Distinct immunological and molecular signatures underpinning influenza vaccine responsiveness in the elderly

Abstract

Seasonal influenza outbreaks, especially in high-risk groups such as the elderly, represent an important public health problem. Prevailing inadequate efficacy of seasonal vaccines is a crucial bottleneck. Understanding the immunological and molecular mechanisms underpinning differential influenza vaccine responsiveness is essential to improve vaccination strategies. Here we show comprehensive characterization of the immune response of randomly selected elderly participants (≥ 65 years), immunized with the adjuvanted influenza vaccine Fluad. In-depth analyses by serology, multi-parametric flow cytometry, multiplex and transcriptome analysis, coupled to bioinformatics and mathematical modelling, reveal distinguishing immunological and molecular features between responders and non-responders defined by vaccine-induced seroconversion. Non-responders are specifically characterized by multiple suppressive immune mechanisms. The generated comprehensive high dimensional dataset enables the identification of putative mechanisms and nodes responsible for vaccine non-responsiveness independently of confounding age-related effects, with the potential to facilitate development of tailored vaccination strategies for the elderly.

Fig. 1. Stratification of vaccinees.

Serum samples derived from vaccinees at baseline and 21 or 70 days after vaccination in the seasons 2014/15 and 2015/16 were used for quantification of influenza antigen-specific antibodies (HAI assay). Venn diagrams depict overlapping responses against the three vaccine antigens (1st study n = 34, 2nd study n = 200 biologically independent samples). Adapted from ref. , copyright by the co-author F.P. Source data are provided as a Source Data file.

Fig. 2. Serum profiling of vaccine responders and non-responders reveals differentially regulated factors.

Cryopreserved serum samples derived from the triple responders and non-responders at all analyzed time points were subjected to a bead-based flow cytometric measurement based on Luminex technology. A Heat map showing the ratio of cytokine concentrations detected before vaccination in responders vs. non-responders. B Heat map depicting the ratio of vaccine-induced serum factors (fold increase over day 0) for triple vaccine responders as compared to non-responders. Orange represents values that show a higher response in responders as compared to non-responders, blue indicates a higher response in non-responders, 2nd study, n = 10 responders, n = 10 non-responders, biologically independent samples). Source data are provided as a Source Data file.

Fig. 3. Pathway analysis of serum factors identified in samples derived from responders and non-responders (see Fig. 2).

A 3 days post-vaccination and B 7 days post-vaccination (Pathview online tool). Input data = log2 values of day X/day 0. Orange indicates factors showing a higher response in responders, blue indicates factors showing a higher response in non-responders (day 3 p = 0.96, day 7 p = 0.99). P value calculated by GAGE.

Fig. 4. Enhanced functionality of T cell populations in vaccine triple responders.

Cryopreserved PBMCs isolated from triple responders and non-responders were re-stimulated with the vaccine antigens overnight, stained for surface antigens and intracellular cytokine production and analyzed by flow cytometry. Frequencies of IFNγ⁺, TNFα⁺, and IL-2⁺ A CD4⁺ T cells (IFNγ day 6/7 triple non-responders vs. responders p = 0.0134, TNFα day 6/7 triple non-responders vs. responders p = 0.0092, IL-2 day 6/7 triple non-responders vs. responders p = 0.0130, IL-2 day 21 triple non-responders vs. responders p = 0.0088), B CD8⁺ T cells (IFNγ day 21 triple non-responders vs. responders p = 0.0234, TNFα day 0 triple non-responders vs. responders p = 0.0130, TNFα day 6/7 triple non-responders vs. responders p = 0.0252, TNFα day 21 triple non-responders vs. responders p = 0.0041, IL-2 day 21 triple non-responders vs. responders p = 0.0245). Violin plots represent the data subtracted for background functionality (1st study, n = 12 (6 non-responders and six responders, biologically independent samples, single missing values, see Source Data file). Violin plots show the mean and the quartiles as dashed and dotted lines, respectively. The shape indicated the data distribution. Asterisks denote statistical significance as calculated by Two-way ANOVA based on nominal p values (uncorrected Fisher’s LSD) comparing triple vaccine responders and non-responders at each time point. Source data are provided as a Source Data file.

