PI3K Orchestrates T Follicular Helper Cell Differentiation in a Context Dependent Manner: Implications for Autoimmunity

T follicular helper (Tfh) cells are a specialized population of CD4+ T cells that provide help to B cells for the formation and maintenance germinal centers, and the production of high affinity class-switched antibodies, long-lived plasma cells, and memory B cells. As such, Tfh cells are essential for the generation of successful long-term humoral immunity and memory responses to vaccination and infection. Conversely, overproduction of Tfh cells has been associated with the generation of autoantibodies and autoimmunity. Data from gene-targeted mice, pharmacological inhibitors, as well as studies of human and mice expressing activating mutants have revealed that PI3Kδ is a key regulator of Tfh cell differentiation, acting downstream of ICOS to facilitate inactivation of FOXO1, repression of Klf2 and induction of Bcl6. Nonetheless, here we show that after acute LCMV infection, WT and activated-PI3Kδ mice (Pik3cdE1020K/+) show comparable ratios of Tfh:Th1 viral specific CD4+ T cells, despite higher polyclonal Tfh cells in Pik3cdE1020K/+ mice. Thus, the idea that PI3K activity primarily drives Tfh cell differentiation may be an oversimplification and PI3K-mediated pathways are likely to integrate multiple signals to promote distinct effector T cell lineages. The consequences of dysregulated Tfh cell generation will be discussed in the context of the human primary immunodeficiency “Activated PI3K-delta Syndrome” (APDS), also known as “p110 delta-activating mutation causing senescent T cells, lymphadenopathy and immunodeficiency” (PASLI). Overall, these data underscore a major role for PI3K signaling in the orchestration of T lymphocyte responses.


INTRODUCTION
Naïve CD4 + T helper (Th) cells play pivotal roles in adaptive immunity through the differentiation into distinct cytokine-producing effector subsets that specifically fight a wide range of pathogens and tumors (1). T follicular helper (Tfh) cells provide help to B cells for the formation of germinal centers (GCs) (2-4), a specialized microenvironment where clonal expansion of B cells, immunoglobulin diversification, affinity maturation, and development of memory B and long-lived plasma cells occur in response to immune challenge (5). The outcome of GC reactions requires proper help provided by Tfh cells (6-9). Most successful human vaccines are based on the generation of long-term protective humoral responses derived from the interactions of Tfh and GC B cells; however, Tfh cells can also promote dysregulated responses and autoimmunity (7, 10, 11). It is, therefore, critical to understand factors that promote or limit Tfh cells to elicit tightly controlled GC responses.
Recent data from gene-targeted mice, as well as mice and humans expressing activating mutants of phosphatidylinositol 3kinase delta (PI3Kδ), suggest that PI3K activity is an essential component of pathways driving Tfh cell and GC formation (12)(13)(14)(15)(16). In this review, we discuss PI3Kδ-mediated pathways involved in the generation, maintenance and function of Tfh cells, including cellular receptors that activate PI3K within T cells, molecular pathways activated, and implications for autoimmunity, with a focus on the genetic disease APDS/PASLI.

