Research ArticleINFECTIOUS DISEASE

Successive annual influenza vaccination induces a recurrent oligoclonotypic memory response in circulating T follicular helper cells

See allHide authors and affiliations

Science Immunology  17 Feb 2017:
Vol. 2, Issue 8, eaag2152
DOI: 10.1126/sciimmunol.aag2152

Remembering TFH cell help

Immunological memory operates on the assumption that if you’re exposed to an infection once, you’re more likely to be exposed again. Chance favors the prepared, and the memory response is bigger, faster, and stronger. Now, Herati et al. examine immunological memory in humans who have received successive annual influenza vaccinations. They find that circulating T follicular helper cells, which provide B cell help, not only respond to influenza vaccination but also form long-lasting memory. These cells may serve as markers for successful vaccination as well as targets for new vaccines.

Abstract

T follicular helper (TFH) CD4 cells are crucial providers of B cell help during adaptive immune responses. A circulating population of CD4 T cells, termed cTFH, have similarity to lymphoid TFH, can provide B cell help, and responded to influenza vaccination. However, it is unclear whether human vaccination-induced cTFH respond in an antigen-specific manner and whether they form long-lasting memory. We identified a cTFH population that expressed multiple T cell activation markers and could be readily identified by coexpression of inducible costimulator (ICOS) and CD38. This subset expressed more Bcl6, c-Maf, and interleukin-21 than did other blood CD4 subsets. Influenza vaccination induced a strong response in the ICOS+CD38+ cTFH at day 7, and this population included hemagglutinin-specific cells by tetramer staining and antigen-stimulated activation-induced marker expression. Moreover, T cell receptor β chain sequencing identified a clonal response in ICOS+CD38+ cTFH that strongly correlated with the increased cTFH frequency and was associated with the circulating plasmablast response. In participants who received successive annual vaccinations, a recurrent oligoclonal response was identified in the ICOS+CD38+ cTFH subset at 7 days after every vaccination. These oligoclonal responses in ICOS+CD38+ cTFH after vaccination persisted in the ICOSCD38 cTFH repertoire in subsequent years, suggesting clonal maintenance in a memory reservoir in the more stable ICOSCD38 cTFH subset. These data highlight the antigen specificity, lineage relationships, and memory properties of human cTFH responses to vaccination, providing new avenues for tracking and monitoring cTFH responses during infection and vaccination in humans.

INTRODUCTION

The development of class-switched, affinity-matured antibody production by B cells is dependent on help from T follicular helper cells (TFH) in the germinal center (GC) (1). A circulating subset of CD4 T cells, termed circulating TFH (cTFH), express CXCR5 and programmed death-1 (PD-1) and share phenotypic, functional, and transcriptional properties with lymphoid TFH (24). Vaccine-induced changes in cTFH have been linked to changes in the influenza-specific antibody production. For example, increased expression of inducible costimulator (ICOS; CD278) by cTFH correlated with vaccine-induced antibodies after influenza vaccination (5, 6). Because of the central role of TFH in antibody development, rational vaccine strategies will benefit from a better understanding of the properties of cTFH during an immune response.

Despite elegant work on the biology of TFH in animal models and humans, our understanding of how antigen-specific TFH are induced and maintained in humans is incomplete. Antigen-specific TFH are generated after an immune stimulus in mice (7) and memory TFH have been identified (8, 9). In humans, tetanus-specific cTFH have been identified at baseline in healthy adults (3), and antigen-specific TFH can be detected after challenge with influenza vaccine (6) and HIV vaccine (10). In humans, protective antibody responses after influenza vaccination target hemagglutinin (HA) (11). However, although HA-specific CD4 T cell responses to vaccination are observed (12), more CD4 T cells respond to the internally conserved influenza proteins than to HA (13, 14). This biased specificity may not be true for TFH, because TFH may preferentially respond to HA rather than nucleoprotein (NP) (15), but the dynamics of antigen-specific TFH responses in humans after vaccination remain poorly understood.

Unfortunately, typical methods to study antigen specificity of CD4 T cells have drawbacks when applied to TFH. Major histocompatibility complex class II tetramer studies require reagents fitting a cohort of human leukocyte antigen (HLA)–matched participants and profile only a fraction of the overall response. Strategies to sort cells based on cytokine production after peptide stimulation have limitations for these analyses, because intrinsic affinity of a T cell receptor (TCR) for its cognate antigen or incomplete cytokine production by the population of interest only reveals a subpopulation of the cells of interest (16). Proliferation as a marker of antigen specificity (17, 18) may also lead to a suboptimal picture because TFH proliferate less than T helper cell 1 (TH1)–type CD4 T cells (19). Next-generation methods such as T cell receptor sequencing (TCRseq) (17, 20, 21) permit broad, detailed analysis of the TCR repertoire, but TCRseq has not previously been used to probe TFH repertoire dynamics.

In this study, we identified a subset of cTFH coexpressing ICOS and CD38 that expressed B cell lymphoma 6 (Bcl6), increased in frequency and clonality after influenza vaccination, and contained influenza-specific cells. Repeated vaccination of the same participants recalled highly overlapping clones in successive years, indicating robust cTFH memory to influenza vaccination. Some clones identified in the expanded ICOS+CD38+ cTFH subset on day 7 were later found in the ICOSCD38 cTFH population, suggesting that ICOSCD38 cTFH form a pool of memory cTFH that can be repeatedly recalled upon subsequent exposures to antigen. These data have implications for repeated influenza vaccination. These results also provide a tractable way to monitor antigen-induced cTFH responses and to identify a pool of memory cTFH that can be repeatedly recalled upon subsequent antigen exposure, providing a platform with which to further interrogate and optimize rational vaccine strategies.

RESULTS

Highly activated phenotype of ICOS+CD38+ cTFH at baseline

We first reasoned that recently activated TFH in the circulation might be identified by the expression of proteins associated with cellular activation. We focused on cTFH, defined as non-naive CD4+CXCR5+PD-1+, that expressed high levels of the costimulatory marker ICOS, which is expressed by lymphoid TFH (1); contributed to in vitro B cell help (3); and were induced in cTFH after influenza vaccination (5, 6). Of circulating cTFH at baseline for participants in cohort 1 (table S1; n = 28), ~5% expressed ICOS and also CD38 (hereafter referred to as ICOS+CD38+ cTFH and compared with ICOSCD38 cTFH). These ICOS+CD38+ cTFH expressed multiple other markers of cellular activation. Directly ex vivo, ICOS+CD38+ cTFH had higher expression of CD25, CD27, CD28, CTLA4, PD-1, Helios, and Ki67 and lower expression of CD127 than ICOSCD38 cTFH (Fig. 1, A to C, and fig. S1A). Higher expression of CD3 and CD4 was also observed in ICOS+CD38+ cTFH compared with ICOSCD38 cTFH, but CXCR5 expression was similar (fig. S1B). The pattern of markers of activation expressed by cTFH in freshly isolated peripheral blood mononuclear cells (PBMCs) (Fig. 1, A to C) was generally similar in cryopreserved PBMCs (fig. S1C). High expression of CD25, CD27, CD39, CTLA4, and Helios was observed even in the absence of Foxp3 (fig. S1D). In sum, ICOS+CD38+ cTFH were a small fraction of all cTFH but expressed many proteins associated with T cell activation.

