AHRR (cg05575921) methylation extent of leukocyte DNA and lung cancer survival

Background Prior studies have shown that AHRR (cg05575921) hypomethylation may be a marker of smoking, lung cancer risk and potentially lung cancer survival (in some lung cancer subtypes). It is unknown if AHRR (cg05575921) hypomethylation is associated with reduced survival among lung cancer patients. Methods In bisulfite treated leukocyte DNA from 465 lung cancer patients from the Copenhagen prospective lung cancer study, we measured AHRR (cg05575921) methylation. 380 died during max follow-up of 4.4 years. Cox proportional hazard models were used to analyze survival as a function of AHRR (cg05575921) methylation. Results We observed the expected inverse correlation between cumulative smoking and AHRR methylation, as methylation (%) decreased (Coefficient -0.03; 95% confidence interval, -0.04- -0.02, p = 8.6x10-15) for every pack-year. Cumulative smoking > 60 pack-years was associated with reduced survival (hazard ratio and 95% confidence interval 1.48; 1.05–2.09), however, AHRR (cg05575921) methylation was not associated with survival when adjusted for sex, body mass index, smoking status, ethnicity, performance status, TNM Classification, and histology type of lung cancer. Conclusion AHRR (cg05575921) methylation is linked to smoking but does not provide independent prognostic information in lung cancer patients.


Introduction
Lung cancer is the leading cause of cancer-related death worldwide [1]. The overall 5-year survival of lung cancer patients remains below 20% [2][3], mainly because tumors are diagnosed at late stages [4]. Patients with advanced stage lung cancer are often treated with systemic chemotherapy or immunotherapy [5] with a limited effect on the prognosis. Introduction of medical therapies targeted to specific oncogenic pathways might have the potential to improve survival for a subset of these patients [6]. However, biomarkers for identification of patients that specifically benefits from therapy are needed.
Alteration in DNA methylation is an epigenetic modification that are common in human malignancies [7]. Tobacco smoke alter DNA methylation and thereby influence disease through complex mechanisms [8][9]. Altered DNA methylation is caused by tobacco smoking with hypomethylation at cg05575921 located at the third intron of the aryl-hydrocarbon receptor repressor (AHRR) [10][11][12][13][14][15][16][17][18]. In the general population, AHRR (cg05575921) hypomethylation of leukocyte DNA is associated with both former and current smoking status, daily cigarette consumption, cumulative smoking and smoking duration in both former and current smokers, and time since quitting in former smokers [19]. This suggests that blood based AHRR (cg05575921) methylation extent can be used as an objective measure of long-term smoking exposure.
Tobacco smoking is a prognostic marker among lung cancer patients overall [20], and after treatment with EGFR-tyrosine kinase inhibitors [21]. Importantly, AHRR (cg05575921) methylation extent has been linked to lung cancer risk [15,17,19], and seems to be a more accurate risk estimator of lung cancer than self-reported smoking [19]. Additionally, DNA methylation plays a role in lung tumor initiation and progression [22][23]. It is unclear if AHRR methylation in itself is a cause of increased lung cancer risk. However, hypomethylation of AHRR has been associated with higher AHRR expression in monocytes [24], lymphoblasts and pulmonary macrophages [25], entailing decreased capability to metabolize and remove harmful agents such as polycyclic aromatic hydrocarbons, that are contained in tobacco smoke, potentially leading to development of tobacco-related lung damage.
We hypothesized that AHRR (cg05575921) hypomethylation, as a strong biomarker of long-term smoking behavior, offers additional information beyond that offered by selfreported smoking to identify lung cancer patients that specifically benefits from therapy. Thus, we examined the association between AHRR (cg05575921) hypomethylation in bisulfite treated leukocyte DNA and survival among lung cancer patients from the Lung Cancer Study (LUCAS), also referred to as the Copenhagen study [26].

