A differentially-methylated-region signature predicts the recurrence risk for patients with early stage lung adenocarcinoma

Materials and Methods

Patient and study design

The study consisted of three cohorts, including marker discovery cohort (n = 39), prognostic model training cohort (n = 117) and prognostic model validation cohort (n = 80). All eligible patients with stage I–IIIA LUAD were enrolled from the First Affiliated Hospital of Southern University of Science and Technology. Tumor tissues and adjacent lung tissues were collected during surgical resection. The diagnosis of LUAD was founded upon the assessment of a pathological specimen subsequent to surgical resection. Researchers assigned pathological stages to all patients, adhering to the 8th edition of the American Joint Committee on Cancer (AJCC) classifications. Patients identified with stage IIIB/IV lung cancer, other types of cancer, anaemia, or autoimmune diseases were excluded from the study. LUAD samples with a tumor fraction less than 40% were excluded from the study. This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the First Affiliated Hospital of Southern University of Science and Technology (AUP-220516-CHR-0390-01). Written informed consents were obtained from all patients. The clinicopathologic characteristics of the three cohorts included in this study are summarized in Table 1. The baseline characteristics including age, sex, smoking history and TNM stage were collected for all the three cohorts and the survival was collected for the model training and validation cohorts only.

Table 1. Baseline characteristics of the cohorts included in the study.

Characteristics	Marker discovery cohort (n = 39)	Model training cohort (n = 117)	Model validation cohort (n = 80)
Sex
Male	24 (61.5%)	61 (52.1%)	39 (48.8%)
Female	15 (38.5%)	56 (47.9%)	41 (51.2%)
Age
<65 yrs	26 (66.7%)	83 (70.9%)	42 (52.5%)
≥65 yrs	13 (33.3%)	34 (29.1%)	38 (47.5%)
Smoking history
Yes	21 (53.8%)	60 (51.3%)	25 (31.3%)
No	17 (43.6%)	57 (48.7%)	55 (68.7%)
NA	1 (2.6%)
AJCC stage
I	39 (100.0%)	46 (39.3%)	60 (75.0%)
II		28 (23.9%)	7 (8.7%)
IIIA		43 (36.8%)	13 (16.3%)

The marker discovery cohort was used to identify LUAD-specific DMRs, the prognostic training cohort was used to develop a risk-modeling signature based on the above DMRs and the prognostic validation cohort was used to validate the developed signature.

Target methylation sequencing and data preprocessing

The procedure for DNA extraction was as previously described [26]. In summary, for tissue samples, the DNA extraction process was carried out utilizing the QIAamp DNA formalin-fixed and paraffin-embedded tissue kit, adhering rigorously to the manufacturer’s prescribed instructions. Subsequently, DNA concentration was quantified using the Qubit double-stranded DNA assay (Life Technologies, Carlsbad, CA, USA).

As for methylation sequencing, a capture-based approach was employed to identify CpG sites. The bisulfite sequencing library was constructed utilizing the brELSATM methodology (Burning Rock Biotech, Guangzhou, China) [27]. Subsequently, the targeted libraries underwent quantitative analysis through real-time PCR and were sequenced on NovaSeq 6000 achieving an average target depth of 500×. With the raw sequencing data, a suite of bioinformatics tools, encompassing Trimmomatic, BWA-meth, and samblaster were deployed to facilitate read alignment and variant calling as part of the downstream analytical pipeline. Notably, as reported, DMRs comprising multiple CpG sites played more pivotal roles in cancer detection compared to individual CpG sites [28]. Methylation blocks were designated as CpG sites exhibiting close genomic proximity and a high degree of correlation in their methylation levels. Utilizing the 80,672 CpG sites, a total of 8,312 such blocks were identified. The score for each block was calculated, incorporating both the depth of coverage and the distance between adjacent CpG sites as follows [29]:

$$MethylationBlockScore=\frac{1}{n}\times {\displaystyle \sum _{i=1}^n\left(\frac{{\displaystyle {\sum }_{i=1}^m{l}_{ij}^2}}{L_i^2}\right)}$$

For a specified methylation block, n represents the aggregate number of reads encompassing several CpG sites, and L_i designates the number of CpG sites covered on ith read. The length of consecutive methylated CpG sites exceeding 1 is denoted as l_ij, and m signifies the total counts on ith read. To normalize the depth variation, the number of reads within each block is employed, thereby constraining the metric within the range of 0 to 1.

