|Home | About | Journals | Submit | Contact Us | Français|
CD4 T cell help is critical for the generation and maintenance of germinal centers (GCs), and T follicular helper (TFH) cells are the CD4 T cell subset required for this process. Signaling lymphocytic activation molecule (SLAM)-associated protein (SAP [SH2D1A]) expression in CD4 T cells is essential for GC development. However, SAP-deficient mice have only a moderate defect in TFH differentiation, as defined by common TFH surface markers. CXCR5+ TFH cells are found within the GC, as well as along the boundary regions of T/B cell zones. In this study, we show that GC-associated T follicular helper (GC TFH) cells can be identified by their coexpression of CXCR5 and the GL7 epitope, allowing for phenotypic and functional analysis of TFH and GC TFH populations. GC TFH cells are a functionally discrete subset of further polarized TFH cells, with enhanced B cell help capacity and a specialized ability to produce IL-4 in a TH2-independent manner. Strikingly, SAP-deficient mice have an absence of the GC TFH cell subset and SAP− TFH cells are defective in IL-4 and IL-21 production. We further demonstrate that SLAM (Slamf1, CD150), a surface receptor that uses SAP signaling, is specifically required for IL-4 production by GC TFH cells. GC TFH cells require IL-4 and -21 production for optimal help to B cells. These data illustrate complexities of SAP-dependent SLAM family receptor signaling, revealing a prominent role for SLAM receptor ligation in IL-4 production by GC CD4 T cells but not in TFH cell and GC TFH cell differentiation.
Long-term humoral immunity is provided by long-lived plasma cells and memory B cells, both of which arise from the germinal center (GC) reaction and are critical components of protective immunity to many pathogens (1–5). GCs require CD4 T cells, and T follicular helper (TFH) cells are the CD4 T cell lineage specialized in B cell help (6, 7). A central property of TFH cells is their ability to migrate to the B cell follicle, allowing them to provide help to B cells to initiate and maintain the GC response. Expression of the chemokine receptor CXCR5 allows TFH cells to home to the B cell follicle (8–10). We and other investigators recently demonstrated that Bcl6 is the master regulator of TFH cell differentiation (11–13). In the absence of TFH cells, as a result of the absence of Bcl6 (11–13) or the presence of B lymphocyte-induced maturation protein (Blimp-1) (11), GC responses are lost, as is most T-dependent Ab production.
SAP (SLAM-associated protein), a small Src homology domain 2-domain adaptor protein encoded by the gene SH2D1A, is integral to T cell-dependent humoral immunity. Like TFH cells, SAP is required for GC formation (14). GCs are absent in SAP− mice postinfection with lymphocytic choriomeningitis virus (LCMV) (14), influenza (15), Schistosoma mansoni eggs (16), or vaccinia virus (17), as well as after protein immunizations (16, 18). SAP expression in CD4 T cells is required for GC responses (14–16, 19). SAP is likely not needed in B cells (19), although this is disputed (20). Studies also showed important roles for SAP in NKT cell development (21, 22), CD4 T cell TH1/TH2 differentiation (23–25), and regulation of NK cell killing (26).
The essential role of SAP in T cell-dependent humoral immunity is conserved in humans. SH2D1A mutations result in the lethal immunodeficiency X-linked lymphoproliferative disease (27). Patients with this disease have severe GC and memory B cell deficiencies (28–30).
The SLAM family of receptors, which bind SAP through their immunotyrosine switch motif, are expressed on T cells, B cells, and other hematopoietic cells (27, 31, 32). The SLAM family consists of nine receptors closely grouped on chromosome 1 in mice and humans, most of which (with the exception of 2B4) form homophilic-binding interactions (27). SLAM (CD150) is the prototypic member of the family. SLAM family receptors are an important genetic susceptibility locus for autoantibody production and systemic lupus erythematosus in mice and humans (33, 34).
SAP has a major role in T cell–B cell adhesion (35). In the absence of SAP expression in CD4 T cells, conjugates between cognate CD4 T cells and B cells were short-lived in vitro and in vivo (35), although the development of CXCR5hiICOShi putative TFH CD4 T cells was not significantly diminished (35). It has not been known which of the SAP-binding SLAM family receptors is required for T cell help to B cells (27, 35) or whether adhesion is the only role for SAP and the SLAM family receptors in the GC.
To better understand the generation of potent GC and Ab responses, we took the complementary approach of studying the roles of SAP and SLAM in the development and function of TFH cells. We found a phenotypically and functionally distinct sub-population of TFH cells that resides within GCs. These GC-resident TFH cells (GC TFH cells) produce IL-4 and are absent in SAP-deficient mice. GC TFH cell production of IL-4 is dependent on SLAM. Furthermore, IL-4 and IL-21 production by GC TFH cells is required for optimal Ab production.
