Sage Journals: Discover world-class research

Abstract

BACKGROUND AND OBJECTIVE:

Chemotherapy remains the basis of the treatment of lung cancer, and screening biomarkers with predictive value for chemotherapy is of great interest. The present study focused on status of genes methylation in NSCLC patients receiving pemetrexed- or gemcitabine-based chemotherapy.

PATIENTS AND METHODS:

Promoter methylation of Ras association domain family (RASSF1A) and short stature homeobox 2 (SHOX2) was examined in bronchoalveolar lavage (BAL) from 117 NSCLC patients treated with chemotherapy. Multivariate analysis was used to identify the predictive value of gene methylation. Progression-free survival (PFS) rather than overall survival (OS) was used as the clinical outcome to minimize the impact of chemotherapy on gene methylation.

RESULTS:

The methylation of RASSF1A and SHOX2 was significantly associated with shorter PFS (RASSF1A: HR $=$ 2.355, 95% CI: 1.533–3.617, $P<$ 0.0001; SHOX2: HR $=$ 2.123, 95% CI: 1.392–3.236, $P=$ 0.0004). After adjusting for confounding factors, RASSF1A methylation was still a predictive factor for PFS (HR $=$ 1.765, 95% CI: 1.064–2.928, $P=$ 0.0278). In the pemetrexed group, unmethylated RASSF1A could be used to predict longer PFS ( $P=$ 0.0001), and no predictive value was found in the gemcitabine group.

CONCLUSION:

Unmethylated RASSF1A is a favorable prognostic indicator for patients receiving pemetrexed doublets. Because of the promoting effect of most chemotherapeutic drugs on gene methylation, unmethylated RASSF1A is not suitable as a predictor for gemcitabine doublets.

Keywords

Lung cancer chemotherapy DNA methylation RASSF1A SHOX2

1. Introduction

Despite recent advances in molecular-targeted therapies and immunotherapy, chemotherapy remains the standard treatment for patients with advanced lung cancer [1, 2]. Docetaxel or pemetrexed are viable options for non-small cell lung cancer (NSCLC) patients who have been unsuccessfully treated with, or are not eligible for, targeted therapy or immunotherapy [3]. To further improve the efficacy of chemotherapy, it is necessary to know the molecular mechanisms influencing lung cancer chemosensitivity and to identify factors associated with improved patient survival after chemotherapy [4, 5, 6, 7, 8, 9, 10].

Methylation of cytosine-guanine (CpG) islands in promoter regions generally blocks gene transcription in a wide variety of tumors [11]. Alterations in gene expression induced by aberrant genomic methylation comprise a major epigenetic feature of multiple tumors, including lung cancer [5, 7]. Considering that these methylation abnormalities are associated with various biological characteristics of lung cancer, the detection of such molecular alterations from specific specimens is helpful for lung cancer diagnosis, treatment efficacy prediction and survival prognosis. Evidences showed that promoter methylation of Ras association domain family 1A (RASSF1A) and short stature homeobox 2 (SHOX2) have been identified as diagnostic and prognostic biomarkers for lung cancer [12, 13, 14, 15]. Among those studies, SHOX2 methylation is mostly used as diagnosis marker for lung cancer. Studies about RASSF1A methylation as prognostic factor focused on early stage lung cancer patients receiving surgical resection [16]. Because chemotherapy was administered in the neoadjuvant setting, the predictive role for chemotherapy could not be assessed accurately. Thus, sufficient confirmatory studies are still lacking to assess the reproducibility and stability of these molecular predictors, especially for advanced stage NSCLC.

In our study, modified sodium bisulfite and real-time methylation-specific PCR with fluorescence hybridization probes were used to determine promoter methylation of RASSF1A and SHOX2 in bronchoalveolar lavage (BAL) from 117 NSCLC patients. Real-time methylation-specific PCR has made it possible to detect gene methylation from ultratrace tumor DNA and has the advantage of increasing specificity via an internally binding probe for each gene [17]. Collected through fiberoptic bronchoscopy operation, BAL represents an alternative source of biomarker detection by minimally invasive methods. Due to its vicinity to tumor cells, BAL is superior for DNA detection compared with traditional analytes, such as plasma or sputum [18]. DNA methylation in non-neoplastic cells from plasma or sputum may make it difficult to interpret the final results.

Although evidences have revealed that methylation of the RASSF1A and SHOX2 were correlated with lung cancer diagnosis or prognosis, the predictive role of these genes’ methylation for prognosis of advanced lung cancer patients receiving chemotherapy has hardly been reported. In the present study we focused on the predictive potential of RASSF1A and SHOX2 methylation in NSCLC patients treated with pemetrexed or gemcitabine doublet chemotherapy.

2. Materials and methods

2.1 Ethics statement

This study was approved by the Medical Ethics Committee of Shanghai Pulmonary Hospital. All patients in our study provided written informed consent. This study was conducted in accordance with the Declaration of Helsinki. The patients’ records were anonymized and de-identified prior to analysis.

2.2 Study subjects and samples

Between August 2016 and December 2018, a total of 129 NSCLC patients from Shanghai Pulmonary Hospital were enrolled in our cohort analysis. The following criteria were used to distinguish patients with lung cancer from those without lung cancer: (a) diagnosis of lung cancer based on histological analysis of a biopsy specimen or (b) cytologic diagnosis of lung cancer was considered to be a positive reference standard if this diagnosis was supported by other clinical findings (i.e., imaging or bronchoscope). Of the 129 patients, 3 patients received radiotherapy, 2 received targeted therapy, 5 received surgical resection and 2 did not receive any treatment. Finally, only 117 patients who received chemotherapy alone were included in the statistical analysis.

Clinicopathology characteristics were obtained from the Oncology Department or Pathology Department. Radiographic tumor responses were evaluated according to the Response Evaluation Criteria in Solid Tumors: complete response (CR), disappearance of all target lesions; partial response (PR), $\leqslant$ 30% decrease in the sum of the target lesion diameters with the summed baseline diameters as a reference; progressive disease (PD), $\leqslant$ 20% increase in the sum of the target lesion diameters with the smallest sum observed during the study serving as reference; and stable disease (SD), insufficient shrinkage to qualify as PR and insufficient expansion to qualify as PD. Progression-free survival (PFS) was calculated from the start of the first chemotherapy date to the date of PD or death from any cause. In surviving patients without any progression, PFS was calculated from the date of chemotherapy to the date of our last contact with them, or to the endpoint date of the study.

