SIPA1L3 methylation modifies the benefit of smoking cessation on lung adenocarcinoma survival: an epigenomic–smoking interaction analysis Ruyang Zhang1,2,3, Linjing Lai1, Xuesi Dong1,4, Jieyu He1, Dongfang You1, Chao Chen1, Lijuan Lin1, Ying Zhu1, Hui Huang1, Sipeng Shen1,2,3 , Liangmin Wei1, Xin Chen1, Yichen Guo2,5, Liya Liu6, Li Su , Andrea Shafer , Sebastian Moran , Thomas Fleischer9 , Maria Moksnes Bjaanæs9, 2,3 7 8 Anna Karlsson10, Maria Planck10, Johan Staaf10,  Aslaug Helland9,11, Manel Esteller8, 1,2,3 1,3,12 Yongyue Wei , Feng Chen and David C. Christiani2,3,7 1 Department of Biostatistics, Center for Global Health, School of Public Health, Nanjing Medical University, China 2 Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, USA 3 China International Cooperation Center for Environment and Human Health, Nanjing Medical University, China 4 Department of Epidemiology and Biostatistics, School of Public Health, Southeast University, Nanjing, China 5 Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, MA, USA 6 Department of Preventive Medicine, Medical School of Ningbo University, China 7 Pulmonary and Critical Care Division, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA 8 Bellvitge Biomedical Research Institute, Institucio Catalana de Recerca i Estudis Avancßats, University of Barcelona, Barcelona, Spain 9 Department of Cancer Genetics, Institute for Cancer Research, Oslo University Hospital, Norway 10 Division of Oncology and Pathology, Department of Clinical Sciences Lund, CREATE Health Strategic Center for Translational Cancer Research, Lund University, Sweden 11 Institute of Clinical Medicine, University of Oslo, Norway 12 Jiangsu Key Lab of Cancer Biomarkers, Prevention and Treatment, Cancer Center, Collaborative Innovation Center for Cancer Personalized Medicine, Nanjing Medical University, China Keywords Smoking cessation prolongs survival and decreases mortality of patients DNA methylation; interaction analysis; with non-small-cell lung cancer (NSCLC). In addition, epigenetic alter- molecular cancer epidemiology; ations of some genes are associated with survival. However, potential inter- non-small-cell lung cancer; overall survival; actions between smoking cessation and epigenetics have not been assessed. smoking cessation Here, we conducted an epigenome-wide interaction analysis between DNA Correspondence methylation and smoking cessation on NSCLC survival. We used a two- F. Chen, Department of Biostatistics, Center stage study design to identify DNA methylation–smoking cessation interac- for Global Health, School of Public Health, tions that affect overall survival for early-stage NSCLC. The discovery Nanjing Medical University, SPH Building phase contained NSCLC patients from Harvard, Spain, Norway, and Swe- Room 412, 101 Longmian Avenue, Nanjing, den. A histology-stratified Cox proportional hazards model adjusted for Jiangsu 211166, China age, sex, clinical stage, and study center was used to test DNA methyla- Tel: +86 25 86868435 E-mail: fengchen@njmu.edu.cn tion–smoking cessation interaction terms. Interactions with false discovery Y. Wei, Department of Biostatistics, Center rate-q ≤ 0.05 were further confirmed in a validation phase using The Can- for Global Health, School of Public Health, cer Genome Atlas database. Histology-specific interactions were identified Nanjing Medical University, SPH Building by stratification analysis in lung adenocarcinoma (LUAD) and lung squa- Room 418, 101 Longmian Avenue, Nanjing, mous cell carcinoma (LUSC) patients. We identified one CpG probe Jiangsu 211166, China (cg02268510SIPA1L3) that significantly and exclusively modified the effect of Tel: +86 25 86868436 smoking cessation on survival in LUAD patients [hazard ratio (HR)interaction = E-mail: ywei@njmu.edu.cn 1.12; 95% confidence interval (CI): 1.07–1.16; P = 4.30 9 10–7]. Further, the effect of smoking cessation on early-stage LUAD survival varied across Abbreviations CI, confidence interval; FDR, false discovery rate; HR, hazard ratio; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; NSCLC, non-small-cell lung cancer; QC, quality control; SD, standard deviation; TCGA, The Cancer Genome Atlas. Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1235 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. Epigenomic–smoking interaction analysis of SIPA1L3 R. Zhang et al. Ruyang Zhang, Linjing Lai contributed patients with different methylation levels of cg02268510SIPA1L3. Smoking equally to the work. cessation only benefited LUAD patients with low methylation (HR = 0.53; David C. Christiani is a senior author who 95% CI: 0.34–0.82; P = 4.61 9 10–3) rather than medium or high methyla- supervised the work. tion (HR = 1.21; 95% CI: 0.86–1.70; P = 0.266) of cg02268510SIPA1L3. (Received 15 November 2018, revised 13 Moreover, there was an antagonistic interaction between elevated methyla- March 2019, accepted 18 March 2019, tion of cg02268510SIPA1L3 and smoking cessation (HRinteraction = 2.1835; available online 17 April 2019) 95% CI: 1.27–3.74; P = 4.46 9 10 3). In summary, smoking cessation ben- efited survival of LUAD patients with low methylation at cg02268510SI- doi:10.1002/1878-0261.12482 PA1L3. The results have implications for not only smoking cessation after diagnosis, but also possible methylation-specific drug targeting. adenocarcinoma (LUAD) and lung squamous cell car- 1. Introduction cinoma (LUSC) cases were used for discovery, and the Lung cancer is a leading cause of cancer mortality findings were independently validated in The Cancer worldwide. In the United States, lung cancer was esti- Genome Atlas (TCGA) data. mated as likely to account for 154 050 deaths in 2018, or one-fourth of all cancer deaths (Siegel et al., 2017). 2. Materials and methods A large proportion of lung cancer cases are attributed to smoking, a well-known risk factor (Flanders et al., 2.1. Study population 2003), and smoking cessation prolongs survival and decreases mortality of lung cancer patients (Balduyck Early-stage (stage I–II) LUAD and LUSC patients et al., 2011; Parsons et al., 2010). However, the under- who were former or current smokers were included in lying mechanisms of these benefits remain largely the study. Never smokers were defined as those who unclear (Bhatt et al., 2015; Parsons et al., 2010). smoked ≤ 100 cigarettes over a lifetime. Current smok- DNA methylation, a reversible epigenetic modifica- ers were defined as those who were smoking within tion, regulates gene expression and provides potential 1 year of diagnosis. Former smokers were defined as cancer biomarkers and therapeutic targets (Egger smokers who quit > 1 year before diagnosis or inter- et al., 2004; Feinberg and Tycko, 2004), including for view (Suk et al., 2006). We encoded the variable smok- non-small-cell lung cancer (NSCLC) (Guo et al., 2018; ing cessation as ‘yes’ for former smokers and ‘no’ for Shen et al., 2018; Wei et al., 2018). Furthermore, as a current smokers. Data were harmonized from five potential mechanistic link between cigarette smoking