Vaccine-induced ICOS+CD38+ circulating Tfh are sensitive biosensors of age-related changes in inflammatory pathways.

Humoral immune responses are dysregulated with aging, but the cellular and molecular pathways involved remain incompletely understood. In particular, little is known about the effects of aging on T follicular helper (Tfh) CD4 cells, the key cells that provide help to B cells for effective humoral immunity. We performed transcriptional profiling and cellular analysis on circulating Tfh before and after influenza vaccination in young and elderly adults. First, whole-blood transcriptional profiling shows that ICOS+CD38+ cTfh following vaccination preferentially enriches in gene sets associated with youth versus aging compared to other circulating T cell types. Second, vaccine-induced ICOS+CD38+ cTfh from the elderly had increased the expression of genes associated with inflammation, including tumor necrosis factor-nuclear factor κB (TNF-NF-κB) pathway activation. Finally, vaccine-induced ICOS+CD38+ cTfh display strong enrichment for signatures of underlying age-associated biological changes. These data highlight the ability to use vaccine-induced cTfh as cellular "biosensors" of underlying inflammatory and/or overall immune health.

Introduction

Influenza vaccination induces strain-specific neutralizing antibodies in healthy adults, but this process is impaired with aging.1, 2, 3, 4 Moreover, immunological aging, which may or may not be directly linked to chronological aging, is associated with increased morbidity and mortality from infectious diseases.5, 6, 7 Studies of immunological aging and vaccines have identified mechanisms, including alterations at the subcellular,8, 9, 10, 11 cellular,, and systems levels,14, 15, 16, 17 that lead to altered vaccine responses. Vaccination strategies that effectively overcome age-associated immune dysfunction are of great interest for rational vaccine design.

The production of class-switched, affinity-matured antibody by B cells is dependent on help from T follicular helper (Tfh) cells in germinal centers (GCs) in lymphoid tissues, but few studies have evaluated the effects of aging on the GC reaction and GC-dependent cellular responses in humans. Suboptimal vaccine responses with aging have been associated with alterations in cellular pathways, such as tumor necrosis factor-nuclear factor-κB (TNF-NF-κB). For example, several studies observed a modest negative correlation between serum TNF and vaccine responses.,20, 21, 22 However, therapeutic anti-TNF antibody paradoxically did not result in improved humoral responses and instead led to reduced vaccine-induced antibody responses to pneumococcal and influenza vaccines.23, 24, 25 Furthermore, mice deficient in TNF receptor 1 or TNF failed to make GCs and were unable to mount sustained immunoglobulin G (IgG) responses to immunization.26, 27, 28, 29 Thus, the role of TNF signaling in GC and Tfh responses remains poorly understood, but it is likely to be critically important.30, 31, 32 Understanding age-associated alterations in pathways such as TNF and others,,, will be necessary for designing effective vaccines for the elderly.

In humans, directly studying the events leading to productive humoral immunity is challenging because lymphoid tissue is not readily available after vaccination. Several studies have examined a subset of CD4 T cells in human peripheral blood, called circulating T follicular helper cells (cTfh), that possesses phenotypic, transcriptional, and functional similarities to lymphoid Tfh.35, 36, 37, 38, 39, 40 Some features of cTfh likely reflect events in lymphoid tissue. For example, vaccine-induced changes in cTfh correlate with changes in influenza-specific antibody production.38, 39, 40 The subset of cTfh expressing high ICOS and CD38 is expanded in blood 7 days after influenza vaccination, correlates with the plasmablast response, and includes recurring influenza-specific T cell receptor clonotypes each year after influenza vaccination. Moreover, we recently demonstrated that ICOS+CD38+ cTfh can be found in lymph and share characteristic features of GC-Tfh. Thus, cTfh may be useful as a cellular biomarker of immunological events, including vaccination, and possibly the broader state of lymphoid compartments from which these cells recently emerged. Moreover, profiling cTfh may provide opportunities to identify the underlying reasons for suboptimal vaccine responses in human populations, including the elderly.

Here, we tested the effects of aging on cTfh after influenza vaccination. Transcriptional profiling of cTfh subsets responding to vaccination revealed key age-associated pathway-level differences in cTfh responses to influenza vaccination. In particular, transcriptional signatures downstream of TNF-NF-κB signaling pathway were enriched in responding ICOS+CD38+ cTfh from elderly adults compared to young adults at 7 days after vaccination. This transcriptional program was associated with a pro-survival effects in these ICOS+CD38+ cTfh. Focusing on activated cTfh responding to vaccination also revealed additional age-related inflammatory pathway differences. In particular, on day 7 after vaccination, ICOS+CD38+ cTfh from older adults displayed transcriptional evidence of major alterations in interleukin-2-signal transducer and activator of transcription 5 (IL-2-STAT5) signaling, IL-6-STAT3 signaling, and other pathways. When examined for general signatures of aging versus youth, ICOS+CD38+ cTfh enriched far better for these signatures than other blood T cell types, suggesting that vaccine-induced or recently activated cTfh are sensitive biomarkers of underlying changes in immune fitness. Thus, these studies identify key changes in cTfh with age that may relate to altered humoral immunity following vaccination.

Results

ICOS+CD38+ cTfh response appears similar with aging

Antibody responses to vaccination depend on Tfh help to B cells, but the effects of aging on Tfh responses remain poorly understood. We and others have demonstrated that blood cTfh can provide insights into Tfh biology relevant to vaccine responses.38, 39, 40, 41 The subset of cTfh expressing ICOS and CD38 are highly activated, expand after influenza vaccination, and contain the influenza-specific clonotypes,; moreover, activated Tfh similar to those in the blood can be found in lymph and have characteristics similar to GC-Tfh, suggesting a direct relationship between ICOS+CD38+ cTfh and events in lymphoid tissue. Here, we focused on this ICOS+CD38+ cTfh subset to understand age-related changes in vaccine-induced immunity.

Our previous work revealed an age-related decrease in total cTfh, defined as CD4+CXCR5+PD-1+ T cells in the blood, but whether there were differences in the more activated, vaccine-responding cTfh cells expressing ICOS together with CD38 and/or Ki67 was less clear. Thus, because the ICOS+CD38+ cTfh subset contains vaccine-responding T cell clones, we sought to investigate this population in more detail in young and elderly adults (Table S1). We again defined cTfh as non-naive CD4+CXCR5+PD-1+ T cells (Figure S1A)., Similar to our previous findings, elderly adults had a 19% lower frequency of total CXCR5+PD-1+ cTfh in circulation and a 1.2-fold higher expression of ICOS at baseline (p = 0.05 and p = 0.06, respectively; Figures S1B and S1C). However, among cTfh, the frequency of cells expressing ICOS and CD38 was similar in young and elderly at baseline (Figures 1A and S1D) and after influenza vaccination (Figures 1B, 1C, and S1E). The vaccine-induced increase in PD-1 expression was also comparable in both cohorts for the ICOS+CD38+ cTfh population (Figures 1D and S1F). In line with the role of Tfh providing help to B cells, the vaccine-induced fold change in the ICOS+CD38+ cTfh response correlated with the vaccine-induced fold change in the plasmablast response but was not different with aging (young, Pearson r = 0.57, p = 3.9 × 10−3; elderly, Pearson r = 0.67, p = 4.8 × 10−4) (Figures 1E and S1G). Thus, the magnitude of the influenza vaccine-induced cTfh responses and their correlation with the plasmablast response was similar in young and elderly adults.

