The origins, function, and regulation of T follicular helper cells

The term “Tfh cell” was first used in the year 2000 to describe a subset of CD4⁺ T cells present in human lymphoid tissues (tonsils) that expresses the chemokine receptor CXCR5 and functions primarily to provide help to B cells (Breitfeld et al., 2000; Schaerli et al., 2000). Expression of CXCR5, together with loss of the T cell zone–homing chemokine receptor CCR7, allows Tfh cells to relocate from the T cell zone to the B cell follicles, where they are positioned to directly support B cell expansion and differentiation (Ansel et al., 1999; Hardtke et al., 2005; Haynes et al., 2007; Fig. 1).

Tfh cells can be distinguished from other CD4⁺ T cell lineages by their low expression levels of cytokines (IFN-γ, IL-4, and IL-17) and transcription factors (T-bet, GATA3, and RORγt) characteristic of Th1, Th2, and Th17 cells, respectively (Chtanova et al., 2004; Kim et al., 2004; Rasheed et al., 2006; Nurieva et al., 2008; Vogelzang et al., 2008; Ma et al., 2009). Furthermore, Tfh cells express a unique combination of effector molecules that are critical for their development and function, including high levels of the surface receptors ICOS, CD40 ligand (CD40L), OX40, PD-1, BTLA and CD84, the cytokine IL-21, the cytoplasmic adaptor protein SLAM-associated protein (SAP), and the transcription factors Bcl-6 and c-Maf (Breitfeld et al., 2000; Schaerli et al., 2000; Chtanova et al., 2004; Vinuesa et al., 2005a; Rasheed et al., 2006; Haynes et al., 2007; Lim and Kim, 2007; Johnston et al., 2009; Ma et al., 2009; Deenick et al., 2010; Kroenke et al., 2012). These molecules play critical roles in promoting the activation, differentiation, and survival of B cells and/or CD4⁺ T cells. For instance, CD40L rescues B cells from apoptosis and promotes their proliferation (Tangye et al., 2012), and IL-21 enhances the differentiation of CD40L-stimulated human B cells, inducing secretion of all Ig isotypes (Bryant et al., 2007; Avery et al., 2008; Ettinger et al., 2008). Engaging ICOS on CD4⁺ T cells induces the production of B helper cytokines, such as IL-10 and IL-21 (Hutloff et al., 1999; Bauquet et al., 2009; Chevalier et al., 2011), and SAP is necessary for CD4⁺ T cells to form stable contacts with cognate B cells to promote their activation and differentiation (Qi et al., 2008; Cannons et al., 2010). The importance of these molecules in Tfh function is underscored by the finding that targeted deletion of Cd40/Cd40l, Icos, Il21/Il21r, or Sh2d1a (encoding SAP) in mice, or loss-of-function mutations in ICOS, CD40L, and SH2D1A in humans (Al-Herz et al., 2011; Crotty, 2011; Tangye et al., 2012), severely compromise the generation of robust GC reactions and subsequent long-lived serological memory against TD Ags (Table 1).

Provision of help to B cells during TD B cell responses requires that CD4⁺ T cells be in close proximity to cognate B cells. This occurs at the T cell–B cell border/interfollicular zones and within GCs (Fig. 1 A). Because of its role in directing cells toward B cell areas, CXCR5 was an early defining marker of Tfh cells. More recently, Bcl-6 has been described as a key molecular feature of these cells. It is now clear, however, that there is marked heterogeneity among CD4⁺CXCR5⁺Bcl-6⁺ T cells with respect to phenotype, function, and anatomical location. Indeed, the CXCR5⁺Bcl-6⁺ population comprises at least two subsets: early Tfh cells, which most likely provide help at the T-B border or in interfollicular zones before GC formation (Kerfoot et al., 2011; Kitano et al., 2011; Lee et al., 2011), and later-arising Tfh cells that are found in the GC (Yusuf et al., 2010; Choi et al., 2011).

In human tonsils, there are two major CD4⁺CXCR5⁺Bcl-6⁺ T cell populations that can be distinguished based on high versus low expression of CXCR5. Those expressing the highest levels of CXCR5 also express the most abundant PD-1, ICOS, CD40L, and Bcl-6, and contain the highest frequencies of cells producing IL-21 and the CXCR5 ligand CXCL13 (Lim and Kim, 2007; Ma et al., 2009; Bentebibel et al., 2011; Kroenke et al., 2012). These cells also express the highest levels of IL-4 (Ma et al., 2009; Bentebibel et al., 2011; Kroenke et al., 2012). Although this finding seems to contradict initial findings showing that IL-4 was not prominently expressed by Tfh cells, these earlier studies examined bulk CD4⁺CXCR5⁺ T cells, rather than distinct subsets (Chtanova et al., 2004; Kim et al., 2004; Rasheed et al., 2006). CD57 is expressed by ∼30% of tonsillar CD4⁺CXCR5^hi T cells (Rasheed et al., 2006; Lim and Kim, 2007; Ma et al., 2009), and this population localizes to GCs (Banerjee and Thibert, 1983; Hutloff et al., 1999; Kim et al., 2001). Importantly, CD4⁺CXCR5^hi T cells are capable of supporting Ig class switching and Ab secretion by naive, memory, and GC B cells in vitro (Kim et al., 2001; Avery et al., 2008; Ma et al., 2009; Bentebibel et al., 2011). Thus, in humans the CXCR5^hiCD57⁺ subset of CD4⁺ T cells correspond to GC Tfh cells. The exact relationship between the CD57⁻ and CD57⁺ subsets of CXCR5^hi CD4⁺ T cells remains to be determined.

