www.aging-us.com AGING 2020, Vol. 12, No. 11 Research Paper Epigenome-wide gene–age interaction analysis reveals reversed effects of PRODH DNA methylation on survival between young and elderly early-stage NSCLC patients Chao Chen1,2,*, Yongyue Wei1,2,3,*, Liangmin Wei1, Jiajin Chen1, Xin Chen1, Xuesi Dong1,4, Jieyu He1, Lijuan Lin1,3, Ying Zhu1, Hui Huang1, Dongfang You1,3, Linjing Lai1, Sipeng Shen1,2,3, Weiwei Duan1,5, Li Su2,3, Andrea Shafer6, Thomas Fleischer7, Maria Moksnes Bjaanæs7, Anna Karlsson8, Maria Planck8, Rui Wang9, Johan Staaf8, Åslaug Helland7,10, Manel Esteller11,12,13,14, Ruyang Zhang1,2,3,9, Feng Chen1,2,15,16, David C. Christiani3,6,# 1 Department of Biostatistics, Center for Global Health, School of Public Health, Nanjing Medical University, Nanjing 211166, Jiangsu, China 2 China International Cooperation Center for Environment and Human Health, Nanjing Medical University, Nanjing 211166, Jiangsu, China 3 Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA 4 Department of Epidemiology and Biostatistics, School of Public Health, Southeast University, Nanjing 210009, Jiangsu, China 5 Department of Bioinformatics, School of Biomedical Engineering and Informatics, Nanjing Medical University, Nanjing 211166, Jiangsu, China 6 Pulmonary and Critical Care Division, Department of Medicine, Massachusetts General Hospital and Harvard Medical School, Boston, MA 02114, USA 7 Department of Cancer Genetics, Institute for Cancer Research, Oslo University Hospital, Oslo 0424, Norway 8 Division of Oncology and Pathology, Department of Clinical Sciences Lund and CREATE Health Strategic Center for Translational Cancer Research, Lund University, Lund 22381, Skane, Sweden 9 Department of Medical Oncology, Jinling Hospital, School of Medicine, Nanjing University, Nanjing 210002, Jiangsu, China 10 Institute of Clinical Medicine, University of Oslo, Oslo 0424, Norway 11 Josep Carreras Leukaemia Research Institute, Badalona, Barcelona 08021, Catalonia, Spain 12 Centro de Investigacion Biomedica en Red Cancer, Madrid 28029, Spain 13 Institucio Catalana de Recerca i Estudis Avançats, Barcelona 08010, Catalonia, Spain 14 Physiological Sciences Department, School of Medicine and Health Sciences, University of Barcelona, Barcelona 08007, Catalonia, Spain 15 State Key Laboratory of Reproductive Medicine, Nanjing Medical University, Nanjing 211166, Jiangsu, China 16 Jiangsu Key Lab of Cancer Biomarkers, Prevention and Treatment, Cancer Center, Collaborative Innovation Center for Cancer Personalized Medicine, Nanjing Medical University, Nanjing 211166, Jiangsu, China *Equal contribution # Senior author Correspondence to: Ruyang Zhang, Feng Chen; email: zhangruyang@njmu.edu.cn, fengchen@njmu.edu.cn Keywords: overall survival, DNA methylation, methylation–age interaction analysis, non-small cell lung cancer, aging Received: November 13, 2019 Accepted: April 27, 2020 Published: June 8, 2020 Copyright: Chen et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY 3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. www.aging-us.com 10642 AGING ABSTRACT DNA methylation changes during aging, but it remains unclear whether the effect of DNA methylation on lung cancer survival varies with age. Such an effect could decrease prediction accuracy and treatment efficacy. We performed a methylation–age interaction analysis using 1,230 early-stage lung adenocarcinoma patients from five cohorts. A Cox proportional hazards model was used to investigate lung adenocarcinoma and squamous cell carcinoma patients for methylation–age interactions, which were further confirmed in a validation phase. We identified one adenocarcinoma-specific CpG probe, cg14326354PRODH, with effects significantly modified by age (HRinteraction = 0.989; 95% CI: 0.986–0.994; P = 9.18×10–7). The effect of low methylation was reversed for young and elderly patients categorized by the boundary of 95% CI standard (HRyoung = 2.44; 95% CI: 1.26–4.72; P = 8.34×10-3; HRelderly = 0.58; 95% CI: 0.42–0.82; P = 1.67×10-3). Moreover, there was an antagonistic interaction between low cg14326354PRODH methylation and elderly age (HRinteraction = 0.21; 95% CI: 0.11–0.40; P = 2.20×10−6). In summary, low methylation of cg14326354PRODH might benefit survival of elderly lung adenocarcinoma patients, providing new insight to age-specific prediction and potential drug targeting. INTRODUCTION However, it remains largely unclear whether alterations of methylation patterns resulting from aging, Population aging has resulted in a rapid increase in lung accumulating environmental exposures throughout life cancer cases as well as corresponding surgeries among [14], and other events also have varied effects on cancer elderly patients [1]. Indeed, the median age at diagnosis survival during aging. Such phenomena may further of lung cancer is 70 years old [2]. Further, lung cancer explain the increased alteration of cancer mortality risk leads as a cause of cancer deaths among men ≥40 years with age and may increase the effectiveness of cancer old and women ≥60 years old [3]. prediction and treatment. Progression of lung cancer is, in part, due to accumulation We hypothesized that the methylation effect on cancer of genomic instability as well as age-related declines in survival changes during aging. Thus, identifying age- system integrity and function [4]. Thus, even for specific signatures will be critical for prognosis individuals exposed to similar levels of risk factors, lung prediction, underpinning potential preventative cancer severity can vary as a function of individual strategies, and improving survival for elderly patients. differences in aging. Therefore, compared to predictive However, most epigenome-wide association studies are guidance for the overall population, effective predictive designed to identify main effects of DNA methylation guidance for age-specific populations, especially elderly and fail to provide knowledge about changes in patients, is needed to better guide postoperative treatment epigenetic effects during aging. Thus, we performed an and improve survival. Developing such guidance epigenome-wide methylation–age interaction analysis to necessitates identifying exclusive prognostic indicators of identify age-specific, prognosis-associated epigenetic lung cancer for the elderly. biomarkers using NSCLC patients from four cohorts, along with an independent population from The Cancer Epigenetic mechanisms represent the molecular Genome Atlas (TCGA) to confirm our results. interface mediating gene–environment interactions throughout the lifecycle [5]. DNA methylation, a RESULTS reversible epigenetic modification, correlates with tumor prognosis in almost all cancers including non- After quality control (QC) procedures, 1,230 lung small cell lung cancer (NSCLC) [6–9]. DNA adenocarcinoma (LUAD) and lung squamous cell methylation events may potentially be cancer bio- carcinoma (LUSC) patients with 311,891 CpG probes markers as well as therapeutic targets to improve cancer remained for subsequent association analysis. There treatment [10]. were 613 (NLUAD = 492; NLUSC = 121) patients in the discovery phase, and 617 (NLUAD = 332; NLUSC = 285) Alterations to DNA methylation often occur during patients in the validation phase. The average age was aging [11]. One of these alterations, known as 66.4 and 66.5 years for patients