Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Sep 10;16(1):124.
doi: 10.1186/s13148-024-01734-7.

DNAm scores for serum GDF15 and NT-proBNP levels associate with a range of traits affecting the body and brain

Danni A Gadd  1 , Hannah M Smith  1 , Donncha Mullin  2   3 , Ola Chybowska  1 , Robert F Hillary  1 , Dorien M Kimenai  4 , Elena Bernabeu  1 , Yipeng Cheng  1 , Chloe Fawns-Ritchie  1   5 , Archie Campbell  1 , Danielle Page  6 , Adele Taylor  6 , Janie Corley  6 , Maria Del C Valdés-Hernández  6   7   8 , Susana Muñoz Maniega  6   7   8 , Mark E Bastin  6   7   8 , Joanna M Wardlaw  6   7   8   9 , Rosie M Walker  8   10 , Kathryn L Evans  1 , Andrew M McIntosh  2 , Caroline Hayward  1   11 , Tom C Russ  2   3 , Sarah E Harris  6 , Paul Welsh  12 , Naveed Sattar  12 , Simon R Cox  6 , Daniel L McCartney  1 , Riccardo E Marioni  13
Affiliations

DNAm scores for serum GDF15 and NT-proBNP levels associate with a range of traits affecting the body and brain

Danni A Gadd et al. Clin Epigenetics. .

Abstract

Background: Plasma growth differentiation factor 15 (GDF15) and N-terminal proB-type natriuretic peptide (NT-proBNP) are cardiovascular biomarkers that associate with a range of diseases. Epigenetic scores (EpiScores) for GDF15 and NT-proBNP may provide new routes for risk stratification.

Results: In the Generation Scotland cohort (N ≥ 16,963), GDF15 levels were associated with incident dementia, ischaemic stroke and type 2 diabetes, whereas NT-proBNP levels were associated with incident ischaemic heart disease, ischaemic stroke and type 2 diabetes (all PFDR < 0.05). Bayesian epigenome-wide association studies (EWAS) identified 12 and 4 DNA methylation (DNAm) CpG sites associated (Posterior Inclusion Probability [PIP] > 95%) with levels of GDF15 and NT-proBNP, respectively. EpiScores for GDF15 and NT-proBNP were trained in a subset of the population. The GDF15 EpiScore replicated protein associations with incident dementia, type 2 diabetes and ischaemic stroke in the Generation Scotland test set (hazard ratios (HR) range 1.36-1.41, PFDR < 0.05). The EpiScore for NT-proBNP replicated the protein association with type 2 diabetes, but failed to replicate an association with ischaemic stroke. EpiScores explained comparable variance in protein levels across both the Generation Scotland test set and the external LBC1936 test cohort (R2 range of 5.7-12.2%). In LBC1936, both EpiScores were associated with indicators of poorer brain health. Neither EpiScore was associated with incident dementia in the LBC1936 population.

Conclusions: EpiScores for serum levels of GDF15 and Nt-proBNP associate with body and brain health traits. These EpiScores are provided as potential tools for disease risk stratification.

Keywords: Brain; Cardiovascular; DNA methylation; Dementia; Diabetes; Epigenetic; GDF15; NT-proBNP; Risk stratification; Stroke.

PubMed Disclaimer

Conflict of interest statement

R.E.M is an advisor to the Epigenetic Clock Development Foundation and has received consultant fees from Optima partners. A.M.M has previously received speaker’s fees from Illumina and Janssen and research grant funding from The Sackler Trust. R.F.H. has received consultant fees from Illumina and Optima partners. D.A.G. has received consultant fees from and is currently employed in part-time capacity by Optima partners. D.L.M. is currently employed in part-time capacity by Optima partners. P.W. reports grant income from Roche Diagnostics in relation to and outside of the submitted work, as well as grant income from AstraZeneca, Boehringer Ingelheim and Novartis, outside the submitted work and speaker fees from Novo Nordisk, and Raisio outside the submitted work. N.S. has consulted for Afimmune, Amgen, AstraZeneca, Boehringer Ingelheim, Eli Lilly, Hanmi Pharmaceuticals, Merck Sharp & Dohme, Novartis, Novo Nordisk, Pfizer and Sanofi and received grant support paid to his University from AstraZeneca, Boehringer Ingelheim, Novartis and Roche Diagnostics outside the submitted work. All other authors declare no competing interest.

