Clonal composition and persistence of antigen‐specific circulating T follicular helper cells

Abstract

T follicular helper (T_FH) cells play an essential role in promoting B cell responses and antibody affinity maturation in germinal centers (GC). A subset of memory CD4⁺ T cells expressing the chemokine receptor CXCR5 has been described in human blood as phenotypically and clonally related to GC T_FH cells. However, the antigen specificity and relationship of these circulating T_FH (cT_FH) cells with other memory CD4⁺ T cells remain poorly defined. Combining antigenic stimulation and T cell receptor (TCR) Vβ sequencing, we found T cells specific to tetanus toxoid (TT), influenza vaccine (Flu), or Candida albicans (C.alb) in both cT_FH and non‐cT_FH subsets, although with different frequencies and effector functions. Interestingly, cT_FH and non‐cT_FH cells specific for C.alb or TT had a largely overlapping TCR Vβ repertoire while the repertoire of Flu‐specific cT_FH and non‐cT_FH cells was distinct. Furthermore, Flu‐specific but not C.alb‐specific PD‐1⁺ cT_FH cells had a “GC T_FH‐like” phenotype, with overexpression of IL21, CXCL13, and BCL6. Longitudinal analysis of serial blood donations showed that Flu‐specific cT_FH and non‐cT_FH cells persisted as stable repertoires for years. Collectively, our study provides insights on the relationship of cT_FH with non‐cT_FH cells and on the heterogeneity and persistence of antigen‐specific human cT_FH cells.

An in‐depth T cell receptor Vβ repertoire analysis of human blood CD4⁺ T helper cells specific for different microbial antigens (i.e. influenza vaccine, tetanus toxoid, and Candida albicans) revealed that circulating T follicular helper cells are heterogeneous in phenotype and persistence.

Introduction

Among memory T helper cells, T follicular helper (T_FH) cells have been considered as a distinct subset due to their unique ability to promote B cell responses and antibody affinity maturation in germinal centers (GCs) [1, 2, 3, 4]. While T_FH cells by definition are fully differentiated effector cells, studies have shown that they are also able to generate long‐lived memory cells residing in lymphoid tissues [5, 6, 7, 8]. In humans, a subset of T helper cells with constitutive expression of the T_FH signature chemokine receptor CXCR5 was found in peripheral blood and termed as circulating T_FH (cT_FH) cells [1, 2, 9]. cT_FH cells represent a heterogeneous population, with some cells being phenotypically closer to GC T_FH than others. In the context of infections or vaccinations, frequency of CCR7^loPD‐1⁺ cT_FH cells able to promote antibody production was shown to correlate with active T_FH cell differentiation in secondary lymphoid organs [10]. Furthermore, frequency of PD‐1⁺CXCR3^– cT_FH cells correlates positively with the humoral immune response against HIV‐1 [11] and HIV‐1‐specific ICOS⁺PD1⁺ cT_FH cells expressing several GC T_FH‐related genes are induced upon HIV‐1 vaccination [12]. ICOS⁺PD1⁺ cT_FH cells are also found upon influenza vaccination and their frequency correlates with the antibody responses [13, 14, 15].

The origin of blood cT_FH cells and their relationship with T_FH cells found in lymphoid tissues or other body fluids remains still poorly defined and is currently subject to intense investigation. Repertoire analysis by TCR Vβ sequencing has revealed clonal overlap between bulk populations of blood cT_FH and tonsillar T_FH cells, thus providing first evidence of a direct link between the two anatomically distinct subsets [12, 16]. A recent study analyzing the composition of T cells in human efferent lymphatic vessels identified a CXCR5⁺PD‐1⁺ T_FH population with an intermediate phenotype between GC T_FH cells and ICOS⁺PD‐1⁺CD38⁺ cT_FH cells [17]. Lymph T_FH cells might represent a circulatory intermediate between GC T_FH and blood cT_FH, in line with a “T_FH on the move” model of cell trafficking across different lymph nodes through the lymphatic and blood systems [18]. The fact that only a proportion of blood cT_FH cells shares functional properties with bona fide GC T_FH raises the question whether all cT_FH cells are directly related to T_FH cells that participated in the GC reactions in lymphoid organs.

In this study, we focused our analysis on antigen‐specific memory cT_FH cells induced in vivo by pathogens or vaccines. We found that cT_FH and non‐cT_FH cells specific for Candida albicans (C.alb) or tetanus toxoid (TT) were functionally diverse but had a largely overlapping TCR Vβ repertoire. In contrast, cT_FH specific for influenza antigens (Flu) had a distinct TCR Vβ repertoire compared to non‐cT_FH cells and, when expressing PD‐1, phenotypically resembled GC T_FH. In‐depth longitudinal TCR Vβ sequencing analysis of serial blood samples unraveled the persistence and stability of Flu‐specific cells, indicating long‐lived cT_FH and non‐cT_FH cell responses. Collectively, our findings demonstrate that human cT_FH cells are a heterogeneous population and their clonal relatedness with other T helper cell subsets and persistence depend upon their antigen specificity.

Results

The clonal relationship between cT_FH and non‐cT_FH cells depends on the nature of the antigen

To define the antigen‐specificities of human cT_FH cells and non‐cT_FH cells, we sorted CD4⁺ memory T cells from PBMCs of healthy donors and divided them into four subsets based on the expression of CXCR5 and CCR6 (Fig. 1A and Fig. S1). T cells were labeled with CFSE and stimulated with autologous monocytes pulsed with Flu, TT or C.alb antigens; frequency of antigen‐reactive CFSE^low proliferating cells was analyzed after 5–6 days, as previously described [19]. While TT‐specific T cells were present in all subsets, Flu‐specific and C.alb‐specific T cells were present mainly in the CCR6^– or CCR6⁺ subsets, respectively (Fig. 1B), consistent with previous data [20, 21]. Within the CCR6^– and CCR6⁺ subsets, C.alb‐specific T cells were significantly more abundant in the non‐cT_FH (CCR6⁺CXCR5^–) subset while Flu‐specific T cells were significantly more abundant in the cT_FH (CCR6^–CXCR5⁺) subset. In all cases, cT_FH cells produced lower amounts of effector cytokines (i.e. IFN‐γ, IL‐4, IL‐17, and IL‐22) than non‐cT_FH cells in response to antigenic stimulation (Fig. S2). Collectively, these data indicate that both cT_FH and non‐cT_FH subsets contain memory T cells specific for common pathogens or vaccines and that, among the antigens tested, influenza virus antigens are able to induce a high frequency of cT_FH cells.