The BAL samples of all patients were used not only for cytological diagnosis, but also for detection of gene methylation. All tumor tissues including primary tumors, lymph nodes and distant metastases were used for pathological diagnosis. In order to verify the reliability of gene methylation detection in BAL samples, gene methylation in primary tumor tissues was also detected, and the results were used as a reference. Methylation status of genes was determined at the time of diagnosis.

2.3 Cytological and pathological analysis

Bronchial biopsy samples were processed and stained with hematoxylin and eosin. BAL samples were collected during bronchoscopy by aspiration with a flexible bronchoscope from the region of the lesion after injecting 40 mL of saline solution. Precipitation of 10 mL BAL was achieved following centrifugation at 20000 rpm for 10 min. The supernatant was removed after centrifugation, and the cellular precipitate was resuspended with 10–15 mL of Scott’s tap water substitute. Next, the specimen was centrifuged again at 2000 rpm for 10 min, and the cellular precipitate was used to make slides using alcohol fixation and hematoxylin-eosin staining.

2.4 DNA Extractions and bisulfite modification of genomic DNA

DNA methylation levels in BAL samples were determined using the Lung-Me ${}^{\text{TM}}$ Assay System (Tellgen Corporation, Shanghai China) certified by the CE and CFDA. The Lung-Me ${}^{\text{TM}}$ Assay System comprises three individual kits including DNA extraction Kit, DNA Preparation Kit for the bisulfite modification, and the real-time PCR Kit for quantitative and sensitive determination of DNA methylation.

Precipitation of 10 mL BAL centrifuged at 10000 rpm or tumor tissues stored in liquid nitrogen were used for DNA extraction. Genomic DNA was isolated using standard phenol and chloroform extraction. Genomic DNA was treated with sodium bisulfite to modify unmethylated cytosine into uracil. The concentration of each sample was quantified by Qubit 3.0 fluorometer_Invitrogen. A total of 200ng DNA was modified with sodium bisulfite and 50ng of modified DNA was used for PCR analysis. DNA samples were analyzed immediately or rapidly frozen in liquid nitrogen and stored at $-$ 80 ${}^{\circ}$ C.

2.5 Methylation specific real-time fluorescent PCR analysis

PCR amplification targeting sequences modified with sodium bisulfite was detected using TaqMan probes. Sequences of primers and probes are as follows: SHOX2 forward primer (5’-TTGTTTTTGGGTT CGGGTT-3’) and reverse primer (5’-CATAACGTAAA CGCCTATACTCG-3’); RASSF1A forward primer (5’- CGGGGTTCGTTTTGTGGTTTC-3’) and reverse pri- mer (5’-CCGATTAAATCCGTACTTCGC-3’). Probe for RASSF1A (FAM-TCGCGTTTGTTAGCGTTTAA AGT-BHQ) and probe for SHOX2 (VIC-ATCGAACA AACGAAACGAAAATTACC-BHQ). The reactions were performed in a 40 $\mu$ L total volume containing 5 $\mu$ L of bisulfite-modified DNA, 250 $\mu$ M dNTP, 0.8 $\mu$ l of each primer (10 $\mu$ M), 1.5–3 mM MgCl ${}_{2}$ , 2 $\times$ Taq buffer (including dNTPs and Taq polymerase) 20 $\mu$ l, ddH ${}_{2}$ O 13.4 $\mu$ l, in a thermocycler with the following cycling parameters: 95 ${}^{\circ}$ C for 10 min; 45 cycles of 95 ${}^{\circ}$ C for 30 s; specific annealing temperature 58 ${}^{\circ}$ C for 35 s, 72 ${}^{\circ}$ C for 30 s and a final extension step at 72 ${}^{\circ}$ C for 8 min.

The unmethylated cytosine bases were modified to uracil base by sulfite, and then the uracil bases were converted to thymine bases in the process of PCR amplification, but the methylated cytosine bases remain, so the methylated and unmethylated cytosine bases can be distinguished. Specific primers targeting sequences after bisulfite modification were designed for PCR. Products of PCR amplification were detected by TaqMan probes (85nM). Plasmid containing corresponding sequence of methylated genes was as positive control and purified water as negative control. The $\beta$ -actin gene (forward primer: 5’-AAGATAGTGTTGTGGGT GTAGGT-3’; reverse primer: 5’-CCTACTTAATACAC ACTCCAAAAC-3’; probe: CY5-ACACCAACCTCA TAACCTTATCACAC-BHQ) was served as internal control for quantification of total input DNA. The baseline position was adjusted (Threshold $=$ 10000; Noise band $=$ 0.8) to obtain the Ct value of the fluorescence signal. Ct value of $\beta$ -actin gene signals should $\geqslant$ 18 and $\leqslant$ 32. When the amplification curve of RASSF1A DNA signals presents a smooth S shape and the Ct value $<$ 35, then the sample is classed as positive for methylation of RASSF1A. When the amplification curve of SHOX2 signals presents a smooth S shape and the Ct value $<$ 32, then the sample is classed as positive for methylation of SHOX2.

Table 1
Consistency of methylation detection between BAL and primary tumor

Methylation in tumor		Methylation in BAL				Total cases		Kappa coefficient	95% CI	McNemar P
		Methylated		Unmethylated
RASSF1A	Methylated	12	(80.00%)	3	(20.00%)	15	(68.18%)	0.7179	0.4334–1.0000	0.0833
	Unmethylated	0	(0.00%)	7	(100.00%)	7	(31.82%)
	Total cases	12	(54.55%)	10	(45.45%)	22	(100.00%)
SHOX2	Methylated	14	(87.50%)	2	(12.50%)	16	(72.73%)	0.6733	0.3349–1.0000	0.5637
	Unmethylated	1	(16.67%)	5	(83.33%)	6	(27.27%)
	Total cases	15	(68.18%)	7	(31.82%)	22	(100.00%)

2.6 Statistical analysis

The methodological consistency was analyzed using McNemar’s test and Kappa test. The characteristics of patients with genetic methylation were compared using Chi-square or Fisher exact tests as appropriate. Correlation was analyzed using Spearman’s correlation coefficient. The Cox model was applied to adjust for potential confounders. Survival curves were generated using the Kaplan-Meier method, and the comparison of curves was performed by the log-rank method. Internal validation of the models was achieved by bootstrap resampling (1,000 samples), with all the steps of the model development process being replicated on each bootstrapped sample.

