The generation of high-affinity Abs is essential for immunity and requires collaboration between B and T cells within germinal centers (GCs). By using novel mouse models with a conditional deletion of the p110δ catalytic subunit of the PI3K pathway, we established that p110δ is required in T cells, but not in B cells, for the GC reaction. We found the formation of T follicular helper (TFH) cells to be critically dependent on p110δ in T cells. Furthermore, by deleting phosphatase and tensin homolog deleted on chromosome 10, which opposes p110δ in activated T cells, we found a positive correlation between increased numbers of TFH cells and GC B cells. These results are consistent with the hypothesis that T cell help is the limiting factor in the GC reaction. P110δ was not required for the expression of B cell lymphoma 6, the downregulation of CCR7, or T cell entry into primary follicles. Instead, p110δ was the critical catalytic subunit for ICOS downstream signaling and the production of key TFH cytokines and effector molecules. Our findings support a model in which the magnitude of the GC reaction is controlled by the activity of the PI3K pathway in TFH cells.
After exposure to Ag via immunization or infection, B cells have the capacity to generate plasma cells and develop an extrafollicular response or, together with follicular dendritic cells, initiate the germinal center (GC) reaction (1). The GC is a microanatomical structure formed in B cell follicles in secondary lymphoid organs where Ag-specific B cells undergo division, isotype switching, somatic hypermutation, and differentiation into memory B cells or plasma cells. Cognate interaction between B and T cells in the GC is essential for the selection of high-affinity Ag-specific B cells, and access to T cell help is thought to be a limiting factor for the positive selection of GC B cells (2).
The PI3K pathway has been implicated in lymphocyte development and activation. It transduces extracellular signals into the production of the second messenger phosphatidylinositol 3,4,5-trisphosphate (PIP3) by phosphorylation of phosphatidylinositol 4,5-bisphosphate. In lymphocytes, PIP3 is generated by the class I PI3K catalytic subunits of which there are four isoforms named α, β, δ, and γ. The α, β, and δ enzymes form heterodimers with one of five Src homology 2 domain-containing regulatory subunits termed p85α, p85β, p55γ, p50α, and p55α, which mediate recruitment to phosphotyrosine-containing signalosomes (3). The levels of cellular PIP3 are also regulated by phosphatases: phosphatase and tensin homolog deleted on chromosome 10 (PTEN) directly opposes PI3K by removing the 3′ phosphate from PIP3; the SHIP enzymes generate the second messenger, phosphatidylinositol 3,4-bisphosphate.
Mice with germline mutations of the p110δ catalytic subunit have demonstrated its importance for the GC reaction. However, it remains unclear whether this is a reflection of the requirement for p110δ in B cells, T cells, dendritic cells (DCs), or other cell types required for the GC response (4, 5). B cells from p110δ-deficient mice display in vitro defects in survival and proliferation that correlate with defective signal transduction following stimulation with Abs via the BCR or CD19 (6–8). Whether these in vitro defects are relevant in vivo is unclear since B cell-specific deletion of PTEN impaired class-switch recombination but not the magnitude of the GC response (9). T cell-specific deletion of the class IA regulatory subunits showed that the GC reaction and Ag-specific Ab titers were reduced, implicating a T cell-intrinsic requirement for class IA PI3K but leaving unresolved the nature of the relevant catalytic subunits (10). Using the p110δD910A mouse model, which carries a point mutation that renders p110δ catalytically inactive, adoptive transfer experiments revealed impaired Th1/Th2 cytokine production and a 2-fold reduction in clonal expansion (5, 11). In contrast, T cell-specific deletion of PTEN removed the requirement for CD28 costimulation (12) and allowed enhanced IL-4 production (13). Gaining a better understanding of the cell intrinsic role for p110δ in Ab responses should prove helpful in determining the mechanism of action of small molecule p110δ inhibitors. This is timely, as these inhibitors have shown promise in animal models of inflammatory and autoimmune disease (14, 15), and human studies have been initiated.
There is an increasing understanding of the nature of GC T cells that are synonymous with the specialized population of T cells called T follicular helper (TFH) cells (reviewed in Refs. 16 and 17). This T cell subset represents a novel lineage of effector T cells that is critical for the provision of help to B cells during an immune response (18). The development of TFH cells is dependent on the presence of Ag-specific B cells and the transcriptional repressor B cell lymphoma 6 (Bcl-6) (19–21). A defining characteristic of TFH cells is their high expression of the chemokine receptor CXCR5 that allows migration along a gradient of CXCL13 into the B cell follicle (22). Recently, ICOS-mediated PI3K signaling has been shown to be an essential regulator of TFH cell numbers and cytokine production (23).