Figure 1

ICOS+CD38+ cTfh responses appear similar with aging

A cohort of young adults (n = 28) and elderly adults (n = 35) received influenza vaccination in Fall 2014.

(A and B) Flow cytometry for days 0 and 7 after vaccination in the frequency of cTfh co-expressing ICOS and CD38 for young (orange, n = 27) and elderly (purple, n = 35) adults.

(C and D) Fold change between days 0 and 7 for the change in ICOS+CD38+ cTfh frequency (C) or geometric MFI of PD-1 (D) in ICOS+CD38+ cTfh for young (orange, n = 27) and elderly (purple, n = 35) adults.

(E) Correlation between ICOS+CD38+ cTfh frequency fold change at day 7 compared to day 0, and the plasmablast fold change at day 7 compared to day 0 for young (orange, n = 25) and elderly (purple, n = 26) adults.

(F) Peripheral blood mononuclear cells (PBMCs) from young (n = 6) and elderly (n = 8) adults were sorted for ICOS+CD38+ cTfh, ICOS−CD38− cTfh, and naive CD4 at days 0 and 7 after vaccination. Log-transformed transcriptional profiling data were queried for selected genes for young (orange) and elderly (purple) adults at day 7 after vaccination. The heatmap is row normalized.

We next hypothesized that transcriptional profiling may be more sensitive than flow cytometric analysis for revealing differences in cTfh responses. We performed RNA sequencing (RNA-seq) on ICOS+CD38+ cTfh, ICOS−CD38− cTfh, and naive CD4+ cells from young (n = 6) and elderly (n = 8) adults before and after influenza vaccination (Figures S1H–S1J). In both age groups, ICOS+CD38+ cTfh had a distinct transcriptomic signature, with a higher expression of genes previously described in Tfh,,44, 45, 46 such as MAF, BCL6, and TIGIT, compared to naive CD4, and ICOS+CD38+ cTfh expressed more CD38, MKI67, and POU2AF1 than ICOS−CD38− cTfh (Figures 1F and S1K). Moreover, the ICOS+CD38+ cTfh displayed greater similarity to lymphoid GC-Tfh than the ICOS−CD38− cTfh subset by gene set enrichment analyses (GSEAs) for human tonsillar GC-Tfh and mouse lymphoid GC-Tfh, datasets (Figures S1L–S1N; Table S2), irrespective of aging. In general, however, although the three CD4 T cell populations examined had distinct transcriptional profiles, the age-related differences within each of these cell types at the level of individual genes, such as POU2AF1, were minimal (Figures 1F and S1O). Thus, the ICOS+CD38+ cTfh response to influenza vaccination was similar in young versus elderly adults, at least at this level of resolution.

TNF-NF-κB signaling is enriched with aging in ICOS+CD38+ cTfh

Despite the findings above, previous studies have suggested that cTfh from older adults may differ qualitatively or functionally from those in younger adults.,, To further interrogate the question of age-related differences in the cTfh response to vaccination, we directly compared the transcriptional profiles of ICOS+CD38+ cTfh from young and elderly adults at day 7 (Figure 2A), when the ICOS+CD38+ cTfh population in blood is enriched for influenza-specific cells. Transcriptional differences were revealed for ICOS+CD38+ cTfh from young compared to elderly adults at day 7 (Figures S2A and S2B). Differentially expressed genes included PLCG2, RNF130, RAB2A, and ACSL3, which were increased in cTfh from elderly adults, whereas CD27, CD6, EIF1AD, EEF2, and MAP2K2 were increased in the young (Figure S2C). CD27, a member of the TNF receptor family, was more strongly expressed at the protein level in the ICOS+CD38+ cTfh from young adults compared to elderly adults (Figure S2D). Overall, however, examining differences in the expression of individual genes did not reveal obvious age-related changes (Figures S1J, S1K, S2A, and S2B) that would explain differences in vaccination outcomes with age, suggesting that differences in cTfh between young and older adults, if they existed, may reside in the coordination of the transcriptional programs.

Figure 2

TNF-NF-κB pathway is enriched with aging in ICOS+CD38+ cTfh

(A) Schematic for RNA-seq analyses.

(B and C) Weighted gene correlation network analysis on ICOS+CD38+ cTfh at day 7 from young (B) and elderly (C) adults.

(D) Module preservation analysis was performed using Fisher’s exact test. Heatmap color and value indicate the −log10(p value) for the overlap in genes in modules.

(E) Genes ranked based on module membership for elderly module EM4. Then, GSEA was performed using the MSigDB HALLMARK collection. Positive enrichment scores indicate enrichment for genes in EM4.

(F) Transcription factors with module membership >0.80 in EM4 displayed as a multiple association network (GeneMANIA without gene prediction).

(G) Ingenuity Pathway Analysis for predicted upstream regulators for ICOS+CD38+ cTfh at day 7 for young and elderly adults.

(H) Differentially expressed NF-κB target genes comparing ICOS+CD38+ cTfh at day 7 from elderly (purple) and young (orange) adults.

(I) GSEA for NF-κB target genes to compare ICOS+CD38+ cTfh from elderly and young adults.

(J) Example plot from 1 young adult and 1 elderly adult for total NF-κB p50 protein from an independent cohort of young and elderly adults. Geometric mean fluorescence intensity (MFI) shown.

(K) Total NF-κB p50 for ICOS+CD38+ cTfh at baseline in young (orange) and elderly (purple) adults (n = 16 in each group), as taken from an independent cohort of young and elderly adults.

(L and M) Gene expression for TNFRSF1A (L) (n = 6 for young; n = 8 for elderly) and TNFRSF1B (M) (n = 8 for elderly) shown at day 7 after vaccine from log2-transformed counts data for young (orange) and elderly (purple) adults.

We, therefore, next hypothesized that there may be broader age-related differences at the pathway level. GSEA identified multiple gene sets potentially relevant to age-related biological changes that preferentially enriched in the ICOS+CD38+ cTfh from the elderly adults compared to young ICOS+CD38+ cTfh at day 7 post-vaccination (Figure S2E). For example, the IL-2/STAT5 signaling pathway was previously demonstrated to have an inhibitory effect on Tfh differentiation via Blimp-1. There was stronger enrichment for the IL-2/STAT5 signaling pathway in ICOS+CD38+ cTfh from the elderly adults (Figure S2E), although this enrichment did not correspond to detectable differences in PRDM1 transcripts in the blood (Figure 1F). Among the inflammatory signatures with the greatest enrichment in the ICOS+CD38+ cTfh from the elderly adults was TNF-NF-κB signaling (Figure S2E). These enrichment patterns were biased to pathway enrichment in ICOS+CD38+ cTfh from elderly adults, with no specific gene sets enriched in ICOS+CD38+ cTfh from the young adults even at a more relaxed false discovery rate (FDR) of <0.20, thereby suggesting more distinct biological changes in the elderly. These data highlighted the potential differential use or engagement of the NF-κB signaling pathway by ICOS+CD38+ cTfh in elderly adults compared to young adults at day 7 post-vaccination.