Statistical significance was set at $P<$ 0.05. Data were analyzed with SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

3. Results

3.1 Evaluation of genes methylation detected in BAL

Of the 117 patients, 51 were diagnosed by cytopathology and the remaining by histopathology. Among those with histopathological biopsies, 22 were from primary tumors, 25 were from lymph nodes and 19 were from distant metastases. All patients were routinely examined by tracheoscopy and alveolar lavage fluid samples were collected.

We tested the promoter methylation pattern of RASSF1A and SHOX2 in 22 primary lung tumors and corresponding matched BAL DNA. The methylation status in tumor tissue and BAL was compared for consistency evaluation. Compared with the methylation detection in tumor specimens, the overall concordance of RASSF1A and SHOX2 promoter methylation were both 86.36% (19/22). The consistency results in BAL for RASSF1A (Kappa $=$ 0.7179, 95% CI: 0.4331–1.0000) and SHOX2 (Kappa $=$ 0.6733, 95% CI: 0.3349–1.0000) was reliable (Table 1).

3.2 Clinicopathological characteristics of patients and genes methylation

After applying a series of exclusion criteria, including cytology or histological examination and treatment regimens, a total of 117 NSCLC patients were included in statistical analysis. The median age at diagnosis was 62 years (range: 33–82 years). The median follow-up time was 10 months, and the median PFS was 5.2 months (95% CI: 4.819–8.381). Overall, RASSF1A methylation was detected in 60.68% of the patients, and SHOX2 methylation was detected in 59.83% of the patients. Demographic factors were similarly distributed in patients with and without methylation, except for histology and smoking status, as squamous cell carcinoma was more frequently observed in patients with RASSF1A methylation (78.57% vs 21.43%, $P=$ 0.0036) or SHOX2 methylation (80.95% vs 19.05%, $P=$ 0.0001). The number of patients who had ever smoked was higher in the RASSF1A methylation group (70.42% vs 29.58%, $P=$ 0.0323). Regarding chemotherapy, frequencies of both RASSF1A and SHOX2 methylation were significantly different between doublet regimens (pemetrexed vs. gemcitabine). There was no significant difference in the frequencies of RASSF1A and SHOX2 methylation between chemotherapy response groups (Table 2).

Table 2
Characteristic of 117 NSCLC patients with genes methylation analysis

				RASSF1A promoter					SHOX2 promoter
Characteristics		All cases		Methylated		Unmethylated		$P^{\text{e}}$	Methylated		Unmethylated		$P^{\text{e}}$
		( $n=$ 117)		( $n=$ 71)		( $n=$ 46)			( $n=$ 70)		( $n=$ 47)
Sex	Male	76	(64.96%)	50	(70.42%)	26	(56.52%)	0.1649	47	(67.14%)	29	(61.70%)	0.5596
	Female	41	(35.04%)	21	(29.58%)	20	(43.48%)		23	(32.86%)	18	(38.30%)
Age	$<$ 60 yr	41	(35.04%)	18	(27.27%)	23	(45.10%)	0.0526	24	(34.29%)	17	(36.17%)	0.8452
	$\geqslant$ 60 yr	76	(64.96%)	48	(72.73%)	28	(54.90%)		46	(65.71%)	30	(63.83%)
Histology ${}^{\text{a}}$	ADC	57	(48.72%)	26	(36.62%)	31	(67.39%)	0.0036	23	(32.86%)	34	(72.34%)	0.0001
	SCC	42	(35.90%)	33	(46.68%)	9	(19.57%)		34	(48.57%)	8	(17.02%)
	Untyped	18	(15.38%)	12	(16.90%)	6	(13.04%)		13	(18.57%)	5	(10.64%)
TNM stage	I stage	3	(2.56%)	1	(1.41%)	2	(4.35%)	0.6288	0	(0.00%)	3	(6.38%)	0.2705
	II stage	8	(6.84%)	5	(7.04%)	3	(6.52%)		4	(5.71%)	4	(8.51%)
	III stage	44	(37.61%)	27	(38.03%)	17	(36.96%)		29	(41.43%)	15	(31.91%)
	IV stage	62	(52.99%)	38	(53.52%)	24	(52.17%)		37	(52.86%)	25	(53.19%)
Smoking	Ever ${}^{\text{b}}$	73	(62.39%)	50	(70.42%)	23	(50.00%)	0.00323	48	(68.57%)	25	(53.19%)	0.1197
	Never	44	(37.61%)	21	(29.58%)	23	(50.00%)		22	(31.43%)	22	(46.81%)
Chemo regimen ${}^{\text{c}}$	Pemetrexed	26	(22.22%)	14	(19.72%)	12	(26.09%)	0.0009	14	(20.00%)	12	(25.53%)	0.0001
1–2 cycles	Gemcitabine	26	(22.22%)	23	(32.39%)	3	(6.52%)		22	(31.43%)	4	(8.51%)
3–4 cycles	Pemetrexed	32	(27.35%)	12	(16.90%)	20	(43.48%)		9	(12.86%)	23	(48.94%)
	Gemcitabine	33	(28.21%)	22	(30.99%)	11	(23.91%)		25	(35.71%)	8	(17.02%)
No. of cycles	1–2 cycles	52	(44.44%)	37	(52.11%)	15	(32.61%)	0.0563	36	(51.43%)	16	(34.04%)	0.0874
	3–4 cycles	65	(55.56%)	34	(47.89%)	31	(67.39%)		34	(48.57%)	31	(65.96%)
Chemo doublet	Pemetrexed	58	(49.57%)	26	(36.62%)	32	(69.57%)	0.0006	23	(32.86%)	35	(74.47%)	0.0001
	Gemcitabine	59	(50.43%)	45	(63.38%)	14	(30.43%)		47	(67.14%)	12	(25.53%)
Response to chemo ${}^{\text{d}}$	PR	15	(12.82%)	9	(12.68%)	6	(13.04%)	0.0579	11	(73.33%)	4	(26.67%)	0.5375
	SD	49	(41.88%)	23	(32.39%)	26	(56.52%)		23	(46.94%)	26	(53.06%)
	PD	53	(45.30%)	39	(54.93%)	14	(30.43%)		36	(67.92%)	17	(32.08%)