C57BL/6J (B6) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). SAP− (36), OTII CD45.1, SLAM−/−SAP−, and SMARTA TCR transgenic (SMtg; LCMV gp66-77 I-Ab specific) (37) CD45.1+ mice were all on a fully B6 background and were bred at the La Jolla Institute for Allergy and Immunology. B6.Sle1b mice were a gift of Edward Wakeland (UT Southwestern Medical Center, Dallas, TX) (38). SLAM mice were generated by Y. Yunagi on the 129 background (24) and backcrossed for 11 generations to the B6 background at Kyushu University and the La Jolla Institute for Allergy and Immunology. Whole-genome microsatellite analysis, through the University of California, Los Angeles, Southern California Genotyping Consortium, verified that the SLAM−/− mice were 99% B6. The remaining 1% was of the Sv129 background around the SLAM locus, incorporating the region between the single nucleotide polymorphism markers mCv22849619 and rs13476259. SAP− mice were >99% B6 by single nucleotide polymorphism analysis, with a small region of the X chromosome remaining Sv129. All animal experiments were conducted in accordance with approved animal protocols.
Cell transfers into host mice were performed by i.v. injection via the retro-orbital sinus. SMtg CD4 T cells were transferred so that each mouse received 5,000 SMtg CD4+ T cells (~12,000 total splenocytes, as determined for each experiment by FACS). For mixed-cell experiments, 25 × 103 wild-type (WT) SMtg cells and 25 × 103 SAP− SMtg cells were transferred into a common CD4-depleted host (50 × 103 SMtg cells/mouse); CD4 depletions were done as previously described (39). LCMV stocks were prepared and quantified as described (39). All infections were done by i.p. injection of 1–2 × 105 PFU LCMV Armstrong per mouse. For OTII CD4+ T cell transfers, naive CD4+ T cells were purified from spleen by negative selection using magnetic beads (Miltenyi Biotec, Auburn, CA); 250 × 103 OTII cells were transferred per mouse. NP-OVA/alum was prepared by mixing NP(19)-OVA (Vector Laboratories, Burlingame, CA) in PBS with alum (Pierce, Rockford, IL) at a 3:1 ratio for 60 min at 4°C. NP-OVA/alum immunizations consisted of 100 μg given i.p.
Single-cell suspensions of spleen were prepared by standard gentle mechanical disruption. Surface staining for flow cytometry used mAbs to B220 (RA3-6B2), GL7 (FITC [BD Pharmingen, San Diego, CA] or purified [eBioscience]) and Fas (BD Pharmingen). Also used were mAbs to CD4 (LT34 and GK1.5), B and T lymphocyte attenuator (BTLA) (6F7), ICOS (7E.17G9), programmed death 1 (PD-1) (J43), CD45.1 (A20), CD45.2 (104), CD62L (MEL-14), IFN-γ (XMG1.2), and CD44 (IM7) (all from eBioscience). Abs against SLAM (TC15-12F12.2) were from BioLegend. FITC-labeled peanut agglutinin was from Vector Laboratories. CXCR5 staining was done using purified anti-CXCR5 (BD Pharmingen), followed by biotinylated anti-rat IgG (Jackson ImmunoResearch Laboratories, West Grove, PA), and then PE- or allophycocyanin-labeled streptavidin (Caltag Laboratories, Burlingame, CA), with each staining step done in PBS + 0.5% BSA + 2% FCS + 2% normal mouse serum on ice; samples were acquired without fixation. Doublets were gated out during FlowJo analysis to exclude possible B–T cell conjugates. For intracellular cytokine staining, splenocytes were stimulated with PMA and ionomycin for 4 h, followed by surface staining. Intracellular staining was performed with Abs against IFN-γ (BD Pharmingen). Anti-SAP mAb(12C4) was a gift of Andre Veillette (Clinical Research Institute of Montreal, Montreal, QC, Canada) and was used as previously described (19) after direct conjugation to Alexa Fluor 647, according to the kit instructions (Invitrogen).
Standard immunofluorescence histology was performed, as described previously (39), using an upright microscope (DM6000B) with an SP5 confocal head (Leica Microsystems, Deerfield, IL). The following Abs were used: GL7 FITC rat anti-mouse (BD Pharmingen), B220 Alexa Fluor 647 (RA3-6B2; eBioscience), and CD45.1 biotin (A20; eBioscience) with Streptavidin Alexa Fluor 555 (Invitrogen).
Based on histology 5–10% of GC cells are CD4 T cells. Mice have ~7.5 × 105 virus-specific GC B cells and 2.9 × 105 virus-specific TFH cells at day 8 after LCMV infection, as measured by flow cytometry. Therefore, the quantity of virus-specific TFH cells was estimated to exceed the number of CD4 T cells inside of GCs by a factor of 3:1.
Splenocytes were isolated, and Ag-specific CD4 T cells were enriched using anti-CD45.1–FITC and magnetic bead purification (Miltenyi Biotec). WT and SAP− TFH cells and non-TFH CD4+CD45.1+TCRβ+CD19−7AAD− cells were sorted on the basis of CXCR5 expression (CXCR5hi versus CXCR5lo), using a FACSAria (BD Biosciences, San Jose, CA). TFH cell gating was confirmed by CXCR5hi and SLAMlo costaining on an aliquot of cells. Approximately 1 × 106 cells from each condition (in duplicate) were sorted directly into RNA later (Ambion, Austin, TX). Naive SMtg CD4 T cells were obtained from intact SMtg mice, sorting for CD4+CD45.1+CD44loCD62Lhi7AAD−. RNA was isolated using RNeasy Mini spin columns (Qiagen, Valencia, CA), including Qiashredder and on-column digestion of genomic DNA. Then, some RNA samples were concentrated using MinElute spin columns (Qiagen).