These observations provoked the hypothesis that inflammatory signals could result in transcriptional network differences in the vaccine-induced cTfh from elderly adults. To investigate this possibility, we next applied weighted gene correlation network analysis (WGCNA) to more deeply interrogate the underlying transcriptional networks of vaccine-induced cTfh from young versus elderly. An advantage of WGCNA is the ability to identify biologically meaningful groups of co-varying genes, even if the individual genes are not statistically differentially expressed. Taking advantage of co-variance often identifies major changes in biological circuits that are not apparent at the level of individual genes. At day 7 after vaccination, 6 transcriptional modules were identified for the ICOS+CD38+ cTfh subset isolated from young adults, whereas 8 transcriptional modules were identified from the elderly (Figures 2B, 2C, S2F, and S2G; Table S3). Although most modules in young adults overlapped with ≥1 modules in the elderly, 2 modules in the elderly (EM1 and EM4) did not have significant overlap with any module in young adults (Figure 2D). To evaluate the underlying biology of these modules, we used GSEA on genes ranked by module membership (Figures 2E and S2H), which revealed TNF-NF-κB signaling to be the highest-scoring pathway associated with EM4. After filtering the EM4 gene list for transcription factors, we identified a dense network of six hub genes—IRAK3, MYD88, TNFAIP3, STAT5A, REL, and TNFSF11—that had direct relevance to the NF-κB signaling pathway (Figures 2F and S2I; Table S4). In addition, Ingenuity Pathway Analysis (IPA) for upstream regulators, predicted lipopolysaccharides (LPS), IL-1B, NF-κB (complex), TNF, and MYD88 to be more activated in ICOS+CD38+ cTfh from elderly adults compared to young adults at day 7 following vaccination, and signatures associated with blocking TNF signaling biased in the opposite direction (Figures 2G and S2J). In addition, several other inflammatory pathways were strongly enriched in EM4, including inflammatory response, transforming growth factor β (TGF-β) signaling, IL-2 STAT5 signaling, and IL-6 JAK STAT3 signaling, as were signatures of hypoxia and complement (Figure 2E). Module EM1 also enriched for inflammatory pathways with type I and type II interferon pathways as the strongest signatures (Figure S2H). These results identified coordinated transcriptional changes reflecting inflammatory signaling in the elderly.

We next asked whether genes downstream of NF-κB were affected, using a list of ∼1,500 NF-κB target genes. We identified 58 targets of NF-κB that were upregulated in ICOS+CD38+ cTfh at day 7 in elderly adults compared with 25 that were upregulated in young adults at Padj < 0.05 (Figures 2H and S2K). GSEA also identified stronger enrichment of the gene targets of NF-κB in the ICOS+CD38+ cTfh at day 7 from elderly compared to young adults (Figure 2I). These data support the observation of increased inflammatory signaling in ICOS+CD38+ cTfh with aging.

Circulating inflammatory cytokines can differ in elderly compared to young adults, which is consistent with the notion of immune dysregulation and chronic inflammation during aging., At baseline, we identified higher levels of TNF, C-X-C motif chemokine 11 (CXCL11), and macrophage inflammatory protein 1β (MIP1β) in the plasma of elderly compared to young adults (Figure S2L). We then used gene set variation analysis (GSVA) to investigate whether the plasma TNF concentrations had any relationship to the increase in the TNF-NF-κB gene set signature in ICOS+CD38+ cTfh from elderly adults at 7 days after immunization, but we found no correlation (Figures S3A and S3B). These data suggested that serum TNF alone was unlikely to be the sole cause of the observed TNF-NF-κB transcriptional signature.

To further investigate a potential role for differences in NF-κB signaling, we next tested whether there was evidence for constitutive upregulation and/or activation of the NF-κB pathway ex vivo. NF-κB1 (p50) protein expression was elevated in ICOS+CD38+ cTfh from elderly adults compared to young adults (Figures 2J, 2K, S3C, and S3D) suggesting the greater activation of, or the ability to activate, this pathway in response to inflammatory mediators in the ICOS+CD38+ cTfh from aged individuals. TNF signaling occurs via the interaction of soluble or membrane-bound TNF with TNF receptors CD120a and CD120b (TNFR1 and TNFR2, respectively), with soluble TNF signaling more strongly through TNFR1 than TNFR2. To test whether cTfh differentially expressed TNFR1 or TNFR2, transcriptional profiling data for CD4 T cell subsets was evaluated for TNFR1 (gene name TNFRSF1A) and TNFR2 (gene name TNFRSF1B). ICOS+CD38+ cTfh had higher gene expression for TNFR1 and TNFR2 than other CD4 subsets at baseline (Figure S3E). Following vaccination, TNFRSF1A expression in ICOS+CD38+ cTfh was increased in elderly compared to young adults, although the expression of TNFRSF1B was similar (Figures 2L, 2M, and S3F–S3H). Thus, several components of the TNF signaling pathway are elevated in vaccine-induced ICOS+CD38+ cTfh from elderly compared to young adults.

TNF signaling promotes Tfh-B cell interactions

TNF-NF-κB signaling with aging has been associated with age-related immune dysfunction. Our transcriptional and network analyses highlighted a role for increased NF-κB signaling with aging within ICOS+CD38+ cTfh at day 7 post-vaccination. We hypothesized that TNF-NF-κB signaling may adversely affect Tfh-B cell interactions. To test this idea, we used cTfh-B cell coculture using autologous naive B cells and CD4 T cells from young adults to interrogate pathways involved in cTfh provision of help to B cells in vitro.35, 36, 37, 38 As expected, cTfh (defined as non-naive CD4+CXCR5+PD-1+) were the major supplier of B cell help, as evidenced by higher IgM and IgG1 in supernatant after staphylococcal enterotoxin B (SEB) stimulation compared to controls (Figures 3A and 3B). To test the impact of TNF on the ability of cTfh to provide help, recombinant TNF protein or α-TNF antibody was added at the beginning of the coculture. Adding TNF did not augment B cell help and antibody production compared to SEB alone (Figures 3A and 3B). In contrast, the addition of α-TNF blocking antibodies reduced supernatant IgG1 concentrations 10-fold compared to SEB alone (Figure 3B). The differentiation of naive B cells into CD19+CD38+CD20lo plasmablasts was reduced in the presence of α-TNF antibodies, but the addition of TNF also had no effect (Figure S4A). IgG1 production by naive B cells stimulated in vitro with α-IgM and soluble trimeric CD40L did not change upon the addition of α-TNF antibodies (Figure S4B), suggesting that reduced antibody production by B cells in the context of the blockade of TNF signaling in the presence of cTfh was through the effects of TNF signaling on Tfh. Live cell counts for T and B cells decreased when α-TNF antibodies were added to the cocultures (Figure S4C) or when the TNF-converting enzyme inhibitor TAPI-0 was used (Figure S4D), suggesting a potential pro-survival role for TNF signaling. Although nonspecific effects of α-TNF antibodies in these cultures are possible, the decreased survival observed with increasing concentrations of TAPI-0 also support a role for soluble TNF signaling. These studies only addressed whether TNF signals could play a role in the Tfh provision of help to B cells in general, but they did not address the question of aging because of potential age-related confounders in the autologous naive B cell compartment from elderly adults. To further explore the role of TNF in cTfh biology in the setting of aging, TNF was assessed in supernatants from previously performed cocultures in which CD4+CXCR5+PD-1+ cells were cocultured at a 1:1 ratio with allogenic naive B cells from a common young donor, thereby allowing the biology of cTfh from older donors to be assessed in vitro. We found a correlation between age and TNF in the supernatant after 7 days of coculture (Figure S4E; Pearson r = 0.35, p = 0.048). Because these cocultures used the same source of naive allogeneic B cells from a single young donor, the greater production of TNF was likely due to cTfh from the elderly adults. These data indicated an age-associated upregulation of the NF-κB signaling pathway in the ICOS+CD38+ cTfh after vaccination that was associated with increased TNF production during Tfh-B cell interactions in the elderly. Moreover, these data suggested a role for TNF-NF-κB signaling in promoting cellular survival in the context of Tfh-B cell interactions.

