William et al. (2002) observed that the splenic autoreactive B cells of autoimmune MRL/lpr mice proliferated and undergo active somatic hypermutation at the T zone-red pulp border rather than in GC. They examined the extrafollicular generation of plasmablasts in AM14 VH transgenic (Tg) mice, which possess rheumatoid factor (RF)-producing B cells with moderate affinity for IgG2a. Intriguingly, AM14 B cells on the MRL/lpr background spontaneously differentiate into extrafollicular plasmablasts and undergo somatic hypermutation at the T zone/red pulp border. In addition, they reported that the extrafollicular plasmablast response is induced by the administration of IgG2a anti-chromatin antibodies, which presumably form immune complexes in vivo with endogenous chromatin (Herlands et al., 2007). This response was found to be T cell independent, although it was totally dependent on MyD88 signaling downstream of Toll-like receptor 7 (TLR7) and TLR9 (Herlands et al., 2008). However, another study revealed that although AM14 B cells can be activated, differentiate, and undergo isotype-switching independent of antigen-specific T helper cells, T cells dramatically enhance the AM14 B cell response via CD40L and IL-21 signaling (Sweet et al., 2011).

Because affinity-enhancing somatic hypermutations are prevalent in autoantibodies, it has long been hypothesized that these autoantibodies are derived from GC. Mouse strains that frequently develop autoimmune diseases (NZB/W F1, BXSB, MRL/lpr, sanroque, and NOD mice) spontaneously form GC-like structures in their spleens, and the onset of autoantibody production correlates with GC formation. Recently, several pieces of evidence have suggested that dysregulated T follicular helper (T_FH) cells significantly contribute to autoimmunity by inducing the aberrant selection of autoreactive B cells. The lupus-like disease that occurs in sanroque mice is caused by Roquin^san/san-induced accumulation of T_FH cells that maintain spontaneously formed GC (Vinuesa et al., 2005). The glomerulonephritis and pathogenic autoantibody production displayed by sanroque mice are ameliorated by Bcl6 haploinsufficiency (Linterman et al., 2009). Moreover, SLAM-associated protein (SAP) deficiency experiments have highlighted the important roles played by T_FH cells in the conditions suffered by sanroque mice. SAP interacts with a conserved tyrosine-based motif that is found in the cytoplasmic tail of SLAM family members, and Sh2d1a (the gene for SAP) deficiency abrogates T_FH formation and GC responses, but not extrafollicular antibody responses. Since SAP deficiency ameliorates the lupus-like phenotype of sanroque mice, it can be assumed that aberrant T_FH cell activation is responsible for the autoimmunity that they display. BXSB mice develop a severe form of lupus caused by the yaa locus, which induces the overexpression of a cluster of X-linked genes that includes Tlr7 gene. Although B6.yaa mice are not overtly autoimmune, the introduction of Sle1, which contains the autoimmune-predisposing Slam/Cd2 haplotype, into their genome causes them to develop fetal lupus (Subramanian et al., 2006). Intriguingly, CD4⁺ T cells from B6.Sle1.yaa mice develop the molecular signature of T_FH cells and also show altered expression levels of various cytokines and chemokines.

As discussed above, dysregulation of the follicular or extrafollicular pathway can cause systemic autoimmune disease in mice. However, the contributions of follicular and extrafollicular checkpoints to the production of disease-associated autoantibodies are more difficult to evaluate in humans than in mice. Levels of autoantibodies do not always correlate with disease activity and response to treatment. For example, the serum concentrations of some autoantibodies correlate with disease activity (i.e., anti-double stranded DNA antibodies and anti-PR3 antibodies), while the titers of other autoantibodies [i.e., anti-ribonucleoprotein (RNP) antibodies and anti-Ro and La antibodies] remain stable irrespective of disease status. The heterogeneous autoantibody effects have also been observed in patients treated with anti-CD20 monoclonal antibody, which depletes B cells and plasmablasts but not long-lived plasma cells (Cambridge et al., 2003, 2006; Lu et al., 2009). In lupus patients, the levels of anti-nucleosome and anti-double stranded DNA antibodies are significantly decreased at 6–8 months after the administration of anti-CD20 monoclonal antibody. In contrast, the same treatment does not significantly alter the levels of anti-Ro, Sm, or RNP antibodies (Cambridge et al., 2006). This suggests that anti-nucleosome and anti-double stranded DNA antibodies are produced through extrafollicular responses, which usually generate short-lived plasma cells, while antibodies to nucleic acid-associated antigens (Ro, Sm, and RNP) are derived from follicular responses, which generate long-lived plasma cells. In RA patients, the levels of IgA-RF, IgG-RF, and IgG anti-cyclic citrullinated peptide (CCP) antibodies are decreased at 6 months after the administration of anti-CD20 monoclonal antibody, and the decreases are proportionately greater than the decreases in their respective total immunoglobulin classes (Cambridge et al., 2003). Plasmablasts and short-lived plasma cells originating from the extrafollicular response might be the major source of RF and anti-CCP antibodies (Looney et al., 2008). Therefore, both extrafollicular- and follicular-mediated antibody productions should be controlled in the treatment of human autoimmune inflammation.