in the discovery and “epigenetic drift”, may further contribute to validation phases, respectively. Most NSCLC patients tumorigenesis in the elderly [12]. Changes in DNA were in stage I (77.5% in discovery; 63.7% in methylation also can contribute to senescence [13]. validation) (Supplementary Table 1). www.aging-us.com 10643 AGING We only observed two significant methylation–age With increased age, there was an increased protective interactions for LUAD patients in the discovery phase effect for low methylation of cg14326354PRODH on (Figure 1, Supplementary Figure 1, Supplementary LUAD survival (Figure 2A, Supplementary Figure 4). Table 2), and none for LUSC patients. Results of the Thus, age was a modifier of the association between epigenome-wide association study are shown in cg14326354PRODH and survival. To better understand the Supplementary Materials 1 and 2. In the validation interaction between DNA methylation and age, patients phase, only one LUAD-specific CpG probe, located in were categorized into young and elderly groups based on proline dehydrogenase 1 (PRODH) (Supplementary the boundary of 95% CI (BoCI) of HR (<57 vs >65 years Table 3), remained significant. in Figure 2A) or the United Nations (UN) standard (≤65 vs >65 years). The BoCI standard provided stable results Low methylation of cg14326354PRODH interacted with in both phases as well as combined data (Supplementary age to affect survival of patients (discovery phase: Table 6), with varied effects of cg14326354PRODH hazard ratio (HR)interaction = 0.982; 95% CI: 0.976– methylation across different age groups. Low methy- 0.989; P = 1.11×10–7; validation phase: HRinteraction = lation of cg14326354PRODH showed a risk effect on 0.981; 95% CI: 0.966–0.997; P = 0.0202; combined survival for young patients (HRBoCI = 1.20; 95% CI: data: HRinteraction = 0.989; 95% CI: 0.986–0.994; P = 1.03–1.40; P = 1.97×10–2; HRUN = 1.10; 95% CI: 0.99– 9.18×10–7). Further, the robustly significant interaction 1.22; P = 8.71×10–2) but benefited survival of elderly effect was confirmed in sensitivity analysis by LUAD patients (HRBoCI = 0.81; 95% CI: 0.75–0.88; P = removing outliers in methylation data (Supplementary 5.38×10–7; HRUN = 0.81; 95% CI: 0.75–0.88; P = Table 4). When using leave-one-out method for 5.38×10–7) (Figure 2B). Kaplan-Meier curves also validation, the interaction remained significant confirmed reversed effect patterns across age groups (Supplementary Figure 2). Moreover, meta-analysis based on BoCI standard (HRyoung = 2.44; 95% CI: 1.26– also exhibited significant (HRinteraction = 0.983; 95% CI: 4.72; P = 8.34×10-3; HRelderly = 0.58; 95% CI: 0.42–0.82; 0.976–0.990; P = 3.95×10–6) and homogenous P = 1.67×10-3), with methylation groups defined by (PHeterogeneity = 0.97) interaction effects across five median values. There was significant heterogeneity of the cohorts (Supplementary Figure 3). Based on stratified low cg14326354PRODH methylation effect between young analysis by smoking status, sex, clinical stage, and and elderly patients (I2 = 93.03%, Q = 14.35, P = study cohort, there was no significant heterogeneity of 1.52×10-4) (Figure 2C). These results indicated that interaction effect between subgroups of any of these elderly LUAD patients had better survival with lower covariates (Supplementary Table 5). methylation of cg14326354PRODH. Figure 1. Flow chart of study design and statistical analyses. www.aging-us.com 10644 AGING In addition, we evaluated the joint effect of 4.35×10−4), and the main effect of elderly age was HR = cg14326354PRODH methylation level (low vs high) and 3.18 (95% CI: 1.85–5.46, P = 2.64×10−5). However, the age (elderly vs young) on LUAD survival (Table 1). joint effect was HR = 1.86 (95% CI: 1.08–3.19, P = The group with the best survival (young patients with 2.42×10−2), which was less than the product of the two high methylation) was used as the reference to evaluate main effects (2.84×3.18 = 9.03). This result indicates an effects of low methylation, elderly age, and their antagonistic interaction between low cg14326354PRODH interaction. The main effect of low cg14326354PRODH methylation and elderly age (HRinteraction = 0.21; 95% methylation was HR = 2.84 (95% CI: 1.59–5.08, P = CI: 0.11–0.40; P = 2.20×10−6). Figure 2. DNA methylation and age interaction on survival of lung adenocarcinoma (LUAD) patients. (A) Hazard ratio (HR) of cg14326354PRODH 5% per decrement of methylation level among different aged patients. The 95% confidence interval (95% CI) band of HR for patients aged <57 or >65 years is statistically significant. Top histogram shows distribution of age. (B) Forest plots of HR of cg14326354PRODH 5% per decrement of methylation level in young and elderly LUAD patients, categorized based on boundary of 95% CI (BoCI) and 1956 United Nations standard. (C) Kaplan-Meier survival curves of low and high methylation groups (categorized by median value) among young and elderly LUAD patients defined using BoCI standard. Pheterogeneity was used to evaluate heterogeneity of HRs across age groups. www.aging-us.com 10645 AGING Table 1. Joint effect and interaction of low methylation and elderly age on the prognosis of early-stage lung adenocarcinoma (LUAD). Effect type a Elderly b Low methylation Number Death Crude mortality HR (95% CI) a Pa No No 75 17 22.67% Ref. Main effect 1 No Yes 70 33 47.14% 2.8398 (1.5876,5.0798) 4.35×10-4 Main effect 2 Yes No 217 98 45.16% 3.1804 (1.8542,5.4553) 2.64×10-5 Joint effect Yes Yes 222 70 31.53% 1.8590 (1.0840,3.1890) 0.0242 Interaction c 0.2058 (0.1070,0.3961) 2.20×10-6 a Patients categorized into two groups (low vs high) by medium of cg14326354PRODH methylation level. Classification criteria of age were based on boundary of 95% confidence interval (CI) standard (young: <57 years; elderly >65 years). b Main effects of low methylation and elderly age and their joint effect and interaction were derived from Cox proportional hazards model adjusted for covariates. c Interaction = Joint effect ÷ (Main effect 1 × Main effect 2). 0.2058 ≈ 1.8590 ÷ (3.1804 × 2.8398). Further, cis-regulation and genome-wide trans- increasing treatment efficiency based on age-specific regulation analyses were conducted in the TCGA drug targeting. cohort. We observed significant correlation between cg14326354PRODH and PRODH expression (r = –0.23; P PRODH is located in chromosome 22q11.2, a region = 3.38 × 10-5) in LUAD patients (Figure 3A), indicating often deleted in various human tumors. This gene that cg14326354PRODH cis-regulated gene expression. encodes a mitochondrial inner membrane-associated Moreover, genome-wide trans-regulation analysis enzyme that acts as a tumor suppressor in vitro and in revealed that expression of 821 genes was significantly vivo [17]. However, PRODH plays a paradoxical role in correlated with methylation level of cg14326354PRODH tumors. Hypoxia and nutrient depletion are important (Supplementary Material 3, Figure 3B). KEGG characteristics of the tumor microenvironment, where enrichment analysis based on 821 trans-regulated genes PRODH may serve as a tumor survival factor [18]. showed several significant immune- or inflammation- Indeed, PRODH supports tumor metastasis formation, related pathways, such as chemokine signaling, T cell and inhibiting its activity impairs cancer cell growth, receptor and B cell receptor signaling, and