| Overweight and obesity,as a major health issue worldwide,is linked to increased risk of cardiovascular diseases,type II diabetes,hyperlipidemia and multiple malignant tumors,including liver cancer,colorectal cancer and postmenopausal breast cancer.However,the association between body mass index(BMI,a measure of overall body fatness)and incident lung cancer has demonstrated inconsistent results.Several prospective cohort studies found the inverse association between BMI and incident lung cancer risk,while others did not observe the significant association between them.Until now,there is no clear explanation about the underlying mechanisms between BMI and lung cancer risk.Obesity could alter the level of DNA methylation.DNA methylation,which refers to the covalent modification of cytosine in the 5'–cytosine–phosphate–guanine–3'(Cp G),is the well–known epigenetic mechanism.The advent of high–throughput DNA methylation assay has greatly advanced the epigenome–wide association study(EWAS)of obesity and its related traits.A number of studies have found numerous BMI–related Cp Gs among Europeans and Americans,which provide the important evidence for obesity epigenetic study.Lung cancer is a multifactorial disease involving both environmental and genetic factors.Alteration of DNA methylation has been found to play a critical role in the development and prognosis of lung cancer.DNA methylation is involved in carcinogenesis mainly via regulating the expression of genes and impeding the stability of genome.Numerous lung cancer–related Cp Gs have been reported in the peripheral blood,which is of great significance to understand the pathogenesis of lung cancer as well as to explore more efficient target sites for the treatment of lung cancer.DNA methylation changes secondarily to the environmental exposure,and acts as a bridge linking environmental and phenotypic changes.With the development of environmental epigenetics,some researchers focus on the obesity–epigenetics–cancer risk.Obesity associated reprogramming of the epigenome via DNA methylation may alter the expression of genes that promote or inhibit tumor progression.However,the role of DNA methylation in the relationship between BMI and lung cancer risk remains to be elucidated.Based on the findings above,we aimed to investigate associations of BMI and its related DNA methylation with lung cancer risk.First,to investigate the association of baseline BMI and incident lung cancer,27 009 participants at baseline and 14 120 new subjects in the first follow–up in 2013 based on Dongfeng–Tongji prospective cohort study,were enrolled in this study and follow up to the end of 2018.Moreover,we searched Pubmed,Embase,China National Knowledge Infrastructure(CNKI)database and Wanfang database.Thirty–two prospective cohort studies were included to perform a Meta–analysis between BMI and incident lung cancer.In order to discover the causal relationship between BMI and lung cancer risk,a two–sample Mendelian randomization study were conducted,based on an aggregated genome–wide association study(GWAS)of BMI in 173 430 Japanese and an aggregated GWAS of lung cancer in 212 453 Japanese.Then a two–stage EWAS analysis in 2266 Chinese participants was performed to discover BMI–related DNA methylation alterations.A total of 1810 participants from the previous work of our research group were designed as the discovery set and two lung cancer case–control studies,including 456 subjects,were designed as the validation set.For the validated BMI–related Cp Gs,we also explored their correlations with the expression of nearby genes.Finally,above mentioned two lung cancer case–control studies were used to evaluate the relationship between BMI–related DNA methylation and lung cancer risk,as well as the mediation effects of DNA methylation on the association of BMI with lung cancer.The diagnostic values of BMI–related DNA methylation on lung cancer were also explored.This paper mainly includes three parts:Part ?.The association of body mass index with incident lung cancerObjectives: To investigate the association of BMI with incident lung cancer risk based on large prospective cohort studies and further explore the causal relationship between BMI and lung cancer by conducting Mendelian randomization analysis.Methods: Based on Dongfeng–Tongji prospective cohort study,we recruited 27 009 retired staffs at baseline in 2008 and 14 120 new participants in the first follow–up in 2013.Of these subjects,who had previous history of cancer(n = 1157),who had lost of follow–up(n = 827)and who had missing or outlier information of BMI(n = 2650)were excluded.Additionally,until the end of 2018,the new incident cases with the other cancer types except lung cancer,were also excluded(n = 2584).Finally,33 911 subjects(including 686 new incident lung cancer cases)were kept for the further analyses.The information about general demographic characteristics,lifestyles and disease history were collected by a questionnaire from all subjects.Height and weight were measured by trained research personnel.The incident lung cancer cases were confirmed by reviewing medical records.We classified subjects based on BMI into the following subgroups: BMI < 18.5,18.5 ~ 24.9,25.0 ~ 29.9,? 