Figure 1.

Variable TCR Vβ repertoire overlap between cT_FH and non‐cT_FH cells specific for different microbial antigens. (A) Flow cytometry analysis on a representative donor showing the expression of CXCR5 and CCR6 on CD4⁺ memory T cells. (B) Four memory T cell subsets were FACS‐sorted according to CXCR5 and CCR6 differential expression, labeled with CFSE, and stimulated with influenza vaccine (Flu), tetanus toxoid (TT) or C. albicans (C.alb) in the presence of autologous monocytes. Shown are the percentages of proliferating CFSE^low cells at 6–7 days post stimulation. In all conditions, 85%‐98% of CFSE^low cells expressed the activation markers CD25 and/or ICOS. Each symbol indicates a different healthy donor (n = 5 donors for Flu, n = 3 for TT, and n = 4 for C.alb). * p‐value < 0.05, ** p‐value < 0.01, *** p‐value < 0.001, **** p‐value < 0.0001, as determined by two‐tailed paired t test. (C) Flu‐, TT‐ and C.alb‐specific CFSE^low T cells from each subset were FACS‐sorted and their repertoire was analyzed by next generation TCR Vβ sequencing. Shown are the cytoscape clonal networks of TCR Vβ repertoire of antigen‐specific T cells isolated from CXCR5⁺CCR6⁺ (purple), CXCR5^–CCR6⁺ (blue), CXCR5⁺CCR6^– (pink), and CXCR5^–CCR6^– (green) T cell subsets. Numbers of total (colored) and shared (black) TCR Vβ clonotypes are indicated. Clonotypes shared among two or three subsets are shown as black dots and clonotypes shared among all four subsets are shown as red dots. Data are from three representative donors (one for each antigen condition, indicated in panel B with blue symbols) out of six donors analyzed. (D) Shown is the fraction of unique and shared TCR Vβ clonotypes in each sample reported in panel C. Only the major responder T cell subsets for each type of antigen are reported. (E) Dendrogram plots of repertoire similarity among the four subsets based on the Morisita‐Horn similarity index between pairs of TCR Vβ repertoires. (F) Pairwise comparisons of TCR Vβ clonotype frequency distribution in antigen‐specific CFSE^low T cell lines isolated from cT_FH and non‐cT_FH cells reported in panel C. Frequencies are reported as a percentage of productive templates. The total number of clonotypes is indicated in the x‐ and y‐axes. The number of clonotypes shared between two samples are reported in the upper right corner. (G) Shown is the cumulative frequency of unique and shared TCR Vβ clonotypes in each sample reported in panel C. Only the major responder T cell subsets for each type of antigen are reported.

The finding that cT_FH and non‐cT_FH cells responded to different extents to antigenic stimulations prompted us to further investigate their lineage relationship. Using a previously established workflow [19, 22], we isolated CFSE^low cells from the antigen‐stimulated cultures and performed next‐generation sequencing to quantify and compare TCR Vβ clonotypes of different antigen‐specific T cell populations. As shown in Fig. 1C, the number of clonotypes was high in the T cell subsets best responding to antigenic stimulation, reflecting a diverse and polyclonal repertoire. Clonotype sharing between cT_FH and non‐cT_FH cells was however diverse, with large overlaps in the case of TT and C.alb but limited overlap in the case of Flu, in particular between CCR6^– cT_FH and CCR6^– non‐cT_FH cells that were the main responding subsets to Flu (Fig. 1C, D). Dendrogram plots based on the Morisita‐Horn similarity index between pairs of TCR Vβ repertoires showed a close clonal relationship between C.alb‐specific cT_FH and non‐cT_FH cells and between TT‐specific cT_FH and non‐cT_FH cells. In contrast, in the case of Flu a closer clonal relationship was detected between CCR6⁺ and CCR6^– cT_FH subsets and between CCR6⁺ and CCR6^– non‐cT_FH subsets (Fig. 1E). Pair‐wise comparisons of the TCR Vβ clonotype frequency distributions further demonstrated that a large proportion of the expanded clonotypes were shared between cT_FH and non‐cT_FH cells responding to C.alb or TT, while highly expanded Flu‐specific clonotypes were not shared and unique to either cT_FH or non‐cT_FH cell subset (Fig. 1F, G).

To further corroborate the above findings, we introduced internal controls by performing TCR Vβ sequencing of technical replicates. Extensive clonotype sharing was measured among replicates of non‐cT_FH cell cultures, which was also evident, although to a lower extent, among replicates of cT_FH cell cultures (Fig. S3A‐D). Data normalization based on the repertoire overlap between CXCR5⁺ technical replicates confirmed the lower overlap in Flu‐specific cT_FH and non‐cT_FH repertoires compared to C.alb‐specific cT_FH and non‐cT_FH cells (Fig. S3E). Collectively, these data indicate that the overlap of TCR Vβ repertoires of antigen‐specific cT_FH and non‐cT_FH cells is variable and depending on the nature of the antigen.

Influenza‐specific cells are enriched in PD‐1⁺ cT_FH and have a distinct repertoire

Previous studies have shown that PD‐1 is expressed on nearly 30% of cT_FH cells and that PD‐1⁺ cT_FH cells, which are also Ki67, CD38, and HLA‐DR negative [10, 11], are close to GC T_FH cells in terms of phenotype and function [10‐12, 14, 23–25]. Based on these earlier findings, we set out to perform antigenic stimulation of cT_FH cells sorted according to PD‐1 expression (Fig. S1B). Flu‐specific cells were present at higher frequencies in the PD‐1⁺ cT_FH subset compared to the PD‐1^– cT_FH and non‐cT_FH subsets, whereas the opposite was observed for C.alb‐specific cells (Fig. 2A). For both specificities, the PD‐1⁺ cT_FH subset produced very little of any of the cytokines tested (Fig. S4). TCR Vβ repertoire analysis showed that PD1⁺ cT_FH cell clonotypes were poorly shared with PD‐1^– cT_FH and non‐cT_FH cell clonotypes in Flu‐ but not C.alb‐specific response (Fig. 2B‐D). Accordingly, Flu‐specific PD‐1⁺ cT_FH and non‐cT_FH cells had the lowest normalized TCR overlap score while C.alb‐specific PD‐1⁺ cT_FH and PD‐1^– cT_FH cells had the highest (Fig. 2E).

