BAFF inhibition in SLE – is tolerance restored?

Summary:

The B cell activating factor (BAFF) inhibitor, belimumab, is the first biologic drug approved for the treatment of SLE, and exhibits modest, but durable, efficacy in decreasing disease flares and organ damage. BAFF and its homolog APRIL are TNF-like cytokines that support the survival and differentiation of B cells at distinct developmental stages. BAFF is a crucial survival factor for transitional and mature B cells that acts as rheostat for the maturation of low-affinity autoreactive cells. In addition, BAFF augments innate B cell responses via complex interactions with the B cell receptor (BCR) and Toll like receptor (TLR) pathways. In this manner, BAFF impacts autoreactive B cell activation via extrafollicular pathways and fine tunes affinity selection within germinal centers (GC). Finally, BAFF and APRIL support plasma cell survival, with differential impacts on IgM- and IgG-producing populations. Therapeutically, BAFF and combined BAFF/APRIL inhibition delays disease onset in diverse murine lupus strains, although responsiveness to BAFF inhibition is model dependent, in keeping with heterogeneity in clinical responses to belimumab treatment in humans. In this review, we discuss the mechanisms whereby BAFF/APRIL signals promote autoreactive B cell activation, discuss whether altered selection accounts for therapeutic benefits of BAFF inhibition, and address whether new insights into BAFF/APRIL family complexity can be exploited to improve human lupus treatments.

1. Introduction

Systemic lupus erythematosus (SLE – lupus) is a severe autoimmune disease in which loss of B cell tolerance to ubiquitous nuclear antigens causes autoantibody-mediated organ damage. Despite the continuing acquisition of new knowledge regarding lupus pathogenesis, numerous biologic drugs have failed to achieve a significant clinical response in rationally designed trials using large patient cohorts ¹. Therefore, much remains to be learned regarding the mechanisms underlying autoreactive B cell activation and strategies to restore B cell tolerance in lupus patients.

The association of high levels of BAFF with lupus susceptibility was fueled by the observation that two independently derived strains of BAFF transgenic mice manifested B cell hyperplasia, anti-dsDNA antibodies and a lupus-like illness with renal immune complex deposits. BAFF levels were found to be elevated in murine lupus models and disease was significantly attenuated by BAFF inhibition ^2–4. Once it was established that patients with SLE also exhibit increased serum BAFF ^{5, 6}, therapeutics that target BAFF and its receptors were developed. This work culminated in the FDA approval of the human monoclonal anti-BAFF antibody, belimumab, for the treatment of SLE - the first approved application of a biologic drug in human lupus.

Despite the reproducible clinical efficacy of belimumab in several trials and convincing longitudinal data showing that it decreases lupus flare rates over long periods of time ⁷, clinical efficacy is only moderate and not all patients respond ^{8, 9}. Thus, many questions remain regarding the mechanism of action of belimumab and about the best strategy to target BAFF, its homologous cytokine APRIL, and the three BAFF family receptors in order to maximize clinical benefit. In this review, we summarize recent insights into the impact of BAFF and APRIL on B cell subsets in SLE, focusing on our own data, and speculate about the potential immune mechanisms underlying BAFF’s roles in lupus pathogenesis.

2. The complexity of the BAFF-APRIL system (Figure 1)

Figure 1: BAFF, APRIL, and their receptors and splice variants:

Both BAFF and APRIL multimerize: BAFF multimerizes into 60-mers via its flap region whereas APRIL multimerizes on cells surfaces by binding to proteoglycans. TACI and BCMA both shed soluble forms that can act as decoy receptors and human TACI has an intracellular short form that preferentially interacts with TLR9. Therapeutic approaches are displayed in purple boxes. Specific targeting of the flap region of BAFF should prevent multimer formation that is required for optimal signaling through all three receptors. This is more likely to target membrane BAFF-TACI and BAFF-BCMA interactions and may help to preserve B cell numbers while allowing the continued function of the soluble decoy receptors ¹³. Given relatively restricted expression on B cells, BAFF family receptors are potential targets for chimeric antigen receptor (CAR) T cell cancer therapies. Specifically, anti-BAFF-R ¹⁴³ and anti-BCMA ¹⁴⁴ CARs are being developed for the treatment of B cell leukemia/lymphoma and multiple myeloma, respectively, while APRIL-expressing CARs target TACI and BCMA expressing cells ¹⁴⁵. Although currently restricted to oncology trials, these therapies could potentially be used to target B cell subsets in SLE.