In this study, we show that, unexpectedly, p110δ activity in B cells is not necessary for the formation or magnitude of the GC response and that the essential function of p110δ in the GC reaction is the T cell-intrinsic regulation of TFH cell formation and function. Moreover, we demonstrate that the p110δ subunit is pivotal for ICOS function, thereby acting as a key regulator of the TFH cell lineage at the level of intracellular signaling. The importance of the PI3K pathway in TFH cells is underscored by our finding that the magnitude of PI3K signaling in T cells regulates TFH cell numbers, the numbers of GC B cells, and the number of high-affinity Ab-secreting cells produced upon secondary immunization.
Materials and Methods
The pik3cd targeting vector was generated using standard methodology. A single LoxP site was introduced into the XhoI site upstream of the first coding exon. A neomycin cassette flanked by FRT sites and a single LoxP site (FRTneoFRT) was inserted into the EcoRV site between exons 9 and 10. Targeted E14 ES clones were identified by Southern blot analysis of EcoRV digests, which yielded a 5.6-kb fragment for nontargeted clones and a 4.2-kb fragment for targeted clones. FLPe was transiently expressed to mediate recombination excision of the neomycin cassette, which was verified by Southern blot analysis. One of the correct clones was selected and injected into blastocysts to produce chimeric mice containing LoxP-flanked exons 1–9 encoding the first 490 aa of p110δ, which will be deleted in the presence of cre. Conditional deletion was achieved by crossing with CD19cre, CD4cre, or Ox40cre mouse strains. Germline p110δ−/− mice were described previously (4). Mice were backcrossed at least five times to C57BL/6. Mice with a conditional allele of PTEN, p110δD910A mice, OT-II transgenic RAG−/− background mice, and anti-hen egg lysozyme (HEL) Ig transgenic mice have been described previously. All mice were bred at the Babraham Institute Small Animal Barrier Unit. Animal husbandry and all procedures involving animals were approved and performed according to United Kingdom Home Office and local Ethical Review Committee guidelines.
Isolation of DNA, RNA, and quantitative PCR
Genomic DNA was prepared using cell lysis and proteinase K digestion (Gentra/Qiagen, Crawley, U.K.). Quantitative PCR was performed to determine the recombination efficiency of genomic DNA using a JOE-labeled probe within exon 2 and a FAM-labeled probe within exon 18 (Supplemental Fig. 1B): p110δ forward exon 18, 5′-CACAGGTCTCATCGAGGTGGTC-3′; p110δ reverse exon 18, 5′-TGGACTTGAGCCAGTTGAGCA-3′; p110δ probe (FAM) exon 18, 5′-CTCGGACACCATCGCCAACATCC-3′; p110δ forward exon 2, 5′-AGCAGCAGGAGTTGGAGGATG-3′; p110δ reverse exon 2, 5′-CCAATGAGGAGGCTGATCTGG-3′; and p110δ probe (JOE) exon 2, 5′-CGACATCCAGCCCTTCCTGCCC-3′.
RNA was extracted from cells using TRIzol (Invitrogen, Carlsbad, CA). DNAse-treated RNA was reverse transcribed using SuperScript II (Invitrogen). Quantitative RT-PCR was performed using mRNA prepared from T cell blasts and quantified using Taqman probe assays (Applied Biosystems, Warrington, U.K.) and normalized to the expression of β2-microglobulin as 2−ΔΔCt.
Adoptive transfer and immunizations
Mice were immunized i.p. with 100 μg of (4-hydroxy-3-nitrophenyl)acetyl (NP)19–keyhole limpet hemocyanin (KLH) or NP36-chicken γ-globulin (CGG) (Biosearch Technologies, Novato, CA) adsorbed to alum or 2 × 108 SRBCs (Harlan, Bicester, U.K.) in saline. Mice were bled at indicated time points before boosting with soluble Ag. In adoptive transfer experiments analyzing T cell help to B cells, 105 OT-II or p110δD910A OT-II T cells isolated from lymph nodes together with 106 carrier C57BL/6 B cells were injected i.v. into recipient B6 mice and recipients were then i.p. immunized with 100 μg of NP13-OVA (Biosearch Technologies) adsorbed to alum. Adoptive transfer experiments into C57BL/6 or p110δ−/− recipient mice were performed as described above using 3 × 105 OT-II T cells per recipient mouse.