cellular indicating PRODH is a potential drug target [19]. A pathways such as cell differentiation and cell cycle metabolic enzyme, PRODH can catabolize proline to (Figure 3C). Notably, these trans-regulated genes were pyrroline-5-carboxylate (P5C). The process can donate also enriched in senescence-related pathways (e.g., electrons that enter the electron transport chain to cellular senescence) and cancer-related pathways (e.g., produce reactive oxygen species (ROS), which then NF-κB signaling). participate in protective autophagy rather than apoptotic cell death [18]. Because tumor mutational burden (TMB) serves as a biomarker to select patients who might benefit from Autophagy, a self-digestion process, plays an important immune checkpoint inhibitors [15], we also evaluated role in maintaining intracellular homeostasis. association between TMB and cg14326354PRODH as Autophagy can clear intracellular abnormally folded well as PRODH expression. TMB of each sample is protein and dysfunctional organelles, inhibit cell stress shown in Supplementary Material 4. cg14326354PRODH response, and prevent genetic damage in early phases of was positively correlated with TMB (r = 0.23; P = tumorigenesis. However, autophagy helps tumor cells 4.04×10-5), while PRODH expression was negatively survive nutritional deficiencies and hypoxic conditions correlated with TMB (r = –0.22; P = 6.62×10-5) when tumors develop and accumulate more mutations (Supplementary Figure 5). to promote malignant progression [20]. Further, tumor- surrounding normal cells, which are active and essential DISCUSSION parts of the microenvironment, support tumor proliferation by autophagy. Besides, autophagy in In this two-stage study using five independent cohorts, distant organs may also support growth of tumor tissue we systematically investigated methylation–age [21]. Additionally, autophagy can act as a mechanism of interactions on an epigenome-wide scale. Our results tumor resistance to chemotherapy agents and lead to show an antagonistic interaction between elderly age antagonistic effects of gefitinib combined with cisplatin and low methylation of cg14326354PRODH, indicating in NSCLC treatment, which may contribute to poor opportunities for epi-drug intervention due to the therapeutic effectiveness and patient prognosis [22, 23]. inherent reversibility of epigenetic events [16] and Further, our results suggest that low methylation of www.aging-us.com 10646 AGING cg14326354PRODH may potentially promote PRODH these cells may protect against transformation into expression, further heighten autophagy to some extent primary tumors [27]. A previous study also shows that [24], and then result in poor prognosis (Figure 4). p53 function declines during aging [28] and might promote tumor growth and decrease cancer survival Age is an independent risk factor for lung cancer [29]. Moreover, senescent cells can promote survival [25]. Individual aging implies a higher reprogramming of tumor stem cells, increase cancer abundance of senescent cells in aged tissues and stemness, and accelerate tumor growth [30]. Thus, reflects an increase in the generation of senescent cells combined with our results, increased generation of [26]. At old age, senescent cells generate a pro- senescent may be relevant to poor NSCLC prognosis tumorigenic microenvironment, though at young age for elderly patients (Figure 4). Figure 3. Scatter plot of cis-regulation, circos plot of genome-wide trans-regulation analysis, and significant pathways of gene enrichment pathway analysis. (A) Correlation between DNA methylation of cg14326354PRODH and expression of PRODH. The r coefficient and P-value were derived from Pearson correlation analysis. Gene expression was log2-transformed before correlation analysis. (B) Circos plot of genome-wide trans-regulation analysis in the TCGA cohort. Blue points ordered by genomic position represent P-values of correlation between gene expression and methylation at cg14326354PRODH. Grey lines represent significant correlations with Bonferroni-adjusted P ≤ 0.05. (C) KEGG gene enrichment analysis of 821 trans-regulated genes correlated with cg14326354PRODH methylation. www.aging-us.com 10647 AGING Autophagy is reduced in aging, likely through several senescence and survival [38, 39], we also analyzed several mechanisms [31]. Lipofuscin produced during aging proliferation-associated genes retrieved from the KEGG can destroy the function of lysosomes, restricting database. Expression of these genes were significantly binding between autophages and lysosomes [32]. In correlated with cg14326354PRODH methylation and addition, expression of lysosome-associated membrane affected LUAD survival, including MKI67, BTG2, glycoprotein (LAMP2a), which assists autophagy, KIAA1524, and CDC123 (Supplementary Table 7). Our decreases during aging and thus can inhibit autophagy previous study of BTG2 expression and methylation [33]. Further, by guaranteeing stability of the cellular already indicated it is a prognostic biomarker of NSCLC proteome and proper organelle turnover, autophagy can [7]. These results also indicate the potential role of prevent or slow down aging and extend lifespan [34]. cg14326354PRODH in indirect induction of autophagy, The antagonistic effect exists between aging and the senescence, and survival. Further functional studies are autophagy level resulting from low methylation of warranted to elucidate the mechanism of cg14326354PRODH, in spite of the harmful effect of both, cg14326354PRODH and age interaction on LUAD survival. which could provide a possible mechanism of the cg14326354PRODH–age interaction (Figure 4). Age represents a complex surrogate for a host of underlying phenomena, although its measurement is The 821 significant trans-regulated genes we identified simple and accurate [40]. A previous study suggested were enriched in KEGG pathways including inflammation that gene–age interactions may partially be surrogates and immune-related pathways. Notably, cellular for gene–gene and gene–environment interactions senescence was involved in these pathways, again [41]. In a study investigating the efficacy of indicting potential indirect induction of cg14326354PRODH metronomic vinorelbine to treat patients with on senescence. Meanwhile, the NF-κB pathway, with the advanced unresectable NSCLC, age was an important ability to upregulate genes responsible for inflammation, factor that decreased treatment efficacy [42]. Our cell survival, proliferation, invasion, angiogenesis, and study might provide a novel explanation of age effects metastasis, often plays a critical role in initiation, on treatment efficacy from the cg14326354PRODH–age promotion, progression, and therapy resistance of cancers interaction perspective. Further clinical studies will [35, 36]. Further, NF-κB family members can activate or provide additional insight into cg14326354PRODH and inhibit signaling pathways, leading to induction of its age-specific effects in tumors, which may lead to autophagy or transcription of certain pro-autophagic- new age-specific biomarkers and therapeutic regulating genes [35], and can induce senescence [37]. strategies that improve prediction accuracy and Because cell proliferation can be associated with both treatment efficacy. Figure 4. Pathway of DNA methylation–age interaction effect on survival