${}^{\text{a}}$ ADC: Adenocarcinoma; SCC: Squamous cell carcinoma; “Untyped NSCLC” indicates no way to further distinguish between adenocarcinoma or squamous cell carcinoma; “Adenocarcinoma” includes bronchioloalveolar carcinoma and adenosquamous histologic features. ${}^{\text{b}}$ “Ever” indicates current or former smoker. ${}^{\text{c}}$ Pemetrexed-based: Pemetrexed plus carboplatin or nedaplatin; Gemcitabine-based: Gemcitabine plus carboplatin or nedaplatin. Pemetrexed, 500 mg/m ${}^{2}$ ; Gemcitabine, 1000 mg/m ${}^{2}$ ; Nedaplatin: 80 mg/m ${}^{2}$ ; Carboplatin: AUC $=$ 5. ${}^{\text{d}}$ PR: partial response; SD: stable disease; PD: progressive disease. ${}^{\text{e}}$ $P$ values were calculated by the Chi-square or fisher exact tests when appropriate.

Since age has been reported to influence patterns of DNA methylation [19], we analyzed the correlation of methylation with age using Spearman’s correlation coefficient. However, no significant correlations were found (RASSF1A methylation and age: Rho $=$ 0.10172, $P=$ 0.2752; SHOX2 methylation and age: Rho $=$ 0.01887, $P=$ 0.8399).

3.3 Association of gene methylation with risk of progression

Before potential confounders were analyzed via Cox regression analysis, we first tested the assumption of hazard proportion. Test results showed that chemotherapy doublet regimen (HR $=$ 2.440, 95% CI: 1.170–5.087, $P=$ 0.0173), number of chemotherapy cycles (HR $=$ 5.019, 95% CI: 1.021–24.059, $P=$ 0.047) and chemotherapy response (HR $=$ 5.021, 95% CI: 1.673–15.072, $P=$ 0.0040) were time-dependent covariates. Therefore, these time-dependent covariates were replaced by the interaction of the covariate with time when analyzed using the Cox regression model. Analysis showed that number of chemotherapy cycles (HR $=$ 0.423, 95% CI: 0.319–0.565, $P=$ 0.0040) and chemotherapy response (HR $=$ 0.105, 95% CI: 0.060–0.184, $P=$ 0.0001 and HR $=$ 0.381, 95% CI: 0.184–0.791, $P=$ 0.0016) were still independent prognostic factors, and chemotherapy regimen was no longer a prognostic factor (Table 3).

Table 3
Predictors on PFS in cox regression analysis for 117 NSCLC patients

Variables	Univariate analysis ${}^{\text{a}}$		Multivariate analysis ${}^{\text{a}}$
	HR (95% CI)	$P$	HR (95% CI)	$P$
Number of chemo cycles (1–2/3–4)	0.436 (0.436–0.549)	$<$ 0.0001	0.423 (0.319–0.565)	0.0001
Response to chemo (PR/SD/PD) ${}^{\text{b}}$	0.104 (0.063–0.173)	$<$ 0.0001	0.105 (0.060–0.184)	0.0001
	0.223 (0.121–0.412)	$<$ 0.0001	0.381 (0.184–0.791)	0.0016
RASSF1A promoter (methylated/unmethylated)	2.355 (1.533–3.617)	$<$ 0.0001	1.765 (1.064–2.928)	0.0278
SHOX2 promoter (methylated/unmethylated)	2.123 (1.392–3.236)	0.0004	–	–

${}^{\text{a}}$ Significance level for entry and significance level for stay were both set as 0.05. ${}^{\text{b}}$ PR: partial response; SD: stable disease; PD: progressive disease.

Table 4

Predication on PFS by interaction between genes methylation and chemotherapy for 117 NSCLC patients

Variables	Univariate analysis ${}^{\text{a}}$		Multivariate analysis ${}^{\text{a}}$
	HR (95% CI)	$P$	HR (95% CI)	$P$
RASSF1A methylation $\|$ Chemotherapy ${}^{\text{b}}$	2.703 (1.481–4.932)	0.0012	2.459 (1.320–4.580)	0.0046
SHOX2 methylation $\|$ Chemotherapy ${}^{\text{b}}$	2.187 (1.267–3.777)	0.0050	–	–

${}^{\text{a}}$ Significance level for entry and significance level for stay were both set as 0.05. ${}^{\text{b}}$ Methylation: unmethylated/ methylated; chemotherapy: pemetrexed doublet/gemcitabine doublet.

The Kaplan-Meier curve of PFS against methylation status showed that patients in unmethylated groups had a significantly longer median PFS than patients with methylated RASSF1A (7.1 vs. 4.0 months, HR $=$ 1.777, 95% CI: 1.115–2.439, $P=$ 0.0001) and methylated SHOX2 (7.5 vs 4.4 months, HR $=$ 1.692, 95% CI: 1.026–2.357, $P=$ 0.0003; Fig. 1). Univariate analysis showed that RASSF1A and SHOX2 methylation statuses were both associated with PFS of NSCLC patients. After adjusting for confounding factors, RASSF1A methylation was still a predictive factor for PFS (HR $=$ 1.765, 95% CI: 1.064–2.928, $P=$ 0.0278), whereas SHOX2 methylation was not (Table 3).

Figure 1.

PFSs according to genes promoter methylation status. A. RASSF1A promoter methylation status; B. SHOX2 promoter methylation status. RASSF1A $-$ and SHOX2 $-$ : unmethylated status; RASSF1A $+$ and SHOX2 $+$ : methylated status.

The prognostic value of RASSF1A was further validated with a resampling bootstrap procedure (1,000 replications), in which the statistical analysis (i.e., univariable selection of genes and selection of and adjustment for clinical confounders) were replicated on each bootstrapped sample. RASSF1A methylation was associated with significantly shorter PFS in 70% of the samples, with 0.76 (95% CI: 0.58–0.68) corrected c-indexes, confirming the predictive accuracy of RASSF1A methylation for worse prognosis.