Study cohort
465 consecutive individuals of the Copenhagen Study [26] who had lung cancer diagnosed at Department of Respiratory Medicine, Bispebjerg University Hospital in Denmark between February 2012 and December 2013. Individuals were excluded if ECOG performance status was above 3 at first assessment by the oncologist. Cancer diagnose was cytologically or histologically verified, and staged according to the TNM Classification of Malignant Tumors (TNM).
At first assessment, individuals were asked about smoking status. Former smoking was defined as no smoking within last 6 months. Smoking status included number of daily consumed cigarettes, cheroots, cigars or pipes, age at smoking initiation and cessation. Pack-years was consumption of 20 cigarettes or equivalent per day for 1 year. Body mass index (BMI) was calculated as measured weight in kilos divided by measured height in meters squared, performance status was assessed according to ECOG classification, and a blood sample were drawn.
The study was approved by Herlev and Gentofte Hospital, the Danish Data Inspection Agency (KF100.2039/91), and was conducted according to the Declaration of Helsinki. Written informed consent was obtained from all participants.

AHRR hypomethylation
The AHRR (cg05575921) methylation extent was measured in duplicate samples using a Taqman assay developed in our own laboratory, and validated against results from pyrosequencing. Bisulphite treated leucocyte DNA drawn from peripheral blood, was amplified using forward and reverse PCR primers, which were designed to bind to DNA around the cg05575921 site on sequences without genetic or possible CpG methylation variation. Details of the AHRR methylation measurements have been published previously [19].

Survival
Date of death was found by linkage with the Danish Civil Registration System until January 2017. There was no loss to follow-up. The underlying cause of death until March 2017 were obtained from the Danish Register of Causes of Death.

Statistical analysis
We used STATA/SE 15.1. All 465 individuals were categorized into ranked quantiles of AHRR (cg05575921) methylation extent (%) with the highest methylation quantile defined as the reference category, presumably representing individuals least exposed to smoking. We showed data in quantiles for simple illustrative presentation of the results; however, analyses on AHRR (cg05575921) methylation extent (%) categorized into deciles and on a continuous scale are also shown. We used Kruskal Wallis test and Pearsons X 2 -test to test for trend of AHRR methylation through all included variables.
Individual follow-up time began at baseline, which was defined as the date of lung cancer diagnosis by biopsy, and ended the day of death or 31 January 2017. Minimum, median and maximum follow-up time were 11 days, 332 days, and 4.4 years for all-cause mortality, giving a total of 688.495 person-years and 380 deaths (81.7%).
Cox proportional hazard regression model using length of follow-up as the time metric estimated hazard ratios (HRs) and 95% confidence intervals (95% CI) for the association between AHRR (cg05575921) methylation (%) and survival in four steps with increasing level of adjustment for potential a priori selected confounders. 1) a crude model, 2) a model adjusted for age at lung cancer diagnosis and sex 3) a multivariable model additionally adjusted for body mass index (kg/m 2 ) (BMI), ethnicity (Europeans/others), Classification of Malignant Tumors (TNM) (Stage I-IIII), histology of lung cancer (small cell lung cancer, adenocarcinoma, squamous-cell carcinoma, other non-small-cell lung carcinoma (NSCLC), performance status (0-3), 4) a smoking plus model additionally adjusted for smoking status (never/former/current smoker) and cumulative smoking (defined as 20 cigarettes/day per year, calculated from smoking intensity (number of cigarettes a day) and smoking duration (years)). Models were applied for the whole cohort as well as stratified by cause of death (lung cancer, cardiac death, respiratory death or other cancer) and treatment. We performed test for trend by using the continuous variable in the Cox model. We checked for the proportional hazards assumption for all variables based on scaled Schoenfeld residuals. Potential interactions between AHRR (cg05575921) methylation (per 1%) and survival were tested for by introducing interaction terms into the Cox model, and tested by the Walds test. Survival curves were generated using Kaplan-Meier analysis method, and the log rank test was used to examine differences in survival between the ranked quantiles of AHRR (cg05575921) methylation.
In a sensitivity analysis, we included 461 individuals without lung cancer from the 1991-1994 examination of the Copenhagen City Heart Study [19], a prospective cohort study of the Danish general population (S1 Table). Individuals were selected after 1:1 matching with the lung cancer patients on body mass index, age, cumulative smoking and sex. Individual followup time for all-cause mortality began at the date of examination, and ended the day of death or 14 November 2014. We did this to assess potential interaction between AHRR (cg05575921) methylation and lung cancer status on survival. Potential interactions were tested for by introducing interaction terms into the Cox model, and tested by the Walds test.
To validate the methylation measurement, we used linear regression analysis to measure expected association between cumulative smoking and AHRR (cg05575921) methylation. To assess the expected association between all-cause mortality and cumulative smoking, treatment, recurrence, ECOG performance status and TNM-status, we used the Cox model.