Figure 2.

Flu‐specific cells are enriched in PD‐1⁺ cT_FH and have a distinct repertoire. (A) Percentages of proliferating CFSE^low cells measured at day 6–7 post stimulation of T cell subsets with Flu or C. al C.alb) in the presence of autologous monocytes. Each symbol indicates a different healthy donor (n = 7 for Flu, and n = 7 for C.alb). * p‐value < 0.05, ** p‐value < 0.01, as determined by two‐tailed paired t test. (B) The TCR Vβ repertoire of Flu‐ and C.alb‐specific CFSE^low T cells from each subset was determined by next generation sequencing. The numbers of TCR Vβ clonotypes detected from each antigen‐specific PD‐1⁺ cT_FH, PD‐1^– cT_FH and CXCR5^– subsets and technical replicates are reported in the black cells of the table. The numbers of clonotypes shared between technical replicates are shown in the grey cells. The numbers of clonotypes shared between any pair of two different subsets are reported in the white cells. Data are from two representative donors (one for each antigen condition, indicated in panel A with blue symbols) out of three donors analyzed. (C) Percentages of unique and shared clonotypes between technical replicates and between different subsets, determined by the Jaccard index calculated on the samples reported in panel B. (D) Cumulative frequencies of unique and shared clonotypes between technical replicates and between different subsets, determined as the average of the cumulative frequencies of the shared clonotypes in each of the two subsets calculated on the samples reported in panel B. (E) Normalized TCR overlap score between pairs of TCR Vβ repertoires of antigen‐specific PD‐1⁺ cT_FH, PD‐1^– cT_FH and non‐cT_FH cells, calculated by Chao‐Jaccard similarity index normalized on technical replicates. Data are multiple pairwise comparisons of the samples reported in panel B. * p‐value < 0.05, ** p‐value < 0.01, *** p‐value < 0.001, **** p‐value < 0.0001, as determined by one‐way ANOVA with Tukey multiple comparisons test. (F) Heat map reporting the frequencies of the top 5% most expanded TCR Vβ clonotypes in the samples reported in panel B (433 Flu‐specific and 201 C.alb‐specific clonotypes, respectively) ranked by the sum frequency of each clone. Dendrograms are based on the Morisita‐Horn similarity index between pairs of TCR Vβ repertoires.

To further corroborate the above findings, we selected the top 5% expanded clonotypes (i.e. 433 Flu‐specific and 201 C.alb‐specific clonotypes) and calculated a dendrogram of repertoire relatedness based on the Morisita‐Horn similarity analysis (Fig. 2F). As expected, technical replicates of each subset clustered together with the exception of C.alb‐specific PD‐1⁺ cT_FH, whose repertoire largely overlapped with PD‐1^– cT_FH cells. Conversely, Flu‐specific PD‐1⁺ cT_FH displayed a distinct TCR Vβ repertoire characterized by low overlap with Flu‐specific PD‐1^– cT_FH and non‐cT_FH populations.

Collectively, these data demonstrate a differential distribution of antigen‐specific cells in the PD‐1⁺ and PD‐1^– cT_FH cell subsets. Flu‐specific T cells were preferentially found in the PD‐1⁺ cT_FH subset, which is considered to be closely related to bona fide GC T_FH cells.

Influenza‐specific PD‐1⁺ cT_FH cells phenotypically resemble GC T_FH cells

To further investigate the phenotype of antigen‐specific PD‐1⁺ cT_FH, PD‐1^– cT_FH, and non‐cT_FH cells avoiding unintended changes that may be caused by prolonged in vitro culture, we adopted a protocol [26] to isolate antigen‐specific cells based on surface overexpression of CD40L shortly after antigenic stimulation (Fig. S5A). Briefly, PD‐1⁺ cT_FH, PD‐1^– cT_FH, and non‐cT_FH cells were stimulated with antigen‐pulsed autologous monocytes in the presence of CD40‐blocking antibodies. After 24 hours, activated T cells that upregulated CD40L were sorted and subject to Nanostring nCounter analysis to quantify the transcripts of 579 immune‐relevant genes. By analyzing the genes that were differentially expressed between PD‐1⁺ cT_FH and PD‐1^– cT_FH, and further comparing those transcripts between Flu‐ and C.alb‐stimulated cells, we identified 40 genes that were up‐regulated in PD‐1⁺ cT_FH cells in both antigen‐stimulated cultures, including HLA‐DR, CXCL13, IL17F, LAG3, ENTPD1 (CD39), CD80, IFNG (Fig. S5B, C). Interestingly, transcripts of 49 genes were uniquely upregulated in Flu‐specific PD‐1⁺ cT_FH, including BCL6 and IL21. Furthermore, nine genes that were upregulated in Flu‐specific PD‐1⁺ cT_FH were downregulated in C.alb‐specific PD‐1⁺ cT_FH (fig. S5B‐D).