Fig. 1 Circulating CXCR5+PD-1+ICOS+CD38+ cTFH express many markers of activation.

(A) Circulating TFH were identified by coexpression of CXCR5 and PD-1 in freshly isolated CD3+CD4+ PBMCs. (B) Representative histograms are shown for direct staining for the indicated proteins, with mean fluorescence intensity for each subset shown on the histograms. (C) Summary plots are shown for CD25 (P = 5.1 × 10−9, paired t test; n = 26), CD27 (P = 3.1 × 10−9, paired t test; n = 27), CD28 (P < 10−16, paired t test; n = 28), CD127 (P < 10−16, paired t test; n = 26), CTLA4 (P = 1.5 × 10−15, paired t test; n = 26), PD-1 (P < 10−16, paired t test; n = 26), Helios (P = 1.7 × 10−13, paired t test; n = 26), and Ki67 (P = 6.0 × 10−13, paired t test; n = 26).

Activated cTFH express Bcl6 and can produce interleukin-10, interleukin-17, and interleukin-21

Bcl6 has a central role in the TFH program (1), but protein expression was not detected in total peripheral blood cTFH in previous studies (2, 3, 5). To address whether ICOS+CD38+ cTFH may resemble lymphoid TFH more than other cTFH, we evaluated expression of Bcl6 protein in the ICOS+CD38+ subset, as well as the expression of other key TH transcription factors and cytokine production. In peripheral blood, Bcl6 protein and mRNA expression was highest in ICOS+CD38+ cTFH compared with ICOSCD38 cTFH or CXCR5 memory CD4 T cells but was lower in splenic GC TFH (Fig. 2A and fig. S2, A to D), suggesting that the ICOS+CD38+ cTFH subset bears closer resemblance to lymphoid TFH than to ICOSCD38 cTFH. Moreover, expression of Bcl6 was highest in the ICOS+CD38+ cTFH with the highest PD-1 expression (Fig. 2B).

Fig. 2 Activated cTFH express Bcl6 and Tbet and can produce IL-10, IL-17, and IL-21.

Transcription factor and cytokine staining was performed on PBMCs after isolation and permeabilization. Not all statistically significant differences are shown. (A) Bcl6 protein expression for ICOS+CD38+ and ICOSCD38 cTFH compared with human splenic CD4 subsets. Summary plots are shown for six participants for PBMCs and four participants for splenocytes, with lines connecting the same participant. Data are representative of three independent experiments [ICOS+CD38+ cTFH versus ICOSCD38 cTFH, P = 2.6 × 10−3; ICOS+CD38+ cTFH versus CXCR5 memory, P = 1.6 × 10−3; ICOS+CD38+ cTFH versus Naïve CD4, P = 3.6 × 10−4; ICOS+CD38+ cTFH versus GC-TFH, P < 10−7; ICOS+CD38+ cTFH, n = 6; ICOSCD38 cTFH, n = 6; CXCR5 memory, n = 6; Naïve CD4, n = 6; GC-TFH, n = 4; Non–GC-TFH, n = 4; CXCR5 memory (spleen), n = 4; naïve CD4 (spleen), n = 4; one-way ANOVA with Tukey’s test]. (B) Bcl6 protein expression for ICOS+CD38+ cTFH that were further divided into PD-1+ and PD-1++ (P = 4.6 × 10−3, paired t test; n = 6). (C) PBMCs were permeabilized and stained for transcription factors (Foxp3: CXCR5 memory versus ICOSCD38 cTFH, P < 10−2, n = 4. c-Maf: CXCR5 memory versus ICOSCD38 cTFH, P < 10−3, n = 4; CXCR5 memory versus ICOS+CD38+ cTFH, P < 10−3, n = 4; ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−2, n = 4. Foxo1: CXCR5 memory versus ICOS+CD38+ cTFH, P < 10−3, n = 4; ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−3, n = 4. GATA3: CXCR5 memory versus ICOS+CD38+ cTFH, P < 10−3, n = 4; ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−3, n = 4; repeated-measures one-way ANOVA). FSC, forward scatter. (D) SPICE analysis of the transcription factors shown for ICOS+CD38+ cTFH and ICOSCD38 cTFH. (E) PBMCs were stimulated with PMA/ionomycin (PMAi) for 5 hours at 37°C, with monensin added for the final 4 hours. Cells were fixed before permeabilization and intracellular staining. Summary plots and examples are shown for each cytokine (IL-10: ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−3, n = 6; IL-17: ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−3, n = 6; IL-21: ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−3, n = 6; IL-2: ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−3, n = 6; IFN-γ: ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−3, n = 6; TNF: ICOSCD38 cTFH versus ICOS+CD38+ cTFH, P < 10−2, n = 5; repeated-measures one-way ANOVA). (F) Polyfunctionality analysis was performed using the cytokines. For all panels, *P < 0.05 and **P < 0.01.

Previous studies have suggested that TH1-, TH2-, and TH17-polarized subsets exist within the cTFH pool (2), but other studies have found low expression of master transcriptional regulators such as Tbet or RORγT in TFH (22, 23). To address whether cTFH expressed these regulators, we profiled protein expression of these transcription factors in cTFH. Most of the activated cTFH lacked detectable expression of any of these lineage-associated transcription factors (Fig. 2C and fig. S2E). Of these, Tbet was most commonly expressed in ICOS+CD38+ cTFH. T follicular regulatory cells (TFR) were identified by expression of Foxp3 and CD25 and represented about 5% of the ICOS+CD38+ cTFH, consistent with the frequency of circulating TFR from mouse studies (24). The transcription factor c-Maf has a critical role in TFH biology (2527) and displayed high expression in ICOS+CD38+ cTFH, as did total Foxo1. GATA3 is critical for development, maintenance, and function of lymphocytes (28) and was higher in ICOS+CD38+ cTFH compared with ICOSCD38 cTFH (Fig. 2C). Coexpression of Tbet and RORγT or GATA3 was observed in a minority of ICOS+CD38+ but not ICOSCD38 cTFH (Fig. 2D and fig. S2F). Thus, the ICOS+CD38+ cTFH subset expressed Bcl6 and c-Maf protein, but few cells expressed Tbet, RORγT, or Foxp3.