Single-cell suspensions were prepared from lymphoid tissues and the peritoneal cavity. Abs against the following surface markers were used: B220, CD4, CD5, CD8, CD21, CD23, CD95, CXCR5, CCR7, ICOS, IgM, programmed cell death 1 (PD-1; BD Biosciences), peanut agglutinin (PNA)-FITC (Vector Laboratories, Peterborough, U.K.), and streptavidin conjugates (eBioscience, Hatfield, U.K.). Dead cells were excluded using DAPI staining. Cells were collected using an LSRII or FACSCalibur (BD Biosciences, Oxford, U.K.) and analyzed using FlowJo software (Tree Star, Ashland, OR). To purify GC cells, B220+CD95+cells were FACSAria-sorted from mice at days 7 and 10 following NP-KLH immunization. To detect CD40L after stimulation, splenocytes were stimulated in vitro using plate-bound anti-CD3 (2C11) at 1 μg/ml and anti-CD28 (37.51) at 10 μg/ml for 5 h, and anti-CD40L-PE (eBioscience) was added for the last 2 h of culture.
Five-micrometer spleen sections were fixed in ice-cold acetone and blocked before incubation with biotinylated primary Abs followed by streptavidin-Alexa 555 (Invitrogen) and then PNA-FITC. The sections were mounted using Aqua-Poly/Mount (Polysciences, Northampton, U.K.). An Olympus IX81 wide-field epifluorescence microscope was used for imagining at ×10–×20 magnification with an Olympus ×10 UPLFL (universal plan fluorite) NA 0.3 objective, and images were taken with a Hamamatsu ORCA ER digital camera at room temperature. Color conversion was performed by the Olympus cell software, and minor adjustments were made using Adobe Photoshop. Five images per mouse were taken, and the numbers of B cell follicles and PNA+ GC structures were enumerated. For quantification of CD4+ cells within the B cell follicle and GC, sections were stained with anti-IgD Alexa 647, PNA-FITC, CD4-bio, and tetramethylrhodamine isothiocyanate-labeled streptavidin (Jackson ImmunoResearch Laboratories, Newmarket, U.K.). Five fields of view were captured per mouse, and the number of CD4+ cells in the GC (PNA+ and IgD− areas) was determined by counting cells relative to a defined area (μm2) using Volocity software (Improvision, Coventry, U.K.).
In vitro IL-21 differentiation
Naive lymph node CD4+ T cells were FACSAria-sorted to >95% purity. CFSE-labeled, naive CD4+ T cells (105+ cells were defined by quadrant gates set using the isotype control staining and the undivided peak of CFSE-labeled cells.
CD4+ T cells were purified by MACS and cultured for 2 d in the presence of plate-bound 10 μg/ml anti-CD3 and soluble 1 μg/ml anti-CD28 and then rested in media for 1 d. The cells were restimulated with 0.5 μg/ml anti-CD3, 2 μg/ml anti-ICOS, and 20 μg/ml anti-hamster IgG for 6 h.
Anti-HEL Ig transgenic B cells were isolated from wild-type and p110δ−/− mice using a MACS B cell isolation kit (Miltenyi Biotec, Bisley, U.K.) according to the manufacturer’s protocol (purity >95% B cells). B cells were cultured with increasing doses of HEL (Sigma-Aldrich, Poole, U.K.) and the I-Ak–restricted HEL-responsive 2G7 T cell line for 24 h at 37°C (24
ELISAs and ELISPOT
ELISA plates (Nunc, Paisley, U.K.) were coated with NP23-BSA or NP33-BSA, or for high-affinity measurement NP2-BSA, or NP3-BSA at 10 μg/ml in carbonate buffer, washed, and blocked. Serum samples were serially diluted and bound Ig was detected using biotinylated anti-IgM or IgG1 and HRP-conjugated streptavidin (SouthernBiotech, Birmingham, AL). Endpoint titers were calculated to give a measure of relative Ab concentration. For ELISPOT, MultiScreen HA mixed cellulose ester plates (Millipore, Watford, U.K.) were coated with NP23-BSA, or NP2
Western blot analysis
CD4+ T cells from C57BL/6 or p110δ−/− mice were MACS-purified and stimulated by anti-CD3 (10 μg/ml) and anti-CD28 (1 μg/ml) for 2 d and then replated in IL-2 (10 ng/ml) (PeproTech) for 2 more days. Before stimulations, the CD4+
Statistical analysis was performed using a nonparametric Mann-Whitney U test, paired t test, or nonparametric ANOVA using Graphpad Instat3 software (GraphPad Software, San Diego, CA). Significance is indicated by *p < 0.05, **p < 0.01, and ***p < 0.001.