By focusing on selected transcripts of interest, we found that Flu‐specific PD‐1⁺ cT_FH were enriched for a GC T_FH gene signature, including overexpression of IL21, CXCL13, BCL6, SLAMF, BTLA, TIGIT, SH2D1A, while Th1‐related genes, such as CXCR3, TBX21, PRDM1 (BLIMP‐1), were expressed at low levels (Fig. 3A). Notwithstanding the expression of CXCL13, SLAMF6 and a few other GC T_FH‐related genes, the enrichment of GC T_FH signature was less evident in C.alb‐specific PD‐1⁺ cT_FH compared to Flu‐specific PD‐1⁺ cT_FH cells. As expected, Th17‐related genes were preferentially expressed by C.alb‐specific non‐cT_FH cells and were mostly negatively enriched in PD‐1⁺ and PD‐1^– cT_FH (Fig. 3A). Ranking GC T_FH‐related genes based on the ratio of their expression in PD‐1⁺ cT_FH to PD‐1^– cT_FH, (Fig. 3B), we found that CXCL13 was the most preferentially overexpressed transcript in both Flu‐ and C.alb‐specific PD‐1⁺ cT_FH cells, consistent with the notion that secretion of high amounts of CXCL13 is an attribute of GC T_FH cells [27, 28, 29]. PD‐1⁺ cT_FH cells were also able to secrete high amounts of CXCL13 in the culture supernatant upon antigenic stimulation (Fig. 3C). Interestingly, transcripts for IL‐21, a key signature cytokine secreted by T_FH cells and promoting B cell proliferation, class switch recombination and plasma cell differentiation [30, 31, 32], were markedly overexpressed only in Flu‐specific PD‐1⁺ cT_FH (4.3 fold) but not in C.alb‐specific PD‐1⁺ cT_FH cells(1.34 fold). The differential expression of IL21 mRNA was further confirmed at the protein level by measuring IL‐21 concentration in the day 7‐stimulated culture supernatants (Fig. 3D).

Figure 3.

Flu‐specific PD‐1⁺ cT_FH cells phenotypically resemble GC T_FH cells. (A) Heat map of the mRNA expression level of selected genes from antigen‐specific PD‐1⁺ cT_FH, PD‐1^– cT_FH, and non‐cT_FH cells. T_FH‐related and Th1‐related genes are shown for Flu‐specific cells (upper panels). T_FH and Th17‐related genes are shown for C. albicans‐specific cells (lower panels). Gene expression is normalized based on the average mRNA counts of each gene from the three T cell subsets. Data from n = 4 healthy donors (HD1‐4) are shown. (B) Shortlist of GC T_FH‐related genes ranked based on the gene expression ratio of PD‐1⁺ cT_FH to PD‐1^– cT_FH specific for influenza (left panel) or C. albicans (right panel). Ratios falling into the range of 0.67‐1.5 were considered not significant. (C) Concentrations of CXCL13 in the day 6–7 culture supernatants of T cell subsets stimulated with Flu or C.alb, as measured by ELISA. For each antigen condition, shown are the experimental replicates (n = 4‐10) from one representative donor out of n = 3 donors analysed. Background cytokine production measured in T cells cultured with monocytes in the absence of the antigen is shown as dashed lines. * p‐value < 0.05, ** p‐value < 0.01, **** p‐value < 0.0001, as determined by two‐tailed unpaired t test. (D) Concentrations of IL‐21 in the day 6–7 culture supernatants of T cell subsets stimulated with Flu or C.alb, as measured by ELISA. For each antigen condition, shown are the experimental replicates (n = 6‐20) from one representative donor out of n = 3 donors analyzed. Background cytokine production measured in T cells cultured with monocytes in the absence of the antigen is shown as dashed lines. ** p‐value < 0.01, as determined by two‐tailed unpaired t test.

Collectively, these data demonstrate that Flu‐specific PD‐1⁺ cT_FH cells overexpress several GC T_FH‐related genes compared to PD‐1^– cT_FH cells. The enrichment for a GC T_FH gene signature is less evident in C.alb‐specific cT_FH cells, thus revealing that the level of cT_FH polarization towards bona fide GC T_FH phenotype is antigen‐dependent.

Influenza‐specific cT_FH and T_EM cells are persistent and phenotypically stable

To comprehensively measure the relationships and evolution of antigen‐specific cT_FH and non‐cT_FH cell repertoires, we performed a longitudinal TCR Vβ repertoire analysis by collecting at multiple time points large blood donations from a healthy donor. At each time‐point, we FACS‐sorted from PBMCs cT_FH and non‐cT_FH memory cells, which for these experiments were further divided into central memory (T_CM) and effector memory (T_EM) subsets based on CCR7 expression (Fig. S1C). The experiment was designed taking into account our previous work showing that Influenza‐specific T cells are present in all three memory subsets [33] and that CCR6 is expressed on both T_CM and T_EM subsets [34]. The CFSE‐labelled T cell subsets were then stimulated with Influenza vaccine and the TCR Vβ repertoire of proliferating CFSE^low T cells was performed by next‐generation sequencing, as described previously. Using this method, we identified at the initial time‐point 3529, 2900, and 3090 TCR Vβ clonotypes in Flu‐specific cT_FH, T_CM, and T_EM cultures, respectively (Fig. 4A). Pairwise comparison of TCR Vβ clonotype frequency distributions in CFSE^low cultures from different subsets revealed extensive sharing of the most expanded clonotypes, particularly between Flu‐specific T_CM and T_EM (Fig. 4A, left). As shown in the Venn diagram in Figure 4A, hundreds of TCR Vβ clonotypes were found in common between different T cell subsets, with 318 clonotypes found shared among all three subsets (Fig. 4A, right). TCR Vβ repertoire overlap was particularly high between Flu‐specific T_CM and T_EM as compared to cT_FH at most time‐points of analysis, as measured by the Chao‐Jaccard index of similarity (Fig. 4B).

Figure 4.