Fig. 5. Vaccine responders show an activated T cell phenotype.

A Single-cell RNA sequencing analysis of re-stimulated PBMCs before and 7 days post-vaccination (2nd study n = 6, three responders and three non-responders, biologically independent samples). The color indicates the expression density of the indicated cytokines. B Frequencies of IFNγ⁺, TNFα⁺, and IL-2⁺ central memory (CM, CD45RA^-CCR7⁺) CD4⁺ T cells (TNFα day 6/7 triple non-responders vs. responders p = 0.0465, IL-2 day 0 triple non-responders vs. responders p = 0.0212, IL-2 day 6/7 triple non-responders vs. responders p = 0.0311, IL-2 day 21 triple non-responders vs. responders p = 0.04785) and effector memory (EM, CD45RA⁻CCR7⁻) CD4⁺ T cells (IFNγ day 6/7 triple non-responders vs. responders p = 0.0220) (1st study, n = 6 non-responders, and n = 6 responders, biologically independent samples, single missing values). Asterisks denote statistical significance as calculated by Two-way ANOVA based on nominal p values (uncorrected Fisher’s LSD) comparing triple vaccine responders and non-responders at each time point. Source data are provided as a Source Data file.

Fig. 6. T_REG and T_FH cells as important factors for vaccine-induced immune responses.

Cryopreserved PBMCs isolated from vaccine triple responders and non-responders were stained for surface antigens and intranuclear expression of FOXP3 ex vivo and analyzed by flow cytometry. A Frequencies of T_REG cells (CD4⁺CD127^lowCD25⁺FOXP3⁺) assessed by flow cytometry (1st study, n = 5 and n = 7 non-responders and responders, respectively, biologically independent samples, single missing values, see Source Data file, day 0 triple non-responders vs. responders p = 0.0390, day 6/7 triple non-responders vs. responders p = 0.0030). Plots show representative tSNE analyses of triple non-responders and responders with the expression density of FOXP3 highlighted (blue = low expression, red = high expression). B Frequencies of T_REG cells assessed by TSDR analysis and correlation with T_REG cells assessed by flow cytometry before, 3 and 70 days post-vaccination (1st study, n = 5 and n = 7 non-responders and responders, respectively, biologically independent samples, Spearman correlation). C Plot (left) depicting the tSNE analysis of 4 triple non-responders and 4 triple responders (gray) with manually gated T_REG cells highlighted and gated accordingly (CD3⁺CD4⁺FOXP3⁺ = purple). tSNE plots of non-responders and responders depicted in pseudo colors indicating low (blue) and high (red) cell densities 3 days post-vaccination. D Mean values of T_REG and T_FH cell frequencies at the indicated time points post-vaccination normalized to day 0 (1st and 2nd study, n = 10 non responders and n = 15 responders, biologically independent samples). Violin plots show the mean and the quartiles as dashed and dotted lines, respectively. The shape indicated the data distribution. Asterisks denote significant nominal p values as calculated by two-way ANOVA or Mixed-effects analysis for data sets with missing single values (uncorrected Fisher’s LSD, see Source Data file) comparing triple vaccine responders and non-responders at a given time point Source data are provided as a Source Data file.

Fig. 7. Predictive potential of data on T cell functionality.

Data derived from the flow cytometric analysis was subjected to mathematical modeling approaches. A A linear model to compute the expansion rates (cells/day) during the first week post-vaccination was computed in responders and non-responders. The bars represent the average fold change increase in expansion rates of responders with respect to non-responders (Ag⁺CD8⁺ p = 0.002, CD8⁺ p = 0.049). B Considering a multiclass naïve Bayes model, the AUC is presented to highlight marker candidates to differentiate between responders vs. non-responders. Asterisks denote significant values as calculated by a two-sample two-tailed t-test and a two-tailed Wilcoxon rank sum test comparing triple vaccine responders and non-responders at the indicated time point.

Fig. 8. B cell responses in vaccine responders and non-responders.