of lung adenocarcinomas (LUAD) patients. www.aging-us.com 10648 AGING Our study has several strengths. First, this is the first have implications for not only age-specific prediction of epigenome-wide study to investigate the interaction cancer survival, but also possible methylation-specific between DNA methylation and age on NSCLC survival, drug targeting. which provides new evidence to account for the missing heritability of complex diseases [43] and may further MATERIALS AND METHODS reveal the role of age in heterogeneity of NSCLC prognosis and treatment efficacy. Second, to identify Lung cancer study populations stable and reliable biomarkers, a two-stage study design along with Bonferroni correction and sensitivity Only early-stage (stage I–II) LUAD and LUSC patients analysis was used to exhaustively search for inter- were included in our study. DNA methylation data was actions, which is quite conservative in controlling for harmonized from five cohorts: Harvard, Spain, Norway, false positives. Finally, with a large sample size to Sweden, and TCGA. analyze DNA methylation–age interactions, our study has improved statistical power to identify complex Harvard associations with small–medium effect size. Since 1992, patients have been recruited at Massachusetts General Hospital (MGH) and Nonetheless, several limitations also need to be histologically confirmed as primary NSCLC [46]. We acknowledged. First, we did not observe robustly profiled 151 early-stage patients from this cohort. significant methylation–age interactions on survival for During curative surgery, tumor specimens were LUSC, which may be due to limited sample size and collected with complete resection and snap-frozen. A thus insufficient power. However, there was no MGH pathologist evaluated each specimen for tumor significant heterogeneous effect between LUAD and cell amount (tumor cellularity > 70%) and quality. LUSC groups (Supplementary Table 8). Second, the Specimens were classified histologically according to association was no longer significant in young LUAD World Health Organization (WHO) criteria. The patients when the analysis used UN standards to define Institutional Review Boards at Harvard T.H. Chan age groups. However, we still observed a significant School of Public Health and MGH approved the study. association in patients <57 years old. This effect might All patients provided written informed consent. be because >62% (240/385) of young patients defined using the UN standard (57–65 years) attenuated the Spain effect of cg14326354PRODH methylation. Therefore, high The Spain cohort included 226 early-stage NSCLC methylation of cg14326354PRODH might benefit survival patients recruited from eight sub-centers in 1991–2009 of young LUAD patients. Third, although widespread [47]. Tumor DNA was extracted from fresh-frozen methylation–age interactions may exist, we only tumor specimens and further checked for integrity and identified one interaction, which may be due to the most quantity. Patients provided written consent, and tumors conservative correction method used in the discovery were surgically collected. The study was approved by phase and limited statistical power in the validation the Bellvitge Biomedical Research Institute Institutional phase due to low event rate of survival time in the Review Boards. TCGA population. We may need longer time to follow- up early-stage patients in TCGA for their events to Norway occur. Nevertheless, the interaction between The Norway study population consisted of 133 early- cg14326354PRODH and age was successfully confirmed, stage NSCLC patients from Oslo University Hospital- indicating it was a conservative and robust association. Riks Hospitalet, Norway, in 2006–2011 [48]. Tumor Fourth, our analysis was based on the assumption of tissues were snap-frozen in liquid nitrogen and stored at linear additive interaction, and new statistical models –80°C until DNA isolation. The project was developed can be used to properly capture non-linear methylation– with approval of the Oslo University Institutional age interactions. Last, the cis-regulation pattern of Review Board and Regional Ethics Committee (S- cg14326354PRODH requires more biological evidence, 05307). All patients provided informed consent. although methylation is believed to play a crucial role in regulating gene expression [44] and further influence Sweden disease gene function [45]. Therefore, our findings Tumor DNA was collected from 103 early-stage NSCLC should be interpreted with caution, and functional patients, including 80 LUAD and 23 LUSC patients, at experiments are warranted to confirm these the Skane University Hospital in Lund, Sweden [49]. The associations. study was developed under the approval of the Regional Ethical Review Board in Lund, Sweden (registration no. In conclusion, low methylation of cg14326354PRODH 2004/762 and 2008/702). All patients provided written benefited survival of elderly LUAD patients. Our results informed consent. www.aging-us.com 10649 AGING TCGA using the R package survival [54]. Hazard ratio (HR) A total of 332 LUAD and 285 LUSC cases with full and 95% confidence interval (CI) were described per DNA methylation, survival time, and covariate data 5% level of methylation decrement. The P-value were included. Level-1 HumanMethylation450 DNA threshold for multiple testing was established using the methylation data of early-stage NSCLC patient were Bonferroni method, which set the significance level to downloaded on October 01, 2015. 0.05 divided by number of tests. This way, overall type I error was controlled at the 0.05 level. In our study, Quality control for DNA methylation data significance level of interaction analysis was defined as 1.60×10-07 = 0.05/311,891. Interactions with P ≤ DNA methylation was assessed using Illumina Infinium 1.60×10-07 were screened out and then further HumanMethylation450 BeadChips (Illumina Inc.). Raw confirmed in the validation phase. Robustly significant image data were imported into Genome Studio probes were retained if they fit both of the following Methylation ModuleV1.8 (Illumina Inc.) to calculate criteria: (i) interaction P ≤ 0.05, and (ii) direction of methylation signals and to perform normalization, interaction effects was consistent across both phases. In background subtraction, and QC. Unqualified probes sensitivity analysis, patients were excluded if their were excluded if they fit any of the following criteria: methylation values were out of range mean ± (i) failed detection (P > 0.05) in ≥5% samples, (ii) 3×standard deviation (SD) on logit 2-transformed scale. coefficient of variance (CV) <5%, (iii) all samples Genome-wide expression correlation analysis was methylated or all unmethylated, (iv) common single performed to identify potential trans-regulation genes in nucleotide polymorphisms located in probe sequence or TCGA. KEGG enrichment analysis of potential trans- in 10-bp flanking regions, (v) cross-reactive probes regulation genes (Bonferroni adjusted P < 0.05) was [50], or (vi) data did not pass QC in all cohorts. conducted using R package clusterProfiler [55]. Samples with >5% undetectable probes were excluded. Methylation signals were further processed for quantile Continuous variables were summarized as mean ± SD; normalization (betaqn function in R package minfi categorical variables