Marker selection for LUAD

DMRs between LUAD samples and matched adjacent tissue samples were identified in the marker discovery cohort (LUAD: n = 39; matched normal tissue: n = 39) with the cutoff values determined as the absolute value of mean difference of methylation levels >0.1 and adjusted P-value < 0.05.

Construction and validation of a DMR signature related to recurrence

To screen DMRs related to the DFS of LUAD patients in the model training cohort (n = 117), a univariable Cox regression analysis was employed. To minimize the risk of overfitting, we used the Least Absolute Shrinkage and Selection Operator (LASSO) penalized Cox regression analysis to construct a prognostic risk model. The LASSO algorithm selects and contracts the variables through R package “glmnet” so that some of the regression coefficients are strictly equal to zero, resulting in an interpretable model. The risk score of each patient was calculated according to the methylation values of selected DMRs and the corresponding regression coefficients:

$$Riskscore={\displaystyle \sum _{i-1}^{n}MB{S}_i\times Coe{f}_i}$$

For each patient, n represents the number of DMRs with prognostic value, MBS_i is the methylation block score of DMR i, and Coef_i represents the regression coefficient of DMR i.

Patients were categorized into high-risk and low-risk groups utilizing the risk score threshold determined from the training set. Survival analysis was performed by R packages “survival” and “survminer” to analyze DFS between the high-risk and low-risk patients. To assess the predictive accuracy of the prognostic DMR signature, a time-dependent receiver operating characteristic (ROC) curve analysis was conducted with “timeROC” R package. Both univariable and multivariable Cox regression analyses were performed, aiming to uncover the independent prognostic significance of the risk score. In addition, the calculation of risk score, grouping of samples, survival analysis, ROC analysis, and univariable and multivariable Cox regression analyses were also performed in the validation cohort (n = 80) for independent validation. The study design is shown in Figure 1 by a flow chart.

Figure 1. Work flow of the study. Abbreviations: LUAD: lung adenocarcinoma; DMR: differentially methylated region; LASSO: least absolute shrinkage and selection operator.

Comparison of risk score between clinical subgroups

Samples in both model training and validation cohorts were divided into subgroups according to several clinical characteristics including sex, age, smoking history, and TNM stage. The comparison of risk scores, derived from the DMR signature, was conducted between different subgroups.

Functional enrichment analysis

Gene enrichment analysis targeting the biological process and biological pathways in the genes corresponding to the DMRs was via the “clusterProfiler” R package combined with Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

Statistical analysis

R software (version 4.2.0; https://www.R-project.org) was used for all statistics. Wilcoxon test was used to screen DMRs. Kaplan-Meier (KM) curves were drawn, and the difference of DFS between risk groups was tested by Log-rank test. The area under ROC curve (AUC) was used to judge the accuracy of predicting. The comparison of risk score between different clinical subgroups was via Wilcoxon test. For all analyses in the study, P < 0.05 was considered statistically significant.

Data availability statement

The relevant data supporting the findings of this study are available within the paper. Due to ethical and privacy concerns, we are unable to publish the patient-level data in our study, of which readers may contact the corresponding authors for access for non-commercial purposes.

Results

Identification of DMRs in early stage LUAD patients

In order to better reflect the LUAD-associated methylation alteration, we first tried to identify DMRs in LUAD by comparing 39 stage I LUAD samples and 39 matched adjacent lung tissue samples in the marker discovery cohort. Investigation of DMRs between the two groups yielded a total of 468 DMRs, which included 438 hypermethylated regions and 30 hypomethylated regions (Figure 2A and Supplementary Table 1). Principal component analysis showed an obvious difference between LUAD tumor and adjacent tissues based on the methylation value of 468 DMRs (Figure 2B). Heat map of the 468 DMRs shows the hypo- and hypermethylated regions in LUAD compared with adjacent lung tissues (Figure 2C). Most of the DMRs located in the CpG islands and transcription start site of promoter (59.1%), and contributed to protein-coding (80.4%) (Figure 2D). In addition, the genes corresponding to the DMRs were mostly enriched in the biological processes including cell fate specification, cell fate commitment, pattern specification process, embryonic organ morphogenesis, and embryonic skeletal system morphogenesis (Supplementary Figure 1A), and the signaling pathways including transcriptional misregulation in cancer, maturity-onset diabetes of the young, and cAMP signaling pathway (Supplementary Figure 1B).

