The response to influenza vaccination is associated with DNA methylation-driven regulation of T cell innate antiviral pathways

Abstract

Background: The effect of vaccination on the epigenome remains poorly characterized. In previous research, we identified an association between seroprotection against influenza and DNA methylation at sites associated with the RIG-1 signaling pathway, which recognizes viral double-stranded RNA and leads to a type I interferon response. However, these studies did not fully account for confounding factors including age, gender, and BMI, along with changes in cell type composition.

Results: Here, we studied the influenza vaccine response in a longitudinal cohort vaccinated over two consecutive years (2019-2020 and 2020-2021), using peripheral blood mononuclear cells and a targeted DNA methylation approach. To address the effects of multiple factors on the epigenome, we designed a multivariate multiple regression model that included seroprotection levels as quantified by the hemagglutination-inhibition (HAI) assay test.

Conclusions: Our findings indicate that 179 methylation sites can be combined as potential signatures to predict seroprotection. These sites were not only enriched for genes involved in the regulation of the RIG-I signaling pathway, as found previously, but also enriched for other genes associated with innate immunity to viruses and the transcription factor binding sites of BRD4, which is known to impact T cell memory. We propose a model to suggest that the RIG-I pathway and BRD4 could potentially be modulated to improve immunization strategies.

Keywords: DNA methylation; RNA sequencing; cell-type deconvolution; influenza vaccine; influenza virus; targeted bisulfite sequencing.

Figure 1

Study schematic and design. (A). Vaccination summary of UGA4 (2019) and UGA5 (2020), the composition for quadrivalent Fluzone influenza vaccine in each cohort. (B) Hemagglutination inhibition assay, targeted bisulfite sequencing, and RNA sequencing schematic. (C) Model illustration, multivariate multiple regression and pseudoinverse were used to address the effects of multiple factors on the epigenome. (D) The proportion of seroprotected subjects at day 0 and 28 days after receiving the vaccination. (E) The proportion of seroconverted subjects 28 days after receiving the vaccination.

Figure 2

Analysis of cell type proportion estimation in two cohorts and their association with phenotypic traits. (A) 37 differentially methylated regions of the recently published methylation atlas, blue indicates hypomethylation and red indicates hypermethylation. (B) Results of cell type deconvolution for cohorts UGA4 and UGA5, assessed at two different time points. 6 cell types were selected which occur in PBMC. (C) Correlation of UGA4 cell type estimates with various phenotypic traits. Blue rows are HAI measurements (U3/4 indicates the vaccine strain is the same in UAG3 and UGA4), yellow rows are log2 fold change of HAI from day 0 to day 28 (sum log2 is the sum of log2 fold change of the vaccine strain used in U4), and pink rows are other factors. (D) Correlation of UGA5 cell type estimates with various phenotypic traits (*Indicates p < 0.05, **Indicates p < 0.01, *** Indicates p < 0.001).

Figure 3

Multivariate multiple regression model LOOCV assessment and HAI significant coefficients. (A) LOOCV of the prediction from the multivariate multiple regression and pseudoinverse. The Pearson correlation between true and predicted of all the traits are all significant. (B) Manhattan plot of the significant coefficients of all the CpG sites for the four vaccine strains. Each dot represents the −log10 of FDR adjusted p values for the coefficients, and the blue line indicates the significant threshold.

Figure 4

Transcription factors binding sites enrichment of significant positive HAI coefficients. (A) Distance from significant positive HAI coefficient sites to transcription start sites percentage. (B) Significant transcription factor binding sites from the Cistrome database, specifically focusing on either blood or immune related sources. (C) Overlap between methylation site and publicly available ChIP-Seq data. Dark blue color marks the BRD4 transcription factor binding sites.

Figure 5

RNA gene expression analysis of UGA5 cohort. (A) Differential gene expression analysis between day 0 and 28 days after vaccine injection, (B) TPM of JAK3 and TYK2 between day 0 and 28 days after vaccine injections. (C) Average Day 0 TPM of Negative Regulation of Immune Response genes (PCBP2, TNFAIP3, FURIN, ATG12, ATG5) from significant HAI positive coefficients enrichment between responder and non-responder.

Figure 6

Generated using Biorender (Additional file 3), the full proposed model of multimodal DNA methylation analysis and differential gene expression analysis of immune responses to influenza vaccine (T cell B cell communication).