Basal T cell activation predicts yellow fever vaccine response independently of cytomegalovirus infection and sex-related immune variations

Abstract

The live-attenuated yellow fever 17D (YF17D) vaccine is a model of acute viral infection that induces long-lasting protective immunity. Among immunocompetent adults, responses to YF17D vary significantly. To understand the sources of this variability, we investigate the influence of sex, age, human leukocyte antigen (HLA) type, and 20 prior infections on basal immune parameters and the cellular and antibody response to YF17D in 250 healthy young individuals. Multivariate regression found that sex and cytomegalovirus (CMV) infection significantly contribute to baseline immune variation but do not affect vaccine responses except for reduced YF17D-specific CD8⁺ frequencies in CMV-infected males. However, the abundance at baseline of non-specific cytokine-expressing T helper cells in circulation is associated with stronger vaccine responses, a state that smoking favors. Additionally, an elevated baseline level of interferon-stimulated CXCL10 is linked to poorer vaccination outcomes. Altogether, YF17D reactivity is conditioned by the baseline immune status independent of sex and CMV-related variations.

Keywords: YF17D; adaptive immunity; flavivirus; human immune variability; immunogenicity; live vaccine; vaccine response; yellow fever vaccine.

Graphical abstract

Figure 1

Effect on baseline immune parameter variation of intrinsic factors and prior infections (A) Multivariate linear regression analysis of baseline immune parameters and 20 prior infections, sex, and age. Only a selection with positive hits is included. Numbers indicate the −log10 of the q-value for the effect of the independent variable. Color scale represents the fold change between positive/negative infection or female/male sex or with a yearly increase of age. Only comparisons passing a multiple testing correction (false discovery rate < 0.1) are depicted for fold change. (B) CMV infection effect on the number of circulating immune cells for CD8⁺, CD4⁺, B, and NK cell parameters. Fold change between infected/uninfected individuals is shown only for comparisons passing a multiple testing correction. Horizontal bars indicate the 2.5%–97.5% confidence intervals. (C and D) Direct comparison of the concentration in blood of different cell populations between CMV-infected and uninfected individuals for CD4⁺ (C) and CD8⁺ (D) subpopulations (n = 158 CMV-negative and n = 78 CMV-positive donors). (E) Representative flow cytometer identification of CD8⁺ subpopulations. Phenotypic markers of the following T cell populations named in the figure: TSCM (CD45RA⁺CCR7⁺CD95⁺), naive (CD45RA⁺CCR7⁺CD95⁻), CM (CD45RA-CCR7⁺), EM (CD45RA-CCR7⁻), EM1 (CD45RA-CCR7-CD27⁺), EM2 (CD45RA-CCR7-CD27⁻), EMRA (CD45RA⁺CCR7⁻), pE (CD45RA⁺CCR7-CD27⁺), E (CD45RA⁺CCR7-CD27⁻), effector (CD45RA⁺CD27⁻), cTfh (CXCR5⁺), Treg (FoxP3⁺CD25⁺CD127⁻), Th1 (CCR6-CXCR3⁺), Th17 (CCR6⁺CXCR3⁻), Th1-17 (CCR6⁺CXCR3⁺), and Th2 (CCR6-CXCR3-CCR4⁺). Phenotypic markers of the following B cell populations named in the figure: AM (CD27⁺CD21⁻), RM (CD27⁺CD21⁺), TLM (CD27⁻CD21⁻), IM (CD27⁻CD21⁺), plasmablasts (CD20⁻CD38⁺), naive (IgD⁺CD27⁻), memory (CD27⁺), DN (IgD⁻CD27⁻), DN1 (DN, CXCR5⁺CD21⁺), DN2 (DN, CXCR5⁻CD21-), memory switched (CD27⁺IgD-IgM⁻), and memory pre-switched (CD27⁺IgD-IgM⁺). Boxplots show a horizontal line indicating the median and lower and upper hinges corresponding to the first and third quartiles. The lower and upper whiskers extend to 1.5x IQR (interquartile range) from the respective hinge. Statistical significance in (C) and (D) was estimated with a Wilcoxon rank-sum test with the designation: ns (non-significant), ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001. See Tables S1 and S2 for linear model sample sizes and results. See also Figures S1–S5.

