Long-term DNA methylation changes mediate heterologous cytokine responses after BCG vaccination

Abstract

Background: Epigenetic reprogramming shapes immune memory in both innate (trained immunity) and adaptive immune cells following Bacillus Calmette-Guérin (BCG) vaccination. However, the role of dynamic DNA methylation changes in post-vaccination immune responses remains unclear.

Results: We established a cohort of 284 healthy Dutch individuals, profiling genome-wide DNA methylation and cytokine responses to ex vivo stimulation at baseline, 14 days, and 90 days post-BCG vaccination. We identified distinct patterns of DNA methylation alternations in the short- and long-term following BCG vaccination. Moreover, we established that baseline DNA methylation profiles exert influence on the change in interferon-γ (IFN-γ) production upon heterologous (Staphylococcus aureus) stimulation before and after BCG vaccination. Specifically, we identified the regulation of kisspeptin as a novel pathway implicated in the modulation of IFN-γ production, and this finding has been substantiated through experimental validation. We also observed associations between BCG-induced DNA methylation changes and increased IFN-γ and interleukin-1 β (IL-1β) production upon S. aureus stimulation. Interestingly, by integrating with genetic, epigenetic, and cytokine response data from the same individuals, mediation analysis demonstrated that most of the identified DNA methylation changes played a mediating role between genetic variants and cytokine responses; for example, the changes of cg21375332 near SLC12 A3 gene mediated the regulation of genetic variants on IFN-γ changes after BCG vaccination. Sex-specific effects were observed in DNA methylation and cytokine responses, highlighting the importance of considering sex in immune studies.

Conclusions: These findings provide deeper insights into immune response mechanisms, crucial for developing effective epigenetic-based medical interventions for personalized medicine.

Keywords: BCG vaccination; Cytokines response; DNA methylation; Systems biology; Trained immunity.

Fig. 1

Study design and data exploration. A Overview of the study design. B Main analysis in this study. C Heatmap of association between covariates, estimated cell proportions, and DNA methylation at T0, which are represented as the top 30 PCs, capturing 35% of the variance. The association was performed with a univariable linear regression model; different colors in the figure indicate different levels of significance. BCG, Bacillus Calmette–Guérin; PC, principal component

Fig. 2

BCG vaccine induced short-term and long-term methylation changes. A Manhattan plot showing DNA methylation changes over time after BCG vaccination. Differentially methylated CpG sites (N = 11) with false discovery rate (FDR) < 0.05 were highlighted and labeled with the probeID. B Heatmap of hierarchical clustering on changes of identified 11 CpG sites at T14 and T90 compared to T0, which are represented as the differences in DNA methylation beta values. C Bar plot of the number of CpG sites with significant change over time with P < 1 × 10⁻⁵ in different groups. Down_down group indicates DNA methylation was decreased at both T14 and T90 compared with T0; similarly, down_up, decreased at T14 and increased at T90, up_down, increased at T14 and decreased at T90, and up_up, increased at both T14 and T90. D–F Change patterns and pathway enrichment analysis of CpG sites identified which assessed the DNA methylation changes at day 14 and day 0 (T14–T0, D), day 90 and day 0 (T90–T0, E), as well as day 90 and day 14 (T90–T14, F) with P < 1 × 10.⁻⁵. Each panel includes line charts showing the patterns of changes with two plots showing decrease (upper) and increase (lower) change, respectively, and the top enriched pathways and chromatin state were labeled right to each line chart. The pathway enrichment analyses were performed by CPDB (http://cpdb.molgen.mpg.de/) an online tool of gene set analysis; the chromatin enrichment analysis was performed by eFORGE (https://eforge.altiusinstitute.org/). G Bar plot showing the number of decreased and increased CpG sites identified from T14–T0, T90–T0, and T90–T14, respectively. Blue represents hypomethylated sites and yellow represents hypermethylated sites. The P value on the top of the figure shows the enrichment of hypomethylated/hypermethylated sites in increase/decrease status (Fisher’s exact test)

Fig. 3

Baseline epigenetic markers and ex vivo cytokine production changes. A Manhattan plot showing the association of baseline DNA methylation level and ex vivo immune response, represented by fold-change of TNF-α, IL-1β, IL6, and IFN-γ production stimulated by S. aureus 90 days after BCG vaccination compared to baseline (from middle to outside). Genome-wide significant CpG sites (FDR < 0.05) are highlighted in red. B Bar plot showing the variation in ex vivo IFN-γ response explained by data from various omics layers, including general information (age and sex), genetics (PRS), baseline DNA methylation, baseline inflammatory proteins, and baseline plasma metabolites. C Experimental validation. Chr, chromosome; AUC, area under the curve; CI, confidence interval; PRS. polygenic risk score; DNA methylation: DNA methylation; protein: inflammatory proteins

Fig. 4

Causal relationship inference by bidirectional mediation analysis. A Framework of bidirectional mediation analysis between SNPs, DNA methylation changes (DNAm-C) and cytokine production changes (TI). B Number of CpG sites that were significant in mediation results of Direction1 (from DNAm-C to TI), Direction2 (from TI to DNAm-C) and both, for IFN-γ (left) and IL-1β (right). C Sankey diagram showing the inferred causal relationship network of Direction1 with mediation P value < 0.05. D–F Example of causal relationships between SNP, DNAm-C, and TI inferred by bidirectional mediation analysis. The beta coefficient and significance are labeled at each edge and the proportions of mediation effect are labeled at the center of ring charts. See also Table S14

Fig. 5

Sex-specific effect on BCG vaccination and the association with ex vivo cytokine production changes. A Miami plot showing the short-term DNA methylation changes upon BCG vaccination (T14–T0) in females (upper) and males (lower). CpG sites with P < 1 × 10⁻⁵ were highlighted by red (female) and blue (male). B Scatter plot showing the consistency of results from males and females regarding the short-term effect. The x-axis represents the − log10 P value with the sign of change direction in females, and the y-axis showing the same value in males. The dots are significant CpG sites identified from female (red dots) and male (blue dots) respectively with P < 1 × 10.⁻⁵. (C) Miami plot showing the long-term DNA methylation changes upon BCG vaccination (T90–T0) in females (upper) and males (lower); the only CpG site that passed FDR significance was highlighted by green. D Scatter plot showing the consistency of results from males and females regarding the long-term effect. E–F Dot plot showing the pathway enrichment of genes annotated to the identified CpG sites in short-term (E) and long-term (F), with two panels for females (left) and males (left) respectively. G Miami plot showing the association between baseline DNA methylation and ex vivo production changes of IFN-γ in females (upper) and males (lower) with FDR < 0.05. H Scatter plot showing the consistency of the results from G in males and females. The x-axis represents the − log10 P value with the sign of the correlation coefficient in females, and the y-axis showing the same value in males. I Heatmap showing the Spearman correlation between TI (IFN-γ) associated CpG sites and baseline hormone levels of androstenedione (ADNC), cortisol (CORC), and 11-deoxycortisol (DESC). Only significant correlations (FDR < 0.05) are shown in this heatmap. Cell colors indicate the − log10(P) with the sign of the correlation coefficient of Spearman correlation (rho), and the asterisks indicate the significance of the correlation (* FDR < 0.05, ** FDR < 0.01)

Fig. 6

Summary of findings of this study. Figure was created using BioRender.com