international study centers, which have been previously and disease, DNA methylation changes can result described (Guo et al., 2018; Shen et al., 2018; Wei from various environmental exposures and may et al., 2018; Zhang et al., 2019). All patients provided explain part of the association between smoking and written informed consent, and the study methodologies cancer recurrence or mortality (Lee and Pausova, conformed to the standards set by the Declaration 2013; Shui et al., 2016). of Helsinki and received approval by its respective Progression of complex diseases, such as cancer, institutional review board. results from interactions between clinical, environmen- tal, genetic, and epigenetic factors (Lacombe et al., 2.1.1. Harvard 2016; Mcnerney et al., 2017). However, most epigen- ome-wide association studies are designed to identify The Harvard Lung Cancer Study cohort was described main effects using a standard marginal test (Karlsson previously (Suk et al., 2006). Briefly, all patients were et al., 2014) while ignoring epigenetic–environment recruited at Massachusetts General Hospital (MGH) interactions. These traditional mining procedures may from 1992 to present and had newly diagnosed, histo- reduce the power to identify new epigenomic biomark- logically confirmed primary NSCLC. We included 133 ers (Slade and Kraft, 2016). early-stage LUAD and LUSC patients who were for- In this study, we hypothesized that epigenetic and mer or current smokers for the current study. DNA smoking cessation interactions may affect NSCLC sur- was extracted from tumor specimens that were evalu- vival. Epigenome-wide DNA methylation data com- ated by an MGH pathologist for amount (tumor cellu- posed of four study cohorts containing lung larity > 70%) and quality of tumor cells and 1236 Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. R. Zhang et al. Epigenomic–smoking interaction analysis of SIPA1L3 histologically classified using World Health Organiza- quality control (QC). Beta values, which range from 0% tion criteria. The study protocol was approved by the (unmethylated) to 100% (methylated), were used to mea- Institutional Review Boards at Harvard School of sure methylation level of each probe. Unqualified probes Public Health and MGH. were excluded if they met any one of the following QC criteria: (a) failed detection (P > 0.05) in > 5% samples; (b) coefficient of variance of < 5%; (c) methylated or 2.1.2. Spain unmethylated in all samples; (d) common single nucleo- The Spain study population was reported previously tide polymorphisms located in probe sequence or 10-bp (Sandoval et al., 2013) and included 196 LUAD and flanking regions; (e) cross-reactive or cross-hybridizing LUSC patients recruited at eight subcenters from 1991 to probes (Chen et al., 2013); or (f) did not pass QC in all 2009. In brief, tumor DNA was extracted from fresh-fro- centers. Samples with > 5% undetectable probes were zen tumor specimens that were collected by surgical excluded. Methylation signals were further processed for resection, and the median clinical follow-up was quantile normalization, design bias correction for type I 7.2 years. The study was approved by the Bellvitge and II probes, and batch effect adjustment using ComBat Biomedical Research Institute institutional review board. correction (Marabita et al., 2013). We performed QC procedures above in each center separately and then merged all data together before association analysis. 2.1.3. Norway Details of QC processes are described in Fig. S1. The Norway cohort consisted of 116 LUAD patients with operable lung cancer tumors who were seen at Oslo 2.3. Gene expression data University Hospital, Rikshospitalet, Norway, in 2006– 2011 (Bjaanæs et al., 2016). Tumor tissues obtained dur- Expression and mRNA sequencing data were available ing surgery were snap-frozen in liquid nitrogen and for 281 LUAD and 277 LUSC patients of the TCGA stored at 80 °C until DNA isolation. The project was dataset (Table S1). TCGA mRNA sequencing data approved by Oslo University institutional review board processing and QC were done by the TCGA work- and regional ethics committee (S-05307). group. Raw counts were normalized using RNA sequencing by expectation maximization. Level 3 gene quantification data were downloaded from the TCGA 2.1.4. Sweden data portal (https://tcga-data.nci.nih.gov; now hosted The Sweden cohort included 85 LUAD and LUSC at https://portal.gdc.cancer.gov) and were further patients. Tumor DNA was collected from early-stage checked for quality. Gene expression data were lung cancer patients who underwent an operation at extracted and log2-transformed before analysis. the Skane University Hospital, Lund, Sweden (Karls- son et al., 2014). The study was approved by the 2.4. Epigenome-wide DNA methylation–smoking Regional Ethical Review Board in Lund, Sweden cessation interaction analysis (Registration no. 2004/762 and 2008/702). Analysis flow is described in Fig. 1. Patients from the first four study centers (Harvard, Spain, Norway, and 2.1.5. TCGA Sweden) were assigned into the discovery phase. A his- The TCGA database contains 562 early-stage LUAD tology-stratified Cox proportional hazards model was and LUSC patients who have full information of sur- used to test the interaction item, which was the inter- vival time and covariates. Level 1 HumanMethyla- action effect between DNA methylation of each CpG tion450 DNA methylation data (image data) for each probe and smoking cessation (CpG probe 9 smoking patient were downloaded on October 1, 2015. cessation) on overall survival. The model was adjusted for age, sex, smoking cessation, clinical stage, and study center. Hazard ratio (HR) and 95% confidence 2.2. Quality control procedures interval (CI) were described per 1% methylation incre- DNA methylation was profiled using Infinium ment. Multiple testing corrections were performed HumanMethylation450 BeadChips (Illumina Inc., San using the false discovery rate method (FDR, measured Diego, CA, USA) for all patients. Raw image data were by FDR-q value) by the Benjamini–Hochberg proce- imported into GenomeStudio Methylation Module V1.8 dure. CpG probes with interaction FDR-q ≤ 0.05 were (Illumina Inc.) to calculate methylation signals and to replicated in the validation phase using the TCGA perform normalization, background subtraction, and dataset. Robustly significant probes were retained if Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1237 Epigenomic–smoking interaction analysis of SIPA1L3 R. Zhang et al. they met all criteria: (a) interaction P ≤ 0.05 in the val- linear regression model adjusted for the same covari- idation phase; and (b) consistent effect direction in ates mentioned above. Association with FDR-q ≤ 0.05 both discovery and validation phases. We performed was considered significant. Additionally, we tested the stratified analysis for robustly significant CpG probes association between gene expression and overall sur- in LUAD and LUSC patients. Finally, CpG probes vival using a Cox proportional hazards model adjusted with a significant interaction with smoking cessation in for the same covariates. Genes involved in significant both phases were identified as histology-specific associations with both methylation and NSCLC sur- probes. vival were filtered. 