RNA quality of all samples was confirmed by BioAnalyzer Nano gel (Agilent, Palo Alto, CA), and then probes were generated by single round linear amplification using the Ovation Pico system (Nugen) and used on Affymetrix 430 2.0 chips. Data were analyzed using Genespring (Agilent), Microsoft Excel (Microsoft, Redmond, WA), and Prism 5.0 (GraphPad, San Diego, CA). Raw microarray signal data have been submitted to National Center for Biotechnology Information GEO database (www.ncbi.nlm.nih.gov/geo/) under accession numbers GSE21379 and GSE21381.
cDNA synthesis was performed using SuperScript II Reverse Transcriptase (Invitrogen) with oligonucleotide dT and random hexamer primed reactions, which were then pooled. Quantitative PCR (qPCR) reactions were performed in triplicate using iTaq SYBR Green Supermix with ROX (Bio-Rad, Hercules, CA) on a Roche LightCycler 480 (Roche, Mannheim, Germany). Primers were as described (11). For GATA3, two independent primer sets were used for confirmation: one primer set within the GATA3 coding region and one primer set within the GATA3 3′ untranslated region. β-actin was used as the reference for normalization, and expression levels were then normalized to the naive CD4 T cell control.
For Figs. 1, ,2,2, and and4,4, CD4 T cells were purified by negative selection using magnetic beads (Miltenyi Biotec). Enriched cells were sorted as CD4+CD44hi7AAD− and on the basis of CXCR5 and GL7 expression: non-TFH cells (CXCR5loGL7−), TFH cells (CXCR5+GL7−), and GC TFH cells (CXCR5+GL7+). RNA isolation, cDNA synthesis, and qPCR analysis were performed as described above. Raw microarray signal data have been submitted to National Center for Biotechnology Information GEO database under accession numbers GSE21380 and GSE21381.
For Fig. 7, purified SLAM−/− and SLAM+/+ CD4 T cells were sorted in triplicate (each replicate consisting of four pooled SLAM−/− or SLAM+/+ spleens) into non-TFH, TFH, and GC TFH populations for qPCR analysis.
Serum was used from B6 mice 30 d after LCMV infection. Anti-LCMV IgG1 and IgG2c were quantified by ELISA using LCMV-infected cell lysate as the capture Ag. Ninety-six–well polysorp microtiter plates (Nunc, Rochester, NY) were coated overnight with LCMV-infected cell lysate in PBS. Following incubation of sample serum, HRP-conjugated goat anti-mouse IgG1 and IgG2c (Jackson ImmunoResearch Laboratories) were used for detection.
Purified CD4+ T cells from B6 mice 8 d post LCMV infection were sorted into GL7hiCXCR5hi, GL7loCXCR5hi, and GL7loCXCR5lo populations. T cells (5 × 104) were cultured for 7 d with purified B cells (5 × 105) in the presence of IL-2 and β-ME. We used the blocking Abs αIL-4 and αIL-21 at 20 and 40 ng/ml, respectively. Supernatants were collected, and total IgG was quantified by ELISA. Goat anti-mouse IgG+IgA+IgM (Caltag Laboratories) was the capture Ab, biotinylated goat anti-mouse IgG (Caltag Laboratories) was the secondary Ab, and HRP Avidin D was added thereafter (Vector Laboratories). For Fig. 8C, purified CD4+ T cells from SLAM−/− and SLAM+/+ mice 8 d post-LCMV infection were sorted for GL7hiCXCR5hi populations. T cells (15 × 104) were cultured for 7 d with purified B cells (5 × 105) in the presence of IL-2 and β-ME. Quantification was done by ELISA, as described above.
Statistical tests were performed using Prism 5.0, and p values were calculated using two-tailed unpaired Student t tests, with a 95% confidence interval. Error bars represent the SEM. The χ2 test (95% confidence) was used to analyze GC CD45.1 CD4 T cell counts by histology versus GL7+ and GL7neg CD45.1 CD4 T cell counts by flow cytometry. A two-tailed Mann–Whitney U test (95% confidence) was used for Fig. 8A.
Murine TFH cells are most frequently defined as a singular population (7), but subpopulations have also been proposed (40, 41). We identified Ag-specific TFH cells after an acute LCMV infection as CXCR5+ CD4 T cells that coexpress a set of TFH surface markers (ICOShiPD-1hiSLAMloBTLAhiCD200hi) validated by our laboratory for transgenic (Fig. 1A, 1B) (11) and nontransgenic CD4 T cells (11) in the context of a viral infection and validated by other laboratories in the context of other infections or immunizations (42–46). TFH cells are required for GC formation in vivo after viral infection (11). However, only ~20% of virus-specific TFH cells were inside GCs (see Materials and Methods for calculations). Therefore, we examined whether the CD4 T cells residing within the GC could be phenotypically identified as a specific sub-population of TFH cells, similar to the model proposed by McHeyzer-Williams et al. (41).