p110δ expression in T cells is critical for the GC reaction
As p110δ is expressed by most cells in the immune system, we used Cre/LoxP technology to identify the cell types that depended on p110δ in the GC reaction. Deletion of p110δfl/fl in T cells using CD4cre (T p110δfl/fl mice) was highly efficient and resulted in a modest reduction in the number of T cells in the spleen, but it did not affect the number of T cells in the lymph nodes (Supplemental Figs. 1, 2). As T lymphocytes are essential for the GC response (25), we evaluated the contribution of p110δ in T cells by immunizing T p110δfl/fl and littermate control p110δfl/fl mice that are CD4cre−. Following immunization with NP-KLH, the number of CD95+B220+ GC B cells was reduced ∼7-fold in T p110δfl/fl compared with control mice at the peak of the GC reaction (Fig. 1A). Also, the ratio of GCs per B cell follicle was significantly reduced (Fig. 1B). In the primary immune response, the NP-specific IgG1 titers detected by binding to highly haptenated BSA (e.g., NP33-BSA) were only slightly decreased in T p110δfl/fl mice. In contrast, the high-affinity NP-specific IgG1 response, detected by binding to lightly haptenated BSA (e.g., NP3-BSA), was severely impaired, showing that affinity maturation was defective in T p110δfl/fl mice (Fig. 1C). To elucidate the role of p110δ in T cells for the generation of a secondary memory response, we boosted the p110δfl/fl and T p110δfl/fl mice using soluble Ag 42 d after primary immunization. Following booster immunization there was no increase in high-affinity Ab, indicating that immunological memory was defective (Fig. 1D). In agreement with this, ELISPOT analysis of NP2- and NP23-binding AFCs in the spleen and bone marrow showed a marked decrease in the number of IgG1-secreting AFCs (Fig. 1D, 1E). To determine whether the enzymatic activity of p110δ is required in a T cell-intrinsic manner for provision of help to B cells, we adoptively transferred catalytically inactive p110δD910A or control wild-type OVA-specific OT-II transgenic T cells into C57BL/6 recipient mice and immunized with NP-OVA. The resulting NP-specific IgG1 titers were 3- to 4-fold lower in recipients of p110δD910A OT-II T cells than in recipients of wild-type OT-II T cells (Fig. 1F). This indicates that the catalytic activity of p110δ is required for T cells to provide optimal help to B cells. Thus, p110δ function is required by T cells for the generation of the GC reaction, affinity maturation, and memory B cells.
p110δ regulates TFH formation
As T cells are known to be critical for the GC reaction (25), the TFH cell population was analyzed in T p110δfl/fl mice by flow cytometry using CXCR5- and PD-1–specific Abs. In T p110δfl/fl mice immunized with NP-KLH, the numbers of TFH cells were reduced compared with p110δfl/fl controls (Fig. 2A). To extend our analysis to a different Ag, we chose SRBCs, which elicit a robust ICOS-dependent expansion of CXCR5+ TFH cells (26). Four days after immunization with SRBC, TFH cell formation was severely impaired in T p110δfl/fl mice and, by day 6, the TFH cell population in these mice had returned to baseline (Fig. 2B). These data suggest that p110δ is required by TFH cells for their initial expansion or differentiation. Furthermore, the GC reaction elicited by SRBCs was impaired in T p110δfl/fl mice, as the number of B220+PNA+CD95+ cells was reduced at all time points tested (Fig. 2B), and the histological appearance of the spleen showed fewer and smaller GC structures in the SRBC- immunized T p110δfl/fl mice (data not shown).
TFH cells must form cognate interactions with B cells and, therefore, migration into the B cell follicle is important for TFH cell differentiation (21, 27). The numbers of CD4+ T cells in the B cell follicle were slightly lower in T p110δfl/fl mice compared with p110δfl/fl mice. However, the numbers of CD4+ T cells in the PNA+IgD− GC area were significantly reduced (Fig. 2C). Thus, the reduced number and size of GCs are reflected by reduced numbers of TFH cells. As CCR7 downregulation is essential for follicular entry and T cell help to B cells (27), we determined the levels of CCR7 on CD4+ T cells. The expression of CCR7 was higher on PD-1− T p110δfl/fl CD4+ T cells compared with p110δfl/fl, which is consistent with a previous study showing that p110δ regulated CCR7 on CD8+ T cells (28). We found that PD-1+ TFH cells downmodulated CCR7 on their surface similarly in T p110δfl/fl and p110δfl/fl mice (Fig. 2D). Taken together, these data suggest that CD4+ TFH cells residing in the GC were reduced in T p110δfl/fl mice, but this reduction is unlikely to be a consequence of failure to downregulate CCR7.