To investigate the cytokine profile of cTFH, we next performed ex vivo stimulation and intracellular cytokine staining for cytokines associated with T helper subsets and important for GC reactions (1), including interleukin-17 (IL-17) (29) and IL-10 (30). Upon stimulation with phorbol 12-myristate 13-acetate (PMA)/ionomycin, ICOS+CD38+ cTFH produced a unique combination of IL-10, IL-17, and IL-21 compared with other CD4 subsets (Fig. 2E and fig. S2, G to I), a pattern also observed after staphylococcal enterotoxin B stimulation (fig. S2J). ICOS+CD38+ cTFH were the strongest producers of IL-21 compared with other subsets. The frequency of IL-21–producing cells in the ICOS+CD38+ cTFH subset was twofold higher than that in the ICOSCD38 cTFH and threefold higher than that in CXCR5 memory CD4. Overall, about one-third of ICOS+CD38+ cTFH expressed none of the assayed cytokines, about one-third expressed tumor necrosis factor (TNF) alone, and the remainder coexpressed multiple cytokines, primarily IL-10 and interferon-γ (IFN-γ) (Fig. 2F and fig. S2H). Most IL-17+ ICOS+CD38+ cTFH also produced IFN-γ and TNF. There was greater polyfunctionality in the ICOS+CD38+ cTFH than in the ICOSCD38 cTFH, consistent with their activation state. Although most ICOS+CD38+ cTFH were capable of production of TNF, this subset produced more IL-21, supporting the hypothesis that ICOS+CD38+ cTFH are more similar to lymphoid TFH than ICOSCD38 cTFH.

Influenza-specific cells can be identified within ICOS+CD38+ cTFH subset after influenza vaccination

We and others previously observed increased expression of ICOS among cTFH after influenza vaccination (5, 6). We next wanted to investigate whether protein expression of other markers of activation changed after immunization. Participants in cohort 1 (table S1) received inactivated influenza vaccine during the fall of 2014 (n = 28). In an influenza challenge study, a subset of CD4 coexpressing CD38 and Ki67 was observed 1 week after infection (13). In our cohort, 7 days after vaccination, we observed increased CD38 expression among non-naive Ki67+ CD4 T cells (Fig. 3A and fig. S3A). Before vaccination, cTFH represented ~25% of the Ki67+CD38+ CD4 T cells. After vaccination, the proportion of cTFH in this Ki67+CD38+ CD4 T cell compartment increased between 4 and 60% (Fig. 3, A and B). Instead, by gating directly on non-naive CXCR5+PD-1+ CD4, we observed a vaccine-induced increase of 20 to 60% in this subset of ICOS+CD38+ cTFH (Fig. 3C and fig. S3B). Expression of PD-1 also increased in the ICOS+CD38+ cTFH after vaccination (fig. S3C), but no changes were detected with respect to expression of CD27, CD28, CD127, CTLA4, Helios, or Ki67 (fig. S3D). Thus, influenza vaccination was associated with increased frequency of the ICOS+CD38+ cTFH population.

Fig. 3 ICOS and CD38 identify a cTFH subset that is induced by vaccination.

Adults were vaccinated with seasonal inactivated influenza vaccine and PBMCs isolated at days 0 and 7. (A) Expression of CD38 in non-naive Ki67+ CD4 at day 7 after vaccination (P = 2.9 × 10−4, paired t test; n = 17). (B) Circulating frequency of CXCR5+PD-1+ cells as a proportion of the non-naive Ki67+CD38+ CD4 population from PBMCs (P = 0.046, paired t test; n = 17). Example flow plots are shown for one participant. (C) Expression of ICOS and CD38 for the cTFH subset at day 7 after vaccination for one participant (left) and for the full cohort (right; P = 0.019, paired t test; n = 28). (D) Major histocompatibility complex class II tetramers were loaded with HA306–318 or HA398–410 peptides. At day 7 after vaccination, PBMCs from two HLA-DRB*0401 participants were stained for surface proteins and tetramer, followed by magnetic enrichment. (E) PBMCs were stimulated for 18 hours with overlapping peptide pools for influenza proteins HA 1, HA 3, NP, and M, followed by cell staining and acquisition. Example plots are shown for expression of CD69 and CD200 for different CD4 subsets. (F) Summary plots for CD4 subsets for AIMs after stimulation (ICOS+CD38+ cTFH, P = 0.084; ICOSCD38 cTFH, P = not significant; CXCR5 memory, P = not significant; paired t test; n = 5). For all panels, *P < 0.05 and **P < 0.01.

We next hypothesized that the day 7 ICOS+CD38+ cTFH population would be enriched for influenza-specific cells. Previous studies have identified CD4 T cells specific for NP, matrix 1 (M), and HA peptides after vaccination (13, 14, 31). However, these methods typically rely on cytokine production after stimulation. We previously observed production of TNF, IFN-γ, or IL-2 by only a minority of CD4 after stimulation with overlapping peptide pools for influenza proteins (31), and IL-21 was produced by a subset of cTFH after PMA/ionomycin stimulation (Fig. 2E). Here, an independent approach using class II tetramers was used (32). A second cohort of participants was recruited (cohort 2, n = 12; table S1), and a similar ICOS+CD38+ cTFH response was observed 1 week after vaccination (fig. S2E). We examined HLA class II/peptide tetramer+ CD4 T cells at 1 week after vaccination in two HLA-DRB1*04:01 participants. CD4 T cells specific for either HA residues 306 to 318 (HA306–318) or HA residues 398 to 410 (HA398–410) (12, 32, 33) were identified in these participants, including HA-specific cTFH (Fig. 3D). Of these HA tetramer+ cTFH, 50 to 91% coexpressed ICOS and CD38. After influenza vaccination, most of these tetramer+ cTFH also expressed CXCR3 (fig. S3F), consistent with previous reports (6). Thus, the identification of influenza HA–specific cells in the cTFH subset 1 week after inactivated influenza vaccine suggests that antigen-specific cTFH were induced by vaccination and could be found in the ICOS+CD38+ cTFH subset.

Although tetramer staining demonstrated that influenza-specific CD4 T cells are contained in the ICOS+CD38+ cTFH population after vaccination, a tetramer-based approach is likely to underestimate the total influenza vaccine–induced population because few epitopes were interrogated. Expression of activation-induced markers (AIMs) has been used to identify antigen-specific CD4 T cells after antigen stimulation (34, 35). This approach has the advantage of not being restricted to a single epitope specificity and offers a broader view of antigen-specific cTFH populations. Thus, to complement the tetramer approach, PBMCs before and after vaccination were stimulated with overlapping influenza peptide pools for NP, M, H1, and H3 for 18 hours. After stimulation of PBMCs from day 7 after vaccination, we observed a clear population of antigen-specific ICOS+CD38+ cTFH identified by up-regulation of CD69 and CD200 (Fig. 3, E and F). This population was not present in unstimulated controls (fig. S3G) and was 2- to 29-fold higher at day 7 after vaccination compared with day 0 in the ICOS+CD38+ population (Fig. 3, E and F). Together, these data demonstrate the presence of robust influenza-specific responses within the ICOS+CD38+ cTFH population at day 7 after influenza vaccination.