2.5. Sensitivity analysis for significant CpG 2.7. Statistical analysis probes Continuous variables were summarized as mean  s- Due to the complex tumor microenvironment—includ- tandard deviation (SD), and categorized variables were ing noncancerous components, which might alter anal- described by frequency (n) and proportion (%). ysis of tumor samples (Aran et al., 2015)—we assessed Kaplan–Meier survival curves were used to compare tumor purity with InfiniumPurify (Zhang et al., 2015) survival difference among subgroups. Statistical analy- using methylation array data from TCGA samples. ses were performed using R version 3.4.4 (The R Foun- Tumor purity was included as an additional covariate dation for Statistical Computing). in the Cox regression model for sensitivity analysis. 3. Results 2.6. Genome-wide methylation transcription After QC, epigenome-wide DNA methylation data analysis including 311 891 CpG sites from 1092 tumor samples For robustly significant histology-specific prognostic of early-stage (stage I–II) NSCLC patients were CpG probes, we also performed genome-wide methyla- retained. There were 530 patients (NLUAD = 413 and tion transcription analysis using mRNA sequencing NLUSC = 117) in the discovery phase and 562 patients data from TCGA. The correlation between DNA (NLUAD = 285 and NLUSC = 277) in the validation methylation and gene expression was tested using a phase. Table 1 details demographic and clinical Fig. 1. Flowchart of study design and statistical analyses. 1238 Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. R. Zhang et al. Epigenomic–smoking interaction analysis of SIPA1L3 information for the study population. There were 37% P = 0.266). We observed significant heterogeneity of and 27% current smokers in the discovery and valida- smoking cessation effect across the three groups tion phases, respectively. (P = 0.014; Fig. 2B), and Kaplan–Meier curves con- In the discovery phase, 15 methylation–smoking ces- firmed these results (Fig. 2C). sation interactions were identified with FDR-q ≤ 0.05 These results also indicated that LUAD patients who (Fig. S2A), and the Manhattan plot also showed the did not quit smoking (current smokers) had the poorest results for main effect additionally (Fig. S2B). Only 1 prognosis if their methylation of cg02268510SIPA1L3 interaction remained statistically significant in the vali- was in a low level. So we combined the medium and dation phase under the most stringent criteria high methylation groups and performed further (Table S2). This site, cg02268510, is located in signal- analysis. Current smokers in the low methylation group induced proliferation-associated 1-like 3 (SIPA1L3). had 1.94 times the mortality risk compared with the Further histology-stratified analysis showed that medium or high methylation group (Fig. 3A), but cg02268510SIPA1L3 is a LUAD-specific CpG probe that there was no statistically significant difference interacts with smoking cessation to affect patient sur- between groups for former smokers (Fig. 3B). The vival in the discovery phase (HRinteraction = 1.10; 95% results also indicated that smoking cessation was quite CI: 1.05–1.16; P = 2.95 9 10–5), the validation phase urgent for LUAD patients with low methylation of (HRinteraction = 1.17; 95% CI: 1.02–1.35; P = 0.0255), cg02268510SIPA1L3. and the combined data (HRinteraction = 1.12; 95% CI: In addition, we evaluated the joint effect of CpG 1.07–1.16; P = 4.30 9 10–7). Moreover, fixed-effect methylation level (medium–high vs low) and smoking meta-analysis of five centers also remained signifi- cessation (Yes vs No) on LUAD survival (Table 2). cant (HRinteraction = 1.09; 95% CI: 1.05–1.13; We used the poorest-prognosis group (current smokers P = 6.66 9 10–6; Fig. S3). As presented in Fig. 2A, with low methylation) as the reference to evaluate with decreased methylation level of cg02268510SIPA1L3, effect of elevated methylation level, smoking cessation, there was an elevated benefit effect size of smoking and their interaction. In the combined dataset, the cessation on LUAD survival. Thus, there was a modi- effect of smoking cessation was HR = 0.5506 (95% fication effect of cg02268510SIPA1L3 on the association CI: 0.36–0.84; P = 5.62 9 10 3) and the effect of between smoking cessation and survival. medium–high methylation of cg02268510SIPA1L3 was After including tumor purity as an additional HR = 0.5214 (95% CI: 0.34–0.81; P = 3.48 9 10 3). covariate in sensitivity analysis, DNA methylation However, the joint effect was HR = 0.6268 (95% CI: at cg02268510SIPA1L3 retained a significant interaction 0.43–0.92; P = 1.84 9 10 2), which was greater than with smoking cessation on LUAD survival the product of the two individual protective effects (HRinteraction = 1.18; 95% CI: 1.02–1.36; P = 0.024). (0.5506 9 0.5214 = 0.2871). The joint effect of two The interaction P-value was still significant but slightly protective factors was less protective than expected, inflated due to (a) the smaller sample size (51% of indicating an antagonistic interaction between elevated original) of the sensitivity analysis, which was only methylation of cg02268510SIPA1L3 and smoking cessa- performed in TCGA; and (b) low tumor purity tion (HRinteraction = 2.1835; 95% CI: 1.27–3.74; (~ 60%) for NSCLC samples in TCGA due to mixed P = 4.46 9 10 3). cell types (Zheng et al., 2017). A growing body of research has reported potential To better illustrate the interaction pattern between associations of DNA methylation with age and smok- DNA methylation and smoking cessation, patients ing (Fraga and Esteller, 2007; Wan et al., 2012; Zagh- were categorized into low, medium, and high groups lool et al., 2015). Therefore, we also tested the based on tertiles of cg02268510SIPA1L3 methylation. association between methylation of cg02268510SIPA1L3 The effect of smoking cessation varied across LUAD and age, as well as smoking-related variables: pack- patients with different DNA methylation levels. year of smoking, years of smoking, and years of smok- Smoking cessation only benefited LUAD patients ing cessation using a linear regression model adjusted with low methylation of cg02268510SIPA1L3 for age, sex, clinical stage, and study centers. Smok- (HRlow = 0.53; 95% CI: 0.34–0.82; P = 4.61 9 10–3). ing-related characteristics of former and current smok- However, there was no significant association ers in early-stage LUAD are described in Table S3. between smoking cessation and survival in LUAD There was no significant association between methyla- patients with medium–high methylation of tion of cg02268510SIPA1L3 and age (b = 0.01; cg02268510SIPA1L3 (HRmedium = 1.12; 95% CI: 0.67– P = 0.521) or years of smoking (b = 0.03; P = 0.210), 1.87; P = 0.665; HRhigh = 1.29; 95% CI: 0.80–2.07; but pack-year of smoking (b = 0.02; P = 3.42 9 10 3) P = 0.293; HRmedium–high = 1.21; 95% CI: 0.86–1.70; as well as years of smoking cessation (b = 0.06; Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1239 Table 1. Demographic and clinical characteristics of former and current smokers with early-stage NSCLC in five study centers. 