Frozen sera derived from vaccinees 21 days after vaccination in the two seasons were used for quantification and qualification of humoral immune responses by HAI and MN assays, respectively. A Correlation of the HAI and MN titer. The shown data is derived from the 1st and the 2nd study (n = 33 biologically independent samples). Asterisks denote significant relationships as calculated by Spearman correlation (two-tailed). Cryopreserved PBMCs isolated from vaccine triple responders and non-responders were stimulated with the vaccine antigens and stained for surface antigens identifying B memory B cells (CD27⁺CD19⁺CD3^-), transitional B cells (CD24^hiCD38^hiCD19⁺CD3^-, day 3 triple non-responders vs. responders p = 0.0039, day 6/7 triple non-responders vs. responders p = 0.0386) and plasma blasts (CD20^-CD38⁺CD27⁺CD19⁺CD3^-). Heat maps show the means of frequencies (% of CD19⁺CD3^- B cells) of re-stimulated samples (2nd study, n = 6 non responders and n = 7 responders, biologically independent samples). C Fold increase of B_REG cells (IL-10⁺CD24⁺CD38⁺CD19⁺CD3^-) at the indicated time points as compared to day 0 (2nd study, n = 6 non responders and n = 7 responders, biologically independent samples, day 6/7 triple non-responders vs. responders p = 0.0199, day 21 triple non-responders vs. responders p < 0.0001, day 70 triple non-responders vs. responders p < 0.0001). Columns represent the mean ± SEM of data from re-stimulated samples with individual values depicted as dots. Asterisks denote significant values as calculated by Two-way ANOVA without correction for multiple comparisons (uncorrected Fisher’s LSD) comparing triple vaccine responders and non-responders at a given time point. n.s. = not significant. Source data are provided as a Source Data file.

Fig. 9. Single-cell RNA sequencing of samples derived from vaccine responders and non-responders.

Cryopreserved PBMCs were re-stimulated with the vaccine antigen mixture and subjected to sorting for CD45⁺ cells prior to further measurements. A Identified clusters and B comparison of cells derived from responders and non-responders according to the time point of sample drawing (light green/blue = responders/non-responders day 0, dark green/blue = responders/non-responders day 7 post-vaccination). C Gene signatures identifying the cells within the clusters. D Cell abundancies (rel. counts [%]) of identified cell populations, significant changes between responders and non-responders at either D0 or D7 were marked. n = 6 (three responders (R)/three non-responders (NR), at baseline (D0) and 7 days post (D7) vaccination, biologically independent samples). The lower and upper hinges of the box plots correspond to the first and third quartiles (the 25th and 75th percentiles). The central line represents the second quartiles (the 50th percentiles). The upper whisker extends from the hinge to the largest value. The lower whisker extends from the hinge to the smallest value. Statistically significant differences were calculated by Dirichlet-multinomial regression adjusted with the Benjamini–Hochberg method (CD8⁺ naïve T cells day 7 triple non-responders vs. responders p = 0.0001, CD4⁺ naïve T cells day 7 triple non-responders vs. responders p = 0.005, NK cells day 0 triple non-responders vs. responders p = 0.0001, NK cells day 7 triple non-responders vs. responders p = 0.0047).

Fig. 10. Functional analysis of differentially expressed genes in responder transcriptomes (bulk genome sequencing).

A The heatmap displays the top 500 ranked differentially expressed genes among responders (n = 10 biologically independent samples) and non-responders (n = 10 biologically independent samples) at different time points (V1=green, V2=purple, V3=orange, V4=yellow, V5=blue) based on adjusted p values (Benjamini-Hochberg). A generalized linear model (GLM) is applied to detect statistical differences at gene level. A z-score based color range (red-white-blue) is used to show relative gene expression (high – median – low). Genes showing similar expression profiles are clustered into five gene sets (horizontal color bars). B Each gene set is applied to gene set enrichment analysis (GSEA) using the following databases: GO, KEGG and Reactome. Only pathways (terms) with corrected p values below 0.05 are displayed. Terms are connected among each other to create a larger module based on their functional known interrelations using ClueGo app (part of cytoscape visualization tool)^,. Color-filled sections in circles (terms) represent the ratio of genes detected and genes belonging to the corresponding term.