Interindividual variability and co-regulation of DNA methylation differ among blood cell populations

Abstract

DNA methylation regulates gene expression in a cell-type specific way. Although peripheral blood mononuclear cells (PBMCs) comprise a heterogeneous cell population, most studies of DNA methylation in blood are performed on total mononuclear cells. In this study, we investigated high resolution methylation profiles of 58 CpG sites dispersed over eight immune response genes in multiple purified blood cells from healthy adults and newborns. Adjacent CpG sites showed methylation levels that were increasingly correlated in adult blood vs. cord blood. Thus, while interindividual variability increases from newborn to adult blood, the underlying methylation changes may not be merely stochastic, but seem to be orchestrated as clusters of adjacent CpG sites. Multiple linear regression analysis showed that interindividual methylation variability was influenced by distance of average methylation levels to the closest border (0 or 100%), presence of transcription factor binding sites, CpG conservation across species and age. Furthermore, CD4+ and CD14+ cell types were negative predictors of methylation variability. Concerns that PBMC methylation differences may be confounded by variations in blood cell composition were justified for CpG sites with large methylation differences across cell types, such as in the IFN-γ gene promoter. Taken together, our data suggest that unsorted mononuclear cells are reasonable surrogates of CD8+ and, to a lesser extent, CD4+ T cell methylation in adult peripheral, but not in neonatal, cord blood.

Figure 1. Location of CpG sites in the analyzed genes. Numbering of CpGs is based on the TSS (+1), represented by a bent arrow. All CpGs within 1,000 nucleotide bases upstream to 500 bases downstream of the TSS (open boxes) and CpGs included in the methylation study (closed boxes) are shown.

Figure 2. Average DNA methylation levels in cord blood and in adult blood. Average methylation levels of about half of the analyzed CpGs were significantly lower in PBMCs than in CBMCs. Stars denote significant differences in average DNA methylation (p < 0.001, Bonferroni-corrected threshold) tested by t-test or Mann-Whitney rank sum test as appropriate. Error bars represent standard deviation (n = 30 donors). Adult blood, open bars; cord blood, closed bars.

Figure 3. Interindividual variability of DNA methylation in blood cell subpopulations. Interindividual variability of DNA methylation levels (expressed as statistical variance) across all CpG sites in different cell types from (A) cord or (B) adult blood donors. Cell types are ordered by decreasing interindividual variability. Boxes display median (horizontal bars), interquartile ranges (lower and upper limits of boxes), 95% interval (whiskers) and outliers (circles). Median values are shown above each box plot. CD19+ and CD14+ cells display a significantly reduced variability in comparison to most other cell types tested by ANOVA on ranks with Tukey’s multiple comparison post-hoc test. Stars denote significant differences (p < 0,05).

Figure 4. DNA methylation of TNF-α, KIR2DL4, and IFN-γ promoters in adult blood. (A) In the TNF-α promoter, the high methylation variability of CpGs -245 and -239 (filled circles) in PBMCs is present in all subpopulations (CD4⁺, CD8⁺, CD19⁺ and CD56⁺ cells) except CD14⁺ cells, which are the main producers of TNF-α. CD4⁺ cells are shown as an example of a subpopulation with high methylation variability. (B) In the KIR2DL4 promoter, the outlier methylation values of CpG -228 (filled circles) in PBMCs are present in all subpopulations (CD4⁺, CD8⁺, CD14⁺ and CD19⁺ cells) except CD56⁺ cells, which are the main expressers of KIR2DL4. CD4⁺ cells are shown as an example of a subpopulation with high methylation variability. (C) In the IFN-γ promoter, CD8⁺ cells present a substantially higher interindividual variability than all other cell types, including CD4⁺ cells which are the main producers of IFN-γ.

Figure 5. Correlations between methylation levels of individual CpG sites. Spearman correlations between methylation levels in (A) CB, (B) AB, and (C) IFN-γ, KIR2DL4 and TNF-α promoters in different blood cell populations. Clusters of intercorrelated CpG methylation levels were identified by visual analysis of both magnitude and significance (p value) of the Spearman’s correlation coefficient. Light gray squares: 0.6 < rho < 0.8; dark gray squares: rho ≥ 0.8, Dotted squares correspond to p values < 0.000031 (Bonferroni-corrected threshold). Intragenic clusters of intercorrelated methylation levels are delimited by bold black borders. Intergenic clusters are drawn with dashed borders. See Tables S4 and S5 for a detailed view of rho and p values.

Figure 6. The influence of CD56+ cell frequency on IFN-γ and KIR2DL4 methylation in PBMCs. Scatterplot showing the correlation between frequency of CD56+ cells and (A) IFN-γ +122 or (B) KIR2DL4 -228 methylation in adult PBMCs.

Figure 7. Methylation levels in blood cell subpopulations. Upper subplots display heatmaps of average DNA methylation levels in each CpG position and in each cell type in (A) CB and (B) AB. The lower subplots show p values for the comparison of median methylation levels between each subpopulation and total C/PBMCs using paired Mann-Whitney rank sum tests. The dashed line indicates the significance threshold of a p value < 0.05.

Figure 8. Correlation of methylation levels between subpopulations and C/PBMCs. (A) Median Pearson correlations between methylation levels of each CpG site in subpopulations and total C/PBMCs. Boxes display median (horizontal bars), interquartile ranges (lower and upper limits of boxes), 95% interval (whiskers) and outliers (circles). C/PBMC methylation levels best correlated with CD4⁺ and CD8⁺ cell methylation. (B) Methylation of TLR4 CpG site +205 in CB. Symbols indicate the different individuals. The ranking of the individuals’ methylation levels differs between cell types.