were described as n (%). Kaplan- [51]), type I and II probe correction (BMIQ function in Meier survival curves were used to illustrate survival R package lumi [52]), and adjusted for batch effects difference between patients in low and high methylation (ComBat function in R package sva) [53]. Details of QC groups (categorized by median value). We used two process are described Supplementary Figure 6. classification criteria to define young and elderly patients: (i) the UN standard (1956) of 65 years old as Quality control for gene expression data the cut-off value to distinguish elderly and young people [56], (ii) and a cut-off value calculated based on The TCGA workgroup completed mRNA sequencing BoCI of HR of the CpG probe. All statistical analyses data processing and QC. Raw counts were normalized were performed in R version 3.5.1 (The R Foundation). using RNA-sequencing by expectation maximization (RSEM). Level-3 gene quantification data were Abbreviations downloaded from the TCGA data portal and were further checked for quality. Expression of genes was NSCLC: non-small cell lung cancer; LUSC: lung extracted and log2-transformed before analysis. squamous cell carcinoma; LUAD: lung Normalization results were then evaluated through adenocarcinoma; TCGA: The Cancer Genome Atlas; boxplots of the distribution of gene expression across all BoCI: boundary of 95% CI; UN: United Nations; QC: samples (Supplementary Figure 7). quality control; HR: hazard ratio; CI: confidence interval; SD: standard deviation; PRODH: proline Statistical analysis dehydrogenase. Statistical analysis flow is presented in Figure 1. AUTHOR CONTRIBUTIONS Patients from Harvard, Spain, Norway, and Sweden study cohorts were assigned to the discovery phase, C.C., Y.W., R.Z., F.C., and D.C.C. contributed to the while patients in TCGA were assigned to the validation study design. R.Z., L.S., A.S., T.F., M.M.B., A.K., phase. M.P., J.S., Å.H., M.E., and D.C.C. contributed to data collection. C.C., Y.W., L.Lai, R.Z., and D.C.C. In the discovery phase, histology-stratified analysis and performed statistical analysis and interpretation and Cox proportional hazards model adjusted for age, drafted the manuscript. L.W., J.C., X.C., X.D., J.H., smoking status, sex, clinical stage, and study center L.Lin, Y.Z., H.H., D.Y., S.S., W.D., L.S., and R.W. were applied to test the methylation–age interaction revised the manuscript. All authors contributed to effect on overall survival in LUAD and LUSC patients critical revision of the manuscript and approved its final www.aging-us.com 10650 AGING version. Financial support and study supervision were affects healthy aging and promotes age-related health provided by R.Z., Y.W., F.C., and D.C.C. problems. Clin Epigenetics. 2011; 2:349–60. https://doi.org/10.1007/s13148-011-0042-6 ACKNOWLEDGMENTS PMID:22704347 6. Wei Y, Liang J, Zhang R, Guo Y, Shen S, Su L, Lin X, The authors thank all participants enrolled in this study. Moran S, Helland Å, Bjaanæs MM, Karlsson A, Planck M, Esteller M, et al. Epigenetic modifications in KDM CONFLICTS OF INTEREST lysine demethylases associate with survival of early- stage NSCLC. Clin Epigenetics. 2018; 10:41. The authors declare no potential conflicts of interest. https://doi.org/10.1186/s13148-018-0474-3 PMID:29619118 FUNDING 7. Shen S, Zhang R, Guo Y, Loehrer E, Wei Y, Zhu Y, This study was funded by National Key Research and Yuan Q, Moran S, Fleischer T, Bjaanaes MM, Development Program of China (2016YFE0204900 to Karlsson A, Planck M, Staaf J, et al. A multi-omic F.C.), US National Institutes of Health (CA209414, study reveals BTG2 as a reliable prognostic marker CA092824, and ES000002 to D.C.C.), National Natural for early-stage non-small cell lung cancer. Mol Science Foundation of China (81530088 to F.C. and Oncol. 2018; 12:913–24. 81973142 to Y.W.), Natural Science Foundation of https://doi.org/10.1002/1878-0261.12204 Jiangsu Province (BK20191354 to R.Z.), Natural PMID:29656435 Science Foundation of the Jiangsu Higher Education Institutions of China (18KJB310011 to R.Z.), China 8. Zhang R, Lai L, He J, Chen C, You D, Duan W, Dong X, Postdoctoral Science Foundation (2018M633767 to Zhu Y, Lin L, Shen S, Guo Y, Su L, Shafer A, et al. EGLN2 R.Z.), and Priority Academic Program Development of DNA methylation and expression interact with HIF1A Jiangsu Higher Education Institutions. R.Z. was to affect survival of early-stage NSCLC. Epigenetics. partially supported by the Outstanding Young Teachers 2019; 14:118–29. Training Program of Nanjing Medical University. https://doi.org/10.1080/15592294.2019.1573066 PMID:30665327 REFERENCES 9. Zhang R, Lai L, Dong X, He J, You D, Chen C, Lin L, Zhu Y, Huang H, Shen S, Wei L, Chen X, Guo Y, et al. SIPA1L3 1. Miao R, Ge C, Zhang X, He Y, Ma X, Xiang X, Gu J, Fu Y, methylation modifies the benefit of smoking cessation Qu K, Liu C, Wu Q, Lin T. Combined eight-long on lung adenocarcinoma survival: an epigenomic- noncoding RNA signature: a new risk score predicting smoking interaction analysis. Mol Oncol. 2019; prognosis in elderly non-small cell lung cancer patients. 13:1235–48. Aging (Albany NY). 2019; 11:467–79. https://doi.org/10.1002/1878-0261.12482 https://doi.org/10.18632/aging.101752 PMID:30924596 PMID:30659574 10. Jones PA, Issa JP, Baylin S. Targeting the cancer 2. Miller KD, Siegel RL, Lin CC, Mariotto AB, Kramer JL, epigenome for therapy. Nat Rev Genet. 2016; Rowland JH, Stein KD, Alteri R, Jemal A. Cancer 17:630–41. treatment and survivorship statistics, 2016. CA Cancer J https://doi.org/10.1038/nrg.2016.93 Clin. 2016; 66:271–89. PMID:27629931 https://doi.org/10.3322/caac.21349 PMID:27253694 11. Day K, Waite LL, Thalacker-Mercer A, West A, Bamman MM, Brooks JD, Myers RM, Absher D. Differential DNA 3. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. methylation with age displays both common and CA Cancer J Clin. 2019; 69:7–34. dynamic features across human tissues that are https://doi.org/10.3322/caac.21551 PMID:30620402 influenced by CpG landscape. Genome Biol. 2013; 4. Levine ME, Hosgood HD, Chen B, Absher D, Assimes T, 14:R102. Horvath S. DNA methylation age of blood predicts https://doi.org/10.1186/gb-2013-14-9-r102 future onset of lung cancer in the women’s health PMID:24034465 initiative. Aging (Albany NY). 2015; 7:690–700. 12. Zheng SC, Widschwendter M, Teschendorff AE. https://doi.org/10.18632/aging.100809 Epigenetic drift, epigenetic clocks and cancer risk. PMID:26411804 Epigenomics. 2016; 8:705–19. 5. Liu L, van Groen T, Kadish I, Li Y, Wang D, James SR, https://doi.org/10.2217/epi-2015-0017 Karpf AR, Tollefsbol TO. Insufficient DNA methylation PMID:27104983 www.aging-us.com 10651 AGING 13. Fraga MF, Esteller M. Epigenetics and aging: the Juhász G, Bilder D, Brech A, Stenmark H, Rusten TE. targets and the marks. Trends Genet. 2007; 23:413–18. Microenvironmental autophagy promotes tumour https://doi.org/10.1016/j.tig.2007.05.008 growth. Nature. 2017; 541:417–20. PMID:17559965 https://doi.org/10.1038/nature20815 14. Christensen BC, Houseman EA, Marsit CJ, Zheng S, PMID:28077876 Wrensch MR, Wiemels JL, Nelson HH, Karagas MR, 22. Amaravadi RK, Lippincott-Schwartz J, Yin XM, Weiss Padbury JF, Bueno R, Sugarbaker DJ, Yeh RF, Wiencke WA, Takebe N, Timmer W, DiPaola RS, Lotze MT, White JK, Kelsey KT. Aging and environmental exposures alter E. Principles and current strategies for targeting tissue-specific DNA methylation dependent upon CpG autophagy for cancer treatment. Clin Cancer Res. 2011; island context. PLoS Genet. 