We performed a search for markers of GC TFH cells and identified GL7 as a specific surface marker (Fig. 1C). A subpopulation of TFH cells in LCMV-infected mice stained with the GL7 mAb, whereas the majority of TFH cells did not (Fig. 1C). A similar GL7+ CXCR5hi PD-1hi staining pattern was seen within CD44hi OTII CD4 T cells after NP-Ova immunization (Fig. 5D, left panel). CD4 T cells within GCs stained positive with GL7 by histology (Fig. 1E), and GL7+ Ag-specific (CD45.1+) CD4 T cells were not present outside of GCs (0 versus 2372; p < 0.0001 by χ2 test) (Fig. 1F).
GL7 is a rat mAb that recognizes α2,6-linked N-acetylneuraminic acid (Neu5Ac) on glycan chains (47, 48). Intriguingly, Neu5Ac is highly expressed on GC B cells, and GL7 has long been used as a marker for GC B cells. Theoretically, Neu5Ac+ (GL7+) CD4 T cells could acquire the Neu5Ac sugar from the surface of adjacent Neu5Ac+ GC B cells. We examined the capacity of GC TFH cells to synthesize Neu5Ac. Cmah is the enzyme that converts Neu5Ac to the more abundant sugar Neu5Gc, which is not bound by GL7 (48). A decreased expression of Cmah results in higher levels of Neu5Ac (48). Gene-expression analysis of naive, non-TFH, TFH, and GC TFH cells revealed that GC TFH cells had lower Cmah expression than the other CD4 T cell populations (p < 0.0001) (Fig. 1G). Therefore, the increased Neu5Ac GL7 epitope on GC TFH cells is the result of metabolic changes within the CD4 T cells. Neu5Gc is a potent CD22 ligand that strongly inhibits B cell activation (48–50). The absence of Neu5Gc on GC B cells (48) and GC TFH cells (Fig. 1) eliminates the tonic negative signaling through CD22 during GC T–B interactions, thereby likely providing an elegant mechanism to enhance GC B cell proliferation.
After protein immunizations, Cyster and colleagues (10) identified a CXCR5+PD-1int TFH cell population and a higher PD-1–expressing GC-associated T cell population. In studying the CD4 T cell response to LCMV, we were unable to distinguish these two TFH-related populations based on PD-1 and CXCR5 expression alone (Fig. 1B), likely as a result of the broad upregulation of PD-1 on T cells after a viral infection (Fig. 1D) (51). Backgating analysis revealed that PD-1 expression was highest within the GL7+ GC TFH population (Fig. 1D).
Having identified GC TFH cells and non-GC TFH cells as distinguishable populations of Ag-specific CD4 T cells, distinct from non-TFH cells, we assessed the phenotypic properties of these cells. Assessment of the expression of the TFH master regulator transcription factor Bcl6 and its antagonist Blimp-1 (11) indicated that GC TFH cells are a further polarized state of TFH differentiation (Fig. 1H, 1O). Expression of IL-21, PD-1, ICOS, CXCR5, BTLA, and CD200 were all increased in TFH cells versus non-TFH cells, and expression of these TFH-associated molecules was further increased in GC TFH cells (Fig. 1I–N), corroborating that GC TFH cells are a further polarized state of TFH cell differentiation. An exception to this pattern was IL-4, which was uniquely expressed by GC TFH cells but not GL7− TFH cells (Fig. 1P).
Although IL-4 was reported to be expressed by TFH or GC-associated CD4 T cells, those findings were in the context of TH2-biased infections or immunizations (46, 52, 53). The observation that GC TFH cells produced IL-4 after an LCMV infection was surprising, because CD4 T cell responses to viral infections, including LCMV, exhibit a strong TH1 bias (11, 54–57). It was also intriguing that the expression was exquisitely selective to the GC TFH cells. This indicated that GC TFH cells are a specific differentiation state of TFH cells, and the ability to produce IL-4 is a delineating feature of GC TFH cells, independent of the type of immunization or infection.
Interestingly, the expression of IL-4 by the GC TFH population was TH2 independent, because GATA3 expression was not increased above basal levels in any virus-specific CD4 T cells, including the GC TFH subpopulation (Fig. 2). C-Maf expression was also unchanged among non-TFH, TFH, and GC TFH cells (Fig. 2G). Consistent with the absence of GATA3, TH2 cytokine IL-5 mRNA was undetectable (data not shown).
The strong TH1 bias of the antiviral immune response was confirmed by several experimental approaches. The majority (95%) of CXCR5−/lo (non-TFH) virus-specific CD4 T cells produced IFN-γ (Fig. 2A, 2B). Most (57%) CXCR5hi TFH cells also produced IFN-γ (Fig. 2A, 2B). Both populations also expressed high levels of T-bet (non-TFH cells versus naive CD4 T cells, 400-fold, p < 0.001; TFH cells versus naive, 100-fold, p < 0.001) (Fig. 2C). Consistent with the expression of IFN-γ by the non-TFH and TFH populations, GC TFH cells (CXCR5hiGL7+) also expressed substantial levels of IFN-γ (Fig. 2D). The Ab response to LCMV is highly dominated by IgG2c (Fig. 2E), consistent with the heavily TH1-biased IFN-γ–producing antiviral CD4 T cell response.