1240 Discovery phase Validation phase Combined data Variable Harvard (N = 133) Spaina (N = 196) Norway (N = 116) Sweden (N = 85) Discovery: All (N = 530) TCGA (N = 562) Overall (N = 1092) Age (years), 68.04  9.39 65.69  10.41 64.89  8.94 65.65  9.69 66.10  9.78 66.55  9.39 66.33  9.58 mean  SD Sex, n (%) Female 56 (42.11) 78 (39.80) 59 (50.86) 42 (49.41) 235 (44.34) 217 (38.61) 452 (41.39) Male 77 (57.89) 118 (60.20) 57 (49.14) 43 (50.59) 295 (55.66) 345 (61.39) 640 (58.61) Smoking cessation, n (%) * No 52 (39.10) 71 (36.22) 42 (36.21) 31 (36.47) 196 (36.98) 168 (26.89) 364 (33.33) Yes 81 (60.90) 120 (61.22) 74 (63.79) 54 (63.53) 329 (62.08) 376 (66.90) 705 (64.56) Unknown 0 5 0 0 5 18 23 Race, n (%) White 133 (100) 196 (100) 116 (100) 85 (100) 530 (100) 444 (79.00) 974 (89.19) Black or African 0 0 0 0 0 54 (9.61) 54 (4.95) American Epigenomic–smoking interaction analysis of SIPA1L3 Asian 0 0 0 0 0 5 (0.89) 5 (0.46) Unknown 0 0 0 0 0 59 59 TNM stage, n (%) * I 87 (65.41) 160 (81.63) 80 (68.97) 79 (92.94) 406 (76.60) 358 (63.70) 764 (69.96) II 46 (34.59) 36 (18.37) 36 (31.03) 6 (7.06) 124 (23.40) 204 (36.30) 328 (30.04) Histology, n (%) * LUAD 79 (59.40) 155 (79.08) 116 (100) 63 (74.12) 413 (77.92) 285 (50.71) 698 (63.92) LUSC 54 (40.60) 41 (20.92) 0 22 (25.88) 117 (22.08) 277 (49.29) 394 (36.08) Chemotherapy, n (%) * No 125 (93.98) 157 (80.10) 86 (74.14) 51 (60.00) 419 (79.06) 178 (31.67) 597 (54.67) Yes 8 (6.02) 15 (7.65) 30 (25.86) 6 (7.06) 59 (11.13) 53 (9.43) 112 (10.26) Unknown 0 24 0 28 52 331 383 Radiotherapy, n (%) No 115 (86.47) 161 (82.14) 115 (99.14) 57 (67.06) 448 (84.53) 220 (39.15) 668 (61.17) Yes 18 (13.53) 11 (5.61) 1 (0.86) 0 30 (5.66) 11 (1.96) 41 (3.75) Unknown 0 24 0 28 52 331 383 Adjuvant therapy, n (%) * No 110 (82.71) 148 (75.51) 85 (73.28) 51 (60.00) 394 (74.34) 172 (30.60) 556 (51.83) Yes 23 (17.29) 24 (12.24) 31 (26.72) 6 (7.06) 84 (15.85) 59 (10.50) 143 (13.10) Unknown 0 24 0 28 52 331 383 Survival year Median (95% CI) 6.44 (4.93–7.44) 6.53 (5.06–8.70) 7.34 (6.70–7.98)c 5.89 (4.18–8.82) 7.12 (6.23–7.95) 4.54 (3.69–5.41) 6.23 (5.55–7.20) Censored rateb, % 14.29 54.59 68.97 40.00 45.30 76.16 61.17 a Spain is a collaborative study center, containing samples from Spain, Italy, the UK, France, and the United States. b Censored rate is the proportion of samples lost to follow-up or alive at the study end. c Restricted mean survival time is provided because median was not available. R. Zhang et al. Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. *Statistically significant difference (P ≤ 0.05) was observed between combined discovery set and validation set (TCGA). R. Zhang et al. Epigenomic–smoking interaction analysis of SIPA1L3 Fig. 2. DNA methylation and smoking cessation interaction on survival of LUAD patients. (A) HR of smoking cessation estimated based on methylation level of cg02268510. The shallow area represents 95% CI, with red, gray and blue areas indicating low, medium and high methylation, respectively. Histogram on the top shows the distribution of methylation. (B) Forest plots of the effects of smoking cessation among combined LUAD populations with low, medium, or high methylation of cg02268510. Pheterogeneity was used to evaluate heterogeneity of HRs across groups. (C) Kaplan–Meier survival curves of current and former smokers among LUAD patients with varying methylation levels. Fig. 3. Kaplan–Meier survival curves of LUAD patients categorized into low and medium–high methylation groups according to tertiles in different smoking cessation groups: (A) current smokers (No) and (B) former smokers (Yes). HR, 95% CI, and P-value were derived from the Cox proportional hazards model adjusted for age, sex, clinical stage, and study center. P = 5.08 9 10 3) in former smoker LUAD patients TCGA dataset. We observed a significant association (Fig. S4). between cg02268510SIPA1L3 and SIPA1L3 expression Further, because cg02268510SIPA1L3 maps to (b = 0.02; P = 0.015) in LUAD patients (Fig. 4), SIPA1L3, the association between cg02268510SIPA1L3 indicating that cg02268510SIPA1L3 cis-regulates gene and SIPA1L3 expression was evaluated using the expression. Moreover, genome-wide methylation Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1241 Epigenomic–smoking interaction analysis of SIPA1L3 R. Zhang et al. Table 2. Joint effect and interaction of elevated methylation and smoking cessation on the prognosis of early-stage LUAD. Effect typea Medium–high methylationb Smoking cessation Number HR (95% CI)a Pa No No 71 Ref. Main effect1 No Yes 157 0.5506 (0.3609, 0.8400) 5.62 9 10 3 Main effect2 Yes No 158 0.5214 (0.3369, 0.8070) 3.48 9 10 3 Joint effect Yes Yes 299 0.6268 (0.4251, 0.9243) 1.84 9 10 2 Interactionc 2.1835 (1.2747, 3.7401) 4.46 9 10 3 a Patients were categorized into two groups (medium–high vs low) by tertiles of cg02268510SIPA1L3 methylation level. b Main effects of elevated methylation and smoking cessation and their joint effect and interaction were derived from the Cox proportional hazards model adjusted for covariates. c Interaction = Joint effect  (main effect1 9 main effect2). 2.1835 = 0.6268  (0.5506 9 0.5214). knowledge, this is the first study with a large sample size to investigate interactions between DNA methyla- tion and smoking behavior on lung cancer survival, and it provides new evidence to account for the missing heritability of complex diseases (Trerotola et al., 2015). Our results show that the effect of smoking cessation on early-stage LUAD patient survival varies with methylation level of cg02268510SIPA1L3. Smoking cessa- tion only benefits LUAD patients with low methyla- tion, rather than medium or high methylation, of cg02268510SIPA1L3. Further, there is an antagonistic interaction between elevated methylation of cg02268510SIPA1L3 and smoking cessation. We found that in LUAD patients with low methyla- tion of cg02268510SIPA1L3, current smokers with more accumulative exposure had worse survival than former smokers. However, for a population with medium– high methylation, the prognosis of current smokers Fig. 4. Association between DNA methylation of cg02268510 and was similar to that of former smokers. The effect of expression of corresponding gene SIPA1L3. The b coefficient and smoking cessation is therefore modified by DNA P-value were based on linear regression analysis adjusted for age, methylation level, indicating opportunities for epi-drug sex, smoking status, and clinical stage. Gene expression was log2- intervention due to the inherent reversibility of epige- transformed before analysis. netic events (Wright, 2013). Up to 50% of lung cancer patients are estimated to transcription analysis revealed that expression of 633 keep smoking after diagnosis or to frequently relapse genes was significantly