost equivalent in the current study, and higher than that in other diseases such as tuberculous pleurisy, empyema and simple pleurisy, which were not associated with AE. In addition, we demonstrated an association between methylation status and the duration of AE in the current study. These results validated the recent findings of the association between methylation and AE reported by Christensen et al. ^{( 19 )} It is well‐known that chronic inflammation is a primary tissue response to AE.^{( 36 )} Based on these findings, we suppose that hypermethylation in MPM and BAP in this study was associated with asbestos‐induced chronic inflammation of the pleura. These results raise the hypothesis that methylation detected in MPM in the current study might not be due to malignant alteration. More evidence is needed to clarify the specific molecular event involved in the development of MPM. This clarification would support not only the differential diagnosis between MPM and BAP, but also the early detection of MPM among patients with past AE.

In conclusion, we have demonstrated that identification of aberrant promoter hypermethylation of the tumor suppressor genes RASSF1A, p16 ^INK4a and RARβ in pleural fluid DNA could be a valuable marker to aid in differentiating MPM from LC. Pleural fluid can be obtained more easily than tumor tissue, which must be obtained with thoracoscopic biopsy. We believe that molecular analysis using pleural fluid is a promising tool suitable to clinical practice. Further studies are warranted to search for the genes methylated specifically in each disease and the best combination of genes and other diagnostic tests.

Disclosure Statement

We declare that we have no conflict of interest.