Flu‐specific cT_FH and T_EM cells induced by infection or vaccination are persistent and phenotypically stable. (A) Serial blood donations from a Flu vaccinated individual were collected and next generation TCR Vβ sequencing was performed on CFSE^low proliferating T cells isolated from Flu‐stimulated cT_FH, T_CM, and T_EM subsets (initial input 5×10⁵ cells per subset) at each time‐point. Shown is the pairwise comparison of TCR Vβ clonotype frequency distribution in Flu‐specific cT_FH, T_CM, and T_EM subsets at the initial time‐point of analysis (t1: May 2014). Frequencies are shown as percentage of productive templates. The total number of clonotypes is indicated in the x‐ and y‐axes. Values in the upper right corner represent the number of clonotypes shared between two samples. The Venn diagrams show the number of unique and shared Flu‐specific TCR Vβ clonotypes between cT_FH, T_CM and T_EM subsets. (B) Sample overlap between pairs of Flu‐specific TCR Vβ repertoires was calculated using the Chao–Jaccard similarity index. The bar histograms show the Chao‐Jaccard index between pairs of Flu‐specific cT_FH, T_CM, and T_EM TCR Vβ repertoires obtained from serial blood donations of a healthy donor (t1: May 2014, t2: Feb 2015, t3: Sep 2015, t4: Jan 2016, t5: Nov 2017). *p < 0.05 as determined by two‐tailed paired t test. (C) Number of productive TCR Vβ clonotypes detected by next generation sequencing of Flu‐specific cT_FH (red circles), T_CM (green squares), and T_EM (blue triangles) cells at each time‐point of the longitudinal analysis. (D and E) The Flu‐specific TCR Vβ clonotypes found at the initial time‐point (t1) in CFSE^low cultures from cT_FH (3529 clonotypes), T_CM (2900 clonotypes), and T_EM (3090 clonotypes) were followed in a longitudinal analysis of serial blood donations. The fraction of TCR Vβ clonotypes detected in CFSE^low cultures at each subsequent time‐point is reported in (D) as percentage of start for cT_FH (red circles), T_CM (green squares), and T_EM (blue triangles). Dates of seasonal Flu vaccination are reported as dashed lines. The stacked barplots in (E) show a longitudinal analysis of the phenotype distribution of Influenza‐specific clonotypes originally detected at the initial time‐point (t1) in CFSE^low cultures from T cell subsets (cT_FH plotted on the left, T_CM in the center, T_EM on the right). Each sector reports the percentage of TCR Vβ clonotypes that are found within the original subset (cT_FH, T_CM and T_EM for the plots on the left, center and right, respectively) or transitioned to other subsets at each time‐point. Data are expressed as percentage of start, and subsets are color‐coded (cT_FH in red, T_CM in green, and T_EM in blue). (F) Within the Flu‐specific TCR Vβ clonotypes found at the initial time‐point (t1) only in CFSE^low cultures from cT_FH (cT_FH‐only: 2541 clonotypes), T_CM (T_CM‐only: 1743 clonotypes), or T_EM (T_EM‐only: 1819 clonotypes), the top 5% clonotypes were shortlisted (top 5% cT_FH‐only: 127 clonotypes, top 5% T_CM‐only: 87 clonotypes, top 5% T_EM‐only: 91 clonotypes) and followed in CFSE^low cultures from subsequent time‐points. The stacked barplots report a longitudinal analysis of the phenotype distribution of the shortlisted TCR Vβ clonotypes detected at each subsequent time‐point (top 5% cT_FH‐only plotted on the left, top 5% T_CM‐only in the center, top 5% T_EM‐only on the right). Each sector reports the percentage of TCR Vβ clonotypes that are found within the original subset (cT_FH, T_CM and T_EM for the plots on the left, center and right, respectively) or transitioned to other subsets at each time‐point. Data are expressed as percentage of start, and subsets are color‐coded (cT_FH in red, T_CM in green, and T_EM in blue). (G and H). The Flu‐specific TCR Vβ clonotypes found at the initial time‐point (t1) shared among CFSE^low cultures from cT_FH, T_CM and T_EM were shortlisted (318 clonotypes) and followed in subsequent time‐points. The fraction of TCR Vβ clonotypes detected at each subsequent time‐point is reported in (G) as percentage of start. Dates of seasonal Flu vaccination are reported as dashed lines. Panel (H) shows the frequency distribution of TCR Vβ clonotypes from total cT_FH, T_CM, and T_EM subsets sequenced directly after ex vivo isolation from the same individual at t1 (May 2014, plots on the left) and t5 (Nov 2017, plots on the right). Frequencies are shown as percentage of productive templates. Colored circles mark the TCR Vβ clonotypes belonging to the shortlist of 318 shared clonotypes. The total number of TCR Vβ clonotypes and the number of retrieved clonotypes from the group of 318 shared clonotypes are reported on top of each graph in black and orange, respectively. Dotted lines in the graphs indicate the frequency threshold of the top 5% expanded clonotypes.

To get insights on the evolution of the clonal architecture of antigen‐specific T cell repertoires, we followed Flu‐specific TCR Vβ clonotypes in samples spanning a period of 42 months. The total number of Flu‐specific TCR Vβ clonotypes detected from different T cell subsets was consistently high at each time point analyzed, spanning from 2196 to 7920 clonotypes (Fig. 4C) and demonstrating a polyclonal T cell response sustained by vaccination or infection. We then selected the whole TCR Vβ repertoires of CFSE^low Flu‐specific cultures at the initial time‐point (t1, composed by 3529, 2900, and 3090 clonotypes in cT_FH, T_CM and T_EM, respectively), tracked their clonotypes in the CFSE^low fractions obtained at subsequent time‐points (Fig. 4D) and quantified the proportion of the t1 TCR Vβ clonotypes transitioning towards other subsets at each time‐point (Fig. 4E). Flu‐specific clonotypes from cT_FH and T_EM were mostly detected in the original subset at all time points, whereas a high frequency of Flu‐specific T_CM clonotypes were found in the cT_FH and T_EM subsets and became undetectable in the T_CM subset at later time points (Fig. 4E).

At the initial time‐point (t1) the Flu‐specific TCR Vβ repertoires comprised many low frequency‐clonotypes found uniquely in one subset and not shared with others, and a few highly expanded clonotypes that were shared among cT_FH, T_CM and T_EM (Fig. 4A, right). To better decipher the role of each of those components to the overall stability and evolution of Flu‐specific repertoires, we selected and followed longitudinally each of them. In particular at the t1, 2541 clonotypes were found uniquely in cT_FH (hereafter named “cT_FH‐only”), 1743 uniquely in T_CM (“T_CM‐only”), and 1819 uniquely in T_EM (“T_EM‐only)”, that corresponded to 72%, 60% and 59% of the Flu‐specific TCR Vβ clonotypes from cT_FH, T_CM and T_EM at t1, respectively. To exclude potential biases in detection due to low expansion, we selected the top 5% expanded t1 TCR Vβ clonotypes of each class (127 clonotypes for cT_FH‐only cells, 87 for T_CM‐only, and 91 for T_EM‐only) and checked for their presence in the CFSE^low cultures from subsequent time points (Fig. 4F). A strong persistence was observed for Flu‐specific t1 clonotypes from cT_FH‐only and T_EM‐only, with almost 50% of them being detected throughout 42 months. Most cT_FH‐only and T_EM‐only t1 clonotypes were preferentially found again at each subsequent time‐point in the original subset (Fig. 4F), indicating rare transitions toward other subsets during the analysis timeframe. Conversely, Flu‐specific t1 clonotypes from T_CM‐only showed a drop in detection in subsequent time‐points, which is consistent with the contraction of certain clonotypes below the sampling and sequencing detection limits. In line with the longitudinal analysis of the whole subset, also T_CM‐only clonotypes showed a higher proportion of transitions toward other subsets (Fig. 4F), thus indicating phenotypic plasticity of Flu‐specific T_CM cells.