Figure 2. Identification of DMRs in the patients with early stage LUAD. (A) Difference analysis yielded a total of 468 DMRs, including 438 hypermethylated regions and 30 hypomethylated regions in the marker discovery cohort (LUAD: n = 39; normal: n = 39). (B) Principal component analysis between LUAD and normal lung samples based on the methylation value of 468 DMRs. (C) Heatmap of the 468 DMRs in the marker discovery cohort. (D) Annotation information of the 468 DMRs, including location and the function of corresponding genes.

Construction of a 15-DMR prognostic signature in early stage LUAD patients

We then aimed to develop a signature with prognostic value in the model training cohort based on the above identified DMRs. We performed univariable Cox regression analysis and 41 DMRs were found to be related to the DFS of 117 stage I–IIIA LUAD patients, which was followed by LASSO penalized Cox regression analysis. A total of 15 DMRs were ultimately incorporated into the recurrence risk predicting model construction (Figure 3A). Hazard ratio with 95% confidence interval (CI) and P-value of the 15 DMRs are shown in Figure 3B. Among them, the methylation value of Chr7: 121956469–121956812 was highly negatively associated with the risk of recurrence (P = 0.03, HR = 0.60, 95% CI = 0.38–0.95; Supplementary Figure 2A), while the methylation value of Chr12: 52214765–52214877 was highly positively associated with the risk of recurrence (P = 0.02, HR = 1.69, 95% CI = 1.07–2.69; Supplementary Figure 2B). In addition, the gene annotation of the 15 DMRs is summarized in Supplementary Table 2.

Figure 3. Construction of a 15-DMR prognostic signature in the model training cohort. (A) LASSO penalized Cox regression analysis yielded 15 DMRs for the recurrence risk predicting model construction. (B) Hazard ratio with 95% confidence interval (CI) and P-value of the 15 DMRs in the univariable Cox regression analysis for disease-free survival (DFS). (C) Heatmap of the methylation value of the 15 DMRs in low-risk (n = 45) and high-risk (n = 72) groups. (D) Kaplan-Meier (KM) curves for DFS in the model training cohort. (E) Time-dependent receiver operating characteristic (ROC) curves in the model training cohort. (F) Comparison of the risk score in the different clinical subgroups including age, sex, smoking history, and TNM stage.

LUAD samples in the model training cohort were then segregated into low-risk and high-risk groups utilizing a risk score threshold of −1.44. A comparison was then conducted, examining the methylation values of these 15 DMRs between the two risk groups. The value of Chr10: 101295771–101295802, Chr5: 72678348–72678362, Chr1: 91184901–91185174, Chr6: 78172227–78172500, Chr15: 89920989–89921000, Chr7: 121956469–121956812, Chr19: 35630031–35630097, Chr19: 35630099–35630164, and Chr3: 138679199–138679210—which were negatively associated with the risk of recurrence—were significantly higher in the low-risk group, while the value of Chr12: 52214765–52214877, Chr5: 16179933–16180189, Chr16: 22825309–22825446, Chr12: 130823681–130823754, Chr12: 130823619–130823681, and Chr14: 51560315–51560326—which were positively associated with the risk of recurrence—were significantly higher in the high-risk group (P < 0.05; Figure 3C). The KM curve distinctly demonstrated a significantly shorter DFS among high-risk patients compared to their low-risk counterparts (P < 0.001, HR = 4.32, 95% CI = 2.39–7.80; Figure 3D). To assess the predictive proficiency for recurrence risk, time-dependent ROC curve analyses were conducted, and the AUC was 0.775 at 1 year, 0.773 at 3 years and 0.805 at 5 years, collectively indicating a robust predictive performance of the DMR signature (Figure 3E). Risk scores had no significant difference between the different age, sex, and smoking history groups, but were obviously higher in stage II and IIIA compared to stage I (stage II vs. stage I: P = 0.008, stage IIIA vs. stage I: P = 0.02; Figure 3F).

To further assess the prognostic significance of the DMR signature, univariable and multivariable Cox regression analyses were conducted on the most relevant clinical variables of LUAD patients in the training set. Results showed that the risk score was significantly associated with the DFS of patients adjusted for age, sex, smoking history, and TNM stage (univariable: P < 0.001, HR = 4.32, 95% CI = 2.39–7.80; multivariable: P < 0.001, HR = 4.46, 95% CI = 2.46–8.11; Table 2). In conclusion, the results above suggested that the 15-DMR signature had a good performance in predicting the risk of recurrence for patients with early stage LUAD in the model training cohort.