correlated with methylation after smoking cessation (Park et al., 2012; Walker level of cg02268510SIPA1L3 (Fig. S5A). Among them, et al., 2006). Our results indicated that smoking cessa- expression of only seven genes was significantly associ- tion was urgent especially for LUAD patients with low ated with overall survival: growth arrest and DNA methylation of cg02268510SIPA1L3. On the other hand, damage-inducible gamma (GADD45G), maturin reduced methylation of cg02268510SIPA1L3 might (MTURN), TMEM200B, RGS20, RELT-like 1 strengthen the protective effect of smoking cessation (RELL1), PGM2, and receptor-interacting serine/thre- on survival. onine kinase 2 (RIPK2; Fig. S5B–H). Many studies have reported significant associations between smoking cessation and overall survival (Koshiaris et al., 2017; Nia et al., 2005), while other 4. Discussion studies have reported negative results (Baser et al., In this study, we systematically evaluated all pairwise 2006; Parsons et al., 2010). Based on our interaction DNA methylation–smoking cessation interactions on analysis, we suspected that epigenetic modifications an epigenome-wide scale and further confirmed these might account for this inconsistent phenomenon. interactions in an independent population. To our Because the effect of smoking cessation varies across 1242 Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. R. Zhang et al. Epigenomic–smoking interaction analysis of SIPA1L3 populations with different methylation levels of drug to treat NSCLC (Besse et al., 2014). Our results cg02268510SIPA1L3, the effect could be neutralized in a suggest that low methylation at cg02268510SIPA1L3 population of patients with mixed cg02268510SIPA1L3 might promote SIPA1L3 expression, further leading to methylation levels. Thus, the traditional marginal test Rap1 activation and resulting in poor prognosis for association between smoking cessation and cancer (Fig. 5). survival inherently loses statistical power to report sig- Many of the deleterious effects of smoking are due nificant findings due to complex association patterns. to induction of inflammatory responses that contribute SIPA1L3, the gene in which cg02268510 is located, to lung cancer progression (Crusz and Balkwill, 2015; encodes GTPase-activating proteins (GAPs) specific for Walser et al., 2008). In vitro experiments in human the GTP-binding protein Ras-associated protein-1 umbilical vein endothelial cells demonstrate that nico- (RAP1), which is implicated in regulation of cell adhe- tine stimulates cellular inflammatory responses by acti- sion, cell polarity, and cytoskeletal organization vating the NF-jB transcription factor axis by a second (Kooistra et al., 2007). SIPA1L3 is a member of the messenger pathway (Ueno et al., 2006). Activation of SPA1 family of RapGAPs, which play a crucial role in NF-jB, one of the most investigated transcription fac- spatiotemporal control of Rap1 activation in cells tors, controls multiple cellular processes in cancer, (Mochizuki et al., 2001). Rap1 plays many roles during including inflammation, transformation, proliferation, cell invasion and metastasis in different cancers (Zhang angiogenesis, invasion, metastasis, chemoresistance, et al., 2017). Additionally, overexpression of RAP1 and radioresistance (Chaturvedi et al., 2010). Nicotine may desensitize NSCLC cells to cisplatin, a first-line protects NSCLC cells against chemotherapy-induced Fig. 5. Diagram for pathway of DNA methylation–smoking cessation interaction effect on survival for LUAD patients. Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1243 Epigenomic–smoking interaction analysis of SIPA1L3 R. Zhang et al. apoptosis and serum deprivation-induced apoptosis studies are warranted to elucidate the mechanism of through NF-jB, and NF-jB activity is also directly cg02268510SIPA1L3 and smoking cessation interaction stimulated by nicotine (Anto et al., 2002; Tsurutani on LUAD survival. et al., 2005). Therefore, for current smokers, nicotine Meanwhile, we observed that some genes trans-regu- in tobacco stimulates activation of NF-jB, induces lated by cg02268510SIPA1L3 are involved in DNA dam- inflammatory responses, and is relevant to poor age response and cell growth (GADD45G) (Guo et al., patient prognosis (Fig. 5). 2013), immune cell functions (MTURN) (Sun et al., Moreover, Rap1 is an essential modulator of NF- 2014), tumor cell migration [regulator of G protein sig- jB-mediated pathways. NF-jB is induced by ectopic naling 20 (RGS20)] (Yang et al., 2016), apoptosis expression of Rap1, whereas its activity is inhibited (RELL1 and RIPK2) (Chin et al., 2002; Cusick et al., by Rap1 depletion (Teo et al., 2010). Furthermore, 2010), and innate and adaptive immunity (RIPK2) levels of Rap1 are positively regulated by NF-jB, (Jaafar et al., 2018). and human breast cancers with NF-jB hyperactivity GADD45G is a member of the GADD45 family, show elevated levels of cytoplasmic Rap1 (Teo et al., which plays an essential role in cellular stress 2010). Thus, positive feedback mechanisms might response, survival, senescence, and apoptosis regula- exist between Rap1 expression and NF-jB activation tion (Liebermann et al., 2011). GADD45G has been (Fig. 5). In terms of cg02268510SIPA1L3 and smoking reported to be a tumor suppressor in multiple cancer cessation interaction, keeping smoking was associated types and can inhibit cell growth and induce apopto- with poor prognosis only in LUAD patients with low sis (Ying et al., 2005). Patients with high GADD45G methylation, rather than medium or high methyla- expression had a better prognosis in our study. tion, possibly because high activation of both Rap1 MTURN is a neural progenitor differentiation regula- and NF-jB may only occur in patients with low tor homolog. 12-O-tetradecanoylphorbol-13-acetate methylation. (TPA) is an effective cancer therapeutic reagent for We also found that methylation level of myelocytic leukemia patients (Han et al., 1998), and cg02268510SIPA1L3 increased along with long pack-year MTURN is TPA-responsive and may promote both of smoking, but decreased with long years of smoking leukemic and normal megakaryocyte differentiation cessation. As presented in Fig. 5, low methylation and (Sun et al., 2014). Indeed, differentiation therapy by keeping smoking resulted in the worst prognosis, which forced differentiation of cancer cells has been success- might be due to the positive feedback in Rap1 and NF- ful in curing acute promyeloid leukemia (Chen et al., jB. However, methylation of cg02268510SIPA1L3 2011). Similarly, LUAD patients with high MTURN increased with the cumulative amount of smoking. But, expression had favorable survival in our study. high methylation of cg02268510SIPA1L3 resulted in low RGS20 is suggested to promote cellular characteristics SIPA1L3 expression that was hard to active Rap1 and that contribute to metastasis, including enhanced cell NF-jB, and then might weaken the harmful effect of aggregation, motility, and invasion. Selective inhibi- smoking, which also indicated an antagonistic