Results
A priori we calculated the power of the study, and the study has 80% power to detect reduced survival of at least 6% every time AHRR (cg05575921) hypomethylation is reduced by 1% (SD = 2.17%).
Compared to patients in the highest AHRR methylation quartile there was no association with survival for individuals in the lowest quartile (1.00; 0.75-1.32) or the lowest decile (1.18; 0.67-2.09) versus the highest quartile or decile. In continuous models using AHRR methylation (per 1%) as exposure, results were similar (Table 4 and S6 Table).
In addition, we observed no associations after stratification by specific causes of death; lung cancer, cardiac disease, respiratory disease or other cancer (S7 Table).   In order to further investigate the lack of association between AHRR methylation extent and reduced survival in lung cancer patients, we also analyzed the association in individuals without lung cancer (Table 5). After multivariable adjustments including age, sex, ethnicity, body mass index and cumulative smoking, hazard ratio for death was 1.49 (95% CI; 1.16-1.93) for the lowest quantile of AHRR methylation extent versus the highest. There was no interaction between lung cancer status and AHRR methylation with survival.

Discussion
In this study of 465 lung cancer cases, we observed no association between AHRR (cg05575921) methylation and survival.
With 81.7% mortality after a median follow up time of 332 days, the overall prognosis for the patient was as expected for lung cancer [27]. Further, since reduced survival was strongly associated with high cumulative tobacco consumption, poor performance status, and with advanced TNM we expect data to be in line with the usual finding. Also, the measurement of AHRR methylation was precise enough to pick up the expected association with cumulative smoking. Therefore, since both exposure and outcome variables were presumably accurately measured and we observed the expected association between AHRR methylation and reduced survival among individuals without lung cancer, we assume that the lack of association is correct, and not caused by measurement or registration errors. We speculate that the lack of association could be due to individual threshold phenomena, where smoking can only reduce AHRR methylation to a certain low level in each individual, and the level of this threshold is not determined by factors associated with reduced survival. Having reached this threshold, more intensive smoking reduces survival, but will not reduce methylation further as suggested by the very narrow range of AHRR methylation in the lung cancer patients compared with the matched individuals from the general population without lung cancer. Further, since AHRR methylation only reverts very slowly over decades after smoking cessation [6], the prognostic benefit from quitting within years before blood sampling, cannot be monitored by measuring AHRR methylation, especially in cohorts with poor short-term prognosis, i.e. lung cancer patients. Further, we found that AHRR hypomethylation behaves different on reduced survival in lung cancer patients than in individuals without lung cancer. Along with prior studies [15,17,19] showing association between AHRR hypomethylation and increased lung cancer risk, these findings suggest that smoking-associated methylation changes are more involved in initiation than progression of lung cancer.