Table 2. Univariable and multivariable Cox regression analysis in the training cohort.

Parameter	Model training cohort (n = 117)
	Univariable Cox analysis		Multivariable Cox analysis
	HR (95% CI)	P-value	HR (95% CI)	P-value
Age (≥65 vs. <65)	0.95 (0.58–1.58)	0.86	0.92 (0.55–1.55)	0.77
Sex (male vs. female)	1.25 (0.79–1.98)	0.33	1.49 (0.87–2.57)	0.15
Smoking history (yes vs. no)	0.97 (0.62–1.53)	0.90	0.78 (0.46–1.33)	0.36
Stage (stage II–IIIA vs. stage I)	3.77 (2.18–6.54)	<0.001*	2.96 (1.68–5.23)	<0.001*
Risk score (high vs. low)	4.32 (2.39–7.80)	<0.001*	3.37 (1.82–6.23)	<0.001*
Note: *represents a statistically significant difference.

Performance of the 15-DMR signature in the model validation cohort

To validate the stability of the DMR signature, utilizing the identical algorithm, the risk score for each patient was further computed within the model validation cohort. 80 stage I–IIIA LUAD samples were also stratified into high-risk and low-risk groups by the same cutoff value. Consistent with the training cohort, the value of Chr10: 101295771–101295802, and Chr3: 138679199–138679210 were significantly increased in the low-risk group, while the value of Chr12: 52214765–52214877, Chr12: 130823619–130823681, Chr12: 130823681–130823754, Chr16: 22825309–22825446, and Chr14: 51560315–51560326 were significantly increased in the high-risk group (P < 0.05; Figure 4A). In addition, a notable difference was observed that high-risk patients exhibiting significantly shorter DFS compared to their low-risk counterparts (P = 0.009, HR = 9.08, 95% CI = 1.20–68.80; Figure 4B). The AUC for predicting risk of recurrence reached 0.757 and 0.686 at 1 and 3 years, respectively (Figure 4C). Risk scores were not statistically different between clinical subgroups (Figure 4D).

Figure 4. Performance of the DMR signature in the model validation cohort. (A) Heat map of the methylation value of the 15 DMRs in low-risk (n = 27) and high-risk (n = 58) groups. (B) KM curves for DFS in the model validation cohort. (C) Time-dependent ROC curves in the model validation cohort. (D) Comparison of the risk score in the different clinical subgroups including age, sex, smoking history, and TNM stage.

Furthermore, both univariable and multivariable Cox regression analysis indicated a robust association between the risk score and the likelihood of recurrence in patients with early stage LUAD (univariable: P = 0.03, HR = 9.08, 95% CI = 1.20–68.80; multivariable: P = 0.04, HR = 8.50, 95% CI = 1.08–66.56; Table 3). In addition to risk score, as expected, TNM stage is an independent risk factor for prognosis. Thus, a nomogram incorporating the risk score and stage was devised to better identify patients of their risk of recurrence (Supplementary Figure 3). Altogether, we developed and validated a 15-DMR signature that was associated with prognosis of patients with LUAD.

Table 3. Univariable and multivariable Cox regression analysis in the validation cohort.

Parameter	Model validation cohort (n = 80)
	Univariable Cox analysis		Multivariable Cox analysis
	HR (95% CI)	P-value	HR (95% CI)	P-value
Age (≥65 vs. <65)	2.02 (0.73–5.56)	0.17	1.82 (0.63–5.27)	0.27
Sex (male vs. female)	2.46 (0.86–7.09)	0.10	2.56 (0.59–11.07)	0.21
Smoking history (yes vs. no)	2.25 (0.84–6.00)	0.11	0.88 (0.22–3.57)	0.86
Stage (stage II–IIIA vs. stage I)	5.18 (1.91–14.04)	0.001*	4.73 (1.63–13.69)	0.004*
Risk score (high vs. low)	9.08 (1.20–68.80)	0.03*	8.50 (1.08–66.56)	0.04*
Note: *represents a statistically significant difference.

Author Contributions

Conception and design: HL, FL, GL, and DW. Collection and assembly of data: HL, FL, XS, CL, and GW. Data analysis and interpretation: HL, FL, XS, CL, GW, YH, LL, CX, WW, and SC. Manuscript writing: All authors. Final approval of manuscript: All authors.