p110δ function in APCs and B cells is not essential for Ab responses
Although the T p110δfl/fl mice revealed a T cell-intrinsic role for p110δ in the GC reaction, the initial observations on the severe defects GC formation and humoral immunity in the p110δ−/− or p110δD910A knock-in mice could also reflect a role for p110δ in APCs or B cells (4, 5). To determine whether p110δ plays a role in the initial priming of naive T cells by APCs, we used adoptive transfer of transgenic OT-II cells and NP-OVA immunization. The ability of wild-type OT-II T cells to stimulate the production of NP-specific IgG1 was assessed 7 d after adoptive transfer into C57BL/6 and p110δ−/− recipients (Fig. 3A). The NP-specific IgG1 titers were not significantly different between the two groups, suggesting that initial T cell priming by APCs is independent of p110δ and, moreover, that provision of extrinsic T cell help through adoptive transfer can rescue defects in humoral immunity in the p110δ−/− mice.
To explore the B cell-intrinsic role of p110δ, we used CD19cre-mediated deletion of p110δ in B p110δfl/fl mice. Deletion was less efficient than that obtained with CD4cre but led to reductions in the number of marginal zone (MZ) and peritoneal B1, but not in follicular B cells (Supplemental Figs. 1, 2); this is a phenotype similar to germline p110δ knockout and p110δD910A mice. The absence of p110δ in B cells did not cause any reduction in GC B cell numbers after immunization (Fig. 3B). Following NP-CGG immunization, IgM and high affinity NP-specific IgG1 titers were not significantly decreased in the B p110δfl/fl mice compared with p110δfl/fl mice (Fig. 3C). However, we found that the Ag-specific IgE titers were increased 30-fold after immunization in B p110δfl/fl compared with p110δfl/fl mice (Fig. 3C). This is in agreement with evidence that PI3K negatively regulates IgE (29, 30) and indicates that this effect is B cell-intrinsic. In the B p110δfl/fl mice, there was a trend toward impairment in affinity maturation compared with p110δfl/fl mice (Fig. 3D), indicating that p110δ has a modest role in affinity maturation or the survival or functionality of plasma cells. Alternatively, the reduction in IgG1 may be secondary to the enhanced class-switching to IgE. Analysis of high-affinity NP-specific IgG1 titers after booster immunization showed no difference in memory response between B p110δfl/fl and p110δfl/fl mice (Fig. 3E). The GC imposes a powerful selective environment upon B cells, which should result in accumulation of p110δ-sufficient cells if this enzyme is critically required for GC B cell selection or survival. However, there was no significant increase in the proportion of nondeleted p110δ genomic DNA in sorted GC B cells analyzed 7 or 10 d after NP-KLH immunization compared with follicular B cells (Fig. 3F).
MHC class II–mediated presentation of peptide Ag by B cells to GC T cells is essential for successful selection of high-affinity B cells. Therefore, we tested the ability of p110δ-deficient B cells to process and present Ag using an in vitro T cell stimulation assay. B cells from p110δ−/− mice bearing the anti-HEL Ig transgene (MD4) (31) showed an increased capacity to present Ag over a range of HEL concentrations compared with p110δ-sufficient controls (Fig. 3G). Taken together, these data suggest that p110δ in B cells plays an essential role in limiting class-switching to IgE, but it is dispensable for the GC response.
TFH cell and GC B cell numbers are regulated by p110δ and PTEN activity in T cells
The deletion of p110δ in developing T cells by CD4cre would be expected to influence early activation events that were dependent on this lipid kinase. To investigate whether p110δ was required for the GC reaction subsequent to the initial activation of T cells, we generated mice in which deletion of p110δ was mediated by OX40cre. The costimulatory molecule OX40 is expressed following TCR- and CD28-mediated T cell activation and has been found to be expressed on CXCR5+CD4+ T cells, suggesting that the OX40cre transgene will be expressed in TFH cells (32, 33). Following immunization, we analyzed TFH cell numbers at the peak of the GC reaction in the OX40cre-p110δfl/fl model and found that TFH cells were significantly diminished in these mice compared with p110δfl/fl controls (Fig. 4A). The number of GC B cells and the ratio of GCs to B cell follicles were concomitantly reduced (Fig. 4B). The B cell memory response was analyzed at day 7 after boosting in the OX40cre models. In the OX40cre-p110δfl/fl mice, the high-affinity Ab titers were modestly reduced and there was a significant reduction in the numbers of NP2-binding AFCs in the bone marrow and spleen (Fig. 5A).