Figure 3

TNF signaling promotes Tfh-B cell interactions

(A and B) PBMC from young adults were freshly isolated and sorted for coculture of T cell subsets (naive, gated as CD4+CD45RA+CD27+; CXCR5− memory, non-naive CD4+CXCR5−; cTfh, non-naive CD4+CXCR5+PD-1+) with autologous naive B cells (CD3−CD19+CD27loIgD+). Supernatant IgM (A) and IgG1 (B) were measured after 7 days, as shown for the following conditions: unstimulated (gray), SEB alone (0.5 μg/mL, green), SEB with recombinant human TNF (125 ng/mL, tan), or SEB with α-TNF antibodies (2 μg/mL, sienna) (1-way repeated-measures ANOVA with Holm-Sidak’s test; n = 12 per group).

(C and D) Correlation shown for the TNFRSF1A in ICOS+CD38+ cTfh at day 7 compared with the H1N1-specific (C) hemagglutinin inhibitory (HAI) titer (n = 14) or H3N2-specific (D) HAI titer (n = 14) in young (orange) and elderly (purple) adults.

(E) Correlation of apoptosis and NF-κB gene set GSVA scores for ICOS+CD38+ cTfh at day 7 for young (orange, n = 6) and elderly (purple, n = 8) adults.

(F) Pearson coefficients for BCL family member genes compared to GSVA scores for TNF-NF-κB signaling in ICOS+CD38+ cTfh at day 7 for young (x axis) and elderly (y axis) adults. Genes classified as either pro-survival (dark blue) or apoptotic (sea green).

Given the coculture data, we hypothesized that TNF-NF-κB signaling may be associated with a beneficial effect on the humoral response in vivo. Influenza vaccination was associated with the higher expression of TNFRSF1A in ICOS+CD38+ cTfh in elderly adults compared to young adults following vaccination (Figures 2L and 2M). Thus, we next compared the expression of TNFRSF1A and TNFRSF1B to the neutralizing antibody responses. Both young and elderly adults produced influenza strain-specific binding antibodies and had high neutralizing antibody titers following vaccination (Figures S4F and S4G). Examining young and elderly adults together, TNFRSF1A expression by ICOS+CD38+ cTfh (but not other CD4 subsets) positively correlated with the fold change in the strain-specific neutralizing antibody titers at day 28 compared to day 0 (Figures 3C, 3D, and S4H). These data support the notion of a positive role for TNF-NF-κB signaling in Tfh-B cell interactions. Moreover, these observations are consistent with observations that clinical TNF blockade resulted in reduced vaccine immunogenicity.,

Since TNF blockade was associated with reduced survival of cTfh and B cells in cocultures in vitro, we next asked whether there was an association between TNF signaling and cell death in vivo. GSVA scores for TNF-NF-κB signaling correlated strongly with apoptosis similarly for both age groups (Figure 3E). This positive correlation seemed paradoxical given the in vitro data above. However, cells did not appear to be undergoing active regulated cell death, as the cellular membrane of ICOS+CD38+ cTfh was intact based on the exclusion of viability dye (Figure S1A), and ICOS+CD38+ cTfh of young and elderly adults did not demonstrate reduced mitochondrial membrane potential compared to other CD4 T cell subsets (Figure S4I). Thus, although pro-apoptotic genes may increase with age, resistance to apoptosis could be linked to the induction of pro-survival genes. The expression of anti-apoptotic genes BCL2A1 and MCL1 correlated with the GSVA scores for TNF-NF-κB signaling in all adults (Figures 3F and S4J). Similarly, the expression of Survivin (gene name BIRC5; Figures 1F and S4K), a cIAP family member that can restrain caspase-3 activation, was high in ICOS+CD38+ cTfh. Thus, together with the reduced viability in vitro with TNF blockade, these data suggest that TNF-NF-κB signaling may be necessary to counterbalance pro-apoptotic effects and foster cell survival circuits.

Vaccination elicits discordant transcriptional pathway responses with aging in ICOS+CD38+ cTfh

To better understand the full spectrum of transcriptional changes in cTfh during aging, we analyzed the vaccine-induced transcriptional signatures in ICOS+CD38+ cTfh on day 7 post-vaccination compared to day 0 for young versus elderly adults. Overall, influenza vaccination was associated with relatively modest changes in transcriptional profiles in ICOS+CD38+ cTfh from young and elderly adults (Figure 4A), suggesting that pre-vaccination ICOS+CD38+ cTfh share many transcriptional characteristics with these cells post-vaccination.

Figure 4

Age-dependent differences in transcriptional profiles of ICOS+CD38+ cTfh

(A) t-Distributed stochastic neighbor embedding (t-SNE) analysis was performed for all CD4 T cell subsets. ICOS+CD38+ cTfh shown before and after vaccination for young (n = 6) and elderly (n = 8) adults.

(B) Aggregated GSEA results for Hallmark gene sets comparing day 7 versus day 0 for ICOS+CD38+ cTfh in young (orange) and elderly (purple) adults.

(C–F) GSEA shown for day 7 versus day 0 for ICOS+CD38+ cTfh from young (orange) and elderly (purple) adults for TGF-β signaling (C), E2F targets (D), IL-2 signaling (E), and TNF-NF-κB signaling (F) gene sets.