2009; 5:e1000602. 17:654–66. https://doi.org/10.1371/journal.pgen.1000602 https://doi.org/10.1158/1078-0432.CCR-10-2634 PMID:19680444 PMID:21325294 15. Rizvi H, Sanchez-Vega F, La K, Chatila W, Jonsson P, 23. Liu JT, Li WC, Gao S, Wang F, Li XQ, Yu HQ, Fan LL, Wei Halpenny D, Plodkowski A, Long N, Sauter JL, W, Wang H, Sun GP. Autophagy inhibition overcomes Rekhtman N, Hollmann T, Schalper KA, Gainor JF, et al. the antagonistic effect between gefitinib and cisplatin Molecular determinants of response to anti- in epidermal growth factor receptor mutant non– programmed cell death (PD)-1 and anti-programmed small-cell lung cancer cells. Clin Lung Cancer. 2015; death-ligand 1 (PD-L1) blockade in patients with non- 16:e55–66. small-cell lung cancer profiled with targeted next- https://doi.org/10.1016/j.cllc.2015.03.006 generation sequencing. J Clin Oncol. 2018; 36:633–41. PMID:25979647 https://doi.org/10.1200/JCO.2017.75.3384 24. Liu W, Phang JM. Proline dehydrogenase (oxidase) in PMID:29337640 cancer. Biofactors. 2012; 38:398–406. 16. Wright J. Epigenetics: reversible tags. Nature. 2013; https://doi.org/10.1002/biof.1036 PMID:22886911 498:S10–11. 25. Torre LA, Siegel RL, Jemal A. Lung Cancer Statistics. In: https://doi.org/10.1038/498S10a PMID:23803942 Ahmad A, Gadgeel S, eds. Lung Cancer and 17. Liu W, Phang JM. Proline dehydrogenase (oxidase), a Personalized Medicine: Current Knowledge and mitochondrial tumor suppressor, and autophagy under Therapies. Cham: Springer International Publishing, the hypoxia microenvironment. Autophagy. 2012; 2016. pp. 1–19. 8:1407–09. 26. Collado M, Blasco MA, Serrano M. Cellular senescence https://doi.org/10.4161/auto.21152 in cancer and aging. Cell. 2007; 130:223–33. PMID:22885468 https://doi.org/10.1016/j.cell.2007.07.003 18. Liu W, Glunde K, Bhujwalla ZM, Raman V, Sharma A, PMID:17662938 Phang JM. Proline oxidase promotes tumor cell survival 27. Schosserer M, Grillari J, Breitenbach M. The dual role in hypoxic tumor microenvironments. Cancer Res. of cellular senescence in developing tumors and their 2012; 72:3677–86. response to cancer therapy. Front Oncol. 2017; 7:278. https://doi.org/10.1158/0008-5472.CAN-12-0080 https://doi.org/10.3389/fonc.2017.00278 PMID:22609800 PMID:29218300 19. Elia I, Broekaert D, Christen S, Boon R, Radaelli E, Orth 28. Feng Z, Hu W, Teresky AK, Hernando E, Cordon-Cardo MF, Verfaillie C, Grünewald TG, Fendt SM. Proline C, Levine AJ. Declining p53 function in the aging metabolism supports metastasis formation and could process: a possible mechanism for the increased tumor be inhibited to selectively target metastasizing cancer incidence in older populations. Proc Natl Acad Sci USA. cells. Nat Commun. 2017; 8:15267. 2007; 104:16633–38. https://doi.org/10.1038/ncomms15267 https://doi.org/10.1073/pnas.0708043104 PMID:28492237 PMID:17921246 20. Sun K, Deng W, Zhang S, Cai N, Jiao S, Song J, Wei L. 29. Wörmann SM, Song L, Ai J, Diakopoulos KN, Kurkowski Paradoxical roles of autophagy in different stages of MU, Görgülü K, Ruess D, Campbell A, Doglioni C, Jodrell tumorigenesis: protector for normal or cancer cells. D, Neesse A, Demir IE, Karpathaki AP, et al. Loss of P53 Cell Biosci. 2013; 3:35. function activates JAK2-STAT3 signaling to promote https://doi.org/10.1186/2045-3701-3-35 pancreatic tumor growth, stroma modification, and PMID:24016776 gemcitabine resistance in mice and is associated with 21. Katheder NS, Khezri R, O’Farrell F, Schultz SW, Jain A, patient survival. Gastroenterology. 2016; 151:180– Rahman MM, Schink KO, Theodossiou TA, Johansen T, 193.e12. www.aging-us.com 10652 AGING https://doi.org/10.1053/j.gastro.2016.03.010 https://doi.org/10.1172/JCI70355 PMID:27003603 PMID:24231352 30. Milanovic M, Fan DNY, Belenki D, Däbritz JHM, Zhao Z, 39. Rosenwald A, Wright G, Wiestner A, Chan WC, Connors Yu Y, Dörr JR, Dimitrova L, Lenze D, Monteiro Barbosa JM, Campo E, Gascoyne RD, Grogan TM, Muller- IA, Mendoza-Parra MA, Kanashova T, Metzner M, et al. Hermelink HK, Smeland EB, Chiorazzi M, Giltnane JM, Senescence-associated reprogramming promotes Hurt EM, et al. The proliferation gene expression cancer stemness. Nature. 2018; 553:96–100. signature is a quantitative integrator of oncogenic https://doi.org/10.1038/nature25167 events that predicts survival in mantle cell lymphoma. PMID:29258294 Cancer Cell. 2003; 3:185–97. 31. Zhang Y, Wang C, Zhou J, Sun A, Hueckstaedt LK, Ge J, https://doi.org/10.1016/s1535-6108(03)00028-x PMID:12620412 Ren J. Complex inhibition of autophagy by mitochondrial aldehyde dehydrogenase shortens 40. Shi G, Gu CC, Kraja AT, Arnett DK, Myers RH, Pankow lifespan and exacerbates cardiac aging. Biochim JS, Hunt SC, Rao DC. Genetic effect on blood pressure Biophys Acta Mol Basis Dis. 2017; 1863:1919–32. is modulated by age: the hypertension genetic https://doi.org/10.1016/j.bbadis.2017.03.016 epidemiology network study. Hypertension. 2009; PMID:28347844 53:35–41. 32. Terman A, Kurz T, Navratil M, Arriaga EA, Brunk UT. https://doi.org/10.1161/HYPERTENSIONAHA.108.12 0071 PMID:19029486 Mitochondrial turnover and aging of long-lived postmitotic cells: the mitochondrial-lysosomal axis 41. Bae HT, Sebastiani P, Sun JX, Andersen SL, Daw EW, theory of aging. Antioxid Redox Signal. 2010; Terracciano A, Ferrucci L, Perls TT. Genome-wide 12:503–35. association study of personality traits in the long life https://doi.org/10.1089/ars.2009.2598 family study. Front Genet. 2013; 4:65. PMID:19650712 https://doi.org/10.3389/fgene.2013.00065 33. Cuervo AM, Dice JF. Age-related decline in PMID:23658558 chaperone-mediated autophagy. J Biol Chem. 2000; 42. D’Ascanio M, Pezzuto A, Fiorentino C, Sposato B, Bruno 275:31505–13. P, Grieco A, Mancini R, Ricci A. Metronomic https://doi.org/10.1074/jbc.M002102200 chemotherapy with vinorelbine produces clinical PMID:10806201 benefit and low toxicity in frail elderly patients affected 34. Pyo JO, Yoo SM, Ahn HH, Nah J, Hong SH, Kam TI, Jung by advanced non-small cell lung cancer. Biomed Res S, Jung YK. Overexpression of Atg5 in mice activates Int. 2018; 2018:6278403. autophagy and extends lifespan. Nat Commun. 2013; https://doi.org/10.1155/2018/6278403 4:2300. PMID:30225260 https://doi.org/10.1038/ncomms3300 PMID:23939249 43. Trerotola M, Relli V, Simeone P, Alberti S. Epigenetic 35. Trocoli A, Djavaheri-Mergny M. The complex interplay inheritance and the missing heritability. Hum between autophagy and NF-κB signaling pathways in Genomics. 2015; 9:17. cancer cells. Am J Cancer Res. 2011; 1:629–49. https://doi.org/10.1186/s40246-015-0041-3 PMID:21994903 PMID:26216216 36. Chaturvedi MM, Sung B, Yadav VR, Kannappan R, 44. Bird A. Perceptions of epigenetics. Nature. 2007; Aggarwal BB. NF-κB addiction and its role in 447:396–98. cancer: ‘One size does not fit all’. Oncogene. 2011; https://doi.org/10.1038/nature05913 30:1615–30. PMID:17522671 https://doi.org/10.1038/onc.2010.566 45. Schübeler D. Function and information content of DNA PMID:21170083 methylation. Nature. 2015; 517:321–26. 37. Salminen A, Kaarniranta K. NF-kappaB signaling in the https://doi.org/10.1038/nature14192 aging process. J Clin Immunol. 