Defective T Memory Cell Differentiation after Varicella Zoster Vaccination in Older Individuals

Abstract

Vaccination with attenuated live varicella zoster virus (VZV) can prevent zoster reactivation, but protection is incomplete especially in an older population. To decipher the molecular mechanisms underlying variable vaccine responses, T- and B-cell responses to VZV vaccination were examined in individuals of different ages including identical twin pairs. Contrary to the induction of VZV-specific antibodies, antigen-specific T cell responses were significantly influenced by inherited factors. Diminished generation of long-lived memory T cells in older individuals was mainly caused by increased T cell loss after the peak response while the expansion of antigen-specific T cells was not affected by age. Gene expression in activated CD4 T cells at the time of the peak response identified gene modules related to cell cycle regulation and DNA repair that correlated with the contraction phase of the T cell response and consequently the generation of long-lived memory cells. These data identify cell cycle regulatory mechanisms as targets to reduce T cell attrition in a vaccine response and to improve the generation of antigen-specific T cell memory, in particular in an older population.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1. Influence of pre-existing VZV immunity on vaccine responses.

(A) VZV-specific T cell frequencies were determined by IFN-γ–specific ELISpot before (day 0) and at days 8±1, 14±1 and 28±3 after vaccination. **p2 = 0.73, p <0.001), but not with initial VZV-specific T cell frequencies (D, p = 0.16). (E and F) Fold change in VZV-specific T cell frequencies from day 0 to day 28 showed no correlation with initial VZV-specific antibodies (E, p = 0.47) or initial T cell frequencies (F, p = 0.12). (G) Fold change in VZV-specific antibodies from day 0 to day 28 did not correlate with fold change in VZV-specific T cell frequencies from day 0 to day 28 (p = 0.38).

Fig 2. Contribution of initial expansion and subsequent contraction of VZV-specific T cells after vaccination to final T cell memory frequencies.

VZV-specific T cell frequencies were measured by IFN-γ–specific ELISpot. Peak ELISpot counts were defined as the highest observed count at either day 8 (25/29 people) or day 14 (4/29 people) after vaccination. Absolute increase to the peak value (expansion) was calculated compared to day 0 counts. Absolute decline from peak (contraction) was calculated compared to day 28 counts. (A) Unsupervised hierarchical clustering analysis based on frequencies of VZV-specific T cells at all three time points. Vaccinees separated into three main clusters [C1 (black), C2 (red), C3 (blue)]. The color code for the different clusters is maintained in Fig 2B–2F. (B) Mean±SD of VZV-specific T cell frequencies at day 0, peak, and day 28 are shown for each cluster. (C-F) Boxplots showing overall increases in VZV-specific T cell frequencies from day 0 to day 28 (C), expansion in VZV-specific T cell frequencies from day 0 to peak value (D), contraction in VZV-specific T cell frequencies from peak to day 28 (E), and age range of individuals in each cluster (F). p-values calculated by one-tailed Wilcoxon-Mann-Whitney test are shown. (G) Vaccinees were grouped according to their age 59 (green) years and T cell frequency trajectories after vaccinations are shown as described in Fig 2B. (H) When corrected for age, VZV-specific T cell expansion showed a weak correlation with overall increase in VZV-specific T cell frequency from day 0 to day 28 that did not reach significance (r = 0.31, p = 0.10). (I) Contraction after peak responses corrected for age inversely correlated with overall increase in VZV-specific T cell frequency from day 0 to day 28 (r = -0.53, p = 0.003).

Fig 3. T cell responses to VZV vaccination are more similar in identical twins than in non-twins.

Pairwise comparisons of fold changes in VZV-specific T cell frequencies (A, C, D) and antibodies (B) were performed for all individuals. Histograms show the frequency distributions of the differences in fold changes for each pair of individuals. Red lines show the position of identical twin comparisons. (A) T cell responses (day 28 to day 0) were more similar between identical twins than between unrelated individuals (p = 0.008). (B) Antibody responses were no more similar between twins than between unrelated individuals (p = 0.44). (C) Effector cell expansion (day 0 to peak) in twins was more similar compared with non-twins with two notable outliers, therefore not reaching significance (p = 0.28). (D) Contraction (peak to day 28) was slightly more similar between twins than between non-twins without reaching significance (p = 0.16).