effect. It tion of RGS20 expression may represent an alterna- is implied that there might be a self-protective mecha- tive means to suppress metastasis (Yang et al., 2016). nism in the human body that prevents the body from Its high expression is significantly associated with pro- receiving excessive damage from exposure. As reported, gression and prognosis of triple-negative breast cancer smoking increases reactive oxygen species (ROS) pro- (Li et al., 2017). Additionally, our study showed simi- duction and is a significant source of oxidative stress lar results in LUAD patients. Though there is a lack (Athanasios et al., 2013), but in vivo, there is a variety of explicit evidence of relevance between these genes of antioxidant defense mechanisms existed to counter- and smoking, what we found may inspire functional act the detrimental effects of ROS by regulating the studies of these potential genes and further help to production of free radicals and their metabolites complete a picture of the mechanism pathway of (Deponte, 2013; He et al., 2017). It may be an adaptive cg02268510SIPA1L3 and smoking cessation interaction defense mechanism to counteract the increased ROS on LUAD survival. production that superoxide dismutase enzyme levels in Our study has some significant strengths. First, this blood and salivary were increased in smokers (Jenifer is the first study to investigate the interaction between et al., 2015). Moreover, a previous study has found DNA methylation and smoking cessation on lung can- that activation of Rap1 serves to attenuate ROS pro- cer survival on an epigenome-wide scale, which pro- duction (Remans et al., 2004) and there is a potential vides new evidence to account for the missing interrelationship between Rap1, ROS, and NF-jB acti- heritability of complex diseases (Trerotola et al., vation (Moon et al., 2011). But further functional 2015). Second, the two-stage study design we used to 1244 Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. R. Zhang et al. Epigenomic–smoking interaction analysis of SIPA1L3 exhaustively search for interactions, as well as the sen- Acknowledgements sitivity analysis, is quite conservative in controlling for false positives. Third, our study included a large sam- The authors thank all participants enrolled in this ple size to analyze DNA methylation–smoking cessa- study. This study was supported by the National Key tion interactions of early-stage NSCLC prognosis, Research and Development Program of China providing an opportunity to identify complex associa- (2016YFE0204900 to FC); the National Natural tions with small–medium effect size. Science Foundation of China (81530088 and 81473070 Despite the strengths of our study, we acknowledge to FC, 81602940 to LLiu); the US National Institutes some limitations. First, data measured categorical of Health (CA209414, CA092824, and ES000002 to smoking cessation rather than smoking pack-years, DCC); the Natural Science Foundation of the Jiangsu which may render less power in the study. Second, Higher Education Institutions of China (18KJB310011 smoking cessation was collected at the time of diagno- to RZ and 14KJA310002 to FC); China Postdoctoral sis and was not reassessed during follow-up. Previous Science Foundation (2018M633767 to RZ); and the studies have found that ‘former smokers’ might more Priority Academic Program Development of Jiangsu accurately represent a mixed exposure status, since Higher Education Institutions. YW and RZ were par- quitters are more likely to relapse (Hughes et al., 2004; tially supported by the Outstanding Young Teachers Walker et al., 2006). Thus, we likely underestimated Training Program of Nanjing Medical University. the benefits of smoking cessation. Third, the associa- tion between cg02268510SIPA1L3 and expression of sev- Conflict of interest eral genes requires more biological evidence, though methylation is believed to play a crucial role in regu- The authors have no conflicts to declare. lating gene expression (Bird, 2007) and further influ- ence disease gene function (Sch€ ubeler, 2015), cell Author contributions differentiation, or reprogramming (Khavari et al., 2010). Thus, functional experiments are warranted to RZ, LLai, YW, and FC contributed to the study confirm these associations, so our findings should be design. RZ, SM, TF, MMB, AK, MP, JS, AH, ME, biologically interpreted with caution thus far. In addi- AS, and LS contributed to data collection. RZ, LLai, tion, our study consisted mainly of a Caucasian popu- and YW performed statistical analysis and interpreta- lation (89.19%), since TCGA data contained only tion and drafted the manuscript. XD, JH, DY, CC, ~ 10% non-Caucasian samples. Our results should LLin, YZ, HH, SS, LW, XC, YG, and LLiu revised therefore be translated with caution for other popula- the manuscript. All authors contributed to critical revi- tions. Lastly, the censored rate of survival time for the sion of the manuscript and approved its final version. TCGA population is relatively high, since early-stage Financial support and study supervision were provided NSCLC patients need longer follow-up time. Thus, the by RZ, FC, and DCC. validation phase using TCGA population had low sta- tistical power. However, we still successfully replicated References one significant interaction, indicating a quite conserva- tive and robust result (Leung et al., 1997; Watt et al., Anto RJ, Mukhopadhyay A, Shishodia S, Gairola CG and 1996). Aggarwal BB (2002) Cigarette smoke condensate activates nuclear transcription factor-kappaB through phosphorylation and degradation of IkappaB(alpha): 5. Conclusion correlation with induction of cyclooxygenase-2. This epigenome-wide DNA methylation–smoking Carcinogenesis 23, 1511–1518. cessation interaction analysis of early-stage Aran D, Sirota M and Butte AJ (2015) Systematic pan- NSCLC identified one LUAD-specific CpG probe, cancer analysis of tumour purity. Nat Commun 6, 8971. Athanasios V, Thomais V, Konstantinos F and Spyridon L cg02268510SIPA1L3, which could significantly modify (2013) Pulmonary oxidative stress, inflammation and effects of smoking cessation on lung cancer survival. cancer: respirable particulate matter, fibrous dusts and Smoking cessation benefited survival of LUAD ozone as major causes of lung carcinogenesis through patients with low methylation at cg02268510SIPA1L3. reactive oxygen species mechanisms. Int J Environ Res These results have implications for not only smoking Public Health 10, 3886–3907. cessation after diagnosis, but also possible methyla- Balduyck B, Sardari NP, Cogen A, Dockx Y, Lauwers P, tion-specific drug targeting. Hendriks J and Van SP (2011) The effect of smoking Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1245 Epigenomic–smoking interaction analysis of SIPA1L3 R. Zhang et al. cessation on quality of life after lung cancer surgery. Egger G, Liang G, Aparicio A and Jones PA (2004) Eur J Cardiothorac Surg 40, 1437–1438. Epigenetics in human disease and prospects for Baser S, Shannon VR, Eapen GA, Jimenez CA, Onn A, epigenetic therapy. Nature 429, 457. Lin E and Morice RC (2006) Smoking cessation Feinberg AP and Tycko B (2004) The history of cancer after diagnosis of lung cancer is associated with a epigenetics. Nat Rev Cancer 4, 143. beneficial effect on performance status. Chest 130, Flanders WD, Lally CA, Zhu BP, Henley SJ and Thun MJ 1784–1790. (2003) Lung cancer mortality in relation to age, Besse B, Adjei A, Baas P, Meldgaard P, Nicolson M, Paz- duration of smoking, and daily cigarette consumption: Ares L, Reck M, Smit EF, Syrigos K and Stahel R results from Cancer Prevention Study II. Cancer Res (2014) 2nd ESMO Consensus Conference on Lung 63, 6556. Cancer: non-small-cell lung cancer first-line/second and Fraga MF and Esteller M (2007) Epigenetics and aging: further lines of treatment in advanced disease. Ann the targets and the marks. Trends Genet 23, 413– Oncol 25, 1475. 