2009; 29:397–405. PMID:25592537 https://doi.org/10.1007/s10875-009-9296-6 46. Heist RS, Zhou W, Chirieac LR, Cogan-Drew T, Liu G, Su PMID:19408108 L, Neuberg D, Lynch TJ, Wain JC, Christiani DC. MDM2 38. Mondal AM, Horikawa I, Pine SR, Fujita K, Morgan KM, polymorphism, survival, and histology in early-stage Vera E, Mazur SJ, Appella E, Vojtesek B, Blasco MA, non-small-cell lung cancer. J Clin Oncol. 2007; Lane DP, Harris CC. P53 isoforms regulate aging- and 25:2243–47. tumor-associated replicative senescence in T https://doi.org/10.1200/JCO.2006.08.8914 lymphocytes. J Clin Invest. 2013; 123:5247–57. PMID:17538168 www.aging-us.com 10653 AGING 47. Sandoval J, Mendez-Gonzalez J, Nadal E, Chen G, 51. Fortin JP, Triche TJ Jr, Hansen KD. Preprocessing, Carmona FJ, Sayols S, Moran S, Heyn H, Vizoso M, normalization and integration of the illumina Gomez A, Sanchez-Cespedes M, Assenov Y, Müller F, et HumanMethylationEPIC array with minfi. al. A prognostic DNA methylation signature for stage I Bioinformatics. 2017; 33:558–60. non-small-cell lung cancer. J Clin Oncol. 2013; https://doi.org/10.1093/bioinformatics/btw691 31:4140–47. PMID:28035024 https://doi.org/10.1200/JCO.2012.48.5516 52. Du P, Kibbe WA, Lin SM. Lumi: a pipeline for processing PMID:24081945 illumina microarray. Bioinformatics. 2008; 24:1547–48. 48. Bjaanæs MM, Fleischer T, Halvorsen AR, Daunay A, https://doi.org/10.1093/bioinformatics/btn224 Busato F, Solberg S, Jørgensen L, Kure E, Edvardsen H, PMID:18467348 Børresen-Dale AL, Brustugun OT, Tost J, Kristensen V, 53. Marabita F, Almgren M, Lindholm ME, Ruhrmann S, Helland Å. Genome-wide DNA methylation analyses in Fagerström-Billai F, Jagodic M, Sundberg CJ, Ekström lung adenocarcinomas: association with EGFR, KRAS TJ, Teschendorff AE, Tegnér J, Gomez-Cabrero D. An and TP53 mutation status, gene expression and evaluation of analysis pipelines for DNA methylation prognosis. Mol Oncol. 2016; 10:330–43. profiling using the illumina HumanMethylation450 https://doi.org/10.1016/j.molonc.2015.10.021 BeadChip platform. Epigenetics. 2013; 8:333–46. PMID:26601720 https://doi.org/10.4161/epi.24008 49. Karlsson A, Jönsson M, Lauss M, Brunnström H, PMID:23422812 Jönsson P, Borg Å, Jönsson G, Ringnér M, Planck M, 54. Therneau T. (2015). A Package for Survival Analysis in Staaf J. Genome-wide DNA methylation analysis of S. version 2.38. lung carcinoma reveals one neuroendocrine and four adenocarcinoma epitypes associated with patient 55. Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R outcome. Clin Cancer Res. 2014; 20:6127–40. package for comparing biological themes among gene https://doi.org/10.1158/1078-0432.CCR-14-1087 clusters. OMICS. 2012; 16:284–87. PMID:25278450 https://doi.org/10.1089/omi.2011.0118 PMID:22455463 50. Chen YA, Lemire M, Choufani S, Butcher DT, Grafodatskaya D, Zanke BW, Gallinger S, Hudson TJ, 56. United Nations Department of Economic and Social Weksberg R. Discovery of cross-reactive probes and Affairs (1956). The aging of populations and its polymorphic CpGs in the illumina infinium economic and social implications. New York: United HumanMethylation450 microarray. Epigenetics. 2013; Nations. 8:203–09. https://doi.org/10.4161/epi.23470 PMID:23314698 www.aging-us.com 10654 AGING SUPPLEMENTARY MATERIALS Supplementary Figures Supplementary Figure 1. Manhattan plot of methylation–age interaction P-values derived from Cox proportional hazards model adjusted for age, smoking status, sex, clinical stage, and study cohort in lung adenocarcinoma (LUAD) patients in the discovery phase. Red line represents Bonferroni adjusted P ≤ 0.05. Blue dot represents one CpG probe identified in this study, with Bonferroni adjusted P ≤ 0.05 in the discovery phase, P ≤ 0.05 in the validation phase, and consistent effect direction across both phases. Supplementary Figure 2. Association results of lung adenocarcinoma (LUAD)-specific cg14326354 PRODH–age interaction using leave-one-out method for validation. Size of each box represents the sample size of each cohort. Hazard ratio (HR) of cg14326354PRODH 5% per decrement of methylation level and P-value of interaction term derived from the Cox proportional hazards model adjusted for age, smoking status, sex, clinical stage, and study cohort in LUAD patients when leaving one cohort out. www.aging-us.com 10655 AGING Supplementary Figure 3. Meta-analysis of interaction between DNA methylation of cg14326354PRODH and age for lung adenocarcinoma (LUAD) patients from five cohorts. Fixed effect and random effect models were both applied, and effect heterogeneity among five cohorts was tested. Size of each box represents the sample size of each cohort, and diamond represents the overall effect of cg14326354PRODH in five cohorts. Hazard ratio (HR) of cg14326354PRODH 5% per decrement of methylation level and P-value of interaction term derived from the Cox proportional hazards model adjusted for age, smoking status, sex, clinical stage, and study cohort in LUAD patients in each of the five cohorts. Supplementary Figure 4. Association results of the cg14326354PRODH effect on lung adenocarcinoma (LUAD) survival in different age subgroups. Size of each box represents the sample size of each cohort. Hazard ratio (HR) of cg14326354PRODH 5% per decrement of methylation level and P-value derived from the Cox proportional hazards model adjusted for age, smoking status, sex, clinical stage, and study cohort in LUAD patients in each age subgroup. www.aging-us.com 10656 AGING Supplementary Figure 5. Results of correlation analysis between tumor mutational burden (TMB) and PRODH methylation as well as expression. Correlation coefficient (r) and P-value were derived from Pearson correlation analysis. TMB as well as gene expression were log2-transformed before correlation analysis due to violation of normal distribution assumption before data transformation. Supplementary Figure 6. Quality control procedures for epigenome-wide DNA methylation data. www.aging-us.com 10657 AGING Supplementary Figure 7. Boxplot of gene expression distribution across all samples after normalization. Gene expression values were log2-transformed. www.aging-us.com 10658 AGING Supplementary Tables Supplementary Table 1. Demographic and clinical characteristics for early-stage non-small cell lung cancer (NSCLC) patients from five cohorts. Validation Combined Discovery phase phase dataset Variable Cohort 1: Cohort 2: Cohort 3: Cohort 4: Discovery: Cohort 5: Overall Harvard Spaina Norway Sweden All TCGA samples (N = 151) (N = 226) (N = 133) (N = 103) (N = 613) (N = 617) (N = 1230) 66.44±10.0 Age (years) 67.67±9.92 65.67±10.58 65.52±9.34 67.54±9.99 66.51±9.47 66.48±9.78 8 Sex, n (%) Female 67 (44.37) 105 (46.46) 71 (53.38) 54 (52.43) 297 (48.45) 255 (41.33) 552 (44.88) Male 84 (55.63) 121 (53.54) 62 (46.62) 49 (47.57) 316 (51.55) 362 (58.67) 678 (55.12) Smoking status, n (%) Never 18 (11.92) 30 (13.57) 17 (12.78) 18 (17.48) 83 (13.65) 55 (9.18) 138(11.22) Former 81 (53.64) 120 (54.30) 74 (55.64) 54 (52.43) 329 (54.11) 376 (62.77) 705 (58.41) Current 52 (34.44) 71 (32.13) 42 (31.58) 31 (30.10) 196 (32.24) 168 (28.05) 364 (30.16) Unknown 0 5 0 0 5 18 23 Clinical stage, n (%) I 104 (68.87) 183 (80.97) 93 (69.92) 95 (92.23) 475 (77.49) 393 (63.70) 868 (70.57) II 47 (31.13) 43 (19.03) 40 (30.08) 8 (7.77) 138 (22.51) 224 (36.30) 362 (29.43) Histology, n (%) LUAD 96 (63.58) 183 (80.97) 133 (100.00) 80 (77.67) 492 (80.26) 332 (53.81) 824 (66.99) LUSC 55 (36.42) 43 (19.03) 0 (0.00) 23 (22.33) 