CXCR5^loBcl-6⁺ CD4⁺ T cells also express low levels of ICOS and are found outside of GCs. These cells provide help to naive and memory, but not GC, B cells (Bentebibel et al., 2011) and may function during the initial phase of B cell activation at the T cell–B cell border. It is also possible that these cells migrate to the GC after initial interactions with naive B cells and mediate the differentiation of GC B cells—a possibility consistent with the finding that CXCR5^loICOS^lo CD4⁺ T cells up-regulate PD-1 and ICOS to levels approximating that on GC Tfh cells after co-culture with naive B cells (Bentebibel et al., 2011). Thus, CD4⁺ T cells expressing low levels of CXCR5 not only provide help to B cells early in the response, but probably also contain precursors from which GC Tfh cells emerge.

Similarly, murine CXCR5^hiBcl-6⁺ CD4⁺ T cells include those that are found in the follicle or within GCs. GL7 has been reported to mark a subset of murine Tfh cells expressing the highest levels of CXCR5, Bcl-6, PD-1, ICOS, CD200, and IL-21 that localize to GCs (Yusuf et al., 2010). Another study, however, found GL7 expressed by CD4⁺CXCR5⁺ T cells in the interfollicular zone (Kerfoot et al., 2011). Additional studies are required to definitively show whether GC Tfh cells can be differentiated from non-GC Tfh cells based on phenotype rather than localization alone. Another subset of murine CD4⁺ T cells provides the requisite help for the early differentiation of B cells into extrafollicular plasmablasts producing low-affinity Ig (Lee et al., 2011). Reminiscent of Tfh cells, these extrafollicular Th cells express Bcl-6 and CXCR5 and provide help to B cells via the production of IL-21. However, they can be distinguished from GC Tfh cells by lower levels of PD-1 and their localization to the T cell–B cell border (Lee et al., 2011). It is likely that these cells represent progenitors of Tfh/GC Tfh cells, and they may well correspond to human CXCR5^loICOS^loBcl-6⁺ CD4⁺ T cells that localize outside of GCs (Bentebibel et al., 2011; Fig. 1 A).

Substantial heterogeneity among Tfh cells was recently revealed by the description of IL-21 reporter mice (Lüthje et al., 2012). Approximately 20–40% of Tfh (i.e., CXCR5⁺PD-1⁺ CD4⁺) cells expressed IL-21 during immune responses against peptide Ag, and these cells also expressed substantially higher levels of IFN-γ, IL-10, and Tbx21 (encoding T-bet) than did IL-21⁻ Tfh cells. On the other hand, Bcl6, GATA3, and Il4 expression, as well as ex vivo proliferation, were comparable between IL-21⁺ and IL-21⁻ Tfh cells. Transfer studies demonstrated that IL-21⁻ Tfh cells could give rise to IL-21⁺ Tfh cells after reexposure to the initiating Ag, demonstrating a precursor/product relationship, at least in the setting of a recall response.

The key features of Tfh cells—production of IL-21, elevated expression of CXCR5, ICOS, and BCL-6, and provision of B cell help—have been used as surrogates to identify cytokines that potentially contribute to their generation.

Several cytokines, including IL-6, IL-21, and IL-27, induce IL-21 production by naive murine and human CD4⁺ T cells in vitro in a STAT3-dependent manner (Huber et al., 2008; Nurieva et al., 2008; Suto et al., 2008; Dienz et al., 2009; Batten et al., 2010; Eto et al., 2011; Lu et al., 2011; Ma et al., 2012). Interestingly, IL-12 also induces IL-21 in human and mouse CD4⁺ T cells (Schmitt et al., 2009; Ma et al., 2009, 2012; Nakayamada et al., 2011; Eto et al., 2011; Diehl et al., 2012). Although IL-12 is often associated with activation of STAT4, its ability to induce IL-21 expression was also largely STAT3 dependent (Nakayamada et al., 2011; Ma et al., 2012). Activation of CD4⁺ T cells with IL-21–inducing cytokines also resulted in elevated expression of CXCR5, ICOS, and Bcl-6, and improved Ab production by co-cultured B cells (Nurieva et al., 2008; Dienz et al., 2009; Ma et al., 2009, 2012; Schmitt et al., 2009; Eto et al., 2011; Nakayamada et al., 2011; Diehl et al., 2012). Collectively, these findings suggested that, akin to Th1, Th2, and Th17 lineages, cytokines present within the stimulating microenvironment in vivo influence the commitment of naive CD4⁺ T cells to the Tfh lineage. However, the in vitro conditions used do not completely recapitulate physiological Tfh differentiation, as in vitro–generated Tfh cells express lower levels of CXCR5, Bcl-6, and ICOS than GC Tfh cells and lack expression of other Tfh-associated molecules (Ma et al., 2009, 2012; Schmitt et al., 2009; Eto et al., 2011). Thus, the differentiation state of in vitro–derived Tfh cells more closely corresponds to pre–Tfh than to GC Tfh cells.