PI3K SIGNALING IN IMMUNITY PI3K Signaling
The PI3Ks are a family of heterodimeric lipid kinases that are activated downstream of a variety of receptors, including growth factor, antigen, costimulatory, cytokine, chemokine, and Toll-like receptors (17,18). Class IA PI3Ks consist of a p85 regulatory and a p110 catalytic subunit that catalyzes the addition of a phosphate to the membrane phospholipid PI(4,5)P 2 , to generate phosphoinositide 3,4,5-triphosphate (PIP 3 ). PIP 3 helps recruit signaling molecules containing pleckstrin homology and other PIP 3 -binding domains to the plasma membrane to propagate signaling cascades (Figure 1). Mammals express three class IA catalytic isoforms: the broadly expressed p110α and p110β, and p110δ, which is expressed primarily by immune cells (17). Notably, PI3Kδ is activated by a variety of cell-surface receptors that are critical for Tfh cell differentiation, localization and function, including the T-cell receptor, CD28, and ICOS coreceptors, and cytokine receptors (17). Activated-PI3K coordinates the recruitment of molecules such as PDK1 that phosphorylates and activates the serine/threonine kinase AKT, which in turn phosphorylates multiple targets. Among these are the FOXO transcription factors, which are then sequestrated outside the nucleus by 14-3-3 proteins and degraded. FOXOs regulate transcription of multiple genes involved in lymphocyte development, differentiation and function (17,20). Another downstream effector of PI3K is the mammalian Target of Rapamycin kinase (mTOR), which forms two complexes, mTORC1 and mTORC2, with different scaffolding partners (21). AKT activates mTORC1, an ancient regulator of metabolism, protein synthesis, and cell growth. mTORC2 is essential to fully phosphorylate and activate AKT, thus contributing to downstream signaling, including FOXO1 inactivation, and actin reorganization (21)(22)(23). PI3K is counteracted by the lipid-phosphatases PTEN and SHIP-1/2 that convert PIP 3 to PI(4,5)P 2 and PI(3,4)P 2 , respectively, (17) (Figure 1).
The importance of PI3Kδ in lymphocyte function is highlighted by the human primary immunodeficiency APDS/PASLI, in which patients are heterozygous for activating mutations in PIK3CD, the gene encoding p110δ. These patients show immunodeficiency and lymphopenia, as well as lymphoproliferation and autoimmunity (14,15,24,25). Four independent groups, including us, have recently generated mouse models expressing the E1020K activating mutant of p110δ, which recapitulate many features of APDS/PASLI (16,(26)(27)(28). Notably, patients and Pik3cd E1020K/+ mice exhibit elevated circulating Tfh cells and GCs associated with autoantibody production (15, 16,24,29). Mice that express constitutively active p110α in T cells (30) or have a T cell-specific deletion of PTEN (13) also have elevated Tfh cell frequencies, supporting a more general connection between PI3K activity and Tfh cells. Nonetheless, the observation that p110δ-inactivation in T cells abrogates Tfh cell generation, supports a non-redundant role of p110δ in this process (12, 13). Together, these data provide strong evidence that PI3Kδ is an important component of pathways driving Tfh cell differentiation.

Tfh Cell Differentiation
The generation of Tfh cells is a multistage process that requires the integration of signals from different cell types (31). In the T cell zone of secondary lymphoid organs, antigen-presenting dendritic cells (DCs) activate T cells to initiate the pre-Tfh cell program, leading to induction of the costimulatory molecule ICOS and chemokine receptor CXCR5, as well as downregulation of CCR7, which together permit migration to the T-B cell border zone (32)(33)(34). Here, activated B cells receive signals from pre-Tfh cells to differentiate either along extra-follicular or GC pathways (5, 35). Cognate interactions with activated B cells help promote the differentiation into GC-Tfh cells (36), identified as CXCR5 hi PD-1 hi Foxp3 − CD4 + T cells that also express high levels of ICOS, CD40L, and the Tfh-master transcription factor BCL-6, which are all critical for Tfh cell differentiation (37,38). In turn, Tfh cells provide signals via costimulatory molecules and cytokines that help establish and maintain GCs. Thus, the generation of Tfh cells and GC reactions requires intimate communication between T and B cells involving multiple receptors that activate PI3K.