We then used pathway enrichment combined with pre-ranked GSEA to compare transcriptional signatures from ICOS+CD38+ cTfh from young versus elderly adults. Clear enrichment patterns were observed for many key gene sets. However, overall, the enrichment patterns were only concordant between young and elderly adults for ∼1/3 of the Hallmark gene sets examined (Figures 4B, and S4L; Table S2), suggesting distinct biology for this cell type in elderly versus young adults. For example, TGF-β signaling displayed reduced enrichment at day 7 compared to day 0 in ICOS+CD38+ cTfh from both young and elderly adults (Figure 4C). In contrast, in young adults, there was positive enrichment (associated with upregulation at day 7) for the E2F target gene set in ICOS+CD38+ cTfh, whereas elderly adults had negative enrichment for this pathway (Figure 4D). The opposite pattern was observed for IL-2-STAT5 (Figure 4E) and TNF-NF-κB signaling (Figure 4F), both of which were enriched in the elderly adults compared to the young adults. In general, ICOS+CD38+ cTfh from the young adults enriched positively for signatures of regulation of proliferation and Myc biology, but negatively for signatures of inflammatory signaling. In contrast, ICOS+CD38+ cTfh from the elderly adults enriched positively for metabolic pathways and inflammatory signaling. These results demonstrate multiple pathway-level differences with aging in ICOS+CD38+ cTfh after influenza vaccination.

Age-related differences preferentially revealed by CD4 T cell subset analysis

We next hypothesized that the age-related differences in cellular pathways may be distinct in different CD4 T cell subsets. We therefore compared enrichment for 9 Hallmark gene sets by GSEA in naive CD4 T cells, ICOS−CD38− cTfh, and ICOS+CD38+ cTfh from young and elderly adults before (day 0) and after (day 7) influenza vaccination (Figures 5A–5I, S5A, and S5B; Table S2). These analyses revealed several key findings. First, consistent with our data in ICOS+CD38+ cTfh (Figure 1), most of the pathways showed only weak differences with aging at baseline. Two exceptions included a moderate enrichment of Myc targets (Figure 5D) in ICOS−CD38− cTfh and ICOS+CD38+ cTfh from the elderly adults at baseline and an enrichment of oxidative phosphorylation (Figure 5F) in the ICOS−CD38− cTfh from the young adults at day 0. Second, the overall similarity at baseline was contrasted by substantial differences following vaccination. Of all of the gene sets, 57% changed in enrichment direction (i.e., from young to elderly or vice versa) for ICOS+CD38+ cTfh between days 0 and 7, compared with 30% for naive CD4 between days 0 and 7. However, one of the most clear patterns was the uncovering of an underlying bias for inflammatory signatures, including TNF-NF-κB, inflammatory response, and IL-2-STAT5 gene sets to be preferentially enriched in the CD4 T cells of the elderly adults, often with the greatest enrichment in the ICOS+CD38+ cTfh subset (Figures 5A–5C, and 5J). Third, ICOS+CD38+ cTfh showed the greatest dynamic range of the 3 T cell subsets, based on the difference with aging in the day 7 and day 0 normalized enrichment scores per subset (Figure 5J). Whether these differences reflect differential systemic induction of inflammation upon vaccination, distinct abilities of T cells to induce transcriptional responses related to these pathways due to aging, or both, remain to be determined.

Figure 5

Age-related differences in signaling pathways are preferentially revealed by subset and time point

(A–I) GSEA performed to compare young versus elderly for each CD4 T cell subset for each time point. Normalized enrichment score shown as circle size and color.

(J) “Delta-NES” score was calculated by subtracting the day 0 NES score from the day 7 NES score with aging for each gene set for each subset. Each dot represents 1 gene set. The red bar indicates median.

The different signatures of ICOS+CD38+ cTfh from young and elderly adults are most apparent at day 7 post-vaccination

Differences in the transcriptional profiles of CD4 T cells with aging were most prominent in ICOS+CD38+ cTfh at the cellular pathway level following influenza vaccination (Figure 5). We next wanted to determine whether a whole-blood baseline gene signature of aging would be reflected in the same manner and whether vaccination was necessary to identify the features of aging.

To test this idea, baseline whole-blood microarray data from the full study cohort (27 young and 35 elderly adults, Immport SDY739) was used to assess differential gene expression with aging, using the same cohort from which the adults for the RNA-seq (Figures 1, 2, 3, 4, and 5) were derived. From these data, we constructed signatures of youth (354 genes upregulated in the young) or aging (232 genes upregulated in the elderly) (Figure S6A; Table S5). These signatures were then used to test whole-blood microarray data from other years’ cohorts from the same study., We found strong enrichment of the signatures of youth and aging in the young and elderly adults, respectively (Figure 6A). The signature of youth had Gene Ontology terms for RNA metabolism, whereas the signature of aging had terms for fatty acid metabolism and cytoskeletal regulation (Figure S6B), in agreement with other signatures of aging.

Figure 6

ICOS+CD38+ cTfh transcriptional profiles reveal signatures of aging

(A) Signatures for youth and aging were constructed from the differential expression of young versus elderly whole-blood transcriptional profiling from Immport study SDY739. These signatures were tested by pre-ranked GSEA for other vaccine years in this study (SDY622, SDY648, and SDY819).

(B) The signatures of youth and aging were validated against publicly available whole-blood transcriptional profiles (GSE accessions GEO: GSE123696, GEO: GSE123697, GEO: GSE123698, and GEO: GSE79396) and Nanostring target profiling (EGAS00001002460, Milieu Intérieur) for young (ages 20–40) and elderly (ages 60 and older) adults. Normalized enrichment score (NES) shown for youth (left) and aging (right) signatures.

(C) CD4 T cell subsets were tested for signatures of youth and aging. Naive (yellow), ICOS−CD38− cTfh (green), and ICOS+CD38+ cTfh (orange) are shown, with days 0 and 7 connected by dotted lines. The yellow region indicates FDR < 0.05.

To further validate these signatures, we identified several published studies,66, 67, 68 that generated whole-blood transcriptional profiles on adults younger than age 40 or older than age 65 (Table S6). From these data, we performed differential expression analysis of young compared to elderly adults in multiple geographically distinct studies and again found robust enrichment for these signatures in the external datasets (Figures 6B and S6C). To test these signatures on continuous (rather than cohorted) data, GSVA scores for the signatures of youth and aging were compared against the Nanostring data for adults in a large cohort (n = 986) across lifespan (Milieu Intérieur cohort). In this cohort, the signature of youth negatively correlated with age (Pearson r = −0.28, p = 8.6 × 10−20, n = 986) and the signature of aging positively correlated with age (Pearson r = 0.31, p = 4.6 × 10−23) (Figure S6D). These results demonstrate conserved gene expression patterns for youth and aging in whole-blood transcriptional profiling data in multiple published studies.

We next tested whether the signatures of youth and aging could be discerned in the CD4 T cell subsets examined above before and after influenza vaccination. For example, the signature of youth was positively enriched in the young cohort in all of the subsets at day 0. However, this signature of youth was most strongly enriched in the ICOS+CD38+ cTfh at day 7 after vaccination (Figures 6C and S6E). Similarly, the signature of aging was evident at day 0, such as in naive CD4 T cells from the elderly adults. Again, however, the strongest enrichment of the aging signature was revealed in the ICOS+CD38+ cTfh after vaccination (Figures 6C and S6E). These data demonstrated that ICOS+CD38+ cTfh are susceptible to immunologic perturbation with influenza vaccination and sensitively reflect general immunological or organismal environment of youth and aging in the cellular transcriptional program.