3.4 Gene methylation, chemotherapy and risk of progression

Among the 117 NSCLC patients, 58 patients (49.57%) were treated with chemotherapy of pemetrexed plus carboplatin (or nedaplatin), and 59 patients (50.43%) were treated with gemcitabine plus carboplatin (or nedaplatin). Overall, chemotherapy partial response (PR) was observed in 15 patients (12.82%), whereas stable disease (SD) was found in 49 (41.88%) patients and progressive disease (PD) was found in 53 patients (45.30%). Response to chemotherapy was similarly distributed in patients with methylated and unmethylated status ( $P=$ 0.0579 for RASSF1A methylation and $P=$ 0.5375 for SHOX2 methylation), suggesting that the methylated status of RASSF1A and SHOX2 was not correlated with overall response.

Figure 2.

PFSs according to chemotherapy regimen and genes methylation status. A. RASSF1A promoter methylation status; B. SHOX2 promoter methylation status. RASSF1A $-$ and SHOX2 $-$ : unmethylated status; RASSF1A $+$ and SHOX2 $+$ : methylated status; PEM: pemetrexed doublets treatment; GEM: gemcitabine doublets treatment.

In univariate analysis, the interaction between methylation status and chemotherapy regimens (pemetrexed doublets vs. gemcitabine doublets) on PFS was statistically significant (HR $=$ 2.703, 95% CI: 1.481–4.932, $P=$ 0.0012 for interaction of RASSF1A; HR $=$ 2.187, 95% CI: 1.267–3.777, $P=$ 0.0050 for interaction of SHOX2) (Table 4). A Stratified Kaplan-Meier curve further revealed the relationships of this interaction. In the pemetrexed doublet group, the PFS of lung cancer patients with unmethylated genes was significantly longer than for patients with methylated genes ( $P=$ 0.0001 for RASSF1A methylation; $P=$ 0.0017 for SHOX2 methylation). However, no significant difference was found in the gemcitabine doublet group. At the same time, we also observed that the curve of patients with methylated genes in the pemetrexed doublet group nearly overlapped with the two curves of patients in the gemcitabine doublet group (Fig. 2).

After adjustment for confounding factors, the statistical interaction between RASSF1A methylation and chemotherapy regimens on PFS remained significant ( $P=$ 0.0046, HR $=$ 2.459, 95% CI: 1.320–4.580), whereas the interaction between SHOX2 and chemotherapy did not (Table 4). Internal validation showed that the RASSF1A variable was present in 76% of the resulting models. Corrected c-index was 0.79 (95% CI: 0.54–0.66) when RASSF1A methylation status was added.

4. Discussion

Platinum-based chemotherapy has long been the standard treatment for advanced-stage NSCLC. The choice of a suitable regimen for NSCLC is mainly dependent on the treatment benefit based on pathological types, which is far from meeting the requirements of individualized precision treatment. At the same time, the nonnegligible cytotoxic effect and the reduction of quality of life limits the application of chemotherapy. Therefore, analysis of molecular factors for predicting the chemosensitivity and the prognosis of NSCLC patients is considered important for selecting better therapeutic strategies [20, 21, 22]. We detected promoter methylation of RASSF1A and SHOX2 in BAL using real-time methylation-specific PCR and analyzed the correlation with PFS in NSCLC patients treated with chemotherapy. Results in our study showed that RASSF1A methylation was an independent predictor of disease progression. In the pemetrexed doublet group, the PFS of patients with unmethylated RASSF1A was significantly longer than those with methylated RASSF1A, whereas no significant difference was found in the gemcitabine doublet group.

In this study, a commercial technique based on real-time fluorescence quantitative PCR was used to detect abnormal methylation of RASSF1A and SHOX2 in BAL. However, the advantages of quantitative analysis are obvious in specimens with uniform distribution of various cells, for the methylated degree of tumor can be truly reflected by quantitative data detected in such specimens as tumor tissue or serum [23, 24]. The distribution of various cells in BAL is greatly affected by external factors, so the quantitative results of methylation can only reflect whether there is abnormal methylation in tumor, but not the degree of methylation. Therefore, the continuous variables obtained by quantitative PCR in our study were converted into binary variables for qualitative analysis. Although the advantages of quantitative detection can not be realized due to the specimens, the technology based on real-time fluorescent PCR still has such advantages as high sensitivity and specificity, totally closed reaction system, integration of detection and analysis, automation and high throughout [17, 25]. In addition, the technique platform in our study was based on multiplexing real-time fluorescent PCR. RASSF1A, SHOX2 and $\beta$ -actin can be detected simultaneously with multiplexing setting. Aside from saving labor and time, the more important reason was that $\beta$ -actin as internal control should be detected in the same tube to facilitate the monitoring of the whole PCR process.

Promoters of genes, including RASSF1A and SHOX2, are frequently methylated in lung cancer [11, 12, 13, 14, 15, 16]. In our study, the methylation of these genes in primary tumors was first compared with corresponding BAL samples. The results showed that methylation detection of RASSF1A and SHOX2 between BAL and tumors have good consistency, which indicated that gene methylation detected in BAL better reflect the methylation status of tumors under our laboratory conditions. We then explored the predictive value of RASSF1A and SHOX2 methylation for chemotherapy response and disease progression. Although no significant association was found between gene methylation and overall response to chemotherapy, RASSF1A methylation status was an independent predictor of PFS for patients treated with chemotherapy. Univariate analysis showed that SHOX2 methylation was associated with PFS in lung cancer patients, but SHOX2 methylation was not an independent predictor of PFS after adjustment for confounding factors. In general, PFS will be further assessed only when the responses to treatment were observed in early clinical trials. Thus, an ideal response to chemotherapy was often demolished by poor PFS [26, 27]. However, unlike clinical trials for assessment of new therapies, our study explored the predictive value of gene methylation for standard chemotherapy. As a tumor suppressor gene [28], methylation-induced silence of RASSF1A may promote the tumor progression, but may not necessarily be related to the pharmacological effect of chemotherapeutic drugs.

RASSF1A, a negative effector of RAS, generally functions as a tumor suppressor in cancer through several distinct pathways, including apoptosis, genomic and microtubule stability, and cell cycle regulation [11]. Promoter methylation of RASSF1A is frequently found in a variety of human cancers, including lung cancer [13]. There have been many studies on RASSF1A methylation as a prognosis factor for lung cancer patients [29]. Previous studies have reported that RASSF1A methylation is a poor prognostic factor for lung cancer, and our findings were generally consistent with these reports [30, 31]. A study by Fischer et al. believed that RASSF1A methylation is a favorable prognostic factor [32]. There may be several reasons that contribute to this contradictory conclusion, e.g., the different subgroups of patients enrolled, varieties of the specimens for methylation detecting and that the different endpoint setting may be more likely to be related to the impact of treatment regimens on the final outcome. As an important epigenetic mechanism, DNA methylation is very sensitive to environmental factors, such as radiation, nutrition and temperature [33, 34]. In particularly, the impact of chemotherapeutic drugs on gene methylation cannot be ignored.