Conflicting results have been obtained regarding the relative importance of these cytokines for Tfh cell induction in vivo. IL-6 or IL-21 deficiency has been reported to have mild, moderate, or no effect on Tfh cell numbers (Nurieva et al., 2008; Vogelzang et al., 2008; Linterman et al., 2010; Poholek et al., 2010; Zotos et al., 2010; Eto et al., 2011; Harker et al., 2011). Despite conjecture regarding the individual roles of IL-6 and IL-21 in driving Tfh formation, these cytokines are likely to collaborate, as concomitant blockade of both cytokines was required to cause a significant decrease in Tfh cells in vivo (Eto et al., 2011). This finding points to redundancy among cytokines in Tfh development such that deficiency in IL-6 or IL-21 is compensated by other factors. Indeed, Tfh generation is also impaired in mice lacking IL-27R (Batten et al., 2010). Thus, in vivo and in vitro findings provide firm evidence for an important role for STAT3 and STAT3-activating cytokines in regulating Tfh development (Table 1).

The findings that cytokines could induce some, but not all, features of Tfh cells indicated that additional signals are required for full Tfh differentiation. Obvious candidates for these signals are specific receptor/ligands pairs on CD4⁺ T cells and APCs.

T cell activation requires engagement of the TCR, CD28, CD40L, and ICOS by peptide/MHC class II, CD80/CD86, CD40, and ICOS-L, respectively. In the absence of signaling though CD28, ICOS/ICOS-L or CD40/CD40L, CD4⁺ T cells fail to up-regulate CXCR5, resulting in reduced numbers of Tfh cells and impaired GC formation (Akiba et al., 2005; Bossaller et al., 2006; Choi et al., 2011; Salek-Ardakani et al., 2011). This is consistent with reduced frequencies of circulating CD4⁺CXCR5⁺ T cells in patients with mutations in ICOS or CD40LG (Bossaller et al., 2006). Interestingly, deficiency of CD86, but not CD80, recapitulates the Tfh defect seen in Cd28-deficient mice (Salek-Ardakani et al., 2011), suggesting that ligation of CD28 by CD86, rather than CD80, has a preferential role in Tfh development (Table 1).

Another important set of molecules involved in T cell–B cell interactions are the SLAM family of receptors—SLAM, CD84, 2B4, Ly9, and NTB-A (Ly108 in mice; Ma et al., 2007)—which elicit intracellular signaling by recruiting SAP. Remarkably, SAP-deficient CD4⁺ T cells cannot support TD B cell responses caused by an inability to form stable conjugates with cognate B cells and differentiate into Tfh cells (Crotty et al., 2003; Qi et al., 2008; Linterman et al., 2009; Cannons et al., 2010; Deenick et al., 2010; Yusuf et al., 2010). The gene encoding SAP (SH2D1A) is mutated in patients with X-linked lymphoproliferative disease, who exhibit impaired Tfh-like function (Ma et al., 2005, 2007). CD84 is also required for Tfh differentiation, as CD4⁺ T cells lacking CD84 generate fewer Tfh cells in vivo, although the deficit is less severe than in Sh2d1a-deficient mice (Cannons et al., 2010). Thus, although CD84 represents one of the SAP-associating receptors required for Tfh formation, SAP-dependent signaling downstream of additional SLAM family receptors, such as Ly108 (Cannons et al., 2010), is also important (Table 1).

Although Tfh cells direct the differentiation of Ag-specific B cells, it is clear that reciprocal signals from B cells are also important for Tfh formation. Specifically, mice lacking B cells or those with B cells deficient in functional molecules (CD19, CD40, MHC class II, ICOS-L, and CD86) exhibited decreased numbers of Tfh cells after immunization or pathogen infection (Haynes et al., 2007; Johnston et al., 2009; Zaretsky et al., 2009; Deenick et al., 2010; Poholek et al., 2010; Salek-Ardakani et al., 2011). This suggests that Tfh generation is critically dependent on interactions between activated CD4⁺ T cells and Ag-presenting B cells. Indeed, the Tfh deficiency seen in the absence of SAP reflected the requirement for SAP in the formation of stable conjugates between CD4⁺ T and B cells, but not DCs (Qi et al., 2008; Cannons et al., 2010). Moreover, ICOS expression specifically on B cells is required for GC function (Nurieva et al., 2008), and B cells are uniquely able to induce Tfh features in co-cultured CD4⁺ T cells (Chevalier et al., 2011).