Figures

Fig. 1
Fig. 1
Study design for this assessment of GDF15 and NT-proBNP EpiScores as biomarkers. Disease associations and epigenome-wide association studies (EWAS) for each protein were first characterised in the full Generation Scotland sample. EpiScores for each protein were initially trained and tested in subsets of the population. This allowed EpiScores to be compared with measured proteins in associations with the four incident diseases profiled in the test set. EpiScores were then retrained on the full sample and tested externally in the LBC1936 Wave 4 population, which had measures of both proteins and DNAm available. EpiScores were projected into the larger LBC1936 Wave 1 population (that has DNAm but no protein measures) and profiled in associations with brain health traits, cross-sectionally and longitudinally. Consent for dementia linkage was available from Wave 2 of the LBC1936; therefore, we also tested whether EpiScores were associated with time-to-dementia. EpiScores were modelled with polygenic risk scores (PRS) for the proteins. CpG: cytosine-phosphate-guanine. IHD: ischaemic heart disease
Fig. 2
Fig. 2
Disease associations for GDF15 and NT-proBNP in Generation Scotland (N ≥ 16,963). Fully adjusted hazard ratios from Cox PH mixed effects regression models between protein levels and incident diseases are plotted with 95% confidence intervals. The six associations in red had FDR P < 0.05 in basic and P < 0.05 in fully adjusted models, whereas associations that had P > 0.05 are shown in black. Hazard ratios are plotted per 1 SD increase in the rank-base inverse normalised levels of each marker. Fully adjusted models controlled for age, sex, relatedness and common health and lifestyle factors (smoking, alcohol intake, BMI, social deprivation and years of education)
Fig. 3
Fig. 3
Comparison of EpiScores versus measured protein equivalents in fully adjusted associations with incident diseases in the Generation Scotland test sample (N ≥ 2808). For each disease, the protein-disease association is plotted, with the equivalent protein EpiScore-disease association shown directly beneath it for comparison. Hazard ratios are plotted per 1 SD increase in the rank-based inverse normalised levels of each marker. Nine associations (red) had FDR P < 0.05 in basic and P < 0.05 in fully adjusted Cox proportional hazards mixed effects models in the test samples. Fully adjusted models adjusted for age, sex, relatedness and common lifestyle risk factors (smoking, alcohol intake, BMI, social deprivation and years of education). Associations that were non-significant (P > 0.05 in fully adjusted models) are shown in black
Fig. 4
Fig. 4
External assessment of the GDF15 and NT-proBNP EpiScores in LBC1936. a Measurements available across the Waves of the LBC1936 external population. b Correlation plots between measured protein levels and GDF15 (orange) and NT-proBNP (red) EpiScores in the LBC1936 Wave 4 external test set (NGDF = 322, NNT-proBNP = 500). Pearson correlation coefficients are annotated in each instance. c, Standardised beta coefficients derived from structural equation models (SEMs) between EpiScore levels at LBC1936 Wave 1 (N = 895 with DNAm, N = 1091 total) and cross-sectional measures of brain health traits that had FDR P < 0.05 in basic (age and sex adjusted) models and P < 0.05 after adjustment for further lifestyle covariates. All associations had a negative beta coefficient (blue). Twenty EpiScore-trait associations were tested in total: 10 cross-sectionally and 10 assessing longitudinal change in brain traits
See this image and copyright information in PMC

References

    1. Hageman SHJ, McKay AJ, Ueda P, Gunn LH, Jernberg T, Hagström E, et al. Estimation of recurrent atherosclerotic cardiovascular event risk in patients with established cardiovascular disease: the updated SMART2 algorithm. Eur Heart J. 2022;43(18):1715–27. 10.1093/eurheartj/ehac056 - DOI - PMC - PubMed
    1. SCORE2 working group and ESC Cardiovascular risk collaboration. SCORE2 risk prediction algorithms: new models to estimate 10-year risk of cardiovascular disease in Europe. Eur Heart J. 2021;42(25):2439–54. - PMC - PubMed
    1. Pence BD. Growth differentiation factor-15 in immunity and aging. Front Aging. 2022;0:8. - PMC - PubMed
    1. Wang Z, Yang F, Ma M, Bao Q, Shen J, Ye F, et al. The impact of growth differentiation factor 15 on the risk of cardiovascular diseases: two-sample Mendelian randomization study. BMC Cardiovasc Disord. 2020;20(1):1–7. 10.1186/s12872-020-01744-2 - DOI - PMC - PubMed
    1. Tanaka T, Basisty N, Fantoni G, Candia J, Moore AZ, Biancotto A, et al. Plasma proteomic biomarker signature of age predicts health and life span. eLife. 2020;9:1–24. - PMC - PubMed

MeSH terms