ICOS-PI3K Pathways in Tfh Cells
One of the key costimulatory receptors expressed by Tfh cells is ICOS, a CD28 family member. CD28 and ICOS both activate PI3Kδ and are required for Tfh cell development and function. CD28-CD80/CD86 interactions are involved in early T cell activation, including initial induction of ICOS, BCL-6, and CXCR5 (39), which are necessary for Tfh cell formation; Cd28 −/− mice show a total absence of Tfh cells and thymus-dependent (TD) germinal centers (40)(41)(42). ICOS is upregulated on activated T cells shortly after TCR stimulation and interacts with ICOS-ligand (ICOS-L) on antigen presenting cells including DCs and B cells (43). Icos −/− and Icosl −/− mice display severely reduced humoral response to TDantigens characterized by a lack of immunological memory and defective GC formation (44)(45)(46)(47)(48)(49). Patients lacking ICOS display a common variable immunodeficiency (CVID) with FIGURE 1 | PI3K signaling pathways in Tfh cell differentiation. Class IA PI3Ks (PI3Kα, PI3Kβ, PI3Kδ) are lipid kinases composed by a regulatory (p85) and a catalytic (p110) subunit. Multiple receptors activate PI3K in CD4 + T cells, including TCR, CD28, ICOS, OX-40 and IL-21R, leading to PI3K recruitment to the plasma membrane and conversion of the membrane lipid PI(4,5)P 2 to PI(3,4,5)P 3 . In T cells, chemokine receptors, including CXCR5, preferentially drive the activation of the class IB PI3Kγ (19). PI3K activity is counteracted by the inhibitor receptor PD-1 that blocks CD28 signal transduction through SHP-2 recruitment and PTEN induction. The phosphatases PTEN and SHIP-1/2 counteract PI3K signaling by converting PIP 3 to PI(4,5)P 2 and PI(3,4)P 2 , respectively. PIP 3 recruits to the plasma membrane proteins containing pleckstrin homology domains, such as AKT and PDK1. The serine/threonine kinase AKT gets activated by phosphorylation by PDK1 (at Thr308) and mTORC2 (at Ser473). In turn, activated pAKT phosphorylates inhibitors of mTORC1 leading to its activation. mTORC1 phosphorylates several factors including S6-kinase (S6K), that phosphorylates S6, driving protein synthesis and cell proliferation, important events for Tfh cell differentiation. pAKT also phosphorylates the FOXO-1 transcription factor leading to its export outside the nucleus and degradation by binding 14-3-3 proteins. FOXO-1 represses Bcl6, essential for Tfh cells, and promotes expression of multiple genes including Sell, Tcf7, Ccr7, and Klf2. While KLF-2 restrains Tfh cell program through multiple mechanisms, TCF-1 promotes Tfh cell formation by inhibiting Il-2rα, Blimp1, Ifng. At the same time, it has been shown that mTORC2-pAKT may also support TCF-1 activity through the inactivation of GSK3β, an inhibitor of β-catenin and TCF-1. Overall, PI3K pathways drive BCL-6 + Tfh cell differentiation that coordinates GC responses and humoral immunity after infections and vaccination through the generation of memory B cells and long-lived plasma cells (LLPC). reduced circulating Tfh cells (50,51). Conversely, mutations affecting Roquin, a negative post-transcriptional regulator of ICOS mRNA, increase Tfh cells and drive autoimmunity (52,53). ICOS helps drive multiple stages of Tfh cell differentiation, including the early generation of CXCR5 high T cells, modulation of other chemokine and homing receptors through regulation of the KLF2 transcription factor (39,54), and T:B cell non-cognate interactions that promote T cell motility at the T:B cell border (55). ICOS-ICOS-L interactions are also critical for localization and maintenance of GC-Tfh cells (9, 39, 54).
The essential role of PI3K in ICOS function was highlighted by data showing that mutation of the p85-binding site, which selectively abrogates PI3K recruitment, led to defects in Tfh cell formation similar to ICOS-deficiency (56). Inhibition of p110δ also prevented ICOS-mediated changes in cell migration and morphology in vitro (55). Conversely, we found that activated-PI3Kδ mice show T cell-intrinsic increases in Tfh cell differentiation, even in the presence of blocking anti-ICOS-L antibody, therefore bypassing the requirement for ICOS for Tfh cell development (16). Thus, PI3K appears to be a major effector of ICOS, required for Tfh cell formation and maintenance.
Nonetheless, despite increased GC-Tfh cell differentiation, Pik3cd E1020K/+ mice show disorganized GCs with increased Tfh cell infiltration and impaired class-switched antigen-specific responses to immunization (16,27,28). Multiple factors may contribute to these poor responses, including impaired B cell selection due to increased Tfh cells (7) and Tfh cell mislocalization (16), or intrinsic B cell defects. Indeed, although deletion of p110δ in B cells only minimally affected GC formation and T cell-dependent humoral responses after protein immunization (13), activated-PI3Kδ drove B cell-intrinsic increases in GC B and plasma cells, as well as impaired classswitched antibody production (16,28). Increased GC B cells may in turn further drive expanded Tfh cell numbers, contributing to immune dysregulation. Additionally, increased GCs that fill the follicular dendritic cell network at baseline, may prevent new GC formation as mice age (16). Whether and how FOXO1 contributes to defects in antigen-specific responses or whether additional downstream effectors of PI3K are involved remain intriguing questions. It should also be noted that additional receptors expressed by Tfh cells, including OX-40, and IL-21R activate PI3K (64) and may contribute to expanded Tfh cell populations in these mice; in contrast, PD-1, an inhibitory receptor highly expressed by Tfh cells (65), counteracts PI3K by blocking CD28 signaling and increasing PTEN expression (66)(67)(68)(69) (Figure 1).