121 (19.74) 285 (46.19) 406 (33.01) Chemotherapy, n (%) No 142 (94.04) 177 (90.77) 102 (76.69) 67 (90.54) 488 (88.25) 194 (76.98) 682 (84.72) Yes 9 (5.96) 18 (9.23) 31 (23.31) 7 (9.46) 65 (11.75) 58 (23.02) 123 (15.28) Unknown 0 31 0 29 60 365 425 Radiotherapy, n (%) 74 No 132 (87.42) 184 (94.36) 132 (99.25) 522 (94.39) 239 (94.84) 761 (94.53) (100.00) Yes 19 (12.58) 11 (5.64) 1 (0.75) 0 (0.00) 31 (5.61) 13 (5.16) 44 (5.47) Unknown 0 31 0 29 60 365 425 Adjuvant therapyb, n (%) No 127 (84.11) 168 (86.15) 101 (75.94) 67 (90.54) 463 (83.73) 187 (74.21) 650 (80.75) Yes 24 (15.89) 27 (13.85) 32 (24.06) 7 (9.46) 90 (16.27) 65 (25.79) 155 (19.25) Unknown 0 31 0 29 60 365 425 Survival year 6.66 7.12 7.36 7.39 7.39 4.54 6.60 Median (95% CI) (5.41-7.87) (5.06-9.63) (6.77-7.95)* (4.98-9.12) (6.50-8.23) (3.68-5.41) (5.84-7.35) Dead (%) 122 (80.79) 101 (44.69) 42 (31.58) 58 (31.58) 323 (52.69) 142 (23.01) 465 (37.80) a Cohort 2: Spain is a collaborative cohort, recruiting samples from Spain, Italy, UK, France, and USA. b Adjuvant therapy includes chemotherapy or radiotherapy. * Restricted mean survival time is given since median is not available. LUAD: lung adenocarcinoma; LUSC: lung squamous cell carcinoma www.aging-us.com 10659 AGING Supplementary Table 2. Results for two lung adenocarcinoma (LUAD)-specific methylation–age interactions identified from a two-stage epigenome-wide association study. Discovery phase Validation phase Combined data Variable HR 95% CI P HR 95% CI P HR 95% CI P cg14326354 2.986 1.921 4.640 1.16×10-06 3.442 1.146 10.337 2.76×10-02 1.885 1.429 2.487 7.16×10-06 Age 0.860 0.809 0.915 1.74×10-06 0.854 0.728 1.002 5.36×10-02 0.926 0.890 0.964 1.53×10-04 Interaction 0.982 0.976 0.989 1.11×10-07 0.981 0.966 0.997 2.02×10-02 0.989 0.986 0.994 9.18×10-07 cg08700284 62.927 15.807 250.507 4.20×10-09 3.336 0.024 455.52 6.31×10-01 8.960 3.779 21.243 7.16×10-06 Age 0.349 0.237 0.513 9.31×10-08 0.718 0.175 2.947 6.46×10-01 0.591 0.466 0.751 1.62×10-05 Interaction 0.944 0.925 0.964 6.55×10-08 0.981 0.910 1.057 6.17×10-01 0.971 0.959 0.984 9.18×10-07 HR: hazard ratio; 95% CI: 95% confidence interval Supplementary Table 3. Annotation information for significant lung adenocarcinoma (LUAD)-specific CpG probe. Relation to CpG CpG probe CHR BP Region islands Gene descriptiona cg14326354 22 18900453 3'UTR S_Shelf proline dehydrogenase 1 (PRODH) a Hyperlinks provide literature-based evidence for each gene from DAVID (https://david.ncifcrf.gov). CHR: chromosome; BP: basepair Supplementary Table 4. Results of interaction for sensitivity analysis of one significant lung adenocarcinoma (LUAD)- specific CpG probe. Discovery phase Validation phase Combined data Variable HR 95% CI P HR 95% CI P HR 95% CI P cg14326354 3.056 1.950 4.792 1.11×10-06 3.522 1.156 10.73 2.67×10-02 3.048 2.043 4.547 4.77×10-08 Age 0.858 0.806 0.914 1.61×10-06 0.853 0.726 1.003 5.36×10-02 0.861 0.814 0.911 1.80×10-07 Interaction 0.982 0.976 0.989 1.03×10-07 0.981 0.966 0.997 2.09×10-02 0.982 0.976 0.988 3.18×10-09 In sensitivity analysis, patients were excluded if their methylation values were out of range mean± 3×standard deviation on logit2 transformed scale. HR: hazard ratio; 95% CI: 95% confidence interval Supplementary Table 5. Results of heterogeneity test of the interaction effect between subgroups categorized by covariates. Covariate Subgroup HRinteraction 95% CI Pinteraction Qheterogeneity Pheterogeneity Never or former smoker 0.982 0.974 0.989 8.42×10-7 Current smoker 0.982 0.971 0.992 7.85×10-4 Male 0.983 0.974 0.990 1.57×10-5 Female 0.978 0.968 0.988 2.10×10-5 I 0.985 0.978 0.992 9.56×10-5 II 0.977 0.965 0.988 1.65×10-4 Harvard 0.979 0.964 0.995 7.92×10-3 Norway 0.987 0.968 1.006 1.72×10-1 Spain 0.985 0.973 0.997 1.75×10-2 Sweden 0.983 0.959 1.008 1.77×10-1 TCGA 0.981 0.966 0.997 2.02×10-2 HR: hazard ratio; 95% CI: 95% confidence interval www.aging-us.com 10660 AGING Supplementary Table 6. Results of low cg14326354PRODH methylation effect on lung adenocarcinoma (LUAD) survival in young and elderly populations defined using boundary of 95% confidence interval (BoCl) standard. Discovery phase Validation phase Combined data Population HR 95% CI P HR 95% CI P HR 95% CI P Young (age <57 years) 1.182 1.005 1.389 4.29×10-2 1.294 1.023 1.636 3.18×10-2 1.200 1.030 1.401 1.97×10-2 Elderly (age >65 years) 0.810 0.742 0.885 3.12×10-6 0.858 0.754 0.976 1.99×10-2 0.814 0.751 0.882 5.38×10-7 Patients from Harvard, Spain, Norway, and Sweden cohorts were assigned to discovery phase; patients in TCGA were assigned to validation phase. HR: hazard ratio; 95% CI: 95% confidence interval Supplementary Table 7. Correlation analysis of association between cg14326354PRODH methylation and proliferation- associated gene expression in lung adenocarcinoma (LUAD) patients using The Cancer Genome Atlas data, as well as survival analysis of proliferation-associated genes (from KEGG database). Correlation analysis Survival analysis Gene r 95% CI P HR 95% CI P BTG2 -0.313 -0.408 -0.212 6.80×10-09 0.704 0.564 0.878 1.83×10-03 NPDC1 -0.266 -0.364 -0.163 9.88×10-07 0.937 0.755 1.164 0.558 KIAA1524 0.243 0.139 0.342 8.40×10-06 1.383 1.103 1.733 4.94×10-03 MKI67 0.199 0.093 0.301 2.84×10-04 1.440 1.115 1.859 5.14×10-03 BTG1 -0.188 -0.290 -0.081 6.29×10-04 1.257 0.869 1.819 0.225 BOP1 0.122 0.014 0.227 2.72×10-02 1.170 0.910 1.505 0.222 BTG4 -0.118 -0.223 -0.010 3.29×10-02 0.995 0.769 1.288 0.971 CDC123 0.111 0.003 0.217 4.46×10-02 1.798 1.045 3.097 3.41×10-02 PA2G4 0.091 -0.017 0.197 9.94×10-02 MTCP1NB 0.079 -0.030 0.185 1.55×10-01 C8orf22 -0.067 -0.174 0.042 2.29×10-01 HEY1 0.061 -0.048 0.168 2.71×10-01 PPDPF -0.058 -0.166 0.050 2.92×10-01 SIPA1L2 0.055 -0.054 0.162 3.24×10-01 MTCP1 0.047 -0.061 0.155 3.94×10-01 SIPA1L3 -0.039 -0.146 0.070 4.84×10-01 URGCP 0.036 -0.073 0.144 5.17×10-01 SAV1 -0.032 -0.139 0.077 5.69×10-01 BTG3 -0.024 -0.132 0.084 6.63×10-01 PDS5B -0.023 -0.131 0.085 6.74×10-01 SIPA1 -0.008 -0.117 0.100 8.80×10-01 PEA15 -0.006 -0.114 0.103 9.20×10-01 SIPA1L1 0.000 -0.108 0.108 9.99×10-01 Correlation coefficient (r), 95% CI, and P-values were derived from Pearson correlation analysis; survival analysis HR, 95% CI, and P-values were derived from Cox proportional hazards model. HR: hazard ratio; 95% CI: 95% confidence interval. Supplementary Table 8. Results of heterogeneity test of the interaction effect between lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) populations. LUAD LUSC Heterogeneity CpG probe HR 95% CI HR 95% CI Q P cg08470135 0.944 0.925 0.964 0.975 0.946 1.004 2.97 0.0847 cg14326354 0.982 0.976 0.989 0.982 0.969 0.995 0.00 0.9764 HR: hazard ratio; 95% CI: 95% confidence interval. www.aging-us.com 10661 AGING Supplementary Materials Please browse Full Text version to see the data of Supplementary Materials 1 to 4. Supplementary Material 1. Epigenome-wide association results in LUAD patients. Only summary results of interaction terms were shown in the table. Supplementary Material 2. Epigenome-wide association results in LUSC patients. Only summary results of interaction terms were shown in the table. Supplementary Material 3. Results of genome-wide trans-regulation analysis of cg14326354 methylation using gene expression data and DNA methylation data from TCGA. Supplementary Material 4. Results of Tumor mutational burden (TMB) analysis of each sample from TCGA. www.aging-us.com 10662 AGING