Integrated methylome and phenome study of the circulating proteome reveals markers pertinent to brain health

Abstract

Characterising associations between the methylome, proteome and phenome may provide insight into biological pathways governing brain health. Here, we report an integrated DNA methylation and phenotypic study of the circulating proteome in relation to brain health. Methylome-wide association studies of 4058 plasma proteins are performed (N = 774), identifying 2928 CpG-protein associations after adjustment for multiple testing. These are independent of known genetic protein quantitative trait loci (pQTLs) and common lifestyle effects. Phenome-wide association studies of each protein are then performed in relation to 15 neurological traits (N = 1,065), identifying 405 associations between the levels of 191 proteins and cognitive scores, brain imaging measures or APOE e4 status. We uncover 35 previously unreported DNA methylation signatures for 17 protein markers of brain health. The epigenetic and proteomic markers we identify are pertinent to understanding and stratifying brain health.

Fig. 1. Methylome and phenome study of the plasma proteome in relation to brain health study design.

A total of 4058 plasma proteins (corresponding to 4235 SOMAmers) were measured in 1065 individuals in Generation Scotland. A methylome-wide association study (MWAS) of each plasma protein level was conducted in 774 individuals that represented a nested subset of the full sample that had DNAm measurements available. A phenome-wide protein association study (Protein PheWAS) identified protein levels that were associated with a minimum of one brain health trait (N > = 909). Overlapping the protein MWAS and PheWAS results identified pQTMs that involved protein markers of brain health. The functional roles of proteins and CpGs involved in this subset were explored further, with approaches tailored to interpretation of cis and trans pQTMs. Created with BioRender.com.

Fig. 2. Methylome-wide association studies (MWAS) of 4058 plasma proteins.

a Summary of MWAS results for 4058 protein levels in Generation Scotland (N = 774). The number of protein quantitative trait methylation loci (pQTMs) that had P < 4.5 × 10⁻¹⁰ (Bonferroni threshold for multiple testing adjustment) in the basic, white blood cell proportion (WBC)-adjusted and fully-adjusted models. Cis associations (purple) and trans associations (green) are summarised for each model. Covariates used to adjust DNAm are described for each model. Normalised protein levels were adjusted for age, sex, 20 genetic principal components (PCs), protein quantitative trait loci (pQTLs) and technical variables and scaled to have a mean of 0 and standard deviation of 1. Results were generated through linear regression models. Full summary statistics with full P values can be accessed in Supplementary Data 6. Created with BioRender.com. b Flow diagram showing the distinction between the highly pleiotropic PAPPA and PRG3 protein pQTMs and the 825 pQTMs that involved the levels of a further 189 proteins. TSS: transcriptional start site of the protein gene. The 434 cis pQTMs (purple) lay on the same chromosome and ≤ 10 Mb from the transcriptional start site (TSS) of the protein gene, whereas the 391 trans pQTMs (green) lay >10 Mb from the TSS of the protein gene or on a different chromosome. Created with BioRender.com. c Genomic locations for 825 of the 2928 fully-adjusted pQTMs, excluding highly pleotropic associations for PAPPA and PRG3 protein levels, with cis pQTMs in purple and trans pQTMs in green. Chromosomal location of CpG sites (x-axis) and protein genes (y-axis) are presented. A list of the full association counts for each protein and CpG site can be found in Supplementary Data 8, 9.

Fig. 3. Pleiotropic associations in the fully-adjusted methylome-wide association studies (MWAS).

a pQTMs that had P < 4.5 × 10⁻¹⁰ (Bonferroni threshold for multiple testing adjustment) in the fully-adjusted MWAS are plotted as individual points (dark blue) with chromosomal locations of the 191 protein genes (upper) and the 1837 CpGs (lower) on the x-axis. 19 proteins with ≥10 associations with CpGs are highlighted in turquoise and labelled on the upper plot. Nine CpGs with ≥6 associations with protein levels are highlighted in turquoise on the lower plot. Results were generated through linear regression models. Full summary statistics with full P values can be accessed in Supplementary Data 6. b A selection of CpGs with highly pleiotropic signals in the fully-adjusted MWAS and the corresponding function of the gene the CpGs were located within. Created with BioRender.com.

Fig. 4. Phenome-wide associations studies (PheWAS) of 4058 plasma proteins and brain health.

a Number of protein marker associations with P < 3.5 × 10⁻⁴ for each of the 15 traits related to brain health in the phenome-wide protein association studies (protein PheWAS). These studies included a maximum sample of 1065 individuals with protein measurements from Generation Scotland and tested for associations between 15 phenotypes and the levels of 4058 plasma proteins via linear mixed effects regression. Cognitive score (green), brain imaging (light blue) and APOE e4 status (dark blue) associations are summarised. Full summary statistics for the 405 associations with P values are presented in Supplementary Data 17. All associations were generated through linear regression and were adjusted for multiple testing correction. b Heatmap of standardised beta coefficients for 77 of the 405 protein PheWAS associations (P < 3.5 × 10⁻⁴ indicated by an asterisk). These include three proteins that had associations with both APOE e4 status and one or more cognitive scores, in addition to 22 proteins that had associations with both a brain imaging measure and a cognitive score. Negative and positive beta coefficients are shown in blue and red, respectively. A heatmap describing the full 405 associations for APOE e4 status, cognitive scores and brain imaging measures is available in Supplementary Fig. 6. All associations were generated through linear regression and were adjusted for multiple testing correction.

Fig. 5. Trans pQTMs involving protein markers of brain health.

Circular plot showing 15 trans pQTM associations between DNAm at 11 CpG sites and the levels of nine protein markers of brain health that had P < 4.5 × 10⁻¹⁰ (Bonferroni threshold for multiple testing adjustment) in the fully-adjusted MWAS. Chromosomal positions are given on the outermost circle. Details of the full set of 35 pQTMs for protein markers of brain health are provided in Supplementary Data 20 with P values. Results were generated through linear regression models.

Fig. 6. Exploration of trans pQTMs for protein markers of brain health.

a Three trans associations with the CpG site cg06690548 in the SLC7A11 gene, which encodes a synaptic protein that is involved in glutamate transmission and oxidative stress. cg06690548 has been implicated in methylome-wide studies of Parkinson’s disease and Amyotrophic lateral sclerosis (ALS) risk. b Five trans associations between CpGs in the ITIH3/ITIH1/NEK4 region on chromosome 3 and the levels of RBL2, which was associated with lower Global Grey Matter Volume. The RBL2 gene has been implicated in genome-wide associations studies (GWAS) of cognitive ability, intelligence and educational attainment. The ITIH3/ITIH1/NEK4 region has been implicated in GWAS of Schizophrenia and Bipolar disorder. c Two trans associations between DNAm at cg11294350 in the CHPT1 gene and two proteins with lysosomal-associated function (SMPD1 and HEXB) that were associated with higher Relative Brain Age and lower General Fractional Anisotropy. Associations with a positive beta coefficient are denoted as red connecting lines, whereas associations with a negative beta coefficient are denoted as blue connecting lines. The full 35 pQTMs for protein markers of brain health (15 trans and 20 cis) can be found in Supplementary Data 20. All associations were generated using linear regression and were adjusted for multiple testing. Created with BioRender.com.