Coefficient (95% CI) Age and sex-adjusted coefficient (95% CI) Multivariable adjusted a coefficient (95% CI)
Smoking status  Although AHRR methylation might be useful in assessing lung cancer risk in a population of smokers [19], our results suggest that it cannot identify lung cancer patients with poor prognosis independently of other commonly used prognostic factors.
Strengths of this study was the prospective design with detailed information on many covariates, complete follow-up, and no misclassification of vital status in Denmark. Lung cancer diagnosis was based on pathology confirmation by a trained pathologist.
Some limitations must, however, be acknowledged. 16 (3,4%) patients had >30 days from date of blood sample and date of diagnosis. However, these represented a non-differential sample according to reduced survival. Possible changes in AHRR methylation during follow-up  Table 4 with hazard ratios in Table 5. https://doi.org/10.1371/journal.pone.0211745.t005 could not been taken into account. Also, we do not have information on the exposure of passive smoking. This would have been relevant to include in our analyses since AHRR hypomethylation has also been found in those exposed to passive smoke [28]. Since it is also associated with reduced survival [29], inclusion of passive smoking in the analyses would probably have weakened the association between smoking and AHRR hypomethylation, but likely not the association between AHRR hypomethylation and reduced survival.

Conclusion
AHRR (cg05575921) methylation cannot be used to predict smoking mediated reduced survival in lung cancer patients.
Supporting information S1  Table. Association between AHRR (cg05575921) methylation extent and reduced survival (from all-cause mortality) by oncological treatment for lung cancer, recurrence of lung cancer and total lines of treatment among 465 patients with lung cancer. AHRR, Arylhydrocarbon receptor repressor CI, confidence interval. a Include platinium-based chemotherapy, combination and monotherapy, targeted therapy (ALKI/EGFR-mutation), immunotherapy, no oncological treatment other than surgery. A priori potential confounders were selected, and included in models 1) A crude model, 2) A model, additionally adjusted for age at lung cancer diagnosis and sex. 3) b A model, additionally adjusted for body mass index (kg/m 2 ), ethnicity (European/others), TNM Classification of Malignant Tumors (TNM) (Stage I-IIII), histology of lung cancer (small cell lung cancer, adenocarcinoma, squamous-cell carcinoma, other non-small-cell lung carcinoma (NSCLC), ECOG performance status (0-3), 4) c A model additionally adjusted for smoking status (never/ former/current smoker) and cumulative smoking (defined as 20 cigarettes/day per year, calculated from smoking intensity (number of cigarettes a day) and smoking duration (years). (DOCX) S6 Table. Association between AHRR (cg05575921) methylation extent in deciles (%) and reduced survival (from all-cause mortality) among 465 patients with lung cancer. AHRR, Aryl-hydrocarbon receptor repressor, CI, confidence interval. A priori potential confounders were selected, and included in models 1) A crude model, 2) A model, additionally adjusted for age at lung cancer diagnosis and sex. 3) a A model, additionally adjusted for body mass index (kg/m 2 ), ethnicity (European/others), TNM Classification of Malignant Tumors (TNM) (Stage I-IIII), histology of lung cancer (small cell lung cancer, adenocarcinoma, squamous-cell carcinoma, other non-small-cell lung carcinoma (NSCLC), ECOG performance status (0-3), 4) b A model additionally adjusted for smoking status (never/ former/current smoker) and cumulative smoking (defined as 20 cigarettes/day per year, calculated from smoking intensity (number of cigarettes a day) and smoking duration (years). (DOCX) S7 Table. Association between by AHRR (cg05575921) methylation extent and reduced survival by registered cause of death a (lung cancer, cardiac disease, respiratory disease and other cancers) among 385 patients with lung cancer. AHRR, Aryl-hydrocarbon receptor repressor CI, confidence interval. a One individual is registered with up till nine causes of death, and might be included in analyzes with several causes of death. A priori potential confounders were selected, and included in models 1) A crude model, 2) A model, adjusted for age at lung cancer diagnosis and sex. 3) b A model, additionally adjusted for body mass index (kg/m 2 ), ethnicity (European/others), TNM Classification of Malignant Tumors (TNM) (Stage I-IIII), histology of lung cancer (small cell lung cancer, adenocarcinoma, squamous-cell carcinoma, other non-small-cell lung carcinoma (NSCLC), ECOG performance status (0-3) 4) c A model additionally adjusted for smoking status (never/former/ current smoker) and cumulative smoking (defined as 20 cigarettes/day per year, calculated from smoking intensity (number of cigarettes a day) and smoking duration (years).