However, the absolute requirement for B cells in the formation of early Tfh cells has been questioned by the finding that provision of excess Ag facilitates Tfh cell formation in the absence of B cells, CD40, or MHC class II (Deenick et al., 2010). This suggests that sustained Ag presentation by non–B cell APCs can support Tfh generation, and the previous conclusion that B cells are required for this process may have simply reflected the fact that they become the predominant APC during a TD immune response. This alternate view of Tfh formation has been validated by several recent studies. First, Ag-specific CD4⁺CXCR5⁺Bcl-6⁺ T cells can be detected within 3 d of immunization/infection, preceding the activation of cognate B cells (Choi et al., 2011; Kerfoot et al., 2011; Kitano et al., 2011). In addition, the early appearance of Tfh-like cells in lymphoid follicles can occur in the absence of Ag-specific B cells (Poholek et al., 2010; Baumjohann et al., 2011; Choi et al., 2011; Kerfoot et al., 2011; Kitano et al., 2011). Formation of CD4⁺CXCR5⁺Bcl-6⁺ pre–Tfh type cells is also independent of SAP (Choi et al., 2011), which is reminiscent of the SAP-independent priming of CD4⁺ T cells by DCs (Qi et al., 2008). Finally, direct immunization with MHC class II–restricted Ag-pulsed DCs, which circumvents Ag presentation by B cells, resulted in early Tfh cell generation (Baumjohann et al., 2011; Choi et al., 2011; Goenka et al., 2011).

Collectively, these findings provided convincing evidence that the initial generation of Tfh cells is actually B cell–independent and is instead driven by DCs. Indeed, initial interactions between DC and naive CD4⁺ T cells facilitate the relocation of primed CD4⁺ T cells from the interfollicular and T cell zones to the B cell follicle to become pre–/early Tfh cells (Fig. 1, A and B). However, B cells are important for complete Tfh differentiation, as they provide signals that reinforce the Tfh phenotype and/or promote their survival (Fig. 1 B). Central to this is the expression of ICOS-L on B cells, which ligates ICOS on CD4⁺ T cells and induces the expression of key Tfh molecules including IL-21, Bcl-6, and c-Maf (Bauquet et al., 2009; Gigoux et al., 2009; Chevalier et al., 2011; Choi et al., 2011). SAP and SAP-associating receptors are also critical for full Tfh formation and function (Fig. 1 B), as the short-lived interactions between SAP- or CD84-deficient CD4⁺ T and B cells results in impaired help to B cells and precludes B cells from providing the signals necessary to maintain Tfh cells (Qi et al., 2008; Cannons et al., 2010; Deenick et al., 2010). Cytokines will also contribute to Tfh formation (Fig. 1 B); however, their roles are likely dictated by the type of immune responses generated and the availability of other requisite signals provided by APCs.

A common tenet in immunology is the notion of “master regulators” of lymphocyte differentiation, whereby a single transcription factor is necessary and sufficient for the commitment of naive cells to a specific lineage of effector cells. This is exemplified by the requirement for T-bet, GATA3, RORγt, and FoxP3 for the differentiation of naive CD4⁺ T cells into Th1, Th2, Th17, and T reg cells, respectively (O’Shea and Paul, 2010).

For Tfh cells, a similar “master regulator” function has now been ascribed to Bcl-6, which was originally found to be expressed by a subset of CD4⁺ T cells in GCs in human tonsils (Cattoretti et al., 1995). Studies performed 10–15 yr later revealed that these Bcl-6⁺ CD4⁺ T cells were Tfh cells (Chtanova et al., 2004; Vinuesa et al., 2005a; Rasheed et al., 2006; Ma et al., 2009), and several studies have now shown that Bcl-6 deficiency inhibits, and overexpression promotes, the generation of Tfh cells (Johnston et al., 2009; Nurieva et al., 2009; Yu et al., 2009; Diehl et al., 2012; Kroenke et al., 2012), as well as extrafollicular CD4⁺ Th cells (Poholek et al., 2010; Lee et al., 2011). These findings firmly established a critical role for Bcl-6 in coordinating Tfh formation and validated the prescient observation by Cattoretti et al. (1995) that “Bcl-6 expression in T cells may also be associated with some GC-related function.”

Bcl-6 functions as a transcriptional repressor. Thus, it was initially difficult to invoke a mechanism by which Bcl-6 actively drives a Tfh program in naive CD4⁺ T cells. However, several scenarios have now been proposed. As Bcl-6 suppresses GATA3 expression and Th2 responses (Kusam et al., 2003), it could promote Tfh differentiation by interfering with the expression or function of transcription factors that control alternate Th cell fates (Fig. 2). Indeed, enforced expression of Bcl-6 in naive CD4⁺ T cells cultured under Th1 or Th17 conditions reduces the proportion of IFN-γ− or IL-17–expressing cells (Nurieva et al., 2009; Yu et al., 2009), most likely caused by its ability to bind to consensus sites in the promoters of Tbx21 and Rorc (encoding RORγt), and thus reduce protein expression and/or function. This is consistent with the enhanced generation of Tfh cells from Tbx21-deficient CD4⁺ T cells both in vivo and in vitro (Nakayamada et al., 2011; Oestreich et al., 2012).

Bcl-6 also inhibits the regulatory effects of Blimp-1 and the expression of a cluster of microRNAs that suppress Tfh generation (Yu et al., 2009). Bcl6 and Prdm1, which encodes Blimp-1, are reciprocally expressed in Tfh and effector CD4⁺ T cells (Fazilleau et al., 2009; Johnston et al., 2009; Ma et al., 2009; Choi et al., 2011), and deletion of Prdm1 from murine CD4⁺ T cells enhances Tfh formation in response to lymphocytic choriomeningitis (LCMV). In contrast, Prdm1 overexpression severely blunts Tfh genesis and attenuates GC reactions and Ab responses (Johnston et al., 2009). Thus, Bcl-6 and Blimp-1 function to control CD4⁺ T cell fate by promoting or inhibiting the generation of Tfh cells (Fig. 2 and Table 1).