Increased clonality of ICOS+CD38+ cTFH after influenza vaccination

Compared with other CD4 T cell populations, the ICOS+CD38+ cTFH subset was enriched for influenza-specific cells based on tetramer and response to influenza antigen stimulation in vitro (AIM analysis). To examine the clonotypes in the vaccine response more broadly, we used TCRseq from multiplex polymerase chain reaction (PCR) amplification of genomic DNA (20) on sorted cTFH populations. In most participants, clonality of the ICOS+CD38+ cTFH population increased after influenza vaccination (Fig. 4, A and B). The ICOSCD38 cTFH had reduced clonality after vaccination, but this did not correlate with the increase in clonality in ICOS+CD38+ cTFH (fig. S4, A and B). The day 7 response in ICOS+CD38+ cTFH was typified by the small but clear increase of many clonotypes, rather than dominance of a single expanded clonotype. Another measure of clonal dominance, the Gini index, correlated strongly with clonality and corroborated these results (fig. S4C). Positive correlations were observed between the change in frequency of the ICOS+CD38+ cTFH population and clonality, when examining both day 7 alone and the fold change at day 7 versus day 0 (Fig. 4C and fig. S4, D and E). We did not detect a clear relationship with the hemagglutination inhibition antibody titer and ICOS+CD38+ cTFH clonality (fig. S4, F and G), perhaps reflecting the fact that hemagglutination inhibition antibody is only a small fraction of the total antibody induced by vaccination. However, the change in clonality of ICOS+CD38+ cTFH demonstrated a trend toward correlation with the change in frequency of circulating plasmablasts when assessed in a subset of participants (Fig. 4D), suggesting a relationship between this cTFH subpopulation and the humoral immune response. These data demonstrate a rapid, oligoclonal ICOS+CD38+ cTFH response to vaccination that strongly correlated with the magnitude of the response. Influenza vaccination resulted in antigen-specific clonal expansion of cTFH populations in the PBMCs.

Fig. 4 The ICOS+CD38+ cTFH subset has increased clonality after vaccination.

(A) All in-frame productive clonotypes from ICOS+CD38+ cTFH were plotted before and after vaccination for one participant. Each symbol indicates a unique clonotype. Clonotypes are randomly distributed across the plot, with the size and color of the symbol representing the clonal frequency. (B) Clonality score calculations were performed for in-frame productive clonotypes at both time points for the 2015–2016 vaccination year for ICOS+CD38+ cTFH (P = 0.11, paired t test; n = 11). (C) Day 7 clonality was plotted against the day 7 ICOS+CD38+ cTFH frequency as a percent of all cTFH (left, r = 0.21, P = 0.54, n = 11, Pearson correlation). The fold change in clonality was also plotted against the fold change in ICOS+CD38+ cTFH circulating frequency (right, r = 0.68, P = 0.021, n = 11, Pearson correlation). (D) PBMCs were assayed by flow cytometry at days 0 and 7 after influenza vaccination. Plasmablasts were identified as CD138+CD20 cells that were also CD19+CD27+CD38+. Correlation for the fold change between days 7 and 0 of the plasmablast frequency against the fold change between days 7 and 0 in ICOS+CD38+ cTFH clonality score is shown (r = 0.73, P = 0.099, n = 6, Pearson correlation).

Repeated influenza vaccinations elicit a recurring oligoclonal response

Nearly all humans become seropositive for influenza virus–specific antibodies by the end of childhood (36). Thus, we assumed that, before entry into the study, all adult participants had either been immunized against or been infected with influenza A virus at least once in the past. However, it was unknown whether a recall response with successive vaccinations would elicit the same oligoclonal response, given the passage of time, the annual reformulation of the influenza vaccine, and the possibility of influenza infection before revaccination. For six of the participants in cohort 2, cryopreserved PBMCs were available after influenza vaccination over multiple years. Of these, four had PBMCs available for the 2014–2015 and 2015–2016 vaccination years, and two had PBMCs available for the 2013–2014, 2014–2015, and 2015–2016 vaccination years. Repeated induction of ICOS+CD38+ cTFH was observed at day 7 after each successive vaccination (Fig. 5A and fig. S5A).

Fig. 5 Repeated clonotypic response in ICOS+CD38+ cTFH identified with successive vaccinations.

The subset of participants for whom cryopreserved samples were available was assessed for year-to-year changes. (A) Schematic to show the nomenclature used. Year 1 indicates the participant’s first entry into the study, and year 2 indicates the participant’s following year participation in the study. Day 0 refers to the prevaccination time point for the particular study year, and day 7 refers to the 1-week post-vaccination time point. Flow plots for cTFH are shown for participant 999 at year 1 day 7, year 2 days 0 and 7, and year 3 days 0 and 7. (B) Clonotypes for each subset were compared between year 1 day 0 and year 2 day 0 to determine the overlap score (ICOS+CD38+ cTFH versus ICOSCD38 cTFH, P = 7.8 × 10−5; ICOS+CD38+ cTFH versus CXCR5 memory, P = 0.03; ICOSCD38 cTFH versus CXCR5 memory, P = 0.014; ICOS+CD38+ cTFH, n = 5; ICOSCD38 cTFH, n = 5; CXCR5 memory, n = 4; one-way ANOVA). (C) For each participant, year 1 day 7 was taken as the reference point, and an overlap score was generated for each subsequent time point for ICOS+CD38+ cTFH. (D) For each subset for participant 999, year 1 day 7 was taken as the reference point, and overlap scores were generated. ICOS+CD38+ cTFH (orange), ICOSCD38 cTFH (green), and CXCR5 memory CD4 (gray) subsets are shown. Size of the symbol indicates the −log10(P value) for the Fisher’s exact test, given a theoretical repertoire of 107 clonotypes. (E) Number of unique complementarity-determining region 3 (CDR3) sequences in the recurrent oligoclonal response for each participant. Squares indicate overlap across two successive vaccinations. Circles indicate overlap across three successive vaccinations. (F) Unique clones are indicated for participant 999. The year 1 day 7 time point is maintained on the y axis in all plots, and the x axis varies on the basis of the time point being analyzed. Values indicate number of clonotypes overlapping between each pair of time points for ICOS+CD38+ cTFH (orange, upper right) or ICOSCD38 cTFH (green, upper left). (G) Clonotypic frequency for individual clonotypes that were present at year 1 day 7, year 2 day 7, and year 3 day 7 was plotted for each subset. Dark line is used to indicate the median value at each time point for the given subset. (H) Cumulative frequency of the recurring oligoclonal response clonotypes within the bulk TCRseq data for the different subsets over time. For all panels, *P < 0.05 and **P < 0.01.

We first investigated the stability of the cTFH repertoire over time, independent of vaccination. The overlap score (Materials and Methods) provides an estimate of similarity in repertoire of a population of T cells across different time points. To assess this, we compared the similarity of the T cell receptor β chain (TCRB) repertoire of each subset with itself 1 year after vaccination (i.e., just before revaccination). The ICOS+CD38+ cTFH subset had a low overlap score over 1 year in the same individual, suggesting high turnover of this subset of cTFH in the absence of vaccination (Fig. 5B). In contrast, the ICOSCD38 cTFH and CXCR5 memory CD4 subsets had more stable TCR repertoires over time. Among these populations, the ICOSCD38 cTFH subset had more overlap across 1 year than the CXCR5 memory subset. These observations indicate more dynamic turnover in the clones contained in the ICOS+CD38+ subset compared with the ICOSCD38 subset of cTFH in the steady state.