Conflicts of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Ethical Statement and Consent

This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the First Affiliated Hospital of Southern University of Science and Technology (AUP-220516-CHR-0390-01). Written informed consents were obtained from all patients.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

References

• 1. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin. 2022; 72:7–33. https://doi.org/10.3322/caac.21708 [PubMed]
• 2. Zappa C, Mousa SA. Non-small cell lung cancer: current treatment and future advances. Transl Lung Cancer Res. 2016; 5:288–300. https://doi.org/10.21037/tlcr.2016.06.07 [PubMed]
• 3. Noguchi M, Morikawa A, Kawasaki M, Matsuno Y, Yamada T, Hirohashi S, Kondo H, Shimosato Y. Small adenocarcinoma of the lung. Histologic characteristics and prognosis. Cancer. 1995; 75:2844–52. https://doi.org/10.1002/1097-0142(19950615)75:12<2844::aid-cncr2820751209>3.0.co;2-%23 [PubMed]
• 4. Sugimura H, Nichols FC, Yang P, Allen MS, Cassivi SD, Deschamps C, Williams BA, Pairolero PC. Survival after recurrent nonsmall-cell lung cancer after complete pulmonary resection. Ann Thorac Surg. 2007; 83:409–17. https://doi.org/10.1016/j.athoracsur.2006.08.046 [PubMed]
• 5. Arriagada R, Bergman B, Dunant A, Le Chevalier T, Pignon JP, Vansteenkiste J, and International Adjuvant Lung Cancer Trial Collaborative Group. Cisplatin-based adjuvant chemotherapy in patients with completely resected non-small-cell lung cancer. N Engl J Med. 2004; 350:351–60. https://doi.org/10.1056/NEJMoa031644 [PubMed]
• 6. Sawabata N, Miyaoka E, Asamura H, Nakanishi Y, Eguchi K, Mori M, Nomori H, Fujii Y, Okumura M, Yokoi K, and Japanese Joint Committee for Lung Cancer Registration. Japanese lung cancer registry study of 11,663 surgical cases in 2004: demographic and prognosis changes over decade. J Thorac Oncol. 2011; 6:1229–35. https://doi.org/10.1097/JTO.0b013e318219aae2 [PubMed]
• 7. Balgkouranidou I, Liloglou T, Lianidou ES. Lung cancer epigenetics: emerging biomarkers. Biomark Med. 2013; 7:49–58. https://doi.org/10.2217/bmm.12.111 [PubMed]
• 8. Robertson KD. DNA methylation and human disease. Nat Rev Genet. 2005; 6:597–610. https://doi.org/10.1038/nrg1655 [PubMed]
• 9. Feinberg AP, Tycko B. The history of cancer epigenetics. Nat Rev Cancer. 2004; 4:143–53. https://doi.org/10.1038/nrc1279 [PubMed]
• 10. Ansari J, Shackelford RE, El-Osta H. Epigenetics in non-small cell lung cancer: from basics to therapeutics. Transl Lung Cancer Res. 2016; 5:155–71. https://doi.org/10.21037/tlcr.2016.02.02 [PubMed]
• 11. Tsou JA, Galler JS, Siegmund KD, Laird PW, Turla S, Cozen W, Hagen JA, Koss MN, Laird-Offringa IA. Identification of a panel of sensitive and specific DNA methylation markers for lung adenocarcinoma. Mol Cancer. 2007; 6:70. https://doi.org/10.1186/1476-4598-6-70 [PubMed]
• 12. Licchesi JD, Westra WH, Hooker CM, Herman JG. Promoter hypermethylation of hallmark cancer genes in atypical adenomatous hyperplasia of the lung. Clin Cancer Res. 2008; 14:2570–8. https://doi.org/10.1158/1078-0432.CCR-07-2033 [PubMed]
• 13. Selamat SA, Galler JS, Joshi AD, Fyfe MN, Campan M, Siegmund KD, Kerr KM, Laird-Offringa IA. DNA methylation changes in atypical adenomatous hyperplasia, adenocarcinoma in situ, and lung adenocarcinoma. PLoS One. 2011; 6:e21443. https://doi.org/10.1371/journal.pone.0021443 [PubMed]