The enzymatic function of p110δ is directly opposed by the lipid phosphatase PTEN that converts PIP3 to phosphatidylinositol 4,5-bisphosphate. Deletion of PTEN is associated with hyperresponsiveness of cells to stimulation and malignant transformation (13). To understand the contribution of PTEN in TFH cells, we again used the OX40cre system. Following immunization of OX40cre-PTENfl/fl mice, the TFH population was increased ∼3-fold compared with PTENfl/fl control mice (Fig. 4C). The number of GC B cells was also significantly increased and there was a trend toward a higher ratio of GCs per B cell follicle (Fig. 4D). In the OX40cre-PTENfl/fl mice, the high-affinity Ab titers were increased and the numbers of NP2-binding AFCs tended to be higher (Fig. 5B).
PI3K activity in T cells and the subsequent effect on TFH cell numbers are highly correlated with GC B cell numbers in the T p110δfl/fl and OX40cre mouse models (Fig. 4E), suggesting that PI3K activity in T cells determines the magnitude of the GC response. Thus, the GC reaction and subsequent effects on B cell memory function is influenced by the amount of PI3K activity in T cells.
The expression of TFH effector molecules requires p110δ
The CD40–CD40L receptor–ligand pair is of crucial importance for the GC reaction, and regulation of CD40L expression on activated T cells is central to controlling this molecular interaction (34). We therefore measured the level of CD40L expressed by T p110δfl/fl CD4+ T cells following in vitro stimulation and found it reduced ∼2-fold compared with p110δfl/fl mice (Fig. 6A). IL-21 production is a major effector function of TFH cells, and this cytokine plays a crucial role in GC B cells for generation of long-term humoral immunity and Bcl-6 expression (35, 36). The IL-21 production induced by CD4+ T cells differentiated in the presence of IL-6 was reduced in in vitro cultures using T p110δfl/fl compared with p110δfl/fl cells (Fig. 6B). These results suggest that the defects in CD40L and IL-21 in p110δ-deficient T cells could contribute to the impairment in humoral immunity in T p110δfl/fl mice.
PI3K signaling via ICOS is mediated by p110δ
TFH cell development and maintenance has been linked to the CD28-related costimulatory molecule ICOS, which was originally identified as being highly expressed on human GC T cells (37). ICOS-deficient mice (26) and ICOS ligand-deleted B cells (38) have revealed the crucial importance of this costimulatory molecule for formation of TFH cells. Additionally, the sanroque mouse, which exhibits expansion of the TFH cell population, has enhanced expression of ICOS (39). The intracellular tail of ICOS is associated with PI3K activity (40), and ICOS signaling is impaired after treatment with non-isoform–selective PI3K inhibitors (41). Recent data have revealed that mutation of the tyrosine motif in the cytoplasmic tail of ICOS that recruits PI3K regulatory subunits leads to a severe impairment in TFH cell numbers (23). To explore the role of p110δ in ICOS signaling, we generated CD4+ T cell blasts from p110δ−/− and C57BL/6 mice with equivalent ICOS expression (data not shown). Stimulation of these T cells with a suboptimal dose of anti-CD3 in combination with anti-ICOS Ab allowed the effect of ICOS activation on the phosphorylation of Akt (protein kinase B) to be analyzed as a read-out for the activity of the PI3K pathway. Following activation in this way, the phosphorylation of Akt was strong and sustained in wild-type CD4+ T cells but was barely detectable in p110δ−/− T cells (Fig. 7A). Furthermore, the unresponsiveness of p110δ−/− CD4+ T cell blasts was evident over a range of concentrations of ICOS Ab (Fig. 7B). To substantiate the observations with p110δ−/− CD4+ T cells, we examined the effect of the p110δ isoforms-selective PI3K inhibitor IC87114 on ICOS signaling in T cell blasts from C57BL/6 mice. This inhibitor mimicked the defect in ICOS signaling seen in p110δ−/− T cells (Fig. 7C). To further explore the role of p110δ in ICOS downstream signaling, we analyzed the constitutively active kinase GSK3 that is phosphorylated and thereby inhibited by p-Akt. We found that the levels of phosphorylation of the isoform GSK3β was reduced in p110δ−/− T cell blasts compared with B6 after ICOS-stimulation (Fig. 7D).