At the initial time‐point (t1), 318 TCR Vβ clonotypes were found shared among cT_FH, T_CM, and T_EM. When followed longitudinally, the vast majority (>80%) of those heterogeneous clonotypes were detected again in the CFSE^low fractions from subsequent time points’ samples (Fig. 4G, left). Comparison with the TCR Vβ repertoire of total cT_FH, T_CM, and T_EM sequenced directly ex‐vivo revealed that those shared clonotypes were clonally expanded at both the initial (t1) and the final time‐point (t5) of analysis (Fig. 4H, right), thus indicating the existence of a small pool of expanded and functionally heterogeneous Flu‐specific T cells that persist in vivo.

Discussion

Several studies in humans have established the connections between T_FH cells in secondary lymphoid organs and cT_FH in blood, providing evidence that T_FH cells that have been engaged in GC reaction in B cell follicles can then enter the circulation and possibly persist as memory cT_FH cells [16, 17, 18]. Whether all cT_FH are derived from GC T_FH remains however to be defined. Our study suggests that not all blood cT_FH cells may be the descendent of GC T_FH cells and that some cells are closer to blood non‐cT_FH cells. Several lines of evidence support this notion. Firstly, the TCR Vβ sequencing data indicate substantial repertoire overlap between C.alb‐ or TT‐specific cT_FH and non‐cT_FH cells. Secondly, longitudinal TCR Vβ sequencing showed that a fraction of cells from the CXCR5^– T_CM subset were able to convert into CXCR5⁺ cT_FH cells. Moreover, antigen‐specific cT_FH cells can produce the same types of cytokines as the non‐cT_FH counterparts, albeit to a different extent. On this regard, previous reports have highlighted the plasticity of the T_FH gene expression program by showing that ex‐vivo sorted lymphoid PD‐1⁺ T_FH cells could be polarized in vitro to produce IFN‐γ, IL‐4, and IL‐17, and likewise Th1, Th2, and Th17 cells, could be reprogrammed to acquire T_FH features [35]. Collectively, previous studies and our new data suggest that cT_FH cells are a heterogeneous T cell population with different origins.

In the context of influenza vaccination, a population of activated cT_FH cells characterized by the expression of ICOS, PD‐1 and CXCR3 has been shown to emerge in the blood at day 7 post‐vaccination and to correlate with antibody response [10, 13–15]. Although in two of these studies and in our own investigation cT_FH cells were not sorted based on CXCR3 expression, it is likely that a large proportion of Flu‐specific T cells overexpress CXCR3 [36], thus suggesting that Flu‐specific CXCR3⁺ PD‐1⁺ cT_FH cells are closely related to GC T_FH cells. However, previous reports monitoring the response to HIV‐1 [11] or malaria infection [24] proposed the CXCR3^– PD‐1⁺ cT_FH subset as the most related to bona fide GC‐T_FH cells. Rather than being in conflict, these studies together indicate that cT_FH origin and phenotype may be intimately dependent on the nature of the antigenic trigger.

In healthy individuals, Candida species are harmless commensals that colonize skin and mucosal surfaces, but they can cause serious opportunistic fungal infections under certain conditions when the tissue homeostasis is disrupted. Mucosal homeostasis and host protection from Candida species is mainly ensured by Th17 cells, as demonstrated by the development of chronic mucocutaneous candidiasis in patients with inborn mutations in genes important for IL‐17 immunity [37, 38], while antibodies are considered dispensable since patients with agammagobulinaemia do not show increased susceptibility to Candida infections [39, 40]. Conversely, protection from influenza virus, a rapidly transmissible pathogen which cause acute respiratory infections, mainly relies on neutralizing antibodies acting sometimes together with CD4⁺ and CD8⁺ T cells [41, 42]. It is tempting to speculate that the distinct and highly differentiated antigen‐specific cT_FH cell population that we measured in response to Flu, but not to C.alb, may be tightly linked to the magnitude of the antibody response and to the extent of antibody affinity maturation to Flu antigens, in line with the growing body of literature on cT_FH cells as correlates of the development of neutralizing antibodies [43, 44]. It is also conceivable that the magnitude of T_FH cell response and GC reactions might be different depending on the nature of the pathogen, hence resulting in the differentiation of a variable fraction of effector and memory cT_FH cells. Our study corroborates this notion by demonstrating that the lineage relationship between cT_FH and non‐cT_FH cells is also largely dependent on the nature of the antigenic trigger.

The persistence of cT_FH cells as long‐lived memory cells remains ill defined. In the context of Flu, we previously showed that several T cell clones specific for an immunodominant epitope of influenza A hemagglutinin can be detected in the TCR Vβ repertoires of cT_FH, T_CM and T_EM analyzed several months later [33]. In the present study, we expanded this observation and showed that cT_FH cells specific for Flu antigens persisted as a stable repertoire over a time interval of 42 months similarly to Flu‐specific T_EM cells, indicating clonal maintenance within the cT_FH subset as long‐lived memory cells, in line with findings by Herati and colleagues [15]. These results are also consistent with a previous study showing that vaccinia virus‐specific cT_FH cells were still detectable in vaccinated donors even 25 years after the smallpox eradication [45]. Moreover, our longitudinal study underlines the power of analysing the evolution of antigen‐specific T cell repertoires by serial sampling. The frequent subset transitions we observed in the case of Flu‐specific T_CM are consistent with a model of progressive differentiation of memory T cells that proposes T_CM as an intermediate stage [46, 47].