• 14. Licchesi JD, Westra WH, Hooker CM, Machida EO, Baylin SB, Herman JG. Epigenetic alteration of Wnt pathway antagonists in progressive glandular neoplasia of the lung. Carcinogenesis. 2008; 29:895–904. https://doi.org/10.1093/carcin/bgn017 [PubMed]
• 15. Selamat SA, Chung BS, Girard L, Zhang W, Zhang Y, Campan M, Siegmund KD, Koss MN, Hagen JA, Lam WL, Lam S, Gazdar AF, Laird-Offringa IA. Genome-scale analysis of DNA methylation in lung adenocarcinoma and integration with mRNA expression. Genome Res. 2012; 22:1197–211. https://doi.org/10.1101/gr.132662.111 [PubMed]
• 16. Merlo A, Herman JG, Mao L, Lee DJ, Gabrielson E, Burger PC, Baylin SB, Sidransky D. 5' CpG island methylation is associated with transcriptional silencing of the tumour suppressor p16/CDKN2/MTS1 in human cancers. Nat Med. 1995; 1:686–92. https://doi.org/10.1038/nm0795-686 [PubMed]
• 17. Kashima T, Nakamura K, Kawaguchi J, Takanashi M, Ishida T, Aburatani H, Kudo A, Fukayama M, Grigoriadis AE. Overexpression of cadherins suppresses pulmonary metastasis of osteosarcoma in vivo. Int J Cancer. 2003; 104:147–54. https://doi.org/10.1002/ijc.10931 [PubMed]
• 18. Brock MV, Hooker CM, Ota-Machida E, Han Y, Guo M, Ames S, Glöckner S, Piantadosi S, Gabrielson E, Pridham G, Pelosky K, Belinsky SA, Yang SC, et al. DNA methylation markers and early recurrence in stage I lung cancer. N Engl J Med. 2008; 358:1118–28. https://doi.org/10.1056/NEJMoa0706550 [PubMed]
• 19. Yamamoto H, Toyooka S, Mitsudomi T. Impact of EGFR mutation analysis in non-small cell lung cancer. Lung Cancer. 2009; 63:315–21. https://doi.org/10.1016/j.lungcan.2008.06.021 [PubMed]
• 20. Zhang Y, Wang R, Song H, Huang G, Yi J, Zheng Y, Wang J, Chen L. Methylation of multiple genes as a candidate biomarker in non-small cell lung cancer. Cancer Lett. 2011; 303:21–8. https://doi.org/10.1016/j.canlet.2010.12.011 [PubMed]
• 21. Yang J, Shen Y, Liu B, Tong Y. Promoter methylation of BRMS1 correlates with smoking history and poor survival in non-small cell lung cancer patients. Lung Cancer. 2011; 74:305–9. https://doi.org/10.1016/j.lungcan.2011.03.002 [PubMed]
• 22. Dietrich D, Hasinger O, Liebenberg V, Field JK, Kristiansen G, Soltermann A. DNA methylation of the homeobox genes PITX2 and SHOX2 predicts outcome in non-small-cell lung cancer patients. Diagn Mol Pathol. 2012; 21:93–104. https://doi.org/10.1097/PDM.0b013e318240503b [PubMed]
• 23. Lokk K, Vooder T, Kolde R, Välk K, Võsa U, Roosipuu R, Milani L, Fischer K, Koltsina M, Urgard E, Annilo T, Metspalu A, Tõnisson N. Methylation markers of early-stage non-small cell lung cancer. PLoS One. 2012; 7:e39813. https://doi.org/10.1371/journal.pone.0039813 [PubMed]
• 24. Imai H, Onozato R, Kaira K, Kawashima S, Masubuchi K, Nagashima T, Tajima K, Minato K. Course of postoperative relapse in non-small cell lung cancer is strongly associated with post-progression survival. Thorac Cancer. 2021; 12:2740–8. https://doi.org/10.1111/1759-7714.14119 [PubMed]
• 25. Robinson MD, Kahraman A, Law CW, Lindsay H, Nowicka M, Weber LM, Zhou X. Statistical methods for detecting differentially methylated loci and regions. Front Genet. 2014; 5:324. https://doi.org/10.3389/fgene.2014.00324 [PubMed]