418. Bhatt VR, Batra R, Silberstein PT Jr, Loberiza FR Jr and Guo Y, Zhang R, Shen S, Wei Y, Salama SM, Fleischer T, Ganti AK (2015) Effect of smoking on survival from Bjaanaes MM, Karlsson A, Planck M, Su L et al. non-small cell lung cancer: a retrospective Veterans’ (2018) DNA methylation of LRRC3B: a biomarker for Affairs Central Cancer Registry (VACCR) cohort survival of early-stage non-small cell lung cancer analysis. Med Oncol 32, 1–6. patients. Cancer Epidemiol Biomarkers Prev 27, 1527– Bird A (2007) Perceptions of epigenetics. Nature 447, 396. 1535. Bjaanæs MM, Fleischer T, Halvorsen AR, Daunay A, Guo W, Zhu T, Dong Z, Cui L, Zhang M and Kuang G Busato F, Solberg S, Jørgensen L, Kure E, Edvardsen (2013) Decreased expression and aberrant methylation H and Børresen-Dale AL (2016) Genome-wide DNA of Gadd45G is associated with tumor progression and methylation analyses in lung adenocarcinomas: poor prognosis in esophageal squamous cell carcinoma. association with EGFR, KRAS and TP53 mutation Clin Exp Metastasis 30, 977–992. status, gene expression and prognosis. Mol Oncol 10, Han ZT, Zhu XX, Yang RY, Sun JZ, Tian GF, Liu XJ, 330. Cao GS, Newmark HL, Conney AH and Chang RL Chaturvedi MM, Sung B, Yadav VR, Kannappan R and (1998) Effect of intravenous infusions of 12-O- Aggarwal BB (2010) NF-jB addiction and its role in tetradecanoylphorbol-13-acetate (TPA) in patients with cancer: ‘one size does not fit all’. Oncogene 30, 1615– myelocytic leukemia: preliminary studies on therapeutic 1630. efficacy and toxicity. Proc Natl Acad Sci USA 95, Chen YA, Mathieu L, Sanaa C, Butcher DT, Daria G, 5357. Zanke BW, Steven G, Hudson TJ and Rosanna W He L, He T, Farrar S, Ji L, Liu T and Ma X (2017) (2013) Discovery of cross-reactive probes and Antioxidants maintain cellular redox homeostasis by polymorphic CpGs in the Illumina Infinium elimination of reactive oxygen species. Cell Physiol HumanMethylation450 microarray. Epigenetics 8, 203– Biochem 44, 532–553. 209. Hughes JR, Keely J and Naud S (2004) Shape of the Chen SJ, Zhou GB, Zhang XW, Mao JH, De TH and relapse curve and long-term abstinence among Chen Z (2011) From an old remedy to a magic bullet: untreated smokers. Addiction 99, 29. molecular mechanisms underlying the therapeutic Jaafar R, Mnich K, Dolan S, Hillis J, Almanza A, Logue effects of arsenic in fighting leukemia. Blood 117, 6425– SE, Samali A and Gorman AM (2018) RIP2 enhances 6437. cell survival by activation of NF-KB in triple negative Chin AI, Dempsey PW, Bruhn K, Miller JF, Xu Y and breast cancer cells. Biochem Biophys Res Commun 497, Cheng G (2002) Involvement of receptor-interacting 115–121. protein 2 in innate and adaptive immune responses. Jenifer HD, Bhola S, Kalburgi V, Warad S and Kokatnur Nature 416, 190. VM (2015) The influence of cigarette smoking on Crusz SM and Balkwill FR (2015) Inflammation and blood and salivary super oxide dismutase enzyme levels cancer: advances and new agents. Nat Rev Clin Oncol among smokers and nonsmokers—A cross sectional 12, 584. study. J Tradit Complement Med 5, 100–105. Cusick JK, Mustian A, Goldberg K and Reyland ME Karlsson A, J€ onsson M, Lauss M, Brunnstr€ om H, J€ onsson (2010) RELT induces cellular death in HEK 293 P, Borg A, J€ onsson G, Ringner M, Planck M and epithelial cells. Cell Immunol 261, 1. Staaf J (2014) Genome-wide DNA methylation analysis Deponte M (2013) Glutathione catalysis and the reaction of lung carcinoma reveals one neuroendocrine and four mechanisms of glutathione-dependent enzymes. adenocarcinoma epitypes associated with patient Biochim Biophys Acta 1830, 3217–3266. outcome. Clin Cancer Res 20, 6127–6140. 1246 Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. R. Zhang et al. Epigenomic–smoking interaction analysis of SIPA1L3 Khavari DA, Sen GL and Rinn JL (2010) DNA Parsons A, Daley A, Begh R and Aveyard P (2010) methylation and epigenetic control of cellular Influence of smoking cessation after diagnosis of early differentiation. Cell Cycle 9, 3880–3883. stage lung cancer on prognosis: systematic review of Kooistra MR, Dube N and Bos JL (2007) Rap1: a key observational studies with meta-analysis. BMJ 340, 251. regulator in cell-cell junction formation. J Cell Sci 120, Remans PHJ, Gringhuis SI, van Laar JM, Sanders ME, 17. Zwartkruis FJT, Levarht EWN, Marcela R, Coffer PJ Koshiaris C, Aveyard P, Oke J, Ryan R, Szatkowski L, and Breedveld FC (2004) Rap1 signaling is required Stevens R and Farley A (2017) Smoking cessation and for suppression of Ras-generated reactive oxygen survival in lung, upper aero-digestive tract and bladder species and protection against oxidative stress in T cancer: cohort study. Br J Cancer 117, 1224–1232. lymphocytes. J Immunol 173, 920–931. Lacombe L, Fradet V, Levesque E, Pouliot F, Larue H, Sandoval J, Mendez-Gonzalez J, Nadal E, Chen G, Bergeron A, Hovington H, Caron A, Nguile-Makao Carmona FJ, Sayols S, Moran S, Heyn H, Vizoso M M, Harvey M et al. (2016) Phase II drug-metabolizing and Gomez A (2013) A prognostic DNA methylation polymorphisms and smoking predict recurrence of non- signature for stage I non-small-cell lung cancer. J Clin muscle-invasive bladder cancer: a gene-smoking Oncol 31, 4140. interaction. Cancer Prev Res (Phila) 9, 189–195. Sch€ubeler D (2015) Function and information content of Lee KWK and Pausova Z (2013) Cigarette smoking and DNA methylation. Nature 517, 321. DNA methylation. Front Genet 4, 132. Shen S, Zhang R, Guo Y, Loehrer E, Wei Y, Zhu Y, Yuan Leung KM, Elashoff RM and Afifi AA (1997) Censoring Q, Moran S, Fleischer T, Bjaanaes MM et al. (2018) A issues in survival analysis. Annu Rev Public Health 18, multi-omic study reveals BTG2 as a reliable prognostic 83. marker for early-stage non-small cell lung cancer. Mol Li Q, Jin W, Cai Y, Yang F, Chen E, Ye D, Wang Q and Oncol 12, 913–924. Guan X (2017) Regulator of G protein signaling 20 Shui IM, Wong CJ, Zhao S, Kolb S, Ebot EM, Geybels correlates with clinicopathological features and MS, Rubicz R, Wright JL, Lin DW and Klotzle B prognosis in triple-negative breast cancer. Biochem (2016) Prostate tumor DNA methylation is associated Biophys Res Commun 485, 693–697. with cigarette smoking and adverse prostate cancer Liebermann DA, Tront JS, Sha X, Mukherjee K, outcomes. Cancer 122, 2168–2177. Mohamedhadley A and Hoffman B (2011) Gadd45 Siegel RL, Miller KD and Jemal A (2017) Cancer statistics, stress sensors in malignancy and leukemia. Crit Rev 2018. CA Cancer J Clin 67, 7. Oncog 16, 129. Slade E and Kraft P (2016) Leveraging methylome- Marabita F, Almgren M, Lindholm ME, Ruhrmann S, environment interaction to detect genetic determinants Fagerstr€ombillai F, Jagodic M, Sundberg CJ, Ekstr€ om of disease. Hum Hered 81, 26–34. TJ, Teschendorff AE and Tegner J (2013) An Suk HR, Zhou W, Cogandrew T, Liu G, Su L, Neuberg evaluation of analysis pipelines for DNA methylation D, Lynch TJ, Wain JC and Christiani DC (2006) profiling using the Illumina HumanMethylation450 MDM2 polymorphism and recurrence-free and overall BeadChip platform. Epigenetics 8, 333. survival in early stage non-small cell lung cancer Mcnerney ME, Godley LA and Le BM (2017) Therapy- (NSCLC). J Clin Oncol 24 (Suppl. 