To directly demonstrate IL-4 production by GC TFH cells, we used IL-4 GFP reporter mice. After LCMV infection, IL-4 production was observed by CXCR5+ but not CXCR5−/lo virus-specific CD4 T cells (Fig. 3A). GL7+CXCR5+ CD4 T cells were the primary IL-4 producers (p < 0.0001) (Fig. 3B). No IL-4 was made by non-TFH CD4 T cells, again confirming that the IL-4 production was TFH specific and not TH2 specific (Fig. 3A, 3C). The majority of the IL-4 was produced by the highest CXCR5-expressing cells, even within the TFH population (p = 0.0002) (Fig. 3C). This corroborates the data that GC TFH cells coordinately express the highest levels of GL7, Bcl6, PD-1, CXCR5, and IL-4 (Fig. 1) and correlates with human data showing that GC TFH cells have the highest level of CXCR5 (12, 58).
SAP mRNA expression was increased in TFH and GC TFH cells (p < 0.0001) (Fig. 4A). These data correlated well with data showing increased expression of SAP in human tonsillar TFH cells compared with other T cell subsets (59, 60). We also examined SAP proteins levels in CD4 T cells and found graded increases from naive cells to non-TFH cells to TFH cells to GC TFH cells, with SAP expression elevated 4-fold in GC TFH cells over naive CD4 T cells (p < 0.0001) (Fig. 4B, 4C).
Given that SAP expression by CD4 T cells is a requirement for GC development (14), we examined whether TFH cell differentiation was dependent on SAP. Up to a 50% reduction in SAP− TFH cell differentiation was observed in vivo for SMtg CD4 T cells responding to an LCMV infection (p = 0.0013) (Fig. 4G). Several independent TFH cell markers were used, including ICOS, PD-1, and BTLA (Fig. 4D–F). Overall, CD4 T cell activation and proliferation were normal, because total numbers of SAP− and WT SMtg CD4 T cells were equivalent (data not shown, Supplemental Fig. 1). A partial reduction in TFH cell frequency was also seen in the polyclonal (CD44hi) LCMV-specific CD4 T cell response in SAP− mice (28%; p < 0.05) (Fig. 4H). TFH cell differentiation is unusual in that it is heavily dependent on the presence of cognate B cells (6, 10, 11). Because B cell defects can cause TFH cell deficiencies, we used WT:SAP mixed adoptive transfers to confirm that the SAP− TFH cell differentiation defect was CD4 T cell intrinsic (p < 0.0001) (Supplemental Fig. 1).
Although TFH cell differentiation was reduced in SAP− mice, this ≤2-fold reduction in TFH cell frequency failed to account for the profound impairment of GC B cell development (Fig. 4I). Furthermore, no loss of TFH cell (CXCR5+) differentiation was observed in a second model, protein-immunization experiments with SAP− OTII (Fig. 4J, 4K), indicating that there is not a universal requirement for SAP in TFH cell differentiation. This is consistent with the study of Qi et al. (35), which was published while this work was ongoing.
In light of the identification of the GL7+ GC-associated GC TFH subset of TFH cells, as well as the high levels of SAP observed in GC TFH cells (Fig. 4B, 4C), we hypothesized that GC TFH cell differentiation may be impaired in SAP-deficient mice. We found a striking 4–7-fold reduction in the GC TFH cell population in SAP− mice after LCMV infection (p < 0.0001) (Fig. 5A, 5B). A comparable reduction in SAP− GC TFH cell differentiation was observed after protein immunization (p < 0.0001) (Fig. 5C–E). An almost complete absence of GC TFH cells (GL7+CXCR5hi) also correlated with the absence of the very highest PD-1–expressing CXCR5hi CD4 T cells (Fig. 5F).
We examined the effects of SAP deficiency on TFH cell gene expression. Global gene-expression analysis comparing purified SAP− TFH cells (CXCR5hi) versus WT TFH cells (CXCR5hi) 8 d after LCMV infection (Fig. 6A) revealed that SAP− TFH cells had predominantly normal TFH cell gene expression (WT versus SAP−, R2 = 0.9905) (Fig. 6B), including normal expression levels of the TFH master regulator transcription factor Bcl6 (Fig. 6C), downregulation of the TFH differentiation antagonist Blimp-1 (Fig. 6D), and upregulation of the TFH-related genes ICOS and CD200 (Fig. 6E, 6F). A 2-fold reduction in IL-21 was seen (Fig. 6G). Strikingly, IL-4 expression was completely lost in SAP− TFH cells (Fig. 6H). qPCR confirmed that IL-4 mRNA levels in SAP− TFH cells were reduced to <5% of that produced in WT TFH cells (p < 0.003). These data matched our earlier observation that GC TFH cells, but not TFH cells, produce IL-4 (Fig. 1P). Because SAP− CD4 T cells fail to differentiate into GC TFH cells, the absence of IL-4 and a reduction in IL-21 production by SAP− TFH cells are interpreted as a direct result of the lack of the GC TFH subpopulation.