• 26. Mao X, Zhang Z, Zheng X, Xie F, Duan F, Jiang L, Chuai S, Han-Zhang H, Han B, Sun J. Capture-Based Targeted Ultradeep Sequencing in Paired Tissue and Plasma Samples Demonstrates Differential Subclonal ctDNA-Releasing Capability in Advanced Lung Cancer. J Thorac Oncol. 2017; 12:663–72. https://doi.org/10.1016/j.jtho.2016.11.2235 [PubMed]
• 27. Yang L, Zhang J, Yang G, Xu H, Lin J, Shao L, Li J, Guo C, Du Y, Guo L, Li X, Han-Zhang H, Wang C, et al. The prognostic value of a Methylome-based Malignancy Density Scoring System to predict recurrence risk in early-stage Lung Adenocarcinoma. Theranostics. 2020; 10:7635–44. https://doi.org/10.7150/thno.44229 [PubMed]
• 28. Guo S, Diep D, Plongthongkum N, Fung HL, Zhang K, Zhang K. Identification of methylation haplotype blocks aids in deconvolution of heterogeneous tissue samples and tumor tissue-of-origin mapping from plasma DNA. Nat Genet. 2017; 49:635–42. https://doi.org/10.1038/ng.3805 [PubMed]
• 29. Liang N, Li B, Jia Z, Wang C, Wu P, Zheng T, Wang Y, Qiu F, Wu Y, Su J, Xu J, Xu F, Chu H, et al. Ultrasensitive detection of circulating tumour DNA via deep methylation sequencing aided by machine learning. Nat Biomed Eng. 2021; 5:586–99. https://doi.org/10.1038/s41551-021-00746-5 [PubMed]
• 30. Jones PA, Laird PW. Cancer epigenetics comes of age. Nat Genet. 1999; 21:163–7. https://doi.org/10.1038/5947 [PubMed]
• 31. Li P, Liu S, Du L, Mohseni G, Zhang Y, Wang C. Liquid biopsies based on DNA methylation as biomarkers for the detection and prognosis of lung cancer. Clin Epigenetics. 2022; 14:118. https://doi.org/10.1186/s13148-022-01337-0 [PubMed]
• 32. Feinberg AP, Ohlsson R, Henikoff S. The epigenetic progenitor origin of human cancer. Nat Rev Genet. 2006; 7:21–33. https://doi.org/10.1038/nrg1748 [PubMed]
• 33. Bird AP, Wolffe AP. Methylation-induced repression--belts, braces, and chromatin. Cell. 1999; 99:451–4. https://doi.org/10.1016/s0092-8674(00)81532-9 [PubMed]
• 34. Rountree MR, Bachman KE, Herman JG, Baylin SB. DNA methylation, chromatin inheritance, and cancer. Oncogene. 2001; 20:3156–65. https://doi.org/10.1038/sj.onc.1204339 [PubMed]
• 35. Wade PA. Methyl CpG binding proteins: coupling chromatin architecture to gene regulation. Oncogene. 2001; 20:3166–73. https://doi.org/10.1038/sj.onc.1204340 [PubMed]
• 36. Jones PA, Baylin SB. The fundamental role of epigenetic events in cancer. Nat Rev Genet. 2002; 3:415–28. https://doi.org/10.1038/nrg816 [PubMed]
• 37. Esteller M, Corn PG, Baylin SB, Herman JG. A gene hypermethylation profile of human cancer. Cancer Res. 2001; 61:3225–9. [PubMed]
• 38. Digel W, Lübbert M. DNA methylation disturbances as novel therapeutic target in lung cancer: preclinical and clinical results. Crit Rev Oncol Hematol. 2005; 55:1–11. https://doi.org/10.1016/j.critrevonc.2005.02.002 [PubMed]
• 39. Belinsky SA. Gene-promoter hypermethylation as a biomarker in lung cancer. Nat Rev Cancer. 2004; 4:707–17. https://doi.org/10.1038/nrc1432 [PubMed]
• 40. Aravin AA, Hannon GJ. Small RNA silencing pathways in germ and stem cells. Cold Spring Harb Symp Quant Biol. 2008; 73:283–90. https://doi.org/10.1101/sqb.2008.73.058 [PubMed]
• 41. Liu M, Hu Z, Qi L, Wang J, Zhou T, Guo Y, Zeng Y, Zheng B, Wu Y, Zhang P, Chen X, Tu W, Zhang T, et al. Scanning of novel cancer/testis proteins by human testis proteomic analysis. Proteomics. 2013; 13:1200–10. https://doi.org/10.1002/pmic.201200489 [PubMed]