We next considered the effect of vaccination on the TCRB repertoire of these CD4 T cell populations. There was essentially no overlap when comparing the ICOS+CD38+ repertoire at year 1 day 7 after vaccination with year 2 day 0, indicating that, within the limit of detection of these assays, the clones present in the ICOS+CD38+ cTFH subset after influenza vaccination were no longer detectable in this subset a year later (Fig. 5C). However, on day 7 after vaccination in year 2, clones from the year 1 post-vaccination ICOS+CD38+ cTFH subset reemerged in the ICOS+CD38+ cTFH subset. Neither the ICOSCD38 cTFH nor the CXCR5 memory CD4 subsets showed such a response (Fig. 5D and fig. S5B). The timing of the vaccination in the previous years did not have any apparent effect on the recall responses in ICOS+CD38+ cTFH (fig. S5B). The median number of clonotypes in the recurrent oligoclonal response for all participants was 66 but varied widely from participant to participant and did not appear dependent on the number of years of observation (Fig. 5E). These data strongly suggested a recurrent oligoclonal response in the ICOS+CD38+ cTFH subset to repeated influenza vaccination in all participants.

To investigate these clonality relationships in more detail, we focused on the two participants for whom 3 years of data were available. There was a robust increase in overlapping clonotypes between year 1 day 7 and subsequent day 7 time points, but we observed only a weak overlap with subsequent day 0 time points in the ICOS+CD38+ cTFH subset (Fig. 5F and fig. S5, B and C). Moreover, clones shared at every day 7 time point in ICOS+CD38+ cTFH were absent at every day 0 time point, whereas clones shared at every day 7 time point in ICOSCD38 cTFH or CXCR5 memory were sometimes present at day 0 time points in these subsets as well (Fig. 5G and fig. S5D), suggesting more stability in the clonality of the latter subsets. Rare clones in the ICOSCD38 cTFH and CXCR5 subsets demonstrated the same pattern as the ICOS+CD38+ cTFH subset and increased in frequency after vaccination, suggesting possible influenza vaccine responses in these subsets. For clones shared at every day 7 time point in ICOS+CD38+ cTFH, there was no quantitative change in the median frequency at each day 7 time point, arguing against successive quantitative boosting with each vaccination (fig. S5E). Last, a cumulative frequency was obtained by adding together all the frequencies for the year 1 day 7 clonotypes in the year 2 day 0 time point. The cumulative frequency of the recurrent oligoclonal response clones ranged from 6 to 15% of the total ICOS+CD38+ cTFH subset (Fig. 5H). Together, these data demonstrate considerable clonal dynamics in the ICOS+CD38+ subset of cTFH, suggesting that this subset contains transient clones responding to recent antigenic stimulation. Moreover, these observations support the idea that influenza vaccination induces a repeated recall of conserved clonotypes in the ICOS+CD38+ cTFH after each yearly vaccination.

Influenza-specific clonotypes are contained within the recurring oligoclonal response

The increase in circulating frequency of ICOS+CD38+ cTFH after influenza vaccination was observed consistently in nearly every participant. We hypothesized that clonotypes identified by tetramer or AIM strategies should thus be enriched in the ICOS+CD38+ cTFH subset at 1 week after vaccination. To test this idea, we sorted HA308–316- or HA398–410-positive CD4 T cells at 6 months after vaccination. At this time point, none of the tetramer+ CD4 T cells were ICOS+CD38+ (fig. S6A). We then sequenced TCRB changes of these cells and compared the TCRB sequences with those previously obtained from the total ICOS+CD38+ cTFH on day 7 after vaccination. Despite only obtaining only a small number of TCRB sequences (n = 10 to 35) due to the low frequency of tetramer+ CD4 T cells, there was overlap of the tetramer-specific clonotypes with the bulk ICOS+CD38+ cTFH subset at the day 7 time point each year for both tetramers for one of the two participants examined (Fig. 6A and fig. S6B). Even when only a single HA398–410 clonotype in participant 999 was identified within the recurring ICOS+CD38+ cTFH oligoclonal response, this finding was highly unlikely by chance alone (P = 2 × 10−4, Fisher’s exact test; HA clonotypes, n = 1; recurring oligoclonal response, n = 131). The tetramer clonotypes together comprised 0.37 to 0.70% of the ICOS+CD38+ cTFH population at day 7 after vaccination each year for participant 999 (Fig. 6B). The inability to detect shared clones between tetramer+ CD4 T cells and the ICOS+CD38+ cTFH subset in participant 108 (fig. S6C) was likely due to the low number of tetramer events that could be captured 6 months after vaccination in this participant, suggesting either involvement of other specificities in the cTFH response or that capturing only 10 to 30 TCRB sequences for a single epitope specificity may be at the limit of detection for this type of analysis.

Fig. 6 Tetramer and AIM clonotypes are found in the ICOS+CD38+ cTFH subset.

TCRseq was performed on cells after staining for tetramer or expression of AIMs after stimulation. (A) Tetramer+ cells were sorted 6 months after influenza vaccination in two HLA-DRB1*04:01 participants. Clonotypic frequency for tetramer (x axis) is plotted against the bulk ICOS+CD38+ cTFH subset (y axis) at various time points. Marginals on the outer edges show the single-axis histogram for individual axes. Number of overlapping clones (upper right) and CDR3 sequences for the overlapping clones (bottom right) are indicated. (B) Cumulative frequency of tetramer clones within the bulk TCRseq data for different subsets. Squares indicate HA306–318 tetramer clonotypes, whereas triangles indicate HA398–410 tetramer clonotypes. (C) PBMCs were stimulated for 18 hours with overlapping peptide pools for influenza proteins, and ICOS+CD38+ cTFH that expressed CD69 and CD200 were sorted for TCRseq. Clonotypic frequency for AIM clonotypes is plotted for the bulk ICOS+CD38+ cTFH subset at various time points. Marginals on the outer edges show the single-axis histogram for individual axes. Number of overlapping clones is given in the plot. (D) Cumulative frequency of AIM clones within the bulk TCRseq data for different subsets. Connected symbols show repeated observations for the same participant over time. (E) Percent overlap in unique CDR3 sequences between the recurring oligoclonal response and the AIM clonotypes for each participant. Size of the symbol indicates −log10(P value) as assessed by Fisher’s exact test, assuming 107 possible clonotypes.

To further explore these questions and test whether the ICOS+CD38+ cTFH population contained influenza-specific cells in a manner that did not depend on HLA class II tetramers, we next used the AIM approach described above. As above, PBMCs were stimulated with pools of influenza peptides, and CD69+CD200+ICOS+CD38+ cTFH were sort-purified on day 7 after influenza vaccination. Using this approach, we identified 33 to 216 clonotypes by TCRseq, depending on the participant (fig. S6D). Many TCRB sequences were identified in the influenza-specific AIM approach that were shared with TCR sequences identified in the total ICOS+CD38+ cTFH subset at day 7 after vaccination of each year (Fig. 6C and fig. S6E). Again, similar to the recurring oligoclonal response, the AIM clonotypes showed similar pattern of increased frequency in the ICOS+CD38+ cTFH subset at each day 7 time point in three of the six participants (Fig. 6D). The overlap between the recurrent oligoclonal response and the AIM clonotypes was statistically significant in four of six participants (Fig. 6E and fig. S6F). AIM clones and recurring oligoclonal response clones were not prominently found in other subsets, suggesting strong enrichment for influenza-specific clones in the cTFH compartment (fig. S6, G and H), though because the non-cTFH compartments are substantially larger, undersampling cannot be excluded. Nevertheless, these data confirm the presence of influenza-specific clonotypes within the ICOS+CD38+ cTFH subset 1 week after each influenza vaccination, as well as within the yearly recurring oligoclonal response.