In general, T cells are indispensable sources of help signals, which promote B cell antibody production. Therefore, control of antibody production at the level of T cells is a rational approach to autoantibody suppression. Indeed, several T cell populations are able to suppress B cell antibody production. In humans, CD4⁺CD25⁺CD69⁻ Treg that suppress antibody production in vitro have been found in GC. The fact that these CD4⁺CD25⁺CD69⁻ Treg hardly express CXCR5 suggests that they mainly reside in the T cell-rich zones of secondary lymphoid tissues (Lim et al., 2004). However, T cell activation switches their chemokine receptor expression pattern from CCR7 to CXCR5 and switches their chemotactic responses from CCL19 to CXCL13. Thus, activation might change the migratory behavior of CD4⁺CD25⁺CD69⁻ Treg so that they can migrate to GC. After migrating to GC, CD4⁺CD25⁺CD69⁻ Treg negatively regulate T cell-dependent B cell responses through their suppressive activity toward T cells.

The preferential killing of antigen-presenting B cells by CD4⁺CD25⁺ Treg was reported in C57BL/6 mice (Zhao et al., 2006). B cell death is not mediated by the Fas–Fas ligand pathway, but instead is mediated by a granzyme-dependent, partially perforin-dependent pathway. Direct suppression of B cells by Treg was also reported in chronic systemic autoimmunity (Iikuni et al., 2009). For example, CD4⁺CD25⁺ Treg have been demonstrated to inhibit B cell antibody production in in vitro models of murine and human lupus. Treg use granule exocytosis pathways involving perforin and granzyme to induce contact-dependent apoptosis in B cells. However, in spite of the fact that CD4⁺CD25⁺ Treg from both young and old NZB/W F1 mice retain a capacity to suppress IgG production in B cells, autoantibodies continuously accumulate in these mice. Therefore, whether CD4⁺CD25⁺ Treg could be used to efficiently control autoantibody production in systemic autoimmunity needs to be examined further.

Recently, several groups simultaneously identified mice CD4⁺CD25⁺Foxp3⁺ Treg subpopulations that are able to suppress B cell antibody production (Chung et al., 2011; Linterman et al., 2011;Wollenberg et al., 2011). Chung et al. (2011) identified a subset of Treg cells that express CXCR5 and Bcl6 and localize to GC in mice and humans. The expression of CXCR5 on Treg depends on Bcl6, and CXCR5⁺Bcl6⁺ Treg are absent from the thymus but can be generated from CXCR5⁻Foxp3⁺ natural Treg precursors. A deficiency of CXCR5⁺ Treg results in enhanced GC reactions involving B cells, affinity maturation of antibodies, and plasma cell differentiation. These results demonstrated that the Bcl6–CXCR5 axis of Treg is one mechanism by which GC responses are controlled. In addition, they observed that Foxp3-mutated scurfy mice display a moderate increase in their T_FH population but a markedly increased number of GL7⁺CD95⁺ B cells. Collectively, these observations suggest that Foxp3⁺ follicular regulatory (T_FR) cells are more specialized for controlling the generation of GC B cells. Linterman et al. (2011) also described a population of Foxp3⁺Blimp-1⁺CD4⁺ T cells that accounted for 10–25% of the CXCR5^highPD-1^highCD4⁺ T cells found in immunized GC. In the absence of these T_FR cells, they noted outgrowths of non-antigen-specific B cells in GC and a decreased number of antigen-specific B cells. Therefore, both groups revealed that T_FR play a role in controlling GC reactions by inhibiting the selection of antigen-specific- and non-specific B cells. Because CXCR5-expressiong T_FR localize to GC, T_FR may suppress GC-mediated autoantibody production. However, whether T_FR actually suppress autoantibody production and whether T_FR deficiency results in autoimmunity remain to be addressed.