30.0,and ? 25.0 kg/m2.Multivariable Cox regression model was used to evaluate the association of BMI and incident lung cancer risk,and calculate the hazard ratio(HR)and its 95% confidence interval(95%CI).Furthermore,stratified analyses by age(? 65 years and > 65 years),gender(males and females),smoking status(current smokers,former smokers and never smokers)and drinking status(current drinkers,former drinkers and never drinkers)were also conducted.On the basis of Dongfeng–Tongji cohort study,we further conducted a literature search in the Pubmed,Embase,CNKI and Wanfang database until the September 2020 to perform a Meta–analysis between BMI and incident lung cancer.The categories of BMI were based on the following criteria: < 18.5,18.5 ~ 24.9 and ? 25.0 kg/m2.Normal BMI(18.5 ~ 24.9 kg/m2)was considered as the reference group.When it was reported by different BMI categories in the original literatures,the estimates for alternative comparisons were converted using the method by Der Simonian and Larid.Reporting biases were assessed with Begg's rank correlation method,Egger's weighted regression method and a funnel plot.Heterogeneity was assessed with Cochran's Q test and I2.The pooled risk ratio(RR)and 95%CI of BMI on lung cancer risk were analyzed with random effects model based on the heterogeneity of these literatures.In addition,subgroup analyses were performed according to gender(males and females),smoking status(current smokers,former smokers and never smokers),and lung cancer subtypes(lung adenocarcinoma,lung squamous cell carcinoma and small cell lung cancer).Mendelian randomization analysis uses single nucleotide polymorphism(SNP)as the instrumental variable to estimate causal relationship between risk factors and outcome variables and is free from the effects of confounders and reverse causality.In order to explore the causal relationship between BMI and lung cancer risk,we conducted a two–sample Mendelian randomization study.The summarized GWAS data of BMI and lung cancer with the hitherto largest sample sizes among Asian populations were obtained by using a public GWAS database MRC–IEU(https://gwas.mrcieu.ac.uk/).The GWAS data of BMI was based on 171 430 Japanese populations.And the GWAS data of lung cancer was based on 4050 Japanese lung cancer cases and 208 403 controls.We selected BMI–related SNPs as instrumental variables at the genome–wide significant level(P < 5.0E-08).Inverse–variance weighted(IVW)method was used to infer the causal relationship between BMI and lung cancer.In addition,heterogeneity analysis(Q test)as well as horizontal pleiotropy analyses(MR–Egger regression and MR–PRESSO global test)were also conducted to ensure the reliability of results.Results: In Dongfeng–Tongji cohort study,compared with normal BMI subjects,subjects with BMI during 25.0 ~ 29.9 kg/m2 and BMI ? 30.0 kg/m2 had the proximally significant effects on incident lung cancer risk [HR(95%CI)= 0.88(0.72,1.08);0.61(0.35,1.07)].BMI ? 25.0 kg/m2 had a significantly lower risk of lung cancer [HR(95%CI)= 0.85(0.70,0.99)].Each 5 kg/m2 increase in BMI lead to 23% decrease in the incidence of lung cancer [HR(95%CI)= 0.77(0.66,0.89)].Stratified analysis showed that the effect of BMI ? 25.0 kg/m2 on decreased risk of incident lung cancer was mainly shown among elders(age > 65 years)[HR(95%CI)= 0.84(0.66,1.05)],males [HR(95%CI)= 0.82(0.66,1.01)],current smokers [HR(95%CI)= 0.64(0.48,0.86)] and current alcohol drinkers [HR(95%CI)= 0.66(0.48,0.90)].Thirty–two prospective cohort studies were included in the Meta–analysis,containing 145 808 incident lung cancer cases.The combined RR with 95%CI for BMI ? 25.0 kg/m2 vs.normal category of BMI was 0.81(0.77,0.84).And the inverse association still remained after stratification by gender and smoking status.After stratification by gender,among males,RR(95%CI)= 0.76(0.71,0.83)for BMI ? 25.0 kg/m2;among females,RR(95%CI)= 0.83(0.77,0.89)for BMI ? 25.0 kg/m2.After stratification by smoking status,compared with normal BMI,the RR and 95%CI for BMI ? 