2.1. The BAFF family cytokines and receptors

BAFF:

BAFF is a type II transmembrane protein produced predominantly by stromal and myeloid cells that initially assembles as a homotrimer on the cell membrane but is subsequently cleaved by furin protease to release the soluble trimeric cytokine ¹⁰. BAFF trimers can also multimerize via a loop region of 10 amino acids (flap) to form 60-mers ^{11, 12}. BAFF has three receptors that are expressed predominantly by B cells, BAFF-R, TACI and BCMA. While trimeric BAFF induces signaling via BAFF-R, TACI is only activated by the higher-order multimers. The flap region is not required for binding to any known receptor but it is necessary for optimal signaling after binding of the soluble BAFF trimer to BAFF-R, suggesting that a degree of multimerization is required even for BAFF-R activation ¹³. Alternate BAFF forms include BAFF-APRIL heterotrimers and a splice variant called ΔBAFF, which is inefficiently cleaved from the cell surface and serves as an antagonist via heteromultimerization with BAFF ¹⁴. As detailed below, a human polymorphism of the BAFF gene that increases BAFF expression by ~1.5–2 fold is associated with a modest increase in susceptibility to certain autoimmune diseases, including SLE ¹⁵.

APRIL:

APRIL is a type II transmembrane protein produced by multiple cell types including myeloid cells, epithelial cells, and tumor cells ¹⁶. APRIL is cleaved within the Golgi apparatus and does not usually appear on the cell membrane ¹⁷. However, the APRIL gene is located adjacent to the TWEAK gene such that hybrid molecules comprising the cytoplasmic domain of TWEAK and the extracellular domain of APRIL may be generated and expressed on the cell surface ¹⁸. APRIL also appears to have a membrane-bound splice isoform ¹⁹. APRIL binds to BCMA, TACI, and to surface heparin sulfate proteoglycan (HSPG), which serves to multimerize APRIL and thereby allow downstream signaling following receptor ligation. Additional HSPG binding to TACI facilitates APRIL-TACI-HSPG interactions, resulting in complex impacts on BAFF family receptor activation. While APRIL deficient mice have normal immune development, class switching to IgA and maintenance of serum IgA levels is highly APRIL-dependent ^{20, 21}. In addition, APRIL supports the survival of long-loved plasma cells (LLPC) in the bone marrow, especially during neonatal development ²¹.

BAFF family receptors:

BAFF-R, TACI, and BCMA are the three known surface receptors for BAFF and APRIL. Each receptor exhibits a unique expression pattern on distinct B cell subsets, in part accounting for differential functional roles during an immune response. In addition, alternative splice isoforms of BCMA and TACI have been reported ²². All three of the BAFF/APRIL receptors can be shed from the cell surface by the activity of γ-secretase (BCMA) or ADAM proteases (BAFF-R and TACI) ^{23, 24}. Shedding of BAFF-R and TACI is dependent on receptor ligation, B cell subset, B cell activation state, and, for BAFF-R, the coexpression of TACI ²⁵. Importantly, constitutive cleavage of BCMA and TACI generates functional decoy receptors, whereas BAFF-R shedding does not appear to generate a decoy but rather regulates B cell survival via modulation of surface BAFF-R expression ²⁶.

BAFF-R:

BAFF-R expression is first observed at the transitional stage of B cell maturation. However, low-level BAFF-R expression has been observed on bone marrow B cells, consistent with evidence that BAFF-R engagement mediates positive selection of developing B cells prior to the transitional type 2 (T2) stage, especially those lacking BCR engagement, thereby skewing the emerging repertoire against autoreactivity ^{27, 28}. In contrast, BAFF/BAFF-R interactions are critical for T2 selection and naïve B cell survival. In these cells, BAFF-R and BCR signals cooperate to reinforce each signaling pathway and inhibit apoptosis ^29–31. Unlike other BAFF/APRIL receptors, BAFF-R signals through the alternative NFκB pathway, for which BCR engagement provides the essential intermediary p100 ^{29, 32, 33}. BAFF-R ligation also activates the phosphoinositide-3-kinase-dependent signaling cascade to enhance protein synthesis and mediate metabolic reprogramming in order to facilitate cell survival (reviewed in ²⁶). Importantly, since immature and anergic B cells exhibit lower surface BCR expression, these cells exhibit an increased requirement for BAFF-R mediated survival signals. In this manner, BAFF functions as a rheostat for naïve B cell selection at the transitional stage.

As described in detail below, BAFF-R is downregulated on GC B cells, but is re-expressed on memory B cells ²⁵. While the function of BAFF-R in memory B cells remains incompletely defined, IgM⁺ memory B cells exhibit partial BAFF/BAFF-R dependence, whereas IgG⁺ memory B cells require neither BAFF nor APRIL for their survival ^{34, 35}. Finally, BAFF-R is also expressed on a subset of T cells and may function to modulate T cell activation and cytokine production (reviewed ³⁶).