Discussion

Induction of protective antibody responses is the goal of most vaccines. Many vaccines are less effective in elderly adults, but the underlying causes, including how aging affects Tfh and B cell responses, remain poorly understood. In this study, we examined the cTfh response to vaccination in young and elderly adults to interrogate the effects of aging on Tfh. There were several key observations from these studies. First, despite the similar magnitude of responses to vaccination and in the expression of individual Tfh-associated genes, ICOS+CD38+ cTfh from young and elderly adults differed in network level transcriptional signatures, revealing underlying changes in key biological pathways not apparent at the level of individual genes. Second, signatures of inflammatory responses, IL-2-STAT5 signaling, and TNF-NF-κB signaling in particular were substantially stronger in ICOS+CD38+ cTfh from elderly adults. Third, vaccine-induced ICOS+CD38+ cTfh reflected underlying signatures of aging versus youth more strongly than ICOS−CD38− cTfh or naive CD4 T cells. These latter observations indicate that recently activated cTfh may function as biosensors of underlying inflammatory and/or host physiological environments and reflect underlying features of “immune health.”

Increased baseline inflammation in elderly adults has often been associated with declines in overall health and, in particular, the TNF-NF-κB pathway has been implicated in many pathological changes with aging. However, these correlations are at odds with the apparent detrimental effects of TNF pathway blockade on vaccination23, 24, 25 that suggest that blocking residual TNF signaling impairs, rather than enhances, vaccine-induced antibody responses. Moreover, data from mice indicate a key role for TNF in GC responses.26, 27, 28, 29 Thus, how TNF-NF-κB signaling regulates GC-dependent, vaccine-induced immunity in elderly adults remains incompletely understood. Here, TNF-NF-κB signaling was correlated with quantitatively better antibody responses, suggesting a positive role for TNF signaling in human GC-dependent immune responses. ICOS+CD38+ cTfh displayed a strong transcriptional signature of TNF-NF-κB pathway signaling in all adults, especially the elderly, and blockade of TNF signaling in vitro led to reduced Ig production by B cells. Furthermore, these in vitro studies indicated reduced cellular survival upon blockade of TNF signaling. Our results point to a pro-survival effect of TNF signaling. likely due to the effects of TNF signaling on cTfh and/or TNF production by cTfh. Despite the increased expression of pro-apoptotic Bcl family members such as BID in ICOS+CD38+ cTfh, there was a concomitant increase in the expression of pro-survival factors BCL2A1 and MCL1 that strongly correlated with TNF-NF-κB GSVA scores in ICOS+CD38+ cTfh at day 7, suggesting a TNF-NF-κB-dependent survival signal in activated cTfh. These genes have NF-κB binding sites in their respective promoter sequences, linking TNF signaling to a potential pro-survival circuit in these cells. Moreover, TNFR1 expression in ICOS+CD38+ cTfh at day 7 positively correlated with neutralizing antibody production. These data are in agreement with and may provide an explanation for clinical data from patients receiving therapeutic TNF blockade demonstrating fewer GC reactions and attenuated antibody responses to influenza vaccination. Whether TNF signaling in GC-Tfh regulates proteins involved in providing B cell help such as ICOS, CD40L, and IL-21 will be interesting to investigate in future studies.

Previously, we identified correlations between cTfh responses and influenza-specific IgG antibody responses in young adults but not in elderly adults. Here, we observed a similar magnitude of the total ICOS+CD38+ cTfh response following vaccination in young and elderly adults, and the cTfh responses correlated with the plasmablast response. Nevertheless, vaccine-induced neutralizing antibody titers were slightly lower in elderly compared to young adults. ICOS+CD38+ cTfh preferentially reflected underlying inflammatory and age-related biological signatures 7 days after influenza vaccination, compared to ICOS−CD38− cTfh or naive CD4 T cells. There are several potential reasons for this. For example, the increase in the ICOS+CD38+ cTfh subset at day 7 post-vaccination likely reflects an induction and/or reactivation of influenza-specific CD4 T cells. Thus, these influenza-specific T cells may be qualitatively different from the ICOS+CD38+ cTfh circulating at baseline. However, a second related feature of these vaccine-induced cTfh may be the synchronous nature of the responses. The ICOS+CD38+ cTfh responses at baseline likely reflect ongoing chronically activated cTfh or cTfh in different phases of acute activation. In contrast, the vaccine-induced cTfh are aligned in their activation and differentiation kinetics, perhaps resulting in more focused transcriptional circuitry. Recent activation may also facilitate sensitivity and/or changes in inflammatory cytokine signaling, as has been suggested for NF-κB signaling in GC B cells. In addition to enrichment for TNF-NF-κB signaling and a broader signature of aging, activated cTfh also displayed transcriptional evidence of age-dependent alterations in many other inflammatory pathways, including IL-2-STAT5 signaling, TGF-β pathway signaling, and IL-6-JAK-STAT3 signaling. These observations were notable given the regulatory role of IL-2-STAT5 signaling in Tfh differentiation.,, The data presented here point to the use of vaccine-induced ICOS+CD38+ cTfh as a window into potentially broader underlying host immunological changes.

To aid in understanding the impact of aging on cTfh, we constructed transcriptional signatures of youth and aging based on whole-blood transcriptional profiling from 27 young adults and 34 elderly adults. These signatures were then validated using multiple, geographically distinct cohorts, representing a total of 566 young adult samples and 344 elderly adult samples. Furthermore, the GSVA scores for the signatures of youth and aging correlated with age as a continuous variable in the Milieu Intérieur study of nearly 1,000 adults, with age explaining 7.8%–9.6% of the variance in the GSVA scores of our signatures. Overall, these data indicated that the highest enrichment scores for our signatures of youth and aging were in the day 7 ICOS+CD38+ cTfh subset. These observations suggest that cTfh, and especially the synchronized, vaccine-induced cTfh at day 7 after influenza vaccination, may be sensitive biosensors of underlying features of immune inflammation or immune health.

Studies of vaccine-induced immune responses during aging are needed to understand poor immune responses in elderly adults, who are at heightened risk from infection. These data not only identify roles for inflammatory pathways in alterations of cTfh during aging but also point to vaccine-induced cTfh as a cell type that may provide a window into changes in overall immune health and fitness.

Limitations of study

Aging is a complex phenomenon, and as a result, the number of participants evaluated here in the RNA-seq studies may not fully reflect the subtleties or full heterogeneity of aging across individuals. Moreover, our study recruited community-dwelling, relatively healthy adults. As a result, evaluation of the effects of comorbidities and concomitant medications was not possible here. Only one type of inactivated influenza vaccine was used for participants in these studies, and it may be interesting to extend these observations to other inactivated influenza vaccines or even other types of vaccines. The extent to which a circulating subset, ICOS+CD38+ cTfh, can serve as a cellular biomarker will need to be further assessed in conjunction with studies of age-associated changes in lymphoid Tfh through access to lymph node material or perhaps lymph. Finally, future studies will be needed to fully understand how aging and TNF signaling affect Tfh, and by extension, vaccine responses.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Ramin Sedaghat Herati (ramin.herati@nyulangone.org).

Materials availability

This study did not generate new unique reagents.

Data and code availability

Transcriptional profiling data reported here are available at the Gene Expression Omnibus (GEO) under accession number GEO: GSE134416. R scripts used in analyses and figure generation are available on Github (https://github.com/teamTfh/cTfh_AgingSignature).