ICOSCD38 cTFH may serve as a long-term reservoir of cTFH memory

The TCRB sequence has been used as a “fingerprint” to track clones through different compartments in other studies (3739), although a particular clone could sometimes be found in multiple distinct CD4 subsets (40). Here, to track clones across subsets over time, clones present for each subset at year 1 day 7 were compared against all the other subsets at year 2 (Fig. 7, A to C, and fig. S7A). A cumulative frequency was obtained by adding together all the frequencies for the year 1 day 7 clonotypes in the year 2 day 0 time point. As expected, clones that were originally present in the CXCR5 memory or ICOSCD38 cTFH subsets in year 1 day 7 could be found in the CXCR5 memory (Fig. 7A) or ICOSCD38 cTFH (Fig. 7B) subsets at year 2 day 0. Clones present in the ICOSCD38 cTFH in year 1 day 7 were most frequently found in the ICOSCD38 cTFH subset at year 2 day 0 (Fig. 7B), consistent with the higher overlap scores for ICOSCD38 cTFH at 1 year in Fig. 5B.

Fig. 7 ICOS+CD38+ cTFH clonotypes are later found in other subsets.

Orange indicates the ICOS+CD38+ cTFH subset, green indicates the ICOSCD38 cTFH subset, and gray indicates the CXCR5 memory subset. (A) Cumulative frequency shown for clonotypes that were present in year 1 day 7 for CXCR5 memory, as later identified in different subsets in year 2 for day 0. (B) Cumulative frequency shown for clonotypes that were present in year 1 day 7 for ICOSCD38 cTFH, as later identified in different subsets in year 2 for day 0 (ICOS+CD38+ cTFH versus ICOSCD38 cTFH, P = 1.1 × 10−4; CXCR5 memory versus ICOSCD38 cTFH, P = 5.9 × 10−5; ICOS+CD38+ cTFH, n = 5; ICOSCD38 cTFH, n = 5, CXCR5 memory, n = 4; one-way ANOVA with Tukey’s post-test). (C) Cumulative frequency for clonotypes that were present in year 1 day 7 for ICOS+CD38+ cTFH, as later identified in different subsets in year 2 for day 0 (ICOS+CD38+ cTFH versus ICOSCD38 cTFH, P = 0.069; CXCR5 memory versus ICOSCD38 cTFH, P = 0.057; ICOS+CD38+ cTFH, n = 5; ICOSCD38 cTFH, n = 5, CXCR5 memory, n = 4; one-way ANOVA with Tukey’s post-test). (D) Clonotypes that were present in year 2 day 7 for participants 101 and 999 were assessed in all three subsets for year 3 day 0. Cumulative clonotypic frequency is shown. (E) Proposed directionality of intersubset changes based on (A) to (C). For all panels, *P < 0.05.

Given the relative stability and greater circulating frequency of the ICOSCD38 cTFH subset compared with ICOS+CD38+ cTFH, we hypothesized that ICOSCD38 cTFH represented a long-term “reservoir” for cTFH from which specific clones may be recalled into the ICOS+CD38+ cTFH pool. To test this idea, we asked whether clones present at year 1 day 7 in the ICOS+CD38+ cTFH could be found in other subsets at year 2. Clones present at year 1 day 7 in ICOS+CD38+ cTFH were most frequently found in the ICOSCD38 cTFH subset at year 2 day 0 (Fig. 7C). Similarly, ICOS+CD38+ cTFH clones present at year 2 day 7 were most frequently found in ICOSCD38 cTFH at year 3 day 0 (Fig. 7D and fig. S7B). As noted earlier, influenza vaccination induced an increase in the frequencies of the year 1 day 7–matched clonotypes in the ICOS+CD38+ cTFH (fig. S7A), but this effect was not seen in CXCR5 memory or ICOSCD38 cTFH. Some clones from year 1 day 7 ICOSCD38 cTFH were present in year 2 day 0 ICOS+CD38+ cTFH, suggesting some low-level conversion of ICOSCD38 cTFH to ICOS+CD38+ cTFH.

Together, these data indicate a strong clonal connection between ICOS+CD38+ and ICOSCD38 cTFH. Clones expanded after influenza vaccination may, therefore, preferentially convert into the ICOSCD38 cTFH subset. These results support a model where cTFH cycle back and forth between activated and inactive states after repeated antigen exposure, with the ICOSCD38 cTFH subset representing the reservoir for memory cTFH.

DISCUSSION

Induction of a strong antibody response is a major correlate of protection in nearly all licensed vaccines. Despite the importance for vaccines, our understanding of the TFH response underlying vaccine-induced antibody responses remains incomplete. In this study, we found that cTFH coexpressing ICOS and CD38 increased in frequency and clonality after influenza vaccination. This subset of cTFH expressed markers of activation and contained influenza virus–specific tetramer+ cells at day 7 after vaccination. These data not only identify a proxy for antigen-specific cTFH, allowing monitoring of human vaccine-induced responses, but also provide insights into the dynamics of cTFH memory and recall responses.

Although cTFH are thought to be derived from lymphoid tissue, differences have been reported between lymphoid TFH and cTFH in protein expression, including lack of transcription factor Bcl6 expression, leading to uncertainty regarding the ontogeny of this cTFH population (2, 3, 4143). We observed the highest expression of Bcl6 and the greatest production of IL-21 in the ICOS+CD38+ cTFH compared with other cTFH in peripheral blood, suggesting a relationship to GC events in lymphoid tissues. It is possible that, given the rapid kinetics of appearance of ICOS+ cTFH after vaccination (6) and the highly activated phenotype of ICOS+CD38+ cTFH that we found here, ICOS+CD38+ cTFH are related to memory follicular mantle TFH-like cells (42), rather than originating from cells that have participated directly in GC interactions. Understanding the ontogeny of cTFH and their relationships to other T helper subsets will clarify when and how cTFH can serve as a biomarker of a vaccine response. Nevertheless, in the absence of ready access to human lymphoid tissues in a routine setting, the circulatory ICOS+CD38+ cTFH population may act as a proxy of events occurring in the lymphoid tissue.

Although memory TFH have been identified in mice (8, 9), the ability to broadly identify memory TFH in humans has been limited. Here, use of TCRB sequencing allowed identification of the memory TFH response to influenza vaccination, based on the repeated induction of an oligoclonal response containing influenza-specific cells in the ICOS+CD38+ cTFH. The clonal composition of this response was repeated with each vaccination, despite the annual vaccine reformulation and possible intercurrent influenza infection. Recall of the same clonotypes could reflect the greater immunodominance of relatively conserved internal influenza proteins in CD4 responses (13, 14) or original antigenic sin (44, 45). Moreover, the stability of the recall response repertoire suggests longevity in the influenza-specific CD4 T cell repertoire. It will be interesting to compare the dynamics, quality, and repertoire diversity of this ICOS+CD38+ subset of antigen-induced cTFH for different vaccination approaches.