TACI:

TACI is expressed by all mature peripheral B cells including marginal zone, B1 B cells, and plasma cells. Engagement with BAFF and APRIL multimers promotes TACI signaling via classical NFκB, Mek and Jnk/p38 pathways to counteract apoptosis, drive immunoglobulin class switch recombination, and promote antibody production. The interaction of TACI with APRIL preferentially supports IgA responses ^{21, 37}. TACI is required for T-independent responses but appears dispensable for the initiation of T dependent responses ^{37, 38}. Nevertheless, TACI serves to maintain BLIMP-1 expression and thus supports the differentiation and survival of long-lived plasma cells ³⁹. While TACI signals promote T cell-independent B cell activation, TACI deficient mice exhibit B cell hyperplasia and mild autoimmunity, suggesting a potential negative regulatory role in B cell activation ^{37, 38}. In addition to a direct regulatory function, loss of surface TACI or of the soluble decoy receptor may account for B cell hyperplasia because of reciprocally increased serum BAFF and APRIL levels ^40–42. In contrast to mice, humans express both short and long isoforms of TACI. Functionally, short isoform TACI is induced by TLR9 agonists, has a higher binding affinity for both BAFF and APRIL, promotes plasma cell development and is preferentially expressed intracellularly, whereas the long form is preferentially membrane bound and expressed by resting cells ⁴³.

BCMA:

BCMA is expressed predominantly by antibody-secreting cells and mediates plasma cell survival, particularly in the bone marrow, making it an excellent therapeutic target for plasma cell depleting strategies. Like TACI, BCMA signals though the classical NFκB pathway. It interacts preferentially to APRIL in mice, but in humans can also bind BAFF. Human BCMA is expressed in the Golgi apparatus and on the cell surface, while soluble BCMA can function as a decoy receptor. In addition, alternate splice variants of unknown function have been identified (reviewed ⁴⁴).

2.2. Interactions between BAFF/APRIL and TLR signaling pathways – Figure 2

Figure 2: Crosstalk between endosomal TLRs and BAFF/APRIL:

Excessive load of cell debris, oxidized DNA and/or nucleic acid containing immune complexes can trigger endosomal TLRs. (A) B cells that internalize DNA activate TLR9 and its adapter MyD88 that directly binds to and enhances TACI activity. In a high BAFF environment mature TLR9 stimulated B cells are protected from apoptosis and differentiate to plasma cells. In marginal zone B cells MyD88 enhances plasmablast differentiation through an mTOR pathway. (B) Excess RNA activates TLR7 in pDCs resulting in the release of Type I IFN that induces myeloid DCs to secrete BAFF. RNA recognizing B cells internalize RNA into endosomes through the BCR; activation of TLR7 in these cells induces expression of TACI which in turn amplifies expression of TLR7. In a high BAFF environment autoreactive B cells will therefore preferentially survive and differentiate into plasma cells secreting more autoantibodies.

BAFF is an essential component of the innate immune response. Type I interferons (IFNs) induce BAFF secretion by myeloid DC and neutrophils ⁴⁵, to complement constitutive expression by radiation resistant stromal cells in lymphoid tissues. Functionally, BAFF upregulates Toll-like receptor (TLR) expression in B cells, and cooperates with dual BCR/TLR signals to drive T independent B cell differentiation. Moreover, B cell activation in response to engagement of endosomal TLRs by nucleic acid-containing autoantigens drives the upregulation of BAFF family receptors, in particular TACI ^{46, 47}. As described below, humoral autoimmunity in BAFF transgenic (BAFF-Tg) mice is T cell independent but requires signals through the TLR adaptor molecule MyD88 and TACI, confirming the relevance of this crosstalk in driving breaks in B cell tolerance ^48–51. In addition, whereas combined BCR/TLR9 signaling induces B cell proliferation followed by apoptotic cell death, the provision of excess BAFF allows these cells to differentiate into short-lived plasma cells with pathogenic potential ⁵². Finally, TACI, in particular the short human isoform, can enter endosomal compartments where it associates with the TLR adapter molecule MyD88 through a domain separate from the TLR TIR domain to mediate enhanced NFκB signals that drive plasma cell differentiation. By enhancing the production of the short TACI isoform, TLR9 signals amplify this response ⁴³. Marginal zone B cells express high levels of TACI and in these cells TLR9 and TACI recruit both MyD88 and mTOR to facilitate T-independent class switching and plasma cell differentiation ⁴⁹. These interactions have important implications for autoimmunity because cross talk between type 1 IFN, TLR, BCR, and BAFF signaling preferentially supports the survival, class-switch recombination, and plasma cell differentiation of B cells recognizing nuclear autoantigens.