Fig 4. Correlation of whole blood-derived gene signatures with VZV-specific T cell responses.

(A) Deconvolution of whole blood gene expression for leukocyte subsets was performed. Volcano plots show fold change of gene expression in monocytes (left), lymphocytes (middle) and neutrophils (right) between day 0 and day 1. The genes with significant changes in expression (fold change>1.5, p

Fig 5. Lasso cytokine predictors of antigen-specific T cell expansion and contraction after VZV vaccination.

(A) Principal component analysis of serum concentrations of 51 cytokines on day 0 and day 1 was performed. The box plots show the log transformed fold changes (day 1 to day 0) of PC1 (explaining 60% of variation, left panel) and PC2 (9%, right panel) among individuals. PC2 significantly changes between before and day 1 after vaccination (p = 0.017). (B-E) Panels B and D show the estimated mean square error (y-axis) from a sequence of lasso models in predicting VZV-specific T cell expansion (B) and contraction (D) using the baseline frequencies of VZV-specific T cells and serum cytokine changes between day 0 and day 1 after vaccination. The x-axis represents the log-transformed penalty parameter controlling the model complexity determined by the number of predictors in the model shown on top. Panels C and E plot the predicted vs. the true frequencies after leave-one-out cross-validation for the lasso procedures for VZV-specific T cell expansion (C) and contraction (E). The 45° line is shown for orientation.

Fig 6. Correlation of gene signatures in activated CD4 T cells with VZV-specific T cell responses.

(A) PBMCs of one donor before (day 0) and at days 8 and 14 after vaccination were gated on either total CD4 T cells (grey) or IE63 HLA-DRB1*15:01-tetramer-specific CD4 T cells (red). Expression of CD38 and HLA-DR is shown. (B) Frequencies of CD38+HLA-DR+ cells in total IE63 HLA-DRB1*15:01-tetramer-specific CD4 T cells of three individuals before (day 0) and on days 8 and 14 after vaccination. Time points are compared by paired t-test. (C) CD4+CD38+HLA-DR+ T cells were sorted before (day 0, left panels) and on days 8 (middle panels) and 14 (right panels) after vaccination and arrayed for gene expressions. Expression data were summarized as transcriptional modules and compared to changes in frequencies of VZV-specific T cells determined by IFN-γ ELISpot as shown in Fig 1. Histograms show probability distributions of Pearson correlation coefficients of expressed gene modules (red bars) with decline in VZV-specific T cell frequencies (peak to day 28, upper panels) and with overall increase in frequencies (day 0 to day 28, lower panels). Grey shaded curves represent probability distributions of correlations between random data generated from permutations of module gene labels and sample labels. FDR is calculated using linear regression analysis. The number of significantly associated modules is given as N for p<0.01. (D) The circos plot shows the sharing of significant modules obtained from activated T cells on days 0, 8 and 14. Grey bars show the log2 p-value of correlation between modules and T cell responses. Red and blue circles represent thresholds for modules with significant negative correlation and positive correlation to T cell responses, respectively (p<0.01). Curved lines in the center connect significant modules that are shared. (E) Heat maps show hierarchical clustering of correlation coefficients from the comparisons of gene modules with the fold decrease of peak to day 28 frequencies (A) and the fold increase of day 28 to day 0 frequencies (B). Directionality of correlation coefficients for peak to day 28 contraction was inversed to have equal biological directionality.

Fig 7. Preferential association of cell cycle and DNA repair pathways with VZV-specific T cell responses.

(A) Correlation coefficients of modules that were significantly correlated with the decline in frequencies after peak responses as well as the overall increase from day 0 to day 28 as described in Fig 6, are shown as heat maps. Module nomenclature is from [13]. (B) Networks of genes in modules M22.0 (top left panel), M103 (top right panel) and M4.1 (bottom panel). The sizes of nodes represent the absolute value of correlation coefficients between gene expression level at day 14 after vaccination and peak to day 28 T cell responses. Orange and blue colors indicate positive and negative correlation, respectively. The grey line represents the co-expression relationship as described previously [13]. Gene names for the top genes with the highest correlation coefficient are listed.