Figure 2

Sex effect on baseline immune parameters alone and in concert with CMV infection (A and B) Cell concentrations of baseline immune populations identified in the linear model to be significantly affected by sex. Fold change difference between female and males is shown for the comparisons passing a multiple testing correction. Horizontal bars indicate the 2.5%–97.5% confidence intervals. Fold changes were calculated for all variables in cell concentration units except for cytokines (ng/mL), IgM (g/L), and siglec1 expression (mean fluorescence intensity). (C) Comparison of immune variables between males and females (for B cell data, n = 64 males and n = 152 females; for T cell data and lymphocytes, n = 76 males and n = 162 females; for IgM data, n = 81 males and n = 169 females). (D) Comparison of CD3⁺ and CD4⁺ memory subpopulations across males and females with or without CMV infection. Statistical significance against uninfected males is depicted above each group only when significant. An additional comparison is depicted between female uninfected and infected. (n = 52 CMV negative males, n = 23 CMV-positive males, n = 110 CMV-negative females, and n = 53 CMV-positive females). Phenotypic markers of the following T cell populations named in the figure: TSCM (CD45RA⁺CCR7⁺CD95⁺), naive (CD45RA⁺CCR7⁺CD95⁻), CM (CD45RA-CCR7⁺), EM (CD45RA-CCR7⁻), EM1 (CD45RA-CCR7-CD27⁺), EM2 (CD45RA-CCR7-CD27⁻), EMRA (CD45RA⁺CCR7-), pE (CD45RA⁺CCR7-CD27⁺), E (CD45RA⁺CCR7-CD27⁻), effector (CD45RA⁺CD27⁻), cTfh (CXCR5⁺), Treg (FoxP3⁺CD25⁺CD127⁻), Th1 (CCR6-CXCR3⁺), Th17 (CCR6⁺CXCR3⁻), Th1-17 (CCR6⁺CXCR3⁺), and Th2 (CCR6-CXCR3-CCR4⁺). Phenotypic markers of the following B cell populations named in the figure: AM (CD27⁺CD21⁻), RM (CD27⁺CD21⁺), TLM (CD27⁻CD21⁻), IM (CD27⁻CD21⁺), plasmablasts (CD20⁻CD38⁺), naive (IgD⁺CD27⁻), memory (CD27⁺), DN (IgD-CD27⁻), DN1 (DN, CXCR5⁺CD21⁺), DN2 (DN, CXCR5-CD21⁻), memory switched (CD27⁺IgD-IgM⁻), and memory pre-switched (CD27⁺IgD-IgM⁺). Boxplots show a horizontal line indicating the median and lower and upper hinges corresponding to the first and third quartiles. The lower and upper whiskers extend to 1.5x IQR from the respective hinge. Statistical significance in (C) and (D) was estimated with a Wilcoxon rank-sum test with the designation: ns (non-significant), ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001. See Tables S1 and S2 for linear model sample sizes and results. See also Figures S1–S5.