Viral Infection
Although the connection between ICOS and PI3Kδ provides strong evidence for PI3Kδ driving Tfh cells, the view that PI3K exclusively promotes Tfh cells may be a simplification; this is particularly apparent when looking at the differentiation of Tfh vs. Th1 cells during viral infection. In response to viral or strong Th1 polarizing infections, CD4 + T cells undergo an early bifurcation such that up to 50% of viral-specific T cells express BCL-6 and become Tfh cells, while the rest express BLIMP1 and SLAM, and become IFNγ-producing Th1 cells (54). Although activated-PI3Kδ increased percentages of Tfh cells at baseline and in response to immunization (16), as well as polyclonal CXCR5 + PD-1 + Tfh cells after LCMV infection (Figure 2A), it did not alter Tfh cell percentages, nor Tfh/Th1 ratios, within viral-specific GP66-tetramer + CD4 + T cells in the same mice (Figures 2B,C). We also observed increased percentages of circulating CXCR3 + Tfh1 cells in patients with APDS/PASLI compared to controls (16), suggesting that PI3K can drive both Tfh cells and Type 1 immunity. Thus, PI3K activity may promote multiple effector T cell lineages and the effects of PI3K on Tfh cells may depend on the activating stimuli and microenvironment.

Metabolic Pathways in Tfh vs. Th1 Cells
Other PI3K-mediated signaling pathways that may influence both Tfh and Th1 cells are those involving mTORC1 and mTORC2. During acute LCMV infection, Th1 cells appear more proliferative and bio-energetically demanding with greater glucose metabolism and metabolic respiration than Tfh cells (83). Data suggest that these Th1 cells were more dependent cells (gated on GP66-tetramer + CD4 + B220 − Foxp3 − T cells). We were unable to evaluate GP66 + Tfr cells due to the low numbers of GP66 + Foxp3 + cells (70). However, polyclonal Tfr cells were increased after LCMV infection (data not shown), as we have previously reported in naïve mice (16). (A-C), n = 5-8. Data are representative of three independent experiments and are expressed as mean ± SEM with each dot indicating one mouse. Significance analyzed by Mann-Whitney U-test. ** p < 0.01; *** p < 0.001. on the IL-2-PI3K-AKT-mTORC1 axis, which preferentially promoted BLIMP1 + Th1 cells at the expense of BCL-6 + Tfh cells and humoral immunity (83,84). However, other studies have demonstrated requirements for mTORC1 and mTORC2 in driving Tfh cells in Peyer's Patches at steady state and in the periphery after LCMV infection and immunization (30,85). Mechanistically, Tfh cells were supported by mTORC1promotion of pS6, GLUT1 expression, glycolysis, lipogenesis and overall proliferation; and by mTORC2-pAKT, which decreased FOXO1 activity (30).
While these studies provide conflicting conclusions on the requirements for PI3K and downstream effectors for Tfh cells, this may result from different experimental systems (knockdown vs. knockout) as well as bio-energetic demands during immune challenges. However, there is also evidence that mTOR may be activated independently of PI3K via pathways involving nutrient sensing that may also affect T helper cell differentiation (22,79,86,87).

PI3K-TCF-1 Cross-Talk
Several recent studies revealed that the transcription factor TCF-1 is expressed at high levels in Tfh cells after viral infection and plays an essential role in their generation and maintenance, via repression of Il2ra, and Prdm1 (which encodes BLIMP1), promotion of Bcl6 (55,(88)(89)(90), and possibly repression of Ifng (91). Intriguingly, PI3K has been implicated both positively and negatively in TCF-1 regulation (92,93). In CD8 + T cells, Tcf7, which encodes for TCF-1, is induced by FOXO1 (94), and both are required for memory T cell formation (95)(96)(97)(98). Strong PI3K signaling would therefore be expected to decrease TCF-1 levels (25), as observed in studies of asymmetric cell division (92,93). Conversely, a positive link between PI3K/AKT and TCF-1 has been proposed via β-catenin (85), a coactivator of TCF-1 that is negatively regulated by phosphorylation by Glycogen Synthase Kinase 3β (GSK3β), which is inactivated by pAKT (Figure 1). mTORC2-deficient T cells, which do not fully activate AKT, show reduced β-catenin and TCF-1 (85). Nonetheless, most studies implicating TCF-1 in Tfh cell generation have been done in the context of strong Th1-inducing infections (88)(89)(90), and how these findings relate to Tfh cells in other contexts remains unknown. Thus, the relationship between PI3K and TCF-1, how they affect Tfh cell differentiation, and involvement in possible feedback loops remain intriguing questions.