Overall, our findings suggest that differentiation and maintenance of cT_FH cells largely depend on the antigen context. Additional work will be needed to understand whether the heterogeneity in phenotype and longevity of cT_FH cells specific for different antigens is a correlate of a variable capacity of related bona fide GC T_FH cells to efficiently sustain antibody production and affinity maturation by B cells. Collectively, our study contributes to the understanding of the clonal relationship between antigen‐specific cT_FH and non‐cT_FH cells in the peripheral blood and provides insight on the heterogeneity, origin, and persistence of human cT_FH cells specific for different antigens.

Materials and methods

Cells and cell sorting

The study was approved by the Ethical committees of Cantone Ticino, Switzerland (Ref. 2018–02166/CE 3428). Blood from healthy donors was obtained from the Swiss Blood Donation Center of Basel and Lugano. All blood donors provided written informed consent for participation in the study. Serial blood donations from a healthy donor who underwent annual Influenza vaccination were obtained after written informed consent. Human primary cell protocols were approved by the Federal Office of Public Health (no. A000197/2 to F.S.). Peripheral blood mononuclear cells (PBMCs) were isolated with Ficoll‐Paque Plus (GE Healthcare). Monocytes and total CD4⁺ T cells were isolated by positive selection using anti‐CD14 and anti‐CD4 magnetic microbeads, respectively (Miltenyi Biotech). Total CD4⁺ cells obtained by positive selection were stained for chemokine receptors at 37°C and on ice for additional markers. Memory CD4⁺ T cells were sorted with a FACSAria III (BD Biosciences) to over 98% purity by gating on CD4⁺CD8^–CD14^–CD16^–CD19^–CD25^–CD56^–CD45RA^– cells, and different memory CD4⁺ T cell subsets were isolated as follows: CXCR5⁺CCR6⁺, CXCR5⁺CCR6^–, CXCR5^–CCR6⁺, CXCR5^–CCR6^–, or CXCR5⁺PD‐1⁺, CXCR5⁺PD‐1^–, CXCR5^–. In some experiments, after gating on CD4⁺CD8^–CD14^–CD16^–CD19^–CD25^–CD56^–CD45RA^– cells, cT_FH were sorted as total CXCR5⁺ cells, T_CM were sorted as CXCR5⁻CCR7⁺ cells, and T_EM were sorted as CXCR5^–CCR7^– cells. The following fluorochrome‐labeled mouse monoclonal antibodies were used for staining: CD4–APC (clone 13B8.2; cat. no. IM2468), CD45RA‐Qdot 655 (clone MEM‐56; cat. no. Q10069), CD45RA‐FITC (clone ALB11; cat. no. A07786), CD8‐PE‐Cy5 (clone B9.11; cat. no. A07758), CD16‐PE‐Cy5 (clone 3G8; cat. no. A07767), CD56‐PE/Cy5 (clone N901; cat. no. A07789), CD19‐PE/Cy5 (clone J3‐119; cat. no. A07771), CD25‐PE‐Cy5 (clone B1.49.9; cat. no. IM2646) from Beckman Coulter, CCR6‐BV421 (clone 11A9; cat. no. 562515), PD‐1‐BV785 (clone EH12.2H7; cat. no. 329929), Human CXCR5 (BLR‐1) MAb (Clone 51505; MAB190‐100), Goat Anti‐mouse IgG2b, human ads‐BIOT (SB Cat. No. 1090‐08), Streptavidin PE‐CF594 (Cat. No. 562284), CD4‐PE–Texas Red (clone S3.5; cat. no. MHCD0417) from ThermoFisher, PE/Cy7‐Streptavidin (cat. no. 405206), ICOS‐Pacific Blue (clone C398.4A; cat. no. 313521) from BioLegend.

Microbes and antigens

The influenza seasonal vaccines Influvac (Abbott, trivalent influenza vaccine) and Inflexal V (Crucell, trivalent influenza vaccine) were used at a concentration of 5 μg/ml for stimulation assays. Candida albicans SC5314 strain was processed in house. Briefly, C. albicans was cultured in YPD medium for 16 hours at 30°C, extensively washed in PBS, and heat inactivated at 65°C for 30 minutes. Ratio used for stimulation assays was three particles per monocyte. Lysate was prepared from mixed cultures of conidia and hyphae. Tetanus toxoid was obtained from Novartis Vaccines (Siena, Italy) and used at a concentration of 10 μg/ml for stimulation assays.

T cell stimulation

T cells were cultured in RPMI 1640 medium supplemented with 2 mM glutamine, 1% (vol/vol) nonessential amino acids, 1% (vol/vol) sodium pyruvate, penicillin (50 U/ml), streptomycin (50 μg/ml) (all from Invitrogen) and 5% human serum (Swiss Red Cross). Sorted memory CD4⁺ T cell subsets were labeled with carboxyfluorescein succinimidyl ester (CFSE) and cultured at a ratio of 2:1 with untreated or antigen‐pulsed autologous monocytes. Before co‐culture, autologous monocytes were irradiated with 45 Gy and pulsed for 3–5 hrs at 37°C with C. albicans, influenza vaccine, and tetanus toxoid. Antigen‐stimulated memory T cell cultures were collected at day 6–7 and stained with antibodies to CD25‐PE (clone M‐A251; cat. no. 555432) from BD Biosciences and ICOS‐Pacific Blue (clone C398.4A; cat. no. 313521 from BioLegend). T cell proliferation was determined by measuring the percentage of CFSE^low cells by flow cytometry followed by analysis with the FlowJo software. Proliferating memory T cells were FACS‐sorted as CFSE^lowCD25⁺ICOS⁺ and expanded in vitro in the presence of IL‐2 (50 IU/ml) for subsequent experiments. Before sorting, cell culture supernatants of antigen‐stimulated memory T cells were collected and preserved. Cytokine and chemokine concentration in the culture supernatants was assessed by Luminex bead‐based multiplex assay according to manufacturer's instructions. In preliminary experiments, the frequency of antigen‐specific T cells in the CFSE^low fraction was determined by performing single‐cell cloning and analyzing the ability of the T cell clones to proliferate upon secondary antigenic stimulation and found to be between 90% and 100%.