Figure 3

Sex and CMV do not influence YF17D vaccination outcome (A and D) Radar plot associating main vaccination outcomes: CD4⁺ and CD8⁺ functionality score (FS), yellow fever-specific IgM and IgG antibody titer, neutralizing antibody titer, and frequency of antigen-specific CD4⁺ and CD8⁺ measured by intracellular cytokine staining for CD40L⁺IFNγ⁺ and IFNγ⁺TNFα⁺, respectively, with CMV infection status (A, n = 110 CMV negative and n = 49 CMV positive) and sex (D, n = 105 female and n = 56 male). (B) Boxplot comparison depicting the neutralizing and antigen-specific CD4⁺ and CD8⁺ response in CMV-positive (n = 81 for neutralizing titer and n = 62 for CD8⁺ and CD4⁺ responses) and negative (n = 164 for neutralizing titer and n = 136 for CD8⁺ and CD4⁺ responses) individuals. (C) Boxplot comparison of CMV-positive or negative female and male individuals for the CD8⁺ FS (n = 43 CMV-negative males, n = 17 CMV-positive males, n = 83 CMV-negative females, and n = 42 CMV-positive females) and the frequency of antigen-specific CD8⁺ IFNγ⁺TNFα⁺ T cells (n = 47 CMV-negative males, n = 19 CMV-positive males, n = 89 CMV-negative females, and n = 43 CMV-positive females). (E) Boxplot comparison depicting the neutralizing and antigen-specific CD4⁺ and CD8⁺ response between both sexes (n = 168 female and n = 79 male for neutralizing titer and n = 133 female and n = 67 male for CD8⁺ and CD4⁺ responses). Radar plots were built using scaled data, and units refer to standard deviations from the mean. For IgG and IgM titers, data were scaled separately for TBEV pre-vaccinated and non-vaccinated groups. Boxplots show a horizontal line indicating the median and lower and upper hinges corresponding to the first and third quartiles. The lower and upper whiskers extend to 1.5x IQR from the respective hinge. Statistical significance was assessed with Student’s t test for radar plots and with Wilcoxon rank-sum test for boxplot comparisons and is depicted as ns (non-significant), ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001. See also Figure S6.

Figure 4

Vaccine response categories defined by clustering of vaccination endpoints (A) Hybrid hierarchical k-means clustering of main vaccination endpoints: frequency of antigen-specific CD4⁺ T cells, frequency of CD8⁺ T cells and neutralizing antibody titers. 3 to 4 clusters identify good, average, and poor responders in each category. Every row represents an individual. (B) Cohort grouping into 5 categories based on the clusters identified by clustering in (A). (C and D) Response group allocation for CMV-infected and uninfected individuals (C) and for female and male sex (D). The sample size is indicated below each comparison and statistical significance was evaluated with a chi-square test. See also Figures S7–S9.

Figure 5

Baseline immune parameters associated with YF17D vaccination outcome (A) Effect of baseline parameters on YF17D vaccine responses estimated by multivariate linear regression for good and weak responders (defined in Figure 4B). Data reflect the fold change between good/weak responders for the variables with a p value < 0.05. Associations with adjusted p value < 0.1 are depicted in red. Horizontal bars indicate the 2.5%–97.5% confidence intervals. (B) Plasma levels of CXCL10 and sIL6Ra at baseline in good (n = 48) and weak (n = 37) vaccine responders (as defined in Figure 4B). (C) Radar plot of seven scaled and continuous vaccine response endpoints for the individuals in the high (n = 34) and low (n = 28) quantiles of CXCL10 concentrations in plasma at baseline. (D) T cell response endpoints for the high and low quantiles of CXCL10 concentration in plasma (n = 41 individuals in high and n = 38 individuals in low quantile). (E) Frequency of activated CD4⁺ T cells at baseline (TNFα⁺, CD40L⁺IFNγ⁺, CD40L⁺TNFα⁺, and IFNγ⁺IL-2⁺), activated CD8⁺ T cells (IFNγ⁺CD107a⁺), and total DC cell concentration in circulation in good (n = 43) and weak (n = 36) vaccine responders. Boxplots show a horizontal line indicating the median and lower and upper hinges corresponding to the first and third quartiles. The lower and upper whiskers extend to 1.5x IQR from the respective hinge. Radar plots were built using scaled data and units refer to standard deviations from the mean. Statistical significance was assessed with Student’s t test for radar plots and with Wilcoxon rank-sum test for boxplot comparisons and depicted as ns (non-significant), ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001. See also Table S2 for the linear model results.

