14 Oct 19. Featured Paper

Nature

 

Authors

Janie Corley, Simon R. Cox, Sarah E. Harris, Maria Valdéz Hernandez, Susana Muñoz Maniega, Mark E. Bastin, Joanna M. Wardlaw, John M. Starr, Riccardo E. Marioni & Ian J. Deary 

 

Abstract

Recent advances in genome-wide DNA methylation (DNAm) profiling for smoking behaviour have given rise to a new, molecular biomarker of smoking exposure. 

It is unclear whether a smoking-associated DNAm (epigenetic) score has predictive value for ageing-related health outcomes which is independent of contributions from self-reported (phenotypic) smoking measures. Blood DNA methylation levels were measured in 895 adults aged 70 years in the Lothian Birth Cohort 1936 (LBC1936) study using the Illumina 450K assay. A DNA methylation score based on 230 CpGs was used as a proxy for smoking exposure. Associations between smoking variables and health outcomes at age 70 were modelled using general linear modelling (ANCOVA) and logistic regression.

Additional analyses of smoking with brain MRI measures at age 73 (n = 532) were performed. Smoking-DNAm scores were positively associated with self-reported smoking status (P < 0.001, eta-squared ɳ² = 0.63) and smoking pack years (r = 0.69, P < 0.001). Higher smoking DNAm scores were associated with variables related to poorer cognitive function, structural brain integrity, physical health, and psychosocial health.

Compared with phenotypic smoking, the methylation marker provided stronger associations with all of the cognitive function scores, especially visuospatial ability (P < 0.001, partial eta-squared ɳp² = 0.022) and processing speed (P < 0.001, ɳp² = 0.030); inflammatory markers (all P < 0.001, ranges from ɳp² = 0.021 to 0.030); dietary patterns (healthy diet (P < 0.001, ɳp² = 0.052) and traditional diet (P < 0.001, ɳp² = 0.032); stroke (P = 0.006, OR 1.48, 95% CI 1.12, 1.96); mortality (P < 0.001, OR 1.59, 95% CI 1.42, 1.79), and at age 73; with MRI volumetric measures (all P < 0.001, ranges from ɳp² = 0.030 to 0.052).

Additionally, education was the most important life-course predictor of lifetime smoking tested.

Our results suggest that a smoking-associated methylation biomarker typically explains a greater proportion of the variance in some smoking-related morbidities in older adults, than phenotypic measures of smoking exposure, with some of the accounted-for variance being independent of phenotypic smoking status.

 

Methods:

Participants were from the Lothian Birth Cohort 1936 (LBC1936), a group of relatively heathy community-dwelling subjects in their seventies, enrolled in a longitudinal study of cognitive and brain ageing conducted in Scotland. Most participants had previously taken part in the Scottish Mental Survey of 1947 (SMS1947) at about age 11 years (from which we derived an age 11 IQ score), and subsequently traced and recruited to the study almost 60 years later, at approximately 70 years of age.

Briefly, individuals born in 1936, who were living in the Lothian area of Scotland, were contacted by Lothian Health Board on behalf of the investigators and invited to take part in the study.

In total, 1091 men and women were recruited at Wave 1 (2004–2007, age ∼70 years, n = 1091) with further follow-up waves at ages 73 (n = 866), 76 (n = 697), 79 (n = 550) and 82 (ongoing).

Extensive phenotypic data have been collected, including blood biomarkers, cognitive testing, neuroimaging, and psychosocial, lifestyle, genetic, and health measures.

All participants provided written informed consent before testing.

The LBC1936 study was approved by the Multi-Centre Research Ethics Committee for Scotland (MREC/01/0/56) and the Lothian Research Ethics Committee (LREC/2003/2/29 for Wave 1 and 07/MRE00/58 for Waves 2–5).

Most of data for the present study come from Wave 1 (age 70).

Structural brain imaging was undertaken three years later for 700 participants at Wave 2 (age 73).

Here, a total of 895 individuals had smoking-DNAm data at age 70, and of the 895, 532 had MRI data at age 73.

Following quality control which removed instances in which aberrant surfaces or segmentation errors were removed, additional analyses of cortical thickness were run for 521 participants.

 

Conclusions:

Our study supports the potential utility of a smoking DNAm score, derived from genome-wide data, as a biomarker of lifetime smoking exposure, and for contributing toward the prediction of important ageing-related health outcomes in later life.

In particular, the smoking methylation biomarker better predicted poorer cognitive function and brain structural integrity, chronic inflammation, stroke and mortality in later life, compared with much-used phenotypic measures of smoking.

It may help to identify novel health impacts, improve adjustment for smoking in research studies, and shed light on the molecular mechanisms by which smoking predisposes to chronic mental and physical disease, and less good brain and cognitive health.

In terms of clinical impacts, a methylation marker holds promise for better risk prediction in precision medicine.

A useful implication of the present study is that it suggests that one may obtain an indication of smoking exposure and its implications even in studies which have not collected smoking data.