A caveat to Bcl-6’s role as a master regulator of Tfh cell differentiation is the inability of ectopic Bcl-6 to induce substantial expression of CXCR5 or IL-21—two of the defining features of Tfh cells—in human or murine naive CD4⁺ T cells in vitro (Nurieva et al., 2009; Yu et al., 2009; Diehl et al., 2012; Kroenke et al., 2012). Thus, other transcription factors must also be involved in Tfh differentiation. Indeed, Tfh cells are also diminished in mice lacking c-Maf (Bauquet et al., 2009), BATF (Betz et al., 2010; Ise et al., 2011), and IRF4 (Kwon et al., 2009; Table 1). Consequently, Bcl-6 may not be the “master regulator” of the Tfh lineage.

c-Maf induces IL-21 production in activated CD4⁺ T cells in vivo (Bauquet et al., 2009) and in vitro (Hiramatsu et al., 2010; Kroenke et al., 2012), and c-Maf expression itself is induced in CD4⁺ T cells by IL-6 (Hiramatsu et al., 2010) or IL-27 (the latter requiring IL-21; Pot et al., 2009). Furthermore, IL-27–mediated induction of IL-21 is c-Maf dependent (Pot et al., 2009), suggesting the existence of a positive feedback loop. Ectopic expression of c-Maf in human naive CD4⁺ T cells also yields a greater proportion of cells expressing CXCR5 compared with control-transduced cells, and co-transduction of Bcl-6 and c-Maf resulted in higher levels of expression of ICOS and PD-1 compared with either factor alone, providing evidence of cooperative interplay between these transcription factors (Kroenke et al., 2012; Fig. 2).

BATF deficiency also causes a cell-autonomous defect in Tfh generation that appears to result, at least in part, from a requirement for BATF in the induction of Bcl-6 and c-Maf (Betz et al., 2010; Ise et al., 2011; Fig. 2). Tfh formation and help for GC responses are defective in BATF-deficient CD4⁺ T cells in vivo; this could be improved by expressing Bcl-6 alone or together with c-Maf, but not by c-Maf alone. However, the level of rescue was only ∼60–70% of that achieved by overexpressing BATF in these cells (Ise et al., 2011), suggesting that other, as yet unidentified, BATF targets contribute to Tfh differentiation. It is also unclear which signals induce BATF. It will be important to determine whether IL-6 and IL-27 can induce BATF, and whether cytokine-driven c-Maf expression is secondary to the induction of BATF.

These studies have revealed a hierarchy of the transcriptional regulation of Tfh cells. BATF and STAT3-activating cytokines induce expression of Bcl-6 and c-Maf. Bcl-6 induces signature features of Tfh cells by suppressing both the function of other lineage-specific transcription factors, as well as Blimp-1 and multiple microRNAs, and c-Maf induces IL-21. STAT3 can also directly induce IL-21. The function of IRF4 in Tfh cell formation remains to be determined; however, it is required for IL-21– and IL-6–induced expression of IL-21 in murine CD4⁺ T cells (Huber et al., 2008).

The eradication of infectious pathogens and the success of vaccines require Tfh cell–mediated B cell differentiation and Ab secretion. However, aberrant function of Tfh cells could trigger autoimmunity. Thus, processes must exist that restrict Tfh activity under normal circumstances.

The primary transcription factor that negatively regulates Tfh development is Blimp-1 (Johnston et al., 2009), which directly suppresses Bcl-6 expression (Fig. 2). Blimp-1 is induced by IL-2 in CD4⁺ T cells (Gong and Malek, 2007), and CD25 (IL-2Rα) expression positively correlates with Blimp-1 and negatively correlates with Bcl-6 and CXCR5 (Choi et al., 2011). Consistently, IL-2 blockade or targeted deletion of STAT5, which is activated by IL-2, increased Tfh cell formation (Johnston et al., 2012; Oestreich et al., 2012), whereas overexpression of constitutively active STAT5 reduced Tfh generation in vivo (Johnston et al., 2012). Similar findings were recently reported in a model of influenza virus infection, whereby repeated administration of IL-2 reduced virus-specific humoral immunity by directly suppressing Tfh cell formation (Ballesteros-Tato et al., 2012). The inhibitory effect of STAT5 was Blimp-1–dependent, as it was overcome by deleting Blimp-1 from CD4⁺ T cells (Johnston et al., 2012). STAT5 can also directly bind BCL6 and suppress its expression (Walker et al., 2007). Thus, the IL-2–STAT5 axis is likely to control Tfh generation by suppressing Bcl-6 both directly and indirectly via induction of Blimp-1 (Fig. 2 and Table 1).

T reg cells are a subset of CD4⁺ T cells that control the function of effector cells (Josefowicz et al., 2012). Remarkably, specific subsets of T reg cells can co-opt the differentiation programs of effector cells to selectively regulate distinct classes of immune responses. For instance, expression of T-bet, IRF4 or STAT3 in T reg cells results in cells that temper the activity of Th1, Th2 or Th17 cells, respectively (Josefowicz et al., 2012). The concept that T reg cells suppress Tfh cells is not novel, as T reg cells were identified in B cell follicles and GCs in 2004 (Lim et al., 2004). However, the nature of these T reg cells, their mechanisms of suppression, and their relevance to immune responses in vivo remained unexplored (Lim et al., 2004, 2005).