TCR Vβ deep sequencing

Ex vivo‐sorted memory CD4⁺ T cell subsets and CFSE^low fractions of antigen‐stimulated memory CD4⁺ T cell cultures (2.5‐5×10⁵ cells) were analyzed by deep sequencing. In brief, cells were centrifuged and washed in PBS, and genomic DNA was extracted from the pellet using QIAamp DNA Micro Kit (Qiagen), according to manufacturer's instructions. Genomic DNA quantity and purity were assessed through spectrophotometric analysis. Sequencing of TCR Vβ CDR3 was performed by Adaptive Biotechnologies using the ImmunoSEQ assay (http://www.immunoseq.com). In brief, following multiplex PCR reaction designed to target any CDR3 Vβ fragments, amplicons were sequenced using the Illumina HiSeq platform. Raw data consisting of all retrieved sequences of 87 nucleotides or corresponding amino acid sequences and containing the CDR3 region were exported and further processed. The assay was performed at deep level for ex vivo‐sorted memory CD4⁺ T cell subsets (detection sensitivity, 1 cell in 200,000) and at survey level for CFSE^low antigen‐reactive cultures (detection sensitivity, 1 cell in 40,000). Each clonotype was defined as a unique productively rearranged TCR Vβ nucleotide sequence; data processing was done using the productive frequency of templates provided by ImmunoSEQ Analyzer V.3.0 (http://www.immunoseq.com). The overlap between pairs of TCR Vβ repertoires was quantified by calculating the following indexes. The percentage of shared clonotypes was calculated using the Jaccard index [J = (A∩B)/(A∪B)] as number of shared clonotypes between two subsets divided by the total number of clonotypes present in the same subsets, and normalized by 100. Cumulative frequencies of shared clonotypes were calculated as the average of the cumulative frequencies of the shared clonotypes in each of the two subsets.

To take into account the experimental variability due to the repertoire diversity and undersampling of lower abundant clonotypes, a normalized TCR overlap score based on technical replicates was calculated for each pair of antigen‐specific TCR Vβ repertoires as follows. To measure the overlap between pairs of antigen‐specific TCR Vβ repertoires, the Chao‐Jaccard (CJ) similarity index was calculated using R package “fossil”. For each pairwise comparison of different T cell subsets, a normalization factor was defined as the lowest CJ value measured between the relating pairs of technical replicates. Lastly, CJ index calculated between pairs of different antigen‐specific TCR Vβ repertoires was divided by the normalization factor, to obtain a normalized TCR overlap score. In formula, if A_n and B_n are the TCR Vβ repertoires sequenced from two different T cell subsets A and B, and if A₁ and A₂, and B₁ and B₂ are the corresponding technical replicates:

A normalized TCR overlap score between each pair of A_n and B_n TCR Vβ repertoires is calculated using Chao‐Jaccard (CJ) similarity index, as follows:

$$\begin{array}{c}{\displaystyle \mathrm{Normalized}\mathrm{TCR}\mathrm{overlap}\mathrm{score}=\frac{\mathrm{CJ}\left({\mathrm{A}}_{\mathrm{n}}-{\mathrm{B}}_{\mathrm{n}}\right)}{\min \left[\mathrm{CJ}\left({\mathrm{A}}_1-{\mathrm{A}}_2\right),\mathrm{CJ}\left({\mathrm{B}}_1-{\mathrm{B}}_2\right)\right]}}\hfill \end{array}$$

Gene expression analysis

Antigen‐reactive cells from each subset were sorted 24 h post‐stimulation with antigens based on the upregulation of CD40L according to the protocol [26] and subjected to transcriptional analysis. Cell input from each sample was adjusted within the range from 6500 to 10,000 cells/sample. RNA was extracted and the expression levels for a total of 594 genes were measured by NanoString technology using nCounter Immunology Panel (Human_V2) (https://nanostring.com/products/ncounter-assays-panels/immunology/immunology-panel) according to the manufacturer's instructions. In brief, 8 μl of the hybridization Master Mix (proteinase K added with working concentration at 200 μg/ml), including the Reporter CodeSet, was distributed in 12 tubes containing 4.5 μl cell lysate from each sample. Subsequently, 2 μl of the Capture ProbeSet and 0.5 μl dH₂O were added and, after mixing samples, the hybridization reaction was run at 65°C for 20 h. After the incubation, samples were loaded on the Nanostring cartridge lanes following the manufacturer's instructions. The assay was run on nCounter SPRINT Profiler device. Data were filtered by background thresholding and normalized by the nSolver analyzer software based on positive controls, negative controls, and housekeeping gene controls.

Statistical analysis

Statistical analysis was mainly performed using Prism software 8.0 (GraphPad). Significance was assigned at p‐value < 0.05 unless stated otherwise. Specific tests are indicated in the figure legends for each comparison. Calculation of Chao‐Jaccard similarity index and dendrogram analysis between TCR Vβ repertoires were performed using R software version 3.5.1. Network graphs showing the clonal relationship were performed using Cytoscape software.

Conflict of interest

The authors declare no commercial or financial conflict of interests.

Ethics Approval

The study was approved by the Ethical committees of Cantone Ticino, Switzerland (Ref. 2018–02166/CE 3428). Blood from healthy donors was obtained from the Swiss Blood Donation Center of Basel and Lugano. All blood donors provided written informed consent for participation in the study. Serial blood donations from a healthy donor who underwent annual Influenza vaccination were obtained after written informed consent. Human primary cell protocols were approved by the Federal Office of Public Health (no. A000197/2 to F.S.).

Author contributions

M.H., A.C., and F.S. were associated with conceptualization, methodology, and project administration. M.H., A.C., S.N., M.F., S.J., F.M., and D.J. performed Investigations. M.H., A.C., M.F., and F.S. performed visualization. F.S. acquired funding. A.C., A.L., and F.S. supervised the study. M.H. and A.C. wrote the original draft. A.C., A.L., and F.S. reviewed and edited the final manuscript.

Peer review

The peer review history for this article is available at https://publons.com/publon/10.1002/eji.202250190

Supporting information

Supplementary material

Data availability statement

The data presented in this manuscript are included in the paper and in the Supplementary Information. TCR Vβ sequences have been deposited in the ImmuneAccess database (DOI: 10.21417/MH2022EJI).