One study examined the variations of gene methylation in cancer cells from patients administered with chemotherapy drugs. Among the 14 genes and 800 CpG islands analyzed, nearly half of the genes tested did not show any alterations in DNA methylation level [35]. However, we cannot exclude the possibility that more aberrant methylation occurred when cancer cells were exposed to chemotherapeutic drugs for a long time. It has been reported that promoter methylation of tumor suppressor genes plays an important role in tumor resistance to chemotherapeutic drugs, targeted drugs and radiotherapy [36, 37, 38, 39, 40], which indicates that various therapies, including chemotherapy, do promote gene methylation. Since gene methylation status can be changed by chemotherapy, it is usually difficult to evaluate the prognostic value of gene methylation for chemotherapy. We assumed that a shorter follow-up time may help to minimize the effect of chemotherapeutic drugs on gene methylation [35, 41], though it is not clear whether this effect is time-dependent or dose-dependent. Therefore, PFS rather than overall survival (OS) was used in our study as the primary clinical outcome measure.

The stratified survival curves of PFS in our study showed that the median PFS of patients with methylated genes in the pemetrexed doublet group was similar to that in the gemcitabine doublet group, whereas the median PFS of patients with unmethylated genes in the pemetrexed group was significantly better than that in other patients (Fig. 2). This may be related to the effect of chemotherapeutic drugs on gene methylation. Insufficient intake of folic acid affects the production of S-adenosylmethionine, which is the methyl donor of cytosine methylation [42]. Pemetrexed is a multitargeted antifolate that inhibits several enzymes involved in the folate pathway [43]. In the pemetrexed doublet group, gene methylation induced by chemotherapeutic drugs may be inhibited, so that unmethylated RASSF1A in lung cancer remains unmethylated. However, unmethylated RASSF1A tends to be methylated in the gemcitabine doublet group due to the lack of inhibition on methylation-promoting by chemotherapy. The inference suggests that RASSF1A methylation status is not appropriate as prognostic marker for most chemotherapy [36, 38, 40], except for pemetrexed-based regimen. From the perspective of guiding clinical treatment, methylation status of RASSF1A can be used as an indicator of pemetrexed-based regimen, because those NSCLC patients with unmethylated RASSF1A who received pemetrexed chemotherapy had longer PFS. For NSCLC patients with methylated RASSF1A, whether pemetrexed- or gemcitabine-based chemotherapy has no significant effect on PFS.

It is worth noting that the validity of the hazard proportional assumption is required in the Cox regression model. We tested confounding factors for hazard proportional assumption and found that the number of chemotherapy cycles, the chemotherapy regimen and overall response were all time-dependent covariates. We used the extended Cox regression model to deal with the time-dependent confounding factors, which has received little attention in previous studies on methylation [16, 44]. Due to the small sample size in our study, analysis of data without significant differences should be interpreted with caution. Future studies using larger samples are needed to verify whether SHOX2 methylation is not an independent factor for PFS in patients receiving pemetrexed-based chemotherapy. After all, pemetrexed, as an inhibitor of folic acid metabolism, should not merely target the methylation of specific genes. In further studies, patients treated with more chemotherapy regimens will also be analyzed to expand the applicability of the conclusions.

Our study focused on the predictive value of gene methylation in NSCLC patients receiving pemetrexed or gemcitabine doublet chemotherapy and observed that RASSF1A methylation is an independent predictor of PFS. The median PFS of patients with unmethylated RASSF1A in the pemetrexed doublet group was significantly longer than that of the patients with methylated RASSF1A. No predictive effect of unmethylated RASSF1A was found in gemcitabine doublets group. In summary, unmethylated RASSF1A may predict disease progression for NSCLC receiving pemetrexed doublets though no predictive value was observed for overall response to chemotherapy. BAL is only a non-invasive alternative to lung cancer biopsy in clinical practice, so results in the study still need to be validated with lung cancer tissue as detection specimen. Future study will collect more lung cancer tissues as test samples to assess the predictive value of RASSF1A, SHOX2 and many other genes that were reported frequently methylated in lung cancer.

Footnotes

Acknowledgments

We gratefully acknowledge the contributions of Huangfu for his help in language modification. We also greatly appreciate technicians from Tellgen Corporation for their support of research.

Conflict of interest

The authors declare that they have no competing interest.

References

Schiller

J.H.

Harrington

Belani

C.P.

Langer

Sandler

Krook

Zhu

Johnson

D.H.

and Eastern Cooperative Oncology Group, Comparison of four chemotherapy regimens for advanced non-small-cell lung cancer, N Engl J Med 346 (2002), 92–98.

Ohe

Ohashi

Kubota

Tamura

Nakagawa

Negoro

Nishiwaki

Saijo

Ariyoshi

and Fukuoka

, Randomized phase III study of cisplatin plus irinotecan versus carboplatin plus paclitaxel, cisplatin plus gemcitabine, and cisplatin plus vinorelbine for advanced non-small-cell lung cancer: Four-Arm Cooperative Study in Japan, Ann Oncol 18 (2007), 317–323.

Planchard

Popat

Kerr

Novello

Smit

E.F.

Faivre-Finn

Mok

T.S.

Reck

Van Schil

P.E.

Hellmann

M.D.

Peters

and ESMO Guidelines Committee, Metastatic non-small cell lung cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. ESMO Guidelines Committee, Ann Oncol 29 (2018), iv192–iv237.

Callinan

P.A.

and Feinberg

A.P.

, The emerging science of epigenomics, Hum Mol Genet 1 (2006), R95–101.

Vizoso

Puig

Carmona

F.J.

Maqueda

Velásquez

Gómez

Labernadie

Lugo

Gabasa

Rigat-Brugarolas

L.G.

Trepat

Ramírez

Moran

Vidal

Reguart

Perera

Esteller

and Alcaraz

, Aberrant DNA methylation in non-small cell lung cancer-associated fibroblasts, Carcinogenesis 36 (2015), 1453–1463.

Stelzer

Shivalila

C.S.