A subset of T reg cells that preferentially inhibits Tfh cells was suggested by the finding that ∼10–15% of the Tfh cells in murine and human lymphoid tissues express the T reg cell–specific transcription factor FoxP3 (Chung et al., 2011; Linterman et al., 2011; Wollenberg et al., 2011). These cells share features with both Tfh cells (expression of Bcl-6, CXCR5, PD1, and ICOS) and T reg cells (expression of GITR, CTLA4, KLRG1, CD25, and Blimp-1), but they differ from Tfh cells in that they lack CD40L, IL-4, and IL-21 (Linterman et al., 2011). Follicular T reg cells originate from thymic-derived (i.e., natural) T reg cells, rather than peripherally induced T reg cells or Tfh cells themselves, yet they exhibit requirements similar to conventional Tfh cells for their development and maintenance (Chung et al., 2011; Linterman et al., 2011; Wollenberg et al., 2011). Genetic ablation of Bcl-6 prevented the generation of follicular T reg cells, and CXCR5 deficiency disrupted their positioning within GCs. Both manipulations affected Tfh cell numbers only marginally but resulted in increased numbers of GC B cells and increased Ab production (Chung et al., 2011; Wollenberg et al., 2011). Thus, by hijacking the transcriptional machinery of Tfh cells and acquiring expression of CXCR5, follicular T reg cells can migrate into GCs to exert their regulatory effect (Fig. 3). Because human follicular T reg cells directly suppress Ab secretion by GC B cells, as well as Tfh cell function in vitro (Lim et al., 2004, 2005), it is likely that follicular T reg cells limit GC reactions by suppressing the in vivo responses of both Tfh and GC B cells. It will be interesting to determine whether some of the clinical symptoms in FOXP3 mutant IPEX patients, such as the production of auto-Abs, are caused by an inability of follicular T reg cells to restrain GC responses.

The effects of follicular T reg cells may be complemented by the actions of both regulatory CD8⁺ T cells and PCs. A subset of CD8⁺ T cells restricted by the nonclassical MHC molecule Qa1 was found to acquire expression of CXCR5, migrate to B cell follicles, and attenuate the function of Tfh cells (Kim et al., 2010). Unlike conventional T reg cells, regulatory CD8⁺ T cells lack expression of FoxP3 and CTLA4, and their suppressive capacity required their expression of perforin and a source of IL-15 (Kim et al., 2010). Preventing the interaction between Tfh cells and regulatory CD8⁺ T cells resulted in a dramatic increase in Tfh cells and the development of a fatal lupus-like disease (Kim et al., 2010).

PCs also negatively regulate Tfh cells, as Bcl6 and Il21 expression was abrogated in co-cultures of PC and in vivo–derived Tfh cells (Pelletier et al., 2010). Intriguingly, although PCs presented peptides to naive Ag-specific CD4⁺ T cells and induced T cell activation in vitro, expression of Bcl6 and Il21 (Pelletier et al., 2010) was not induced. This led to the proposal that PC limited the generation of Tfh cells, a scenario supported by increased numbers of Tfh cells in PC-deficient mice (Pelletier et al., 2010). However, this increase could also reflect the lack of Ab-mediated Ag clearance, thereby resulting in sustained T cell activation and Tfh formation. It also remains to be determined where and when PCs would interact with Tfh cells and how PCs would deliver such inhibitory signals.

In humans, a small subset of CD4⁺ T cells in peripheral blood expresses CXCR5 (Förster et al., 1994; Breitfeld et al., 2000; Schaerli et al., 2000). These cells express low levels of Tfh markers (ICOS, OX40, IL-4, IL-21, Bcl-6, and c-Maf; Förster et al., 1994; Breitfeld et al., 2000; Schaerli et al., 2000; Chevalier et al., 2011) and are diminished in individuals unable to form GCs because of mutations in CD40LG or ICOS (Bossaller et al., 2006). For this reason, it has been inferred that circulating CD4⁺CXCR5⁺ T cells derive from GCs and are presumably Tfh cells. Most circulating CD4⁺CXCR5⁺ T cells express CD45RO (Förster et al., 1994; Breitfeld et al., 2000; Schaerli et al., 2000; Bossaller et al., 2006; Chevalier et al., 2011; Ma et al., 2012), which is typically associated with a memory cell phenotype. However, it is important to note that CD45RO expression does not discriminate between long-lived memory cells and recently generated effector cells. Interestingly, in healthy donors that received a tetanus booster >2 yr earlier, circulating tetanus toxoid–specific CD4⁺ T cells resided only in the CXCR5⁻ subset of CD4⁺CD45RO⁺ T cells (Breitfeld et al., 2000). In contrast, CMV or influenza-specific cells were present in both the CXCR5⁻ and CXCR5⁺ subsets (Morita et al., 2011). The presence of virus-specific CD4⁺ T cells in the CXCR5⁺ subset may reflect periodic encounter with Ag, as CMV establishes a persistent, latent infection and exposure to influenza may occur seasonally. Tetanus toxoid, in contrast, is likely to be cleared after vaccination. Thus, circulating CD4⁺CXCR5⁺ T cells in humans may represent recently generated effector cells that have exited lymphoid follicles rather than long-lived memory cells.