To further explore the role p110δ in ICOS signaling, we restimulated ICOS-expressing T cell blasts in vitro with anti-CD3 and anti-ICOS and measured selected cytokine and transcription factor mRNAs in these cells. The transcription factor c-Maf (42) is known to be induced by ICOS (43). In T cells lacking p110δ, the amount of c-Maf, IL-21, and IL-4 mRNA was reduced compared with control p110δfl/fl T cells, showing that p110δ is crucial for ICOS-dependent gene expression (Fig. 7E). The mRNA levels of the mutually antagonistic transcriptional repressors Bcl-6 and B lymphocyte-induced maturation protein 1 (Blimp-1) were analyzed after ICOS restimulation. Bcl-6 was induced by ICOS restimulation in p110δfl/fl T cell blasts, but in the T p110δfl/fl blasts Bcl-6 expression was already high. Blimp-1 expression in CD4+ T cell blasts has previously been shown to be lower in stimulated CD4+ T cells from p110δD910A mice compared with controls (44), and we found that Blimp-1 mRNA levels were also lower in T p110δfl/fl blasts after ICOS restimulation (Fig. 7E).
In this study, we show that the essential cell-intrinsic role of p110δ required for the GC reaction to proceed resides within T lymphocytes. Moreover, our results suggest that the activity of the PI3K pathway in T cells is a critical determinant of GC size and output. The numbers of GC B cells was dependent on PI3K activity in T cells, which is compatible with the idea that access to TFH cell-mediated help is a limiting factor in the GC (2). In contrast, when the GC defect is primarily B cell-intrinsic, such as in the IL-21–deficient mice, there is an excess of TFH cells per Ag-specific GC B cell (36). The correlation between GC B cell and TFH cell numbers was also related to humoral memory, since a reduction in GC B and TFH cells was accompanied by impairments in the memory response, whereas the larger GC cellularity correlated with increased high-affinity Ab titers. Thus, PI3K activity in T cells determines the number of GC B cells and the magnitude of the memory response in a non-cell–autonomous manner. Previous studies, such as on the heart, have emphasized the cell-autonomous role of the PI3K pathway in determining organ size in mammals (45). This is a novel example of PI3K activity determining cell population size as a consequence of regulating signaling in T cells.
When extrinsic T cell help was provided, p110δ was dispensable in other cells for the generation of an isotype-switched Ab response. Furthermore, in B p110δfl/fl mice the magnitude and output from GC reaction was intact. As there was no evidence for selection of nondeleted cells, we conclude that p110δ signaling in B cells is not essential for the GC magnitude. This was unexpected given the many reports, including our own, of defective signaling by p110δ-deficient B cells and challenges the widely held view that p110δ is the main PI3K isoform involved in B cell activation (46). These previous findings may reflect the key role of p110δ in the development of MZ B cells, as the in vitro studies using splenic B cell populations did not take account of the hyperresponsiveness of MZ B cells (47, 48) which were missing in the p110δ mutants. Similarly, previous suggestions of defective Ag presentation by p110δD910A B cells (49) may reflect the paucity of MZ B cells, which are known to be more effective at presenting Ags than follicular B cells (47). Indeed, we found Ag processing and presentation was enhanced in the absence of p110δ when Ag-specific B cells were tested. Our study also revealed that a B cell-intrinsic role of p110δ was to suppress class-switch recombination. This is consistent with PTEN inactivation in B cells inhibiting class-switch recombination (9) and the enhanced switching to IgE when p110δ activity is attenuated in vitro (29, 30). Taken together, the data show the nonredundant role of p110δ in B cells to control the differentiation or survival of the MZ and B1 B cell subsets and during the TD response to suppress Ig class-switch recombination. Other PI3K catalytic subunits, such as the p110α isoform that has been shown to be associated with CD19 after BCR stimulation, may contribute in a redundant manner to the GC reaction (6).
Multiple priming/activation events are required for a successful GC reaction. First, T and B lymphocytes individually need to recognize Ag and initiate signaling through the TCR or BCR, respectively. In a second phase, the T and B cells home to the T cell–B cell border zone within secondary lymphoid tissues, form conjugates, and exchange signals that regulate selection and survival. Subsequently, continued interactions between B and T cells are required for the sustenance of the GC reaction. The T cell-intrinsic requirement for p110δ was characterized by a paucity of the TFH cell subset, which is specialized in providing help to B cells. Although p110δ has previously been shown to be important for FoxP3+ regulatory T cell function (50), the TFH cell defect is unlikely to be secondary to a defect in these cells since adoptively transferred regulatory T cells do not contribute to GC formation in the spleen (51). CD4+ T cells lacking p110δ could enter primary follicles but were substantially reduced in GCs. The downmodulation of surface expression of CCR7, a prerequisite for follicular entry, suggested that some early activation events in T cells were independent of p110δ. The importance of the signaling lymphocyte activation molecule (SLAM) family members CD84 and Ly108 for cognate interactions between TFH cells and GC B cells (52) led us to analyze their presence on CD4+ T cells; however, we found that their expression was p110δ-independent (data not shown). Deletion of p110δ after the initial priming event, by expression of OX40cre, suggested that TFH cells required the continual presence of p110δ. This finding may reflect the key role of p110δ in ICOS signaling.