Soldner

Markoulaki

and Jaenisch

, Tracing dynamic changes of DNA methylation at single-cell resolution, Cell 163 (2015), 218–229.

Belinsky

S.A.

, Gene-promoter hypermethylation as a biomarker in lung cancer, Nat Rev Cancer 4 (2004), 707–717.

Esteller

, Cancer epigenomics: DNA methylomes and histone-modification maps, Nat Rev Genet 8 (2007), 286–298.

Sandoval

Peiró-Chova

Pallardó

F.V.

and García-Giménez

J.L.

, Epigenetic biomarkers in laboratory diagnostics: emerging approaches and opportunities, Expert Rev Mol Diagn 13 (2013), 457–471.

10.

Sandoval

and Esteller

, Cancer epigenomics: beyond genomics, Curr Opin Genet Dev 22 (2012), 50–55.

11.

Baylin

S.B.

and Herman

J.G.

, DNA hypermethylation in tumorigenesis: epigenetics joins genetics, Trends Genet 16 (2000), 168–174.

12.

Dietrich

Kneip

Raji

Liloglou

Seegebarth

Schlegel

Flemming

Rausch

Distler

Fleischhacker

Schmidt

Giles

Walshaw

Warburton

Liebenberg

and Field

J.K.

, Performance evaluation of the DNA methylation biomarker SHOX2 for the aid in diagnosis of lung cancer based on the analysis of bronchial aspirates, Int J Oncol 40 (2012), 825–832.

13.

Zhang

Wang

Zhao

Guo

and Lou

, DNA methylation analysis of the SHOX2 and RASSF1A panel in bronchoalveolar lavage fluid for lung cancer diagnosis, J Cancer 8 (2017), 3585–3591.

14.

Fischer

J.R.

Ohnmacht

Rieger

Zemaitis

Stoffregen

Manegold

and Lahm

, Prognostic significance of RASSF1A promoter methylation on survival of non-small cell lung cancer patients treated with gemcitabine, Lung Cancer 56 (2007), 115–123.

15.

Dietrich

Hasinger

Liebenberg

Field

J.K.

Kristiansen

and Soltermann

, DNA methylation of the homeobox genes PITX2 and SHOX2 predicts outcome in non-small-cell lung cancer patients, Diagn Mol Pathol 21 (2012), 93–104.

16.

de Fraipont

Levallet

Creveuil

Bergot

Beau-Faller

Mounawar

Richard

Antoine

Rouquette

Favrot

M.C.

Debieuvre

Braun

Westeel

Quoix

Brambilla

Hainaut

Moro-Sibilot

Morin

Milleron

Zalcman

and Intergroupe Francophone de Cancérologie Thoracique, An apoptosis methylation prognostic signature for early lung cancer in the IFCT-0002 trial, Clin Cancer Res 18 (2012), 2976–2986.

17.

Toyooka

K.O.

Toyooka

Maitra

Feng

Kiviat

N.C.

Smith

Minna

J.D.

Ashfaq

and Gazdar

A.F.

, Establishment and validation of real-time polymerase chain reaction method for CDH1 promoter methylation, Am J Pathol 161 (2002), 629–634.

18.

Chan

E.C.

Lam

S.Y.

Tsang

K.W.

Lam

J.C.

K.H.

Lam

W.K.

and Kwong

Y.L.

, Aberrant promoter methylation in Chinese patients with non-small cell lung cancer: patterns in primary tumors and potential diagnostic application in bronchoalevolar lavage, Clin Cancer Res 8 (2002), 3741–3746.

19.

Gao

Colicino

Shen

Just

A.C.

Nwanaji-Enwerem

J.C.

Wang

Coull

Lin

Vokonas

Zheng

Hou

Schwartz

and Baccarelli

A.A.

, Accelerated DNA methylation age and the use of antihypertensive medication among older adults, Aging (Albany NY) 10 (2018), 3210–3228.

20.

Friboulet

Olaussen

K.A.

Pignon

J.P.

Shepherd

F.A.

Tsao

M.S.

Graziano

Kratzke

Douillard

J.Y.

Seymour

Pirker

Filipits

André

Solary

Ponsonnailles

Robin

Stoclin

Dorvault

Commo

Adam

Vanhecke

Saulnier

Thomale

Le Chevalier

Dunant

Rousseau

Le Teuff

Brambilla

and Soria

J.C.

, ERCC1 isoform expression and DNA repair in non-small-cell lung cancer, N Engl J Med 368 (2013), 1101–1110.

21.

Bepler

Williams

Schell

M.J.

Chen

Zheng

Simon

Gadgeel

Zhao

Schreiber

Brahmer

Chiappori

Tanvetyanon

Pinder-Schenck

Gray

Haura

Antonia

and Fischer

J.R.

, Randomized international phase III trial of ERCC1 and RRM1 expression-based chemotherapy versus gemcitabine/carboplatin in advanced non-small-cell lung cancer, J Clin Oncol 31 (2013), 2404–2412.

22.

Maus

M.K.

Mack

P.C.

Astrow

S.H.

Stephens

C.L.

Zeger

G.D.

Grimminger

P.P.

Hsiang

J.H.

Huang

Lara

P.N.

Danenberg

K.D.

and Gandara

D.R

, Histology-related associations of ERCC1, RRM1, and TS biomarkers in patients with non-small-cell lung cancer: implications for therapy, J Thorac Oncol 8 (2013), 582–586.

23.

Shi

Wang

M.X.

and Caldwell

C.W.

, CpG islands: their potential as biomarkers for cancer, Expert Rev Mol Diagn 7 (2007), 519–531.

24.

Sidransky

, Emerging molecular markers of cancer, Nat Rev Cancer 2 (2002), 210–219.

25.

Liang

Arora

Zhang

K.L.

Yeh

D.C.

Lai

Pearson

Barnett

Whiley

Sloots

Corrie

S.R.

and Barnard

R.T.

, A new, multiplex, quantitative real-time polymerase chain reaction system for nucleic acid detection and quantification, Methods Mol Biol 1039 (2013), 51–68.

26.

Chen

T.T.

Chute

J.P.

Feigal

Johnson

B.E.

and Simon

, A model to select chemotherapy regimens for phase III trials for extensive-stage small-cell lung cancer, J Natl Cancer Inst 92 (2000), 1601–1607.

27.

Zia

M.I.

Siu

L.L.

Pond

G.R.

and Chen

E.X.