Fazilleau et al. (2007) described a similar population of CD4⁺CXCR5⁺ T cells in draining lymph nodes of mice that persisted after the peak of GC responses (30 d after immunization). As in humans, these cells had down-regulated expression of signature Tfh molecules (ICOS, OX40, IL-4, and IL-21). Although these murine CD4⁺CXCR5⁺ T cells were designated as “memory” Tfh cells, they may result from recent Ag encounter, as complexes of specific Ag–MHC class II were clearly still detectable 70 d after immunization (Fazilleau et al., 2007). IL-21⁺ Tfh cells isolated from mice 30–55 d after immunization or infection and transferred into naive mice gave rise to both IL-21⁺ Tfh cells and conventional CXCR5⁻PD-1⁻IL-21⁻ effector cells (Lüthje et al., 2012) after reexposure to antigen. This suggests that Tfh cells can enter the pool of CXCR5⁻ memory CD4⁺ T cells, as described originally for tetanus-specific human CD4⁺ T cells (Breitfeld et al., 2000). Although these studies have shed some light on Tfh memory, further investigation is required to provide a complete understanding of the relationship between, and kinetics of formation of, conventional memory, Tfh memory, and effector cells.

T cell differentiation was once considered linear and irreversible, but it is now appreciated that this process is flexible, as so-called committed cells can acquire features of alternate effector fates (O’Shea and Paul, 2010). This has raised the question of whether Tfh cells are a distinct lineage of effector CD4⁺ T cells or whether they instead represent a stage of activation of Th1, Th2, and Th17 cells. Initial studies demonstrated that Tfh cells could be generated in vivo under conditions that cripple the formation of Th1, Th2, and Th17 cells (Nurieva et al., 2008), suggesting that Tfh cells are a distinct lineage. However, Tfh cells can arise from Th1 (i.e., CXCR5⁻T-bet⁺IFN-γ⁺) and Th2 (i.e., CXCR5⁻PD-1⁻Bcl6^loGATA3⁺IL-4⁺) precursors in response to infection with LCMV (Th1; Fahey et al., 2011), Heligmosomoides polygyrus (Th2), Schistosoma mansoni (Th2), or Nippostrongylus brasiliensis (Th2; King and Mohrs, 2009; Reinhardt et al., 2009; Zaretsky et al., 2009; Fig. 3). Tfh cells may also arise from Th17 cells. Indeed, Tfh and Th17 cells have several similarities, such as the molecular requirements for their generation (ICOS, c-Maf, IRF4, and STAT3), production of IL-21, and the ability to induce B cell differentiation (Breitfeld et al., 2000; Huber et al., 2008; Ma et al., 2008, 2012; Nurieva et al., 2008; Bauquet et al., 2009; Mitsdoerffer et al., 2010). This scenario, however, is speculative and requires examination in relevant in vitro or in vivo models.

Recent studies have provided potential explanations for this flexibility. First, kinetic analyses of CD4⁺ T cell responses have shown that CXCR5 and Bcl-6 are expressed within 1–3 d of Ag exposure (Kerfoot et al., 2011; Kitano et al., 2011). Early Bcl6 expression is likely facilitated by the open, active conformation of the Bcl6 locus in naive mouse CD4⁺ T cells, and permissive epigenetic marks (i.e., H3K4me3) in this locus in in vitro generated Th1, Th2, and Th17 cells (Lu et al., 2011; Nakayamada et al., 2011). In addition, in vitro stimulation of murine and human naive CD4⁺ T cells under Th1-polarizing conditions (i.e., IL-12) yields a transitional cell population with features of both Th1 (IFN-γ⁺T-bet⁺) and Tfh (IL-21⁺Bcl-6⁺) cells (Schmitt et al., 2009; Nakayamada et al., 2011; Ma et al., 2012). However, Th1-type cells predominated over time under these conditions, as the majority of IL-21⁺Bcl-6⁺ cells were eventually lost because of sustained expression of T-bet (Nakayamada et al., 2011) and its ability to bind to and inhibit Bcl-6 (Oestreich et al., 2012). Indeed, overexpression of T-bet in vitro reduces Bcl-6 expression in naive CD4⁺ T cells cultured with IL-12, whereas Tbx21-deficient CD4⁺ T cells generated more Tfh-like cells in vivo and in vitro (Nakayamada et al., 2011; Oestreich et al., 2012). Collectively, these findings suggest that Bcl6 is primed for rapid expression in activated CD4⁺ T cells, thereby providing the molecular machinery to guide CD4⁺ T cells to a Tfh fate (Fig. 3; Table 1).