• 42. Chen Z, Che Q, Jiang FZ, Wang HH, Wang FY, Liao Y, Wan XP. Piwil1 causes epigenetic alteration of PTEN gene via upregulation of DNA methyltransferase in type I endometrial cancer. Biochem Biophys Res Commun. 2015; 463:876–80. https://doi.org/10.1016/j.bbrc.2015.06.028 [PubMed]
• 43. Liu X, Sun Y, Guo J, Ma H, Li J, Dong B, Jin G, Zhang J, Wu J, Meng L, Shou C. Expression of hiwi gene in human gastric cancer was associated with proliferation of cancer cells. Int J Cancer. 2006; 118:1922–9. https://doi.org/10.1002/ijc.21575 [PubMed]
• 44. Navarro A, Tejero R, Viñolas N, Cordeiro A, Marrades RM, Fuster D, Caritg O, Moises J, Muñoz C, Molins L, Ramirez J, Monzo M. The significance of PIWI family expression in human lung embryogenesis and non-small cell lung cancer. Oncotarget. 2015; 6:31544–56. https://doi.org/10.18632/oncotarget.3003 [PubMed]
• 45. Xie K, Zhang K, Kong J, Wang C, Gu Y, Liang C, Jiang T, Qin N, Liu J, Guo X, Huo R, Liu M, Ma H, et al. Cancer-testis gene PIWIL1 promotes cell proliferation, migration, and invasion in lung adenocarcinoma. Cancer Med. 2018; 7:157–66. https://doi.org/10.1002/cam4.1248 [PubMed]
• 46. Zhang Y, Fan Q, Guo Y, Zhu K. Eight-gene signature predicts recurrence in lung adenocarcinoma. Cancer Biomark. 2020; 28:447–57. https://doi.org/10.3233/CBM-190329 [PubMed]
• 47. Miyamoto K, Asada K, Fukutomi T, Okochi E, Yagi Y, Hasegawa T, Asahara T, Sugimura T, Ushijima T. Methylation-associated silencing of heparan sulfate D-glucosaminyl 3-O-sulfotransferase-2 (3-OST-2) in human breast, colon, lung and pancreatic cancers. Oncogene. 2003; 22:274–80. https://doi.org/10.1038/sj.onc.1206146 [PubMed]
• 48. Dietrich D, Lesche R, Tetzner R, Krispin M, Dietrich J, Haedicke W, Schuster M, Kristiansen G. Analysis of DNA methylation of multiple genes in microdissected cells from formalin-fixed and paraffin-embedded tissues. J Histochem Cytochem. 2009; 57:477–89. https://doi.org/10.1369/jhc.2009.953026 [PubMed]
• 49. Retraction. Aberrant methylation to heparan sulfate glucosamine 3-O-sulfotransferase 2 genes as a biomarker in colorectal cancer. Anticancer Res. 2012; 32:5138. [PubMed]
• 50. Bernal C, Aguayo F, Villarroel C, Vargas M, Díaz I, Ossandon FJ, Santibáñez E, Palma M, Aravena E, Barrientos C, Corvalan AH. Reprimo as a potential biomarker for early detection in gastric cancer. Clin Cancer Res. 2008; 14:6264–9. https://doi.org/10.1158/1078-0432.CCR-07-4522 [PubMed]
• 51. Hwang JA, Kim Y, Hong SH, Lee J, Cho YG, Han JY, Kim YH, Han J, Shim YM, Lee YS, Kim DH. Epigenetic inactivation of heparan sulfate (glucosamine) 3-O-sulfotransferase 2 in lung cancer and its role in tumorigenesis. PLoS One. 2013; 8:e79634. https://doi.org/10.1371/journal.pone.0079634 [PubMed]
• 52. Wang K, Liu Y, Lu G, Xiao J, Huang J, Lei L, Peng J, Li Y, Wei S. A functional methylation signature to predict the prognosis of Chinese lung adenocarcinoma based on TCGA. Cancer Med. 2022; 11:281–94. https://doi.org/10.1002/cam4.4431 [PubMed]
• 53. Dong M, Yang Z, Li X, Zhang Z, Yin A. Screening of Methylation Gene Sites as Prognostic Signature in Lung Adenocarcinoma. Yonsei Med J. 2020; 61:1013–23. https://doi.org/10.3349/ymj.2020.61.12.1013 [PubMed]
• 54. Cai Q, He B, Xie H, Zhang P, Peng X, Zhang Y, Zhao Z, Wang X. Identification of a novel prognostic DNA methylation signature for lung adenocarcinoma based on consensus clustering method. Cancer Med. 2020; 9:7488–502. https://doi.org/10.1002/cam4.3343 [PubMed]