SLAM is expressed on all CD4 T cells (39) and is upregulated on activated CD4 T cells (24, 39, 61–63), and we observed SLAM expression on TH1 and TH2 CD4 T cells in vitro (Fig. 7A). After LCMV infection, SLAM is upregulated on non-TFH CD4 T cells but not TFH CD4 T cells (11) (Fig. 7B). When virus-specific GC TFH cells were examined, an increase in SLAM expression from GL7− TFH cells was noted (Fig. 7D, 7E). SLAM can be involved in IL-4 production in a SAP-dependent manner (23, 24). SLAM is a self-associating receptor and is expressed on B cells. SLAM expression is substantially upregulated on GC B cells (Fig. 7F). Therefore, SLAM–SLAM interaction likely occurs between GC B cells and GC TFH cells, allowing for adhesion and bidirectional signaling. We hypothesized that interaction may induce GC TFH cell differentiation or IL-4 production.
We tested whether SLAM–SAP signaling regulates GC TFH cell differentiation or function, using SLAM-deficient mice (slamf1−/−). Given that all SLAM family receptors are closely located on chromosome 1 and have one or more roles in autoimmunity (27, 33, 34, 64), it was necessary to use 129Sv SLAM family locus (sle1b) B6 congenic mice as the appropriate Slamf1+/+ control (34, 65). We analyzed Slamf1−/− versus Slamf1+/+ CD4 T cells for TFH cell and GC TFH cell differentiation in vivo and found normal frequencies of Slamf1−/− TFH cells and GC TFH cells (Fig. 7G–I). Therefore, although SAP was necessary for the differentiation of GC TFH cells SLAM was not. This was confirmed using SLAM−/−SAP− double-deficient mice, which showed that SAP is epistatic to SLAM for GC TFH cell differentiation and GC development (Supplemental Fig. 2). Therefore, SAP is required for positive signaling and not inhibition of an alternative fate-determination signal through SLAM.
Although Slamf1−/− GC TFH cells were present, IL-4 production by Slamf1−/− GC TFH cells was markedly reduced (77% loss compared with WT; p = 0.0014) (Fig. 7J). As was observed for WT GC TFH cells, no GATA3 mRNA induction was observed in Slamf1−/− GC TFH cells (p 0.05 versus naive CD4 T cells). IL-21 production was normal (Fig. 7K). These Slamf1−/− CD4 T cell data demonstrate that SLAM signaling is required for IL-4 production by GC TFH cell differentiation in a TH2-independent manner.
The production of IL-4 by GC TFH cells led us to hypothesize that IL-4 is required for optimal GC TFH cell function and B cell help. We examined whether GL7− TFH cells and the IL-4–producing GL7+ GC TFH subpopulation provided quantitatively different amounts of B cell help. Purified GL7− TFH cells provided efficient help to cocultured B cells in vitro, inducing significantly greater levels of IgG production than non-TFH cells (p < 0.01) or no T cells (p < 0.001) (Fig. 8A). GC TFH cells were even more potent than TFH cells at stimulating IgG production (p < 0.01) (Fig. 8A). Blocking IL-21 eliminated IgG production (p < 0.001), confirming the critical role of IL-21 in plasma cell differentiation (Fig. 8B). Importantly, blocking IL-4 in the GC TFH–B cell cocultures decreased the levels of secreted IgG by 75% (p < 0.04) (Fig. 8B), a level comparable to that of GL7− TFH–B cell cocultures. These results illustrate that GC TFH cells produce physiologically meaningful levels of IL-4, and TFH and GC TFH cells are functionally distinct.
Based on those findings, we hypothesized that if SLAM engagement is a primary regulator of IL-4 production by GC TFH cells, Slamf1−/− GC TFH cells should exhibit defective B cell help. We tested this using sorted Slamf1−/− GC TFH cells and congenic Slamf1+/+ GC TFH cell controls cocultured with B cells. B cells incubated with SLAM-deficient GC TFH cells produced 74% less IgG than B cells incubated with WT GC TFH cells (p < 0.0001) (Fig. 8C). Slamf1−/− GC TFH cells produced levels of IL-21 comparable to WT cells (Fig. 7K), and the defective Slamf1−/− GC TFH B cell help can be fully accounted for by the decrease in IL-4 production, given that SLAM deficiency or blockade of IL-4 resulted in a comparable loss of GC TFH B cell help activity (Fig. 8B, 8C, Fig. 9).
In this study we report four findings of note: GC TFH cells are a phenotypically and functionally distinct further differentiated subpopulation of TFH cells; the GC TFH cells are specialized producers of IL-4; SAP is not required for TFH cell differentiation, but is required for GC TFH cell differentiation and the resultant IL-4 production; and IL-4 production by GC TFH cells is dependent on SLAM.