18), 7221. related myeloid neoplasms: when genetics and Sun X, Lu B, Hu B, Xiao W, Li W and Huang Z (2014) environment collide. Nat Rev Cancer 17, 513. Novel function of the chromosome 7 open reading Mochizuki N, Yamashita S, Kurokawa K, Ohba Y, Nagai frame 41 gene to promote leukemic megakaryocyte T, Miyawaki A and Matsuda M (2001) Spatio- differentiation by modulating TPA-induced signaling. temporal images of growth-factor-induced activation of Blood Cancer J 4, e198. Ras and Rap1. Nature 411, 1065. Teo H, Ghosh S, Luesch H, Ghosh A, Wong ET, Malik Moon EY, Lee JH, Lee JW, Song JH and Pyo S (2011) N, Orth A, Jesus PD, Perry AS and Oliver JD (2010) ROS/Epac1-mediated Rap1/NF-kappaB activation is Telomere-independent Rap1 is an IKK adaptor and required for the expression of BAFF in Raw264.7 regulates NF-jB-dependent gene expression. Nat Cell murine macrophages. Cell Signal 23, 1479–1488. Biol 12, 758–767. Nia PS, Weyler J, Colpaert C, Vermeulen P, Marck EV Trerotola M, Relli V, Simeone P and Alberti S (2015) and Schil PV (2005) Prognostic value of smoking Epigenetic inheritance and the missing heritability. status in operated non-small cell lung cancer. Lung Hum Genomics 9, 17. Cancer 47, 351–359. Tsurutani J, Castillo SS, Brognard J, Granville CA, Zhang Park ER, Japuntich SJ, Rigotti NA, Traeger L, He Y, C, Gills JJ, Sayyah J and Dennis PA (2005) Tobacco Wallace RB, Malin JL, Zallen JP and Keating NL components stimulate Akt-dependent proliferation and (2012) A snapshot of smokers after lung and colorectal NFkappaB-dependent survival in lung cancer cells. cancer diagnosis. Cancer 118, 3153. Carcinogenesis 26, 1182–1195. Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd. 1247 Epigenomic–smoking interaction analysis of SIPA1L3 R. Zhang et al. Ueno H, Pradhan S, Schlessel D, Hirasawa H and Sumpio Zhang YL, Wang RC, Cheng K, Ring BZ and Li S (2017) BE (2006) Nicotine enhances human vascular Roles of Rap1 signaling in tumor cell migration and endothelial cell expression of ICAM-1 and VCAM-1 invasion. Cancer Biol Med 14, 90–99. via protein kinase C, p38 mitogen-activated Zhang N, Wu HJ, Zhang W, Wang J, Wu H and Zheng X protein kinase, NF-jB, and AP-1. Cardiovasc Toxicol (2015) Predicting tumor purity from methylation 6, 39–50. microarray data. Bioinformatics 31, 3401. Walker MS, Vidrine DJ, Gritz ER, Larsen RJ, Yan Y, Zheng X, Zhang N, Wu HJ and Wu H (2017) Estimating Govindan R and Fisher EB (2006) Smoking relapse and accounting for tumor purity in the analysis of during the first year after treatment for early-stage DNA methylation data from cancer studies. Genome non–small-cell lung cancer. Cancer Epidemiol Biol 18, 17. Biomarkers Prev 15, 2370. Walser T, Cui X, Yanagawa J, Lee JM, Heinrich E, Lee G, Sharma S and Dubinett SM (2008) Smoking and lung Supporting information cancer: the role of inflammation. Proc Am Thorac Soc Additional supporting information may be found 5, 811–815. online in the Supporting Information section at the end Wan ES, Weiliang Q, Andrea B, Carey VJ, Helene B, of the article. Rennard SI, Alvar A, Wayne A, Lomas DA and Fig. S1. Quality control processes for DNA methyla- Demeo DL (2012) Cigarette smoking behaviors and tion chip data. time since quitting are associated with differential Fig. S2. Manhattan plot of DNA methylation–smok- DNA methylation across the human genome. Hum ing cessation interaction P-values (A) and main effect Mol Genet 21, 3073. P-values (B) derived from histology-stratified Cox pro- Watt DC, Aitchison TC, Mackie RM and Sirel JM (1996) portional hazards model in the discovery phase. Survival analysis: the importance of censored Fig. S3. Fixed-effect meta-analysis of interaction observations. Melanoma Res 6, 379–385. Wei Y, Liang J, Zhang R, Guo Y, Shen S, Su L, Lin X, between DNA methylation of cg02268510 and smok- Moran S, Helland A  and Bjaanæs MM (2018) ing cessation for LUAD patients from five centers. Epigenetic modifications in KDM lysine demethylases Fig. S4. Linear regression analysis between methyla- associate with survival of early-stage NSCLC. Clin tion of cg02268510 and age (A) as well as smoking- Epigenetics 10, 41. related variables: pack-year of smoking (B), year of Wright J (2013) Epigenetics: reversible tags. Nature 498, smoking (C), and year of smoking cessation (D), 10–11. adjusted for age, sex, smoking status, clinical stage, Yang L, Lee MM, Leung MM and Wong YH (2016) and study center. Regulator of G protein signaling 20 enhances cancer Fig. S5. Genome-wide methylation transcription analy- cell aggregation, migration, invasion and adhesion. Cell sis results from the TCGA cohort. (A) Circos plot of Signal 28, 1663–1672. genome-wide gene expression. For plots in B–H, left Ying J, Srivastava G, Hsieh W-S, Gao Z, Murray P, Liao panels show correlation of (B) GADD45G, (C) S-K, Ambinder R and Tao Q (2005) The stress- MTURN, (D) TMEM200B, (E) RGS20, (F) RELL1, responsive gene GADD45G is a functional tumor (G) PGM2, and (H) RIPK2 expression (X-axis) with suppressor, with its response to environmental stresses methylation of cg02268510 (Y-axis). Right panels show frequently disrupted epigenetically in multiple tumors. Kaplan–Meier survival plots of gene expression Clin Cancer Res 11, 6442–6449. divided into low and high groups by median value. Zaghlool SB, Al-Shafai M, Al Muftah WA, Kumar P, Table S1. Demographic and clinical characteristics of Falchi M and Suhre K (2015) Association of DNA early-stage NSCLC patients in TCGA dataset. methylation with age, gender, and smoking in an Arab Table S2. Results for 15 methylation–smoking interac- population. Clin Epigenetics 7, 1–12. tions using a two-stage association study. Zhang R, Lai L, He J, Chen C, You D, Duan W, Dong X, Table S3. Smoking-related characteristics of former Zhu Y, Lin L, Shen S et al. (2019) EGLN2 DNA and current smokers in early-stage LUAD. methylation and expression interact with HIF1A to affect survival of early-stage NSCLC. Epigenetics 14, 118–129. 1248 Molecular Oncology 13 (2019) 1235–1248 ª 2019 The Authors. Published by FEBS Press and John Wiley & Sons Ltd.