Experimental model and subject details

Flu vaccine cohorts

In the Fall of 2014, study participants were recruited and consented at the Clinical Research Unit at Duke University Medical Center (Durham, NC, USA), in accordance with the Institutional Review Boards of both Duke University (Protocol Pro00031791) and the University of Pennsylvania (Philadelphia, PA, USA) (Protocol #812170). Adults were classified as young (30-40 years of age) or elderly (65 years of age or older) (Table S1). Adults were eligible if they were community-dwelling and had not received influenza vaccine in the prior 6 months; they were excluded if they had contraindications to influenza vaccine, active substance abuse, HIV/AIDS, clinically active malignancy, immunomodulatory medication need (i.e., chemotherapy, corticosteroids), or active illness (i.e., active respiratory tract infections). Seasonal trivalent influenza vaccine (Fluarix, GlaxoSmithKline) was administered and peripheral venous blood was drawn on days 0, 7, and 28 after vaccination. Blood was collected into heparinized tubes and shipped overnight to Philadelphia, PA. All studies were conducted according to the Declaration of Helsinki principles.

Primary cells

Primary human PBMC were obtained at the University of Pennsylvania from de-identified donors through the Human Immunology Core (see Acknowledgments) and through a separate protocol (Protocol #820151). Samples used in Figure S3E were from adults who provided informed consent in accordance with the Institutional Review Boards of the Louis Stokes Veterans Affairs Medical Center and Case Western Reserve University, as previously reported. All studies were conducted according to the Declaration of Helsinki principles.

Method details

Flow cytometry

PBMC and plasma were isolated using Ficoll-Paque PLUS (GE Healthcare) and stained for surface and intracellular markers. Permeabilization was performed using the Intracellular Fixation/Permeabilization Concentrate and Diluent kit (ThermoFisher). Antibodies and clones are described in Key resources table. Antibodies against CD8 and CD3 were conjugated to QDot 565 and 585 (Fisher), respectively. Cells were resuspended in 1% para-formaldehyde until acquisition on a BD Biosciences LSR II or a BD Symphony A5 cytometer. Fluorescence-minus-one controls were performed in pilot studies. DiOC6 (Fisher) staining was performed according to manufacturer instructions.

Co-culture experiments

Fresh PBMC were sorted for T cell subsets (Naive, CD4+CD45RA+CD27+; CXCR5- memory, non-naive CD4+CXCR5-; cTfh, non-naive CD4+CXCR5+PD-1+) or naive B cells CD3-CD19+CD27loIgDhi. For each co-culture condition, 4.5x104 sorted T cells were combined with autologous 5x104 naive B cells. Co-cultures were either left unstimulated, stimulated with Staphyloccocal Enterotoxin B (SEB, Toxin Technologies, 0.5 μg/mL), SEB (0.5 μg/mL) with human TNF cytokine (Biolegend, 125 ng/mL), or SEB (0.5 μg/mL) with α-TNF antibodies (BD Biosciences, clone MAb11, 2 μg/mL). At the conclusion of the co-culture, supernatant was aspirated and saved. The remaining cells underwent anti-human Fc-blockade (Biolegend, TruStain FcX) for 10 minutes at room temperature, followed by routine staining for flow cytometry. The inhibitor of TNF processing TAPI-0 (Enzo, BML-PI133-0001) was used in Figure S3D. Cultures with B cells only were stimulated using soluble αIgM (10 ug/mL, clone DA4-4, Fisher) with soluble trimeric MegaCD40L protein (100 ng/mL, Enzo Life Sciences). For Figure S3E, samples were taken from previously-reported co-cultures involving CD4+CXCR5+PD-1+ cTfh from young or elderly adults combined with common allogenic naive B cells from a young adult donor in a 1:1 ratio for one week with SEB stimulation.

Multiplex bead assays

Multiplex bead assays for plasma samples were performed in duplicate for TNF, CXCL11, and MIP1β from the MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel (Millipore Sigma). Co-culture supernatants were assayed in duplicate for IgG1 production using the Legendplex Human Immunoglobulin Isotyping Panel (Biolegend) and data acquired on a BD Symphony A5 cytometer, in accordance with manufacturer instructions.

Transcriptomic analyses

PBMC were sorted on a BD Aria II cell sorter, followed by total RNA extraction by RNeasy Micro Plus kit (QIAGEN) and polyA amplification with SMARTer Ultra-Low Input RNA kit v3 (Clontech) according to manufacturer instructions. Libraries were prepared using the Illumina Nextera XT Library Preparation kit and sequenced on an Illumina HiSeq 2000 using 100bp single-read format. FASTQ files were trimmed with Trimmomatic (version 0.32), aligned using STAR (version 2.5.2a) and normalized by PORT (https://github.com/itmat/Normalization/, version 0.8.5) against the GRCh38 reference assembly of the human genome. Variance-stabilizing transformation and differential expression were performed using DESeq2 (version 1.20.0) based on genes with at least 20 counts in at least 25% of all libraries, using the R environment (version 3.5.0). Of the 84 resultant libraries, one sample was excluded from further analyses (Figures S1H and S1I). Plots were made by ggplot2 (version 2.2.1) and heatmaps using the “inferno” color-scheme from the “viridis” library. For differential expression analysis, genes were included if they had at least 20 counts in at least 25% of the samples. Samples from the same person (e.g., inter-subset comparisons or day 0-to-day 7 comparisons) were treated as paired in statistical models for DESeq2. t-SNE maps were produced using Rtsne (version 0.13) from the transformed counts data. Gene set enrichment analyses were performed with at least 10000 permutations of pre-ranked GSEA (https://www.gsea-msigdb.org/gsea/downloads.jsp). Gene ontology was performed in Cytoscape using ClueGo, or Metascape (https://metascape.org/gp/index.html#/main/step1). Ingenuity Pathway Analysis (QIAGEN) was performed by uploading tables of the differential expression results from DESeq2 analyses for log2 fold-change, p values, and p-adjusted values, followed by Expression Analysis core analysis for all genes with p value < 0.10. Upstream Regulator Analysis was used to calculate Z-scores in Figure 2G., A potential regulator (drug, transcription factor, miRNA, etc) is associated with “activation” or “inhibition” effects on a series of downstream genes. These associations are together summarized as a Z-score, in comparison to a model that assigns random regulation direction.

Network analysis

Weighted gene correlation network analysis (WGCNA, version 1.63) was performed on variance-stabilizing transformation data, as calculated by DESeq2 after filtering out genes with less than 5 counts in fewer than 12% of all libraries. Adjacency matrix was calculated using β = 20. Minimum module size was 100. Gene ontology was analyzed using Metascape for all genes with a module membership of at least 0.8. Pre-ranked GSEA was performed using all genes ranked by membership for each module. Transcription factors were identified from a curated list and displayed with functional association data overlaid from GeneMANIA.