The nature of the infection may also dictate the fate of Ag-specific CD4⁺ T cells. For example, Tfh cells are preferentially generated under conditions of pathogen/Ag persistence, such as chronic LCMV infection, whereas both Tfh and Th1 cells are generated in similar numbers and with similar kinetics after acute LCMV infection (Fahey et al., 2011). This deviation from Th1 cells during chronic infection correlated with increased expression of Bcl6 and reduced expression of Tbx21 and Prdm1. Thus, T-bet can constrain Bcl-6 expression under conditions of limiting Ag, leading to reduced Tfh formation (Nakayamada et al., 2011). However, abundant Ag—resulting from persistent/chronic infection—overrides these constraints to promote the Tfh program (Fahey et al., 2011), most likely by maintaining high levels of Bcl-6, which then antagonizes T-bet function and Th1 cell formation (Johnston et al., 2009; Nurieva et al., 2009; Yu et al., 2009; Oestreich et al., 2012; Fig. 3). This model is supported by the findings that high-affinity CD4⁺ T cells preferentially develop into Tfh cells (Fazilleau et al., 2009), and that boosting with peptide Ag promotes Tfh formation (Deenick et al., 2010). In a similar fashion, IL-4–expressing Th2 cells may convert into Tfh cells during infection with Th2-inducing pathogens (King and Mohrs, 2009; Reinhardt et al., 2009; Zaretsky et al., 2009), in which persistent pathogen exposure induces Bcl-6 expression, which in turn suppresses GATA3. Although this has not been formally tested, it is consistent with the severe Th2-mediated inflammation seen in Bcl-6–deficient mice (Kusam et al., 2003).

These studies add to the concept of plasticity in T cell differentiation and infer that T cell differentiation is neither absolute nor solely driven by master regulators. Rather, the ultimate fate of activated CD4⁺ T cells reflects the modular nature of the expression, function, and interplay of antagonistic and cooperative transcription factors.

Although Tfh cells are critical for the generation of effective long-lived protective Ab responses, dysregulation of Tfh function could manifest as various immune dyscrasias. This has indeed been established in both mice and men.

The first suggestion that Tfh cells could underlie immune-mediated diseases was their overabundance in several murine models of SLE (Vinuesa et al., 2005a; Subramanian et al., 2006) and the fact that preventing Tfh formation reduced disease (Hron et al., 2004; Linterman et al., 2009). Circulating Tfh-like cells are also found at increased frequencies in the blood of some patients with SLE, Sjogren’s syndrome (Simpson et al., 2010), and juvenile dermatomyositis (Morita et al., 2011), which are autoimmune conditions characterized by production of pathogenic auto-Abs. Importantly, the expansion of Tfh-like cells in SLE correlated with auto-Abs titers and tissue damage (Simpson et al., 2010). The flip-side to autoimmunity is immunodeficiency. Tfh cells are reduced in patients with mutations in STAT3 (Ma et al., 2012), CD40L, and ICOS (Bossaller et al., 2006), as well as in corresponding gene-targeted mice (Akiba et al., 2005; Bossaller et al., 2006; Nurieva et al., 2008), and are associated with severe defects in protective humoral immunity (Al-Herz et al., 2011; Crotty, 2011; Tangye et al., 2012). Furthermore, although Tfh cells can be generated under some conditions in SAP-deficient humans and mice, they exhibit compromised function (Ma et al., 2005; Deenick et al., 2010; Lu et al., 2011). Thus, excessive generation of Tfh cells likely contributes to the production of pathogenic auto-Abs in several human autoimmune conditions, and insufficient Tfh generation underlies impaired humoral immune responses in primary immunodeficiencies.

Since their formal identification just over a decade ago, our knowledge of Tfh cells has grown exponentially, and we now have a much clearer understanding of the cells, molecules and signaling pathways required for their differentiation, as well as the mechanisms by which Tfh function is induced and constrained. The rapid influx of information has by no means abated, as witnessed by the recent descriptions of subsets of invariant NKT cells that resemble Tfh cells and provide help for primary B cell responses to lipid Ags in vivo (Chang et al., 2012; King et al., 2012; Tonti et al., 2012). Together with previous reports of populations of CD8⁺ and γδ T cells with Tfh-like features in humans (Caccamo et al., 2006; Quigley et al., 2007), these findings illustrate that Tfh function is not restricted to classical TD B cell responses. Rather, subsets of Tfh cells appear to have evolved to promote Ab-mediated protection against specific microorganisms, such as glycolipid-rich pathogens (iNKT Tfh cells; Chang et al., 2012; King et al., 2012; Tonti et al., 2012) and intracellular pathogens such as mycobacteria (e.g., γδ Tfh cells; Caccamo et al., 2006). Mechanisms probably also evolved to ensure that the appropriate type of Tfh response is elicited while deleterious responses are suppressed. With such rapid advances in our understanding of Tfh cells, we are well placed to harness the power of these cells for therapeutic intervention and modulation. Thus, the identification of diverse subsets and functions of Tfh cells provides promise for novel strategies to improve vaccine development for inducing long-lived humoral protection against infectious diseases. Similarly, promoting or attenuating Tfh cell development or function may be an attractive rationale for treating immunological diseases such as immunodeficiency or autoimmunity. Hopefully, the next decade of Tfh research will see some of these possibilities realized.

We thank Shane Crotty for providing data before publication, and helpful discussion.

The authors’ research into Tfh cells is supported by grants and fellowships awarded by the National Health and Medical Research Council of Australia.