Initial evaluation identified many phenotypic differences between ICOS+CD38+ cTFH and ICOSCD38 cTFH. It was unclear whether these subsets were directly related and whether the repertoire dynamics differed between these subsets. We observed notable stability of the repertoire of ICOSCD38 cTFH over 1 year, in contrast to the dynamic changes in the ICOS+CD38+ cTFH. Moreover, the clonotypes from ICOS+CD38+ cTFH and ICOSCD38 cTFH were observed most frequently in the ICOSCD38 cTFH subset at year 2 day 0. Thus, the ICOSCD38 cTFH may function as a long-term reservoir from which clones can be selectively reactivated, leading to reexpression of ICOS and CD38. Curiously, some year 1 day 7 ICOS+CD38+ cTFH clones were still present in the year 2 day 0 ICOS+CD38+ cTFH subset at detectable frequencies. Persistence of clones within the ICOS+CD38+ cTFH subset over many months could reflect ongoing antigenic availability, such as from commensal bacteria, viruses, or other environmental antigens. The ICOS+CD38+ cTFH subset may provide a novel way to interrogate ongoing immune responses in the steady state, with potential application to persisting infections, autoimmunity, and allergy. Overall, these data provide a foundation for tracking cTFH memory and a framework for future studies that should help delineate the relationship of circulating TFH populations to TFH in lymphoid tissues.

It is perhaps remarkable that recurrent oligoclonal responses were detectable in the ICOS+CD38+ subset of cTFH, particularly in the setting of an unadjuvanted seasonal influenza vaccine. For example, assuming that the circulating frequencies of CD4 T cells are similar across the 5 liters of human blood volume, there are ~3 million ICOS+CD38+ cTFH total in circulation. In these studies, we sampled from 40 to 100 ml of blood (or 0.8 to 2.0% of the total) and only recovered productive TCRB sequences from 10 to 20% of the cells analyzed (i.e., sampling only 0.08 to 0.4% of total ICOS+CD38+ cTFH). The likelihood of identifying the same random 50 to 70 clones at two time points with this sampling approach among the 3 million ICOS+CD38+ cTFH is extraordinarily low. Although we cannot sample and interrogate rare clonotypes by these approaches, these data provide strong evidence for the repeated recruitment of common clonotypes into the antigen-specific cTFH response induced by influenza vaccination. Moreover, on the basis of the AIM analysis, we can account for up to 20% of the vaccine-induced ICOS+CD38+ cTFH as antigen-specific [e.g., 11% of ICOS+CD38+ induce CD200 and CD69 (Fig. 3E), and about one-half of the ICOS+CD38+ cTFH population is induced by vaccination (Fig. 3C)]. These data are consistent with the clonotypic analysis that identifies up to ~10 to 20% of the ICOS+CD38+ cTFH (maybe 20 to 40% if one considers only the increase in the ICOS+CD38+ cTFH population after vaccination) as using TCRB sequences repeatedly recalled by influenza vaccination. Perhaps the remainder of this vaccine-induced ICOS+CD38+ cTFH population is nonspecific bystander cTFH activation. Alternatively, these cells could represent heterogeneous rare clonotypes and/or specificities not efficiently assayed in the AIM approach. Future studies should be able to address these questions with deeper TCRB analysis and analysis of additional specificities. Nevertheless, these distinct approaches yield data that not only are internally consistent and demonstrate a robust influenza virus–specific cTFH response but also provide approaches to allow future interrogation of antigen-induced cTFH responses and TFH memory in humans. The relationship of these cTFH to lymphoid TFH in terms of function, clonality, and dynamics remains to be evaluated. However, one interesting possibility is that the cTFH memory populations identified here using repeated yearly vaccination may be a source of systemic memory, allowing new GC reactions to be seeded in any lymphoid tissue depending on where the antigen is encountered.

Future rational vaccine strategies may require eliciting precise T cell responses to provide help for the desired types and specificities of the antibody response. Identification of the specific TCR repertoire expanded by immunization reveals a broader picture of the cTFH response than typically accessible by HLA class II tetramers or cytokine production after antigen stimulation. The identification of cTFH memory and the dynamics of this population upon vaccination provide not only a framework for dissecting how these cells are related to lymphoid tissue responses but also key insights for evaluating and comparing future vaccination regimens for their ability to induce the desired magnitude and quality of cTFH memory.

MATERIALS AND METHODS

Human participants

Participants were eligible if they were community-dwelling and had not received influenza vaccine in the previous 6 months; they were excluded if they had contraindications to influenza vaccine, active substance abuse, HIV or AIDS, clinically active malignancy, immunomodulatory medication need (i.e., chemotherapy and corticosteroids), or active illness (i.e., active respiratory tract infections). Seasonal inactivated influenza vaccine (Fluarix, GlaxoSmithKline) was administered, and peripheral venous blood was drawn on days 0, 7, and 28 after vaccination. For cohort 1, study participants were recruited and consented in the fall of 2014 at the Clinical Research Unit at Duke University Medical Center (Durham, NC), in accordance with the institutional review boards of both Duke University and the University of Pennsylvania (Philadelphia, PA). Blood was collected into heparinized tubes and shipped overnight to Philadelphia, PA. For cohort 2, study participants were recruited and consented in between 2013 and 2015 at the University of Pennsylvania. Samples from 2013 to 2014 were cryopreserved, whereas 2015 samples were used immediately after collection. Human spleen samples from relatively healthy adults were obtained as deidentified excess medical tissue via the Cooperative Human Tissues Network, typically after trauma or incidental splenectomy. Splenocytes were obtained by mechanical dissociation and cryopreserved until needed. All human participants research was performed in accordance with the relevant institutional review boards.

Flow cytometry

Fresh PBMCs and plasma were isolated using Ficoll-Paque PLUS (GE Healthcare) or SepMate isolation (STEMCELL Technologies) and stained for surface and intracellular markers. Permeabilization was performed using the Foxp3 Fixation/Permeabilization Concentrate and Diluent kit (eBioscience). Antibodies and clones are described in table S2. Cells were resuspended in 1% paraformaldehyde until acquisition on a BD Biosciences LSR II cytometer and analyzed using FlowJo (Tree Star) and viSNE (Cytobank). Fluorescence-minus-one controls were performed in pilot studies. Gating controls are shown in fig. S1A. The Bcl6 protein expression analysis was performed on a BD LSR II (18-color instrument) cytometer and repeated using a BD Symphony A5 (28-color) cytometer. The advantage of the latter experiment was that the Bcl6 antibody was the only reagent used on the violet laser, greatly reducing fluorescence spillover from other fluorochromes and substantially improving the signal-to-noise ratio for detection of low protein amounts.