25.0 kg/m2 was 0.78(0.73,0.83),0.80(0.73,0.88)and 0.84(0.77,0.91)for current smokers,former smokers and never smokers,respectively.After further stratification by the histological types,BMI was still inversely associated with the risk of lung adenocarcinoma and lung squamous cell carcinoma [RR(95%CI)= 0.80(0.74,0.87);0.82(0.72,0.95)].However,the significantly inverse association disappeared for small cell lung cancer [RR(95%CI)= 1.05(0.95,1.17)].In the Mendelian randomization analysis of BMI with lung cancer risk,we observed 83 BMI–related SNPs from the GWAS data of BMI,and among them,the genotype information of 78 SNPs could be derived from the GWAS data of lung cancer.After excluding SNPs which showed potentially horizontal pleiotropy and heterogeneity,the remaining 34 SNPs were used as instrumental variables.The MR–Egger regression provided little evidence of horizontal pleiotropy as its intercept was not significantly deviated from zero [b(95% CI)= 1.01(0.97,1.05),P = 0.528].The MR–PRESSO global test also did not show horizontal pleiotropy(P = 0.115).Heterogeneity test did not show evidence for heterogeneity among instrumental variables(Q test P = 0.100).IVW method showed that genetically determined BMI was significantly associated with lung cancer risk [OR(95%CI)= 0.58(0.39,0.88)].Conclusions: Compared with normal BMI,BMI ? 25.0 kg/m2 was inversely associated with incident lung cancer risk.In the Meta–analysis of prospective cohort studies,the significant protective effect of BMI ? 25.0 kg/m2 on lung cancer risk was still observed after stratification by gender and smoking status.However,when stratifying by histological types,the inverse association was mainly shown for the risk of non–small cell lung cancer(lung adenocarcinoma and lung squamous cell carcinoma),and it disappeared for the risk of small cell lung cancer.The Mendelian randomization analysis of BMI and lung cancer risk suggested that there was causal relationship between genetically determined BMI and decreased lung cancer risk.Currently,the mechanisms for the association between BMI and lung cancer risk remain unclear.Further studies should be implemented to explore the underlying mechanisms.Part ?.Epigenome–wide association of DNA methylation with body mass indexObjectives: DNA methylation,as an important epigenetic mechanism,plays a crucial role in gene expression regulation.We aimed to identify BMI–related DNA methylation sites,and investigate their relationship with gene expression.Methods: This part used a two–stage study design.The discovery stage,based on the previous study of our research group,contained 1810 individuals from several Chinese population cohort studies,including 198 acute coronary syndrome patients from Hubei and Guangdong Province,261 participants from Wuhan–Zhuhai Cohort study,137 participants from Wuhan Coke–oven Worker study,191 pairs of acute coronary syndrome nested cases and controls from the Shanghai Women's and Men's Health Study,344 pairs of acute coronary syndrome cases and controls from the Dongfeng–Tongji cohort study and 144 healthy subjects from Shiyan City,Hubei Province.The validation stage contained two lung cancer case–control studies,including 105 lung cancer cases and 103 controls as well as 125 lung cancer cases and 123 controls,respectively.Among them,cases were the lung cancer patients who underwent surgical resection of primary lung cancer during 2012 to 2019 in the Department of Thoracic Surgery,Tongji Hospital in China.And the controls were 1:1 frequency matched with cases at recruitment for age and gender from the physical examination center,Tongji Hospital,during the same periods with patients.The Illumina Human Methylation Bead Chip was used to assay DNA methylation.Multivariate linear regression models,with adjustment for age,gender,smoking status,alcohol drinking status and all surrogate variables calculated by R package Smart SVA,were used to test the association of methylation level at each Cp G with BMI.The result of each group was combined by using a fixed–effect Meta–analysis.The genome–wide significance level was set at FDR < 0.05 in the discovery stage.P value < 0.05 for the association between BMI and DNA methylation was defined as the significance threshold in the validation stage.In order to explore the association between DNA methylation and gene expression,gene expression profiles were assayed using Human HT–12 v4 Bead Chip for the above mentioned 144 participants from Shiyan City,Hubei Province.The correlations between BMI–related CpGs and the expression level of their annotated genes were tested by multivariate linear regression models.Results: The Meta–analysis of DNA methylation and BMI identified a total of 20 Cp Gs at genome–wide significance level in the discovery stage(FDR < 0.05).The methylation levels of 4 Cp Gs(cg12593793,cg17061862,cg11024682,and cg06500161,annotated to LMNA,ZNF143,SREBF1,and ABCG1,respectively)were found to be significantly associated with BMI in the validation stage(P < 0.05).Meta–analysis of both stages confirmed the robust negative associations