Figure 6

Frequency of baseline cytokine-expressing CD4⁺ T cells predicts YF17D vaccination outcome (A) Hybrid hierarchical k-means clustering of the frequency of cytokine-expressing CD4⁺ T cell populations at baseline. 4 clusters define individuals with high CD4⁺ inflammatory status (n = 71), average-high (n = 48), average-low (n = 65), and low (n = 33) inflammatory status. (B) Distribution of individuals classified based on the baseline CD4⁺ inflammatory status in (A) across the vaccine response categories defined in Figure 4. Statistical significance is evaluated with a chi-square test and sample size is indicated below every comparison. (C) Radar plot associating vaccination endpoints across the different pre-vaccination CD4⁺ activation clusters identified in (A) (n = 52 high, n = 27 average high, n = 50 average low, and n = 19 low). Radar plots were built using scaled data and units refer to standard deviations from the mean. Statistical significance was assessed with ANOVA. Statistical significance is depicted with the following designation: ns (non-significant), ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001. See also Figures S10 and S11.

Figure 7

Immunological, environmental, and behavioral factors associated with the activated CD4⁺ T cell signature enhancing YF17D vaccine responses (A) Multivariate linear regression analysis of the 768 baseline immune parameters with the CD4⁺ baseline cytokine expression clusters identified in Figure 6 (high versus average-low and low). Fold change is shown only for comparisons passing a multiple testing correction. Horizontal bars indicate the 2.5%–97.5% confidence intervals. (B) Hybrid hierarchical k-means of the T cell memory populations identified in (A) to be associated with baseline cytokine-expressing CD4⁺ T cells. 3 clusters identify participants with a high frequency of differentiated CD4⁺ T cells (n = 71), with average (n = 94), and with low abundance of differentiated but high of naive T cells (n = 73). (C) Radar plot associating the clusters identifies in (B) with the main vaccine outcome endpoints (n = 47 high and n = 45 low). (D) Grouping of individuals with high/low abundance of differentiated T cells within the good and weak responder classification defined in Figure 4. (E) Direct comparison of the frequency of CCR6⁺ CD4⁺ T cells between weak (n = 37) and good (n = 47) vaccine responders. (F) The effect of symptomatic infection in the 2 weeks before YF17D vaccination on the response to the YF17D vaccine. Distribution of individuals with or without a symptomatic infection across clusters defined by the abundance of differentiated T cells (B), baseline activated CD4⁺ T cells (see Figure 6A), good and weak responders (see Figure 4B), and a direct comparison of the FS of YF17D-specific CD4⁺ T cells at day 28 between individuals with (n = 34) or without (n = 151) a symptomatic infection in the 2 weeks before YF17D vaccination. (G) Effect of smoking on the response to the YF17D vaccine. Smoker distribution across clusters of the abundance of differentiated T cells (B), baseline activated CD4⁺ T cells (see Figure 6A), good and weak responders, and a direct comparison with the neutralizing antibody titer (n = 37 smokers, n = 209 non-smokers). Phenotypic markers of the following T cell populations named in the figure: naive (CD45RA⁺CCR7⁺CD95⁻), CM (CD45RA-CCR7⁺), EM (CD45RA-CCR7⁻), cTfh (CXCR5⁺), Treg (FoxP3⁺CD25⁺CD127 -), Th1 (CCR6 -CXCR3⁺), Th17 (CCR6⁺CXCR3⁻), Th1-17 (CCR6⁺CXCR3⁺), and Th2 (CCR6-CXCR3-CCR4⁺). Phenotypic markers of the following B cell populations named in the figure: memory pre-switched (CD27⁺IgD-IgM⁺). Boxplots show a horizontal line indicating the median and lower and upper hinges corresponding to the first and third quartiles. The lower and upper whiskers extend to 1.5x IQR from the respective hinge. Radar plots were built using scaled data and units refer to standard deviations from the mean. Statistical significance was assessed with Student’s t test for radar plots, chi-square for categorical group comparisons, and with Wilcoxon rank-sum test for boxplot comparisons and depicted as ns (non-significant), ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001. See also Table S2 for the linear model results and Figures S10 and S11.