A recent study reported that Qa-1 restricted CD8⁺ Treg cells directly inhibit Qa-1⁺ T_FH cells. Qa-1 is a non-classical MHC class Ib molecule presenting a peptide derived from the signal sequence of classical MHC class I proteins, named Qa-1 determinant modifier (Qdm), as well as peptides derived from proteins associated with infectious or inflammatory responses (Lu et al., 2006). Previously, a subpopulation of CD8⁺ T cells was reported to suppress T cell help to B cells (Noble et al., 1998), and subsequent studies have shown that Qa-1 restricted CD8⁺ T cells inhibit experimental autoimmune encephalomyelitis (EAE) by targeting autoreactive CD4⁺ cells (Hu et al., 2004). Nevertheless, although Qa-1 deficient mice showed dysregulated immune responses to immunization with self and foreign antigens, Qa-1^−/− mice do not develop spontaneous autoimmunity. Since Qa-1 interacts with both the T cell receptor (TCR) on CD8⁺ T cells and the CD94/NKG2A receptor expressed by activated CD4⁺ T cells, Qa-1 knock-in mice, B6 Qa-1(D227K) mice, were generated. B6 Qa-1 (D227K) mice harbor a Qa-1 amino acid exchange mutation that disrupts the binding of Qa-1 to the TCR/CD8 complex, but has no effect on its binding to the inhibitory NKG2A receptor. Intriguingly, the B6 Qa-1 (D227K) mice exhibit lupus-like systemic autoimmune disease and a fivefold to sixfold increase in their numbers of T_FH cells (Kim et al., 2010).

Analysis of the surface phenotype of Qa-1 restricted CD8⁺ Treg indicated that they express CD44, ICOSL, and CXCR5 and the CD44⁺ICOSL⁺CD8⁺ T cells inhibit the generation of high-affinity antibodies and Qa-1⁺ T_FH cells. This observation provides a clue that might greatly increase our understanding of autoantibody production. However, the antigen-specificity of Qa-1 restricted CD8⁺ Treg during T_FH cell suppression remains unclear because the repertoire of peptides presented by Qa-1 is substantially smaller than the repertoire of classical MHC molecules (Lu et al., 2006). Only a small number of peptides have been identified that bind to Qa-1 and stimulate CD8⁺ T cells, including dominant Qdm as well as peptides from HSP60, insulin, Salmonella GroEL, and TCR Vβ chains. Thus, Qa-1 restricted CD8⁺ Treg might suppress T_FH cells irrespective of the antigen-specificity of the TCR on T_FH cells. Because Qa-1 restricted CD8⁺ Treg express CXCR5 and migrate to lymphoid follicles (Kim et al., 2010), Qa-1 restricted CD8⁺ Treg may suppress GC-mediated autoantibody production.

T_FR and Qa-1 restricted CD8⁺ Treg appear to be important checking mechanisms for antibody production. However, no Treg populations that control autoantibody production and autoimmunity in an antigen-specific manner have yet been identified. Although the importance of T regulatory type I (Tr1) cells for controlling immune responses has been described in a number of reports, anti-CD46-induced IL-10-secreting T cells even enhance antibody production by B cells (Fuchs et al., 2009). Recently, several CD4⁺ T cell populations that possess regulatory activity have been identified (Fujio et al., 2010). CD4⁺CD25⁻LAP⁺ T cells and CD4⁺NKG2D⁺ T cells produce both IL-10 and TGF-β (Oida et al., 2003; Dai et al., 2009), and CD4⁺CD25⁻IL-7R⁻ T cells and CD4⁺CD25⁻LAG3⁺ T cells produce large amounts of IL-10 (Haringer et al., 2009; Okamura et al., 2009). The association between these recently identified Treg and antigen-specific autoantibody suppression should be investigated. In particular, CD4⁺CD25⁻LAG3⁺ T cells, which characteristically express the anergy-linked transcription factor Egr2, might be associated with autoantibody suppression, because T cell-specific Egr2-deficient mice exhibit lupus-like disease (Zhu et al., 2008) and polymorphisms in the EGR2 gene are associated with human SLE susceptibility (Myouzen et al., 2010). Although both T_FR cells and Qa-1 restricted CD8⁺ Treg express CXCR5 and may suppress GC-mediated autoantibody production, Treg populations which suppress extrafollicular response are yet to be identified.