of BMI with cg12593793 and cg17061862(b =-0.72,P = 1.23E-13 and b =-0.92,P = 2.83E-10,respectively),and the significantly positive associations of BMI with cg11024682 and cg06500161(b = 1.07,P = 1.03E-16 and b = 1.35,P = 1.61E-17,respectively).The DNA methylation levels of two BMI–related Cp Gs were significantly associated with the expressions of their annotated genes(cg11024682–SREBF1: b =-7.30,P = 3.43E-03;cg06500161–ABCG1: b =-10.67,P = 7.80E-04 and b =-11.76,P = 1.88E-04).Conclusions: In the two–stage designed study,we identified and validated 4 BMIrelated DNA methylation markers: cg12593793 in LMNA,cg17061862 in ZNF143,cg11024682 in SREBF1,and cg06500161 in ABCG1.Among them,the hypermethylation of cg11024682 and cg06500161 were linked to decreased gene expression levels of SREBF1 and ABCG1,respectively.Part ?.Associations of BMI–related DNA methylation signatures with lung cancer riskObjectives: Previous studies have established the associations of BMI–related DNA methylation with the risk of multiple cancers,including breast cancer,colorectal cancer and endometrial cancer.However,the association between BMI–related DNA methylation with lung cancer risk still remains largely unknown.We aimed to investigate the associations of BMI–related DNA methylation signatures with lung cancer risk in lung cancer case–control studies.We further discovered the mediation effect of DNA methylation on the association of BMI with lung cancer and assessed the diagnostic value of DNA methylation on lung cancer.Methods: Based on two lung cancer case–control studies described in Part II,we used multivariate logistic regression models,with adjustment for age,gender,smoking status,alcohol drinking status and BMI,to investigate the associations of BMI–related DNA methylation with lung cancer risk.Besides,stratified analyses based on age,gender,smoking status and drinking status as well as interaction analyses were also performed by using logistic regression models.The association heterogeneity between two lung cancer subtypes(lung adenocarcinoma and lung squamous cell carcinoma)was testified by using multivariate linear regression models.To evaluate the effects of DNA methylation on the association of BMI with incident lung cancer,causal mediation analyses were conducted based on two statistical models(the mediator model and the outcome model)by using the “%mediation” SAS macro.Finally,the diagnostic performance of Cp Gs and their methylation risk score(MRS)on lung cancer risk were further assessed by using 5–fold cross–validation receiver operating characteristic curve(ROC).Results: In lung cancer case–control studies,by using multivariate logistic regression models,methylation levels of cg12593793,cg11024682 and cg06500161 were observed to be inversely associated with lung cancer risk [OR(95%CI)= 0.22(0.16,0.31),0.39(0.30,0.50)and 0.66(0.53,0.82),respectively].The stratified analyses by age,gender,smoking status and alcohol drinking status did not modify the above associations(Pinteraction>0.05).In secondary analysis of the separate associations of these Cp Gs with risk of lung adenocarcinoma and lung squamous cell carcinoma,we did not find significant heterogeneity between two pathological lung cancer subtypes(P ? 0.05).To further testify the possible mediation effects of above three Cp Gs,we found that the methylation levels of cg11024682 and cg06500161 could mediated a separate 45.3% and 19.5% of the association between BMI and lung cancer risk(P < 0.001 and P = 0.013,respectively).However,we did not observe the mediation role of cg12593793(P = 0.315).MRS based on the above mentioned 3 Cp Gs could mediate 51.3% of the association between BMI and lung cancer risk(P < 0.001).Compared with the risk prediction model based on traditional factors(age,gender,smoking pack–years and BMI),MRS significantly improved 8.5% of the diagnostic performance for lung cancer [MRS + traditional factors vs.traditional factors,area under curve(AUC)= 0.817 vs.0.732)].Conclusions: In lung cancer case–control studies,we found three BMI–related DNA methylation markers(LMNA cg12593793,SREBF1 cg11024682 and ABCG1 cg06500161)were associated with decreased lung cancer risk.The blood methylation levels of cg11024682 and cg06500161 had significant mediation effects on the association of BMI with lung cancer risk.The MRS based on above mentioned 3 Cp Gs combined with traditional factors possessed satisfactory diagnostic value in lung cancer.However,due to the cross-sectional design of current study,no clear causal relationship can be inferred.Further cohort studies and bi–directional Mendelian randomization studies based on large populations are required to infer the causal relationship between BMI–related DNA methylation and lung cancer. |
PDF Full Text Request