Our studies using hapten protein and SRBCs, both of which elicit ICOS-dependent GC responses, implicate p110δ as being an essential catalytic subunit for ICOS signaling. ICOS is a key cell-surface receptor regulating the generation and function of TFH cells. ICOS signals principally by recruitment of PI3K to the tyrosine within its cytoplasmic YFMF motif (40, 53). Mutation of this tyrosine has revealed the requirement for this interaction in the generation and function of TFH cells (23). The similarities between the recently reported ICOS-Y181F mutant and the T p110δfl/fl mice are striking. Both of these mouse models resemble ICOS knockout mice or ICOS inhibitory Ab treatment in terms of induction of TFH cells and the GC response (23, 26). The p110δ dependence of ICOS signaling suggests that p110α and p110β, which are expressed in T cells, are either quantitatively or qualitatively unable to substitute for p110δ. The latter possibility may reflect the growing awareness of selective recruitment and activation of PI3K isoforms by particular receptors. ICOS signaling is known to generate high levels of PIP3 (54) in the immunological synapse via recruitment of the p50α regulatory subunit (55). The selective requirement for p110δ downstream of ICOS signaling may reflect a preferential molecular association between p50α and p110δ, rather than p110α or p110β.
We found that p110δ was required for ICOS-mediated increases in IL-21, IL-4, and c-Maf mRNA. These observations are in accordance with ICOS regulating c-Maf (43), which in turn regulates IL-21 (56) and IL-4 (42). Our data are consistent with results showing ICOS signaling regulates NFAT-dependent expression of cytokine genes in a PI3K-dependent manner (57) and with observations that Kruppel-like factor 2, which is suppressed by the PI3K pathway, attenuates IL-4 expression (58). The kinase GSK3 is known to be inhibited by ICOS costimulation (55) and to negatively regulate at least 50 target proteins, including several transcription factors that are critical for T cell effector functions and also molecules controlling cell growth and apoptosis (reviewed in Ref. 59). Thus, p110δ may orchestrate multiple transcription factor activities downstream of ICOS to control TFH cell function through GSK3β, which may be the nexus between PI3K signaling and downstream targets. The PI3K pathway also regulates the expression of the mutually antagonistic transcriptional repressors Bcl-6 and Blimp-1, which play a central role in TFH cell differentiation. p110δ was not required for the induction of Bcl-6 mRNA, and our data rather suggested that PI3K repressed Bcl-6 mRNA. This may reflect the loss of PI3K inhibition of the FOXO family of transcription factors, as these activate transcription of the Bcl-6 gene (60). The low levels of Blimp-1 mRNA in T p110δfl/fl T cells may further reflect the effect of the PI3K pathway on the Bcl-6–Blimp-1 axis that regulates TFH cell differentiation (21). Although p110δ may negatively regulate Bcl-6 mRNA levels, the increased Bcl-6 mRNA levels in T p110δfl/fl T cells is insufficient to overcome the requirement for p110δ during TFH cell formation in vivo.
In summary, we propose that the amount of PIP3 generated is crucial for TFH cell formation and for optimal T cell help to B cells. This suggestion is supported by our observation that TFH cell numbers were reciprocally regulated by the p110δ-PTEN balance in activated T cells. The p110δ catalytic subunit is absolutely necessary for inducing the TFH cell lineage in vivo, thereby playing a dominant role among signal transduction pathways in the regulation of adaptive immunity.
We thank the animal facility staff and Martin George for technical assistance and Geoff Morgan for FACS sorting. We also thank Drs. Michelle Linterman, Peter D. Katsikis, and Geoffrey Butcher for critically reading the manuscript.
Disclosures The authors have no financial conflicts of interest.
This work was supported in part by the Biotechnology and Biological Sciences Research Council and the Medical Research Council.
The online version of this article contains supplemental material.
Abbreviations used in this paper:
- Ab-forming cell
- B cell lymphoma 6
- B lymphocyte-induced maturation protein 1
- bone marrow
- chicken γ-globulin
- germinal center
- glycogen synthase kinase 3
- hen egg lysozyme
- integrated intensity
- keyhole limpet hemocyanin
- mean fluorescence intensity
- marginal zone
- nonimmunized mice
- programmed cell death 1
- peanut agglutinin
- phosphatase and tensin homolog deleted on chromosome 10
- T follicular helper.
- Received May 24, 2010.
- Accepted July 23, 2010.
- Copyright © 2010 by The American Association of Immunologists, Inc.