, Comparison of outcomes of phase II studies and subsequent randomized control studies using identical chemotherapeutic regimens, J Clin Oncol 23 (2005), 6982–6291.

28.

Wang

Protopopov

Malyukova

Kashuba

Minna

J.D.

Lerman

M.I.

Klein

and Zabarovsky

, Inactivation of RASSF1C during in vivo tumor growth identifies it as a tumor suppressor gene, Oncogene 23 (2004), 5941–5949.

29.

Wang

Chen

and Bi

, The prognostic value of RASSF1A promoter hypermethylation in non-small cell lung carcinoma: a systematic review and meta-analysis, Carcinogenesis 32 (2011), 411–416.

30.

Wang

Lee

J.J.

Wang

Liu

D.D.

Fan

Y.H.

Hong

W.K.

and Mao

, Value of p16INK4a and RASSF1A promoter hypermethylation in prognosis of patients with resectable non-small cell lung cancer, Clin Cancer Res 10 (2004), 6119–6125.

31.

Tomizawa

Kohno

Kondo

Otsuka

Nishioka

Niki

Yamada

Maeshima

Yoshimura

Saito

Minna

J.D.

and Yokota

, Clinicopathological significance of epigenetic inactivation of RASSF1A at 3p21.3 in stage I lung adenocarcinoma, Clin Cancer Res 8 (2002), 2362–2368.

32.

Fischer

J.R.

Ohnmacht

Rieger

Zemaitis

Stoffregen

Manegold

and Lahm

, Prognostic significance of RASSF1A promoter methylation on survival of non-small cell lung cancer patients treated with gemcitabine, Lung Cancer 56 (2007), 115–123.

33.

Abdul

Q.A.

B.P.

Chung

H.Y.

Jung

H.A.

and Choi

J.S.

, Epigenetic modifications of gene expression by lifestyle and environment, Arch Pharm Res 40 (2017), 1219–1237.

34.

Waterland

R.A.

Lin

J.R.

Smith

C.A.

and Jirtle

R.L.

, Post-weaning diet affects genomic imprinting at the insulin-like growth factor 2 (Igf2) locus, Hum Mol Genet 15 (2006), 705–716.

35.

Boettcher

Kischkel

and Hoheisel

J.D.

, High-definition DNA methylation profiles from breast and ovarian carcinoma cell lines with differing doxorubicin resistance, PLoS One 5 (2010), e11002.

36.

Fang

Cardenas

Huang

Jiang

Perkins

S.M.

Zhang

Keer

H.N.

Liu

Nephew

K.P.

and Matei

, Genomic and epigenomic signatures in ovarian cancer associated with resensitization to platinum drugs, Cancer Res 78 (2018), 631–644.

37.

Ramachandran

Speer

Nathanson

Claros

and Singal

, Role of DNA methylation in cabazitaxel resistance in prostate cancer, Anticancer Res 36 (2016), 161–168.

38.

Flanagan

J.M.

Wilson

Koo

Masrour

Gallon

Loomis

Flower

Wilhelm-Benartzi

Hergovich

Cunnea

Gabra

Braicu

E.I.

Sehouli

Darb-Esfahani

Vanderstichele

Vergote

Kreuzinger

Castillo-Tong

D.C.

Wisman

G.B.A.

Berns

E.M.

Siddiqui

Paul

and Brown

, Platinum-based chemotherapy induces methylation changes in blood DNA associated with overall survival in patients with ovarian cancer, Clin Cancer Res 23 (2017), 2213–2222.

39.

Chen

Liu

Mims

Punska

E.C.

Williams

K.E.

Zhao

Arcaro

K.F.

Tsang

A.W.

Zhou

and Furdui

C.M.

, Analysis of DNA methylation and gene expression in radiation-resistant head and neck tumors, Epigenetics 10 (2015), 545–561.

40.

Ogawa

Liggett

T.E.

Melnikov

A.A.

Monitto

C.L.

Kusuke

Shiga

Kobayashi

Horii

Chatterjee

Levenson

V.V.

Koch

W.M.

Sidransky

and Chang

, Methylation of death-associated protein kinase is associated with cetuximab and erlotinib resistance, Cell Cycle 11 (2012), 1656–1663.

41.

Yao

Kerns

Yan

Onitilo

A.A.

Misleh

Young

Lei

Bautista

Mohamed

Mohile

S.G.

Ambrosone

C.B.

Liu

and Janelsins

M.C.

, Impact of chemotherapy for breast cancer on leukocyte DNA methylation landscape and cognitive function: a prospective study, Clin Epigenetics 11 (2019), 45.

42.

G.M.

Presnell

S.R.

and Gu

, Folate deficiency, mismatch repair-dependent apoptosis, and human disease, J Nutr Biochem 14 (2003), 568–575.

43.

Adjei

A.A.

, Pemetrexed (ALIMTA), a novel multitargeted antineoplastic agent, Clin Cancer Res 10 (2004), 4276s–4280s.

44.

Fischer

J.R.

Ohnmacht

Rieger

Zemaitis

Stoffregen

Manegold

and Lahm

, Prognostic significance of RASSF1A promoter methylation on survival of non-small cell lung cancer patients treated with gemcitabine, Lung Cancer 56 (2007), 115–123.

Predictive value of unmethylated RASSF1A on disease progression in non-small cell lung cancer patients receiving pemetrexed-based chemotherapy

Abstract

BACKGROUND AND OBJECTIVE:

PATIENTS AND METHODS:

RESULTS:

CONCLUSION:

Keywords

1. Introduction

2. Materials and methods

2.1 Ethics statement

2.2 Study subjects and samples

2.3 Cytological and pathological analysis

2.4 DNA Extractions and bisulfite modification of genomic DNA

2.5 Methylation specific real-time fluorescent PCR analysis

Table 1 Consistency of methylation detection between BAL and primary tumor

3. Results

3.1 Evaluation of genes methylation detected in BAL

3.2 Clinicopathological characteristics of patients and genes methylation

Table 2 Characteristic of 117 NSCLC patients with genes methylation analysis

Table 3 Predictors on PFS in cox regression analysis for 117 NSCLC patients

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
Consistency of methylation detection between BAL and primary tumor

Table 2
Characteristic of 117 NSCLC patients with genes methylation analysis

Table 3
Predictors on PFS in cox regression analysis for 117 NSCLC patients