PI3K PATHWAYS IN Tfr CELLS
A subset of thymic derived T regulatory cells, defined as T follicular regulatory (Tfr) cells, are localized at the T-B cell border and inside the GC area (99), and directly control the activation and differentiation of Tfh and GC B cells, including the development of autoimmunity (100). Tfr cells are phenotypically similar to Tfh cells and express BCL-6 (101-103), yet lack expression of B-cell-helper molecules, such as CD40L, and express inhibitory molecules CTLA-4, GITR, IL-10, and granzymes (104). Although excessive PI3K-mTOR activity is detrimental for induced-Treg cell differentiation (105-107), Pik3cd E1020K/+ mice show increased Treg and Tfr cells at steady state (16,26). Indeed, although Tfr cells derive from Tregs, Tfr have different requirements for their differentiation and function. For example, IL-2 is necessary for Treg development and suppressive capability (108), yet prevents Tfr cell differentiation in a BLIMP1-dependent manner, similar to IL-2's effects on Tfh cells (109). Tfr cells also display high mTORC1 activity that promotes differentiation and STAT3 phosphorylation, which induces Tcf7 and Bcl6 (110); increased AKT-mTOR activity in Roquin-deficient Treg cells upregulates Tfh cell gene signatures that drive Tfr cell differentiation (111). How PI3K affects ratios of Tfh:Tfr cells, which are important for regulating humoral responses and autoimmunity (104), is less clear; notably, Pik3cd E1020K/+ mice have parallel increases in both cell populations (16).
Recent data suggest that certain autoantibodies cross-react with gut microbiota, supporting links between the microbiome and autoimmunity (123). Interestingly, we found increased local and systemic immune responses against gut commensals in Pik3cd E1020K/+ mice, with evidence for cross-reactivity between anti-self and anti-bacterial antibodies. Furthermore, autoantibodies could be prevented by systemic antibiotic treatment (16). Such data highlight roles for PI3Kδ in modulating T and B lymphocyte activation, including that induced by the microbiota, which can lead to autoimmunity.

CONCLUDING REMARKS
Together, a growing body of evidence supports a connection between PI3Kδ and Tfh cell differentiation, raising the possibility that altered PI3K pathways may contribute to both immunodeficiency and autoimmunity. Nonetheless, results during viral infection suggest that effects of PI3K on Tfh cell differentiation may be context-dependent and that PI3K may promote multiple effector cell lineages. A recent report demonstrated that treatment of APDS/PASLI patients with leniolisib, a selective PI3Kδ inhibitor, showed promising improvements in cellular dysfunction and lympho-proliferation (124) Notably, selective PI3Kδ inhibition reduced serum IgM in vivo (124), while increasing IgG class-switching in vitro (16,27); however, effects on Tfh cells and autoantibodies have not yet been reported. Our results further suggest that evaluation of microbiota composition and systemic responses to gut commensals in APDS/PASLI may provide new opportunities, possibly in association with leniolisib, for managing this and other conditions where Tfh cells and autoantibodies contribute to pathogenesis. Such approaches may also be relevant for autoimmunity induced by checkpoint-blockade therapy, where PI3Kδ-inhibition may provide a selective control of immune responses. Finally, PI3Kδ activation may help improve vaccine responses, although this would have to be carefully assessed. Thus, a more comprehensive understanding of PI3K regulation and signaling in T and B cells is of crucial importance to more effectively improve humoral immune responses while minimizing autoimmunity.

Animal Care and Ethics
Control (C57Bl/6J) and Pik3cd E1020K/+ mice (16) were maintained and treated under specific pathogen-free (SPF) conditions under protocols reviewed and approved by the NINDS (protocol 1295-12) and NHGRI (protocol G98-3) Animal Care and Use Committees at the NIH.

AUTHOR CONTRIBUTIONS
SP designed and performed experiments, analyzed and interpreted data, wrote the manuscript and prepared the figures. BH and JLC performed experiments, edited the manuscript, and contributed to discussions. DBM contributed essential reagents and advice. PLS conceived the project, wrote the manuscript, and provided overall direction of the study. All authors concur with this submission.

FUNDING
This work was supported in part by funds from the intramural programs of the National Human Genome Research Institute, National Institute of Allergy and Infectious Diseases and National Institute of Neurological Disorders and Stroke, NIH.

ACKNOWLEDGMENTS
We thank members of the PLS and DBM labs for their assistance with this work. GP66 tetramer [I-A(b) QVYSLIRPNENPAHK PE] was obtained from the NIH tetramer facility (Emory University).