TFH cells can outnumber GC resident CD4 T cells. Migration studies demonstrated that CXCR5+ CD4 T cells home to B cell follicles and the T/B border zones (8, 10) (Supplemental Fig. 3). We propose that CD4 T cells differentiate to TFH cells after interacting with cognate B cells at the T/B border zones, via B cell induction of Bcl6 expression in the CD4 T cells (6, 11). These TFH cells can provide B cell help at the T/B border zones, or they can migrate into GCs, becoming GC TFH cells and providing survival and differentiation signals to GC B cells. Signals from SLAM family receptor engagement may directly drive GC TFH cell differentiation in a SAP-dependent manner, or adhesion may only be required for prolonged T–B conjugate formation, with GC TFH cell differentiation induced by other signals provided by the cognate B cells. In most cases, with the exception of IL-4 production, GC TFH cells seem to be highly activated TFH cells [enhanced Bcl6, ICOS, and IL-21 (Fig. 1)], consistent with ongoing Ag recognition, which would be explained by their localization in GCs with cognate GC B cells presenting Ag. In addition, intravital microscopy studies indicated that CD4 T cells in GCs are not sessile (9) and regularly exit GCs (35). Incorporating those observations, we infer that GC TFH cells are most likely not terminally differentiated and GC TFH cells exchange with the bulk TFH cell population (Fig. 9). The gene expression and intravital microscopy data suggest that GC TFH cells are in a transient differentiation state induced by interaction with GC B cells. GC TFH cells exit the GC with some frequency and, thereby, presumably return to a GL7−Bcl6intIL-4− TFH state until they re-enter a GC and re-engage cognate B cells. This possibility remains to be experimentally tested. This model highlights the likely interrelatedness of TFH and GC TFH cells. The use of TFH and GC TFH nomenclature (41) seems most appropriate to describe these cell types, given their apparent close relationship by gene expression (Fig. 1) and their putative interconversion (Fig. 9).
The expression of SLAM on GC TFH cells and GC B cells suggests that GC TFH cells initiate SAP signaling though SLAM–SLAM engagement upon T–B conjugate formation, which, in turn, drives IL-4 production. Importantly, the presence of GC TFH cells in Slamf1−/− mice, but not SAP− mice, indicates that GC TFH cell differentiation, although dependent on SAP, is controlled by a SLAM family receptor other than SLAM. Alternatively, SLAM may contribute to GC TFH cell differentiation, with other SLAM family receptors sharing overlapping functions. Qi et al. (35) showed that SAP controls T cell–B cell adhesion. SLAM may participate in that adhesion function, but its role in IL-4 production seems to be distinct and nonoverlapping. Other SAP-binding members of the SLAM receptor family, such as CD84 and Ly108, form considerably higher-affinity homophilic interactions and are likely to be stronger contributors to SAP-mediated T cell–B cell adhesion (66–69). Indeed, it was recently shown that CD84 contributes to optimal SAP-dependent GC development (70). Nonetheless, our data show that SLAM’s role in IL-4 production by GC TFH cells is not redundant with other SLAM family receptors and that SLAM is the primary receptor responsible for driving IL-4 production by GC TFH cells during a viral infection (Figs. 7J, ,8C8C).
The severe GC defect phenotype seen in the absence of SAP has not been recapitulated in any single SLAM family receptor knockout mouse (24, 27, 31, 32, 70–73), including SLAM−/− (39). At least five SLAM family receptors are expressed on CD4 T cells. Functional redundancies in signaling among the SLAM family receptor members can occur, as seen in NKT cell development (22). Overlapping SAP signaling through SLAM family receptors likely occurs in TFH and GC TFH cells during CD4 T cell help to B cells, where normally different receptors have different functions, or are expressed at different stages of the immune response; however, sufficient overlap exists that the loss of any single SLAM family receptor is compensated for by the presence of other SLAM family receptors. Because they are encoded by adjacent genes, double or triple SLAM family mutants cannot be generated by breeding, and redundancies will have to be addressed with alternative approaches.
Non-TFH CD4 T cell expression of SAP is substantially lower than that of GC TFH cells. All CD4 T cells express SLAM (39). This suggests that the coordinated expression of SAP in GC TFH cells (Fig. 4B) is critical for SLAM-dependent IL-4 production (Fig. 9). Collectively, these data define a novel role for SLAM and SAP in GC TFH cell function and development while also providing new insights into the complexity of SAP-dependent SLAM family receptor signaling.
What is the key cytokine produced by TFH? IL-4 was originally identified as a B cell growth and differentiation factor, and it has long been known to be a potent cytokine for murine or human B cells (74). Those observations led to a long-standing conclusion that TH2 cells were the primary CD4 T cells responsible for B cell help in vivo. However, IL-4−/−, IL-4R−/−, or STAT6−/− mice exhibit grossly normal GC development and Ag-specific IgG responses (43, 75–77), although with significant isotype sub-class changes (75, 78). TH2 cells are not required for B cell help in vivo. A newly identified subset of CD4 T cells (TFH cells) is specialized in B cell help (7) and required for GC development in vivo (6, 11–13).
IL-21 has become the leading candidate to be the primary B cell help cytokine produced by TFH cells (7, 79). However, like IL-4, the absence of IL-21 or IL-21R had minimal impact on GC development or TFH cell differentiation in a majority of in vivo models (56, 80–84), although GC and TFH cell defects were observed in some systems (43, 85). Furthermore, IL-21 production is not unique to TFH cells; it is also produced by TH17, TH2, and TH1 cells (86).