HLA class II tetramers

HLA class II tetramers were prepared, as previously described (32). Staining by tetramers against HA residues 306 to 318 (PKYVKQNTLKLAT) or HA residues 398 to 410 (SVIEKMNTQFTAV) in two HLA-DRB1*04:01 participants was performed for 1 hour at room temperature, followed by magnetic bead enrichment at 4°C and fixation. Fixed samples were acquired on a BD Biosciences LSR II cytometer. For sorting experiments, tetramer staining was performed in a similar fashion, but samples were maintained unfixed at 4°C until they could be sorted on a BD FACSAria cell sorter.

AIM stimulations

PBMCs were thawed and rested overnight at 37°C in RPMI with 10% fetal calf serum and 1% l-glutamine. Overlapping peptide pools for HA 1 (A/California/7/2009, catalog NR-19244), HA 3 (A/Perth/16/2009, catalog NR-19266), NP (A/California/7/2009, catalog NR-18976), and M (A/California/7/2009, catalog NR-18977) were obtained from BEI Resources and resuspended in dimethyl sulfoxide. Stimulation was performed in flat-bottom plates with each peptide pool (0.5 μg/ml) for 18 hours, followed by surface staining for 20 min at room temperature. Acquisition and sorting were performed on a BD FACSAria cell sorter.

Hemagglutination inhibition assays

Sera were treated with a receptor-destroying enzyme and then heated for 30 min at 55°C. Sera were serially diluted in 96-well round-bottom plates. Four agglutinating doses of A/California/7/2009 or A/Switzerland/9715293/2013 were added to the sera in a total volume of 100 μl. Turkey red blood cells were added [12.5 μl of a 2% (v/v) solution)], and agglutination was read out 60 min later. Titers are expressed as the inverse sera dilution that inhibited viral agglutination.

Reverse transcription quantitative PCR

PBMCs and splenocytes were thawed and sorted for subsets on a BD FACSAria cell sorter, followed by RNA extraction by a Qiagen Micro Plus kit. Reverse transcription using the High-Capacity Reverse Transcription kit was performed as per the manufacturer’s instructions (Applied Biosystems). Real-time PCR was performed using predesigned primers and hydrolysis probes for B2M (Hs.PT.58v.18759587, Integrated DNA Technologies) and BCL6 (Hs.PT.56a.19673829.g, Integrated DNA Technologies) using the PrimeTime Gene Expression Master Mix on a ViiA7 Real-Time PCR system (Applied Biosystems).

TCR sequencing

PBMCs were sorted on a BD FACSAria, followed by DNA extraction by Qiagen QIAamp DNA Micro Kit. Amplification, library preparation, sequencing, and preliminary bioinformatics analysis were performed by Adaptive Biotechnologies. TCRB was sequenced at Survey level resolution.

TCRseq analysis

Calculations were performed in R for Shannon’s entropy and normalized Shannon’s entropy (46, 47). Gini index was calculated using the R package “ineq” (48). Clonality was calculated as follows: 1 − normalized Shannon’s entropy. The overlap score was calculated as the sum of the counts for all shared sequences observed in both sample A and sample B, divided by the total number of counts for all sequences in samples A and B (shared and unshared) in the two samples. Some figures were generated with packages “ggplot2” and “gplots” (49, 50). All R code and primary TCR sequences used for these analyses are available at http://github.com/Sedmic/cTfh_TCRB_2015-2016FluSpec.

Cytokine production

PBMCs were stimulated for 5 hours in the presence of PMA/ionomycin or left unstimulated as a control. Monensin was added for the final 4 hours of stimulation. Staining was performed at 37°C in the stimulation medium, followed by fixation with 1% paraformaldehyde for 5 min at room temperature. Cells were then permeabilized with the Foxp3 Fixation/Permeabilization Concentrate and Diluent kit, and intracellular staining was performed for 1 hour at room temperature with antibodies against IL-2, IL-10, IL-13, IL-17, IL-21, TNF, and IFN-γ. Pestle and SPICE (51) were used to analyze polyfunctionality.

Statistics

All analyses were performed as two-tailed tests. Comparisons using Student’s t test, paired t test, or one-way analysis of variance (ANOVA) with Tukey’s post hoc analysis were performed with Prism 5 (GraphPad) or with R. Repeated-measures one-way ANOVA and repeated-measures two-way ANOVA were performed with Prism 5 (GraphPad). Correlation analyses were performed as the Pearson correlation. Fisher’s exact test was used to compare overlap in TCR sequences based on 107 theoretical TCRB sequences.

SUPPLEMENTARY MATERIALS

immunology.sciencemag.org/cgi/content/full/2/8/eaag2152/DC1

Fig. S1. ICOS+CD38+ cTFH express many markers of activation.

Fig. S2. ICOS+CD38+ cTFH are distinct from other subsets.

Fig. S3. Influenza vaccine–induced changes in cTFH subsets by flow cytometry.

Fig. S4. Influenza vaccine–induced changes in cTFH subsets by TCRseq.

Fig. S5. Longitudinal analysis of cTFH subsets by TCRseq.

Fig. S6. Comparison of tetramer clonotypes and AIM clonotypes with the cTFH clonotypic response.

Fig. S7. Intersubset relationships.

Table S1. Demographic data for clinical cohorts.

Table S2. Fluorochrome reagents and antibody clones used.

Data file S1. Raw data for Figs. 1 to 7 and figs. S1 to S7.

REFERENCES AND NOTES

Acknowledgments: We thank the members of the Wherry, Su, and Hensley laboratories for their critical comments and review of the manuscript. This publication was made possible through core services and support from the Penn Center for AIDS Research, the University of Pennsylvania Flow Cytometry and Cell Sorting Facility, and the University of Pennsylvania Human Immunology Core. Funding: R.S.H. was supported by NIH grants AI114852 and AG047773. K.E.S. was supported in part by the NIH/National Institute on Aging Claude D. Pepper Older Americans Independence Centers (grant AG028716). This work was supported in part by a grant from the Penn Center for AIDS Research (P30 AI045008). This work was also funded by NIH grants AI113047 and AI108686 (to S.E.H.) and AI112521, AI117950, and AI2010085 (to E.J.W.), as well as by U.S. Broad Agency Announcements grant HHSN272201100018C (to K.E.S. and E.J.W.). This work was also supported in part by the Parker Institute for Cancer Immunotherapy. Author contributions: R.S.H. and E.J.W. conceived the overall design and designed experiments. R.S.H., A.M., L.V., B.B., and J.K. designed and performed TFH experiments. D.D.A. and L.F.S. produced and assisted with tetramer studies. S.A.D. and K.E.S. recruited and vaccinated study participants for cohort 1. P.T. assisted with clinical cohort recruitment for cohort 2. K.P. and S.E.H. performed hemagglutination inhibition assays. All authors analyzed and interpreted data, discussed the results, and commented on the manuscript. R.S.H. performed all statistical analyses. R.S.H. and E.J.W. wrote the manuscript. Competing interests: The authors declare that they have no competing financial interests, patents, patent applications, or material transfer agreements associated with this study.
View Abstract

Navigate This Article