Aging signature

Transcriptional profiling data for whole blood microarray studies was obtained from the Gene Expression Omnibus (GEO accession numbers 123697, 123696, and 123698), from Immport (SDY622, SDY648, SDY739, and SDY819), and from the European Genome-phenome Archive (EGAS00001002460). For each dataset, participants were stratified into cohorts for young adults (< 40 years old) or elderly adults (> 65 years old). Differential expression was calculated using GEO2R or by limma-voom. An aging signature of 354 genes was constructed from genes downregulated in the elderly cohort from SDY739 with p.adj < 0.20 using limma-voom. For analyses of the aging signature in CD4 subsets, pre-ranked GSEA analyses were performed with at least 25000 permutations.

Influenza-specific antibodies

The two influenza A vaccine strains of the 2014/2015 seasonal influenza vaccine, A/California/7/2009 (H1N1) pdm09-like virus and A/Victoria/361/2011 (H3N2)-like virus, were obtained from the Centers for Disease Control and Prevention (Atlanta, GA). Assays to detect hemagglutinin inhibition assay (HAI) titers were performed. Infectious virus was used for neutralizing Ab assays or inactivated by β-propionolactone for H1N1/California- and H3N2/Victoria-specific binding Ab ELISA assays. Nunc Maxisorp plates (Nunc) were coated with 10 mg/mL influenza A/H1N1/California and A/H3N2/Victoria virus along with isotype standards for IgA1, IgG, and IgM (Athens Research and Technology) in bicarbonate buffer overnight at 4°C. Plates were blocked with 3% BSA in PBS and incubated with heat-inactivated sera of young and elderly adults. Abs were detected using alkaline phosphatase-conjugated mouse anti-human IgA1, IgG, and IgM (Southern Biotechnology).

Quantification and statistical analysis

Statistical analyses were performed with Prism 7 (GraphPad) and R. Data was compared using Student’s t test, paired t test, one-way Analysis of Variance (ANOVA) with Tukey post hoc analysis, non-parametric ANOVA with Friedman’s post-test, or Fisher’s Exact test, as indicated. All t tests were performed as two-tailed tests at a 0.05 significance level. Outlier analysis for Figure 1E was performed using Grubb’s test at α = 0.05 and the plot with all points included is shown in Figure S1G. Study schematic in Figure 2A was diagramed using BioRender.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-human CD3 (OKT3)	Biolegend	Cat # 317301; RRID:AB_571926
Anti-human CD4 (S3.5)	Thermo-Fisher	Cat # MHCD0418; RRID:AB_10376013
Anti-human CD8 (RPA-T8)	Biolegend	Cat # 301002; RRID:AB_314120
Anti-human CD14 (M5E2)	Biolegend	Cat # 301819; RRID:AB_493694
Anti-human CD16 (3G8)	Biolegend	Cat # 302017; RRID:AB_314217
Anti-human CD19 (HIB19)	Biolegend	Cat # 302217; RRID:AB_314247
Anti-human CXCR5 (RF8B2)	BD Biosciences	Cat # 563105; RRID:AB_2738008
Anti-human PD-1 (EH12.2)	BD Biosciences	Cat # 562516; RRID:AB_11153482
Anti-human ICOS (C398.4a)	Novus	Cat # NBP1-43065; RRID:AB_10006379
Anti-human CD38 (HIT2)	BD Biosciences	Cat # 563964; RRID:AB_2738515
Anti-human CD27 (O323)	Biolegend	Cat # 356429; RRID:AB_2650750
Anti-human CD45RA (HI100)	Biolegend	Cat # 304133; RRID:AB_11126164
Anti-human NFkB p50 (4D1)	Biolegend	Cat # 616701; RRID:AB_315855
Anti-human Survivin (REA459)	Miltenyi	Cat # 130-106-741; RRID:AB_2653606
Anti-human TNF (Mab11)	BD Biosciences	Cat # 554511; RRID:AB_395442
Anti-human IgM (DA4-4)	Fisher	Cat # NBP234252A; RRID:AB_2890064
Anti-human TNFR1 (W15099A)	Biolegend	Cat # 369908; RRID:AB_2650932
Chemicals, peptides, and recombinant proteins
Ficoll-Paque PLUS	GE Healthcare	Cat # 17-1440-03
Foxp3 Transcription Factor Staining Buffer Kit	Fisher	Cat # A25866A
DiOC6(3) (3,3′-Dihexyloxacarbocyanine Iodide)	Fisher	Cat # D273; CAS 53213-82-4
Staphyloccocal Enterotoxin B	Toxin Technology	Cat # BT202; CAS 11100-45-1
Recombinant human TNF	Biolegend	Cat # 570102
Anti-human TruStain Fc-X	Biolegend	Cat # 422301; RRID:AB_2818986
MegaCD40L	Enzo	Cat # ALX-522-110-C010
TAPI-0 inhibitor	Enzo	Cat # BML-PI133-0001
Qdot 565 ITK Amino (PEG) Quantum Dots	Fisher	Cat # Q21531MP
Qdot 585 ITK Amino (PEG) Quantum Dots	Fisher	Cat # Q21511MP
BD Horizon BB515 Streptavidin	BD Biosciences	Cat # 564453; RRID:AB_2869580
Critical commercial assays
MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel	Millipore	Cat # HCYTOMAG-60K
Legendplex Human Immunoglobulin Isotyping Panel	Biolegend	Cat # 740639
SMARTer Ultra-Low Input RNA kit v3	Clontech	Cat # 634850
Nextera XT Library Preparation kit	Illumina	Cat # FC-131-1096
Deposited data
RNaseq for CD4 subsets from young and elderly adults before and after influenza vaccination	This paper	GEO: GSE134416
Alpert A et al., Longitudinal tracking of whole blood gene-expression in healthy young and old individuals	Gene Expression Omnibus	GEO: GSE123696
Alpert A et al., Longitudinal tracking of whole blood gene-expression in healthy young and old individuals	Gene Expression Omnibus	GEO: GSE123697
Alpert A et al., Longitudinal tracking of whole blood gene-expression in healthy young and old individuals	Gene Expression Omnibus	GEO: GSE123698
Humoral responses to Influenza vaccination in aged populations	Immport	SDY622
Humoral responses to Influenza vaccination in aged populations	Immport	SDY648
Humoral responses to Influenza vaccination in aged populations	Immport	SDY739
Humoral responses to Influenza vaccination in aged populations	Immport	SDY819
Milieu Interieur Consortium study	European Genome-phenome Archive	EGAS00001002460
Software and algorithms
Trimmomatic	Ref76	Version 0.32
STAR	Ref77	Version 2.5.2a
PORT normalization	N/A	Version 0.85; http://bioinf.itmat.upenn.edu/benchmarking/rnaseq/port/index.php
R environment	R Foundation	Version 4.0.0
DESeq2 library	Ref78	Version 1.20.0
Limma library	Ref79	Version 3.44.3
WGCNA library	Ref51	Version 1.63; https://horvath.genetics.ucla.edu/html/CoexpressionNetwork/Rpackages/WGCNA/
Cytoscape	Ref80	https://cytoscape.org/
Custom bioinformatics scripts	N/A	https://github.com/teamTfh/cTfh_AgingSignature
Statistical analysis
Prism 6	Graphpad software	N/A
R environment	R Foundation	version 4.0.0