In this article, we showed that IL-21 and IL-4 are abundantly produced by Ag-specific GC TFH CD4 T cells, and production of those cytokines is defective in the absence of SAP. Several recent studies showed selective IL-4 production by GC TFH cells during TH2-biased parasitic infections (46, 52, 53). Our data indicated that the IL-4+ GC TFH subset of TFH cells that we identified in the context of a TH1-biased viral infection is analogous cells, and the production of IL-4 is controlled by the TFH transcriptional program. Recent human data indicated that human TFH cells exhibit similar phenotypic properties: tonsillar CXCR5int CD4 T cells exhibited many TFH cell characteristics but did not express IL-4, whereas CXCR5hi CD4 T cells expressed the highest levels of TFH cell markers and expressed IL-4 (58).
A reasonable model of TFH cytokines is that IL-21 and IL-4 are the two main TFH B cell help cytokines in vivo, and the absence of one can largely be compensated for by the other. This model stems from the observations that IL-21 and IL-4 are produced by GC TFH cells, whereas IL-21, but not IL-4, is made in large quantities by the whole TFH population during a viral infection, and both cytokines provide B cell help in vitro when produced by GC TFH cells (Fig. 8B). The absence of either cytokine alone had a modest impact on GCs in several in vivo mouse systems. However, the combined absence of IL-4 and IL-21R resulted in a severe loss of GCs and Ab responses (80). The antiapoptotic impact of IL-4 on B cells is well documented (87–89), and recent work by Locksley and colleagues (46) elegantly showed a role for IL-4 in somatic hypermutation and affinity maturation. IL-4 can also contribute to sustaining GCs (52). IL-21 was shown to be a potent inducer of plasma cell differentiation by human B cells (90–93) and murine B cells (94, 95). In our TFH cell studies, blocking IL-21 in GC TFH cell–B cell cultures reduced IgG production to near background levels (Fig. 8B). This is consistent with studies of human TFH cells (92). Therefore, although it seems that the functions of these two cytokines in GCs can largely be compensated for in vivo, they have separable activities on B cells.
How is IL-4 produced in a TH2-independent manner? In this article, we demonstrated that IL-4 production by GC TFH cells is dependent on SLAM and SAP signaling. Previous work showed that a SLAM-SAP-Fyn-PKCθ signaling axis can induce IL-4 production by CD4 T cells in vitro (23, 24, 31, 96). A SAP-Fyn-PKCθ signaling axis also controls the development of unconventional thymocyte-selected CD4 T cells (97). SLAM recruits SAP, which recruits Fyn kinase via R78, and Fyn kinase activity can result in PKCθ recruitment (23, 24). Further work will help to determine whether Fyn and PKCθ are downstream components of IL-4 induction in GC TFH cells. Consistent with previous studies (24, 96), IFN-γ levels were moderately increased in the TH1 (non-TFH), TFH, and GC TFH populations in the absence of SLAM (data not shown). Elegant work by Schwartzberg and colleagues (23) demonstrated that the SAP− IL-4 production defect is independent of the presence of elevated IFN-γ, because SAP−IFN-γ−/− CD4 T cells still displayed a severe IL-4 defect in vitro.
Notably, GATA3 is not induced in the GC TFH cells, and the TH2 cytokine IL-5 is not produced. There are multiple precedents for IL-4 production in the absence of GATA3 induction. NKT cells express IL-4 but have only low levels of GATA3, and NKT cell IL-4 transcription is NFATc1 dependent (98). In conventional CD4 T cells, STAT5 can induce IL-4 production, but not IL-5 transcription, in the absence of GATA3 induction (99).
The relationships among different CD4 T cells lineages have become more and more complex. We recently proposed that although TFH cells are a distinct CD4 T cell lineage with a master regulator transcription factor, specialized gene-expression profile, and unique biological function, differentiation of a cell to a TFH cell does not outright preclude TH1, TH2, or TH17 characteristics, including cytokine and master regulator transcription factor expression (6, 11). Similar ideas of overlapping differentiation programs between TFH cells and other canonical CD4 lineages were proposed by other investigators (41, 46, 100). In this article, we further defined the complexity of TFH cell-differentiation stages and TFH cell cytokine regulation. Our findings highlight that TFH cell differentiation is a multistage process requiring Bcl6 induction (11–13), cognate B cells (11), and then SAP-dependent differentiation and functions (14, 35), including SLAM-dependent IL-4 induction. Additional studies to understand the integration of these signals and the TFH cell to GC TFH cell transition will be important for developing improved vaccine strategies for long-term Ab responses.
We thank Lindsay Crickard, Elisabeth Krow-Lucal, Danelle Eto, Cheryl Kim, Kurt Van Gunst, and Sacha Garcia for technical assistance. We also thank Andre Veillette for the gift of anti-SAP mAb and Edward Wakeland for B6.sle1b mice.
This work was supported by a Pew Scholar award, La Jolla Institute for Allergy and Immunology Institutional Funds, a Cancer Research Institute award, and National Institutes of Health National Institute of Allergy and Infectious Diseases Grants R01 072543 and R01 063107. I.Y. and R.J. were supported by fellowships from the University of California, San Diego/La Jolla Institute for Allergy and Immunology National Institutes of Health Training Grant.
Raw microarray signal data have been deposited at the National Center for Biotechnology Information GEO database (www.ncbi.nlm.nih.gov/geo/) under accession numbers GSE21379–GSE21381.
The online version of this article contains supplemental material.
The authors have no financial conflicts of interest.