3. Regulation of B cell tolerance by BAFF and its receptors – Figure 3

Figure 3: The effect of BAFF inhibition on B cell development:

Stages of B cell development are shown. The red bars indicate known checkpoints for autoreactivity in the bone marrow (1), at the early transitional stage (2), at GC entry (3) and at post-GC differentiation to memory and plasma cells (4). Green arrows indicate the B cell developmental stages at which BAFF inhibition influences B cell development at the T1 to T2 stage (5), follicular vs marginal zone decision point (6), GC entry (7), GC T dependent affinity maturation (8), plasma cell survival (9) or promotes cell anergy or death (10). X indicates inhibition. B1 B cells, memory B cells and age-associated B cells are BAFF independent.

B cell receptors are generated in the bone marrow by random recombination of immunoglobulin heavy and light chain gene segments to yield a broad repertoire capable of recognizing diverse pathogens. To counter the inevitable generation of self-reactive B cells via random VDJ recombination, overlapping tolerance mechanisms restrain autoreactive B cells, including receptor editing, clonal deletion and the induction of functional unresponsiveness or anergy ⁵³. Importantly, excessive purging of self-reactive clones from the B cell compartment would promote “gaps” in the pre-immune repertoire which could be exploited by pathogens. To counter this problem, coordinated positive and negative selection mechanisms ensure a diverse B cell repertoire balancing breadth of response with risk of autoimmunity. Chief among these positive selection mechanisms, serum BAFF levels regulate B cell development and homeostasis ¹⁰. In murine models, either Baff or BAFF-R deletion results loss of peripheral B cells beyond the transitional type 1 stage, indicating a critical role for BAFF/BAFF-R signals in regulating in mature B cell survival and the size of the naïve B cell compartment ⁵⁴. An additional function of BAFF may be to preferentially enhance the generation of IL10-producing regulatory cells ^55–57, however this effect appears to be lost in the setting of inflammation ⁵⁸.

In addition to regulating B cell survival, serum BAFF levels modulate the relative autoreactivity of the naive B cell repertoire. Initial studies indicated that transgenic BAFF overexpression (BAFF-Tg) results in spontaneous humoral autoimmunity, suggesting that excess BAFF broadly disrupts B cell tolerance ³. However, analysis of autoreactive BCR transgenic models indicated that the deletion of high-affinity autoreactive B cells via central tolerance is not impacted by BAFF overexpression, a finding consistent with limited BAFF-R expression on developing BM B cells. Moreover, whereas excess BAFF promotes the survival and follicular entry of high-affinity anergic B cells, the introduction of competing non-autoreactive B cells during development curtails the maturation of these autoreactive B cell clones. Rather, supraphysiologic BAFF exerts a relatively nuanced impact on the stringency of transitional B cell selection, by exerting context specific effects on the survival of lower affinity self-reactive B cells ^59–61.

As detailed above, BAFF-R signals integrate with tonic BCR signaling pathways to modulate transitional B cell selection in competitive environments ^62–64. Since autoreactive B cells downregulate surface BCR, they exhibit greater dependence on BAFF survival signals. Thus, in settings of reduced available BAFF, these cells are more likely to be deleted or anergized than their non-autoreactive counterparts. In contrast, increased serum BAFF enhances the survival and maturation of these lower affinity self-reactive clones. As an example of this dichotomy, the Brink group investigated the effects of excess BAFF on autoreactive B cell survival using the Hen Egg Lysozyme (HEL) transgenic system. In this model, whereas BAFF failed to rescue high-affinity anti-HEL B cells from central deletion, it enhanced the survival, maturation and follicular entry of lower affinity autoreactive B cells. However, if intercellular competition for available BAFF was provided by non-autoreactive B cells, these same cells could no longer be rescued by excess BAFF and were anergized or deleted. In the same competitive environment, B cells expressing the anti-HEL heavy chain alone (i.e. exhibiting lower affinity HEL binding) matured to the follicular compartment but were excluded from the marginal zone. When excess BAFF was provided these cells were then able to reconstitute the marginal zone ^{59, 65}. These studies are highly instructive because they provide a framework for interpreting studies of the effect of BAFF excess or deficiency on naive B cell fate in immunoglobulin transgenic mice based on two major variables, namely affinity of the immunoglobulin for autoantigen and the degree of competition for BAFF provided by non-autoreactive B cells.

Importantly, affinity for antigen is not the only predictor of BAFF dependence since other autoreactive transgenes behave differently. The Manser group has studied two heavy chain knock-in mice that produce large numbers of follicular (FO) B cells that are either weakly or strongly autoreactive with nuclear autoantigens. In these models, BAFF excess expanded the MZ pool without allowing either highly- or weakly-autoreactive B cells to outcompete non-autoreactive cells in the mature compartment ⁶⁶. The reasons for these conflicting data remain incompletely understood, but may relate to the chemical nature, abundance or valence of the relevant self-antigen in these different transgenic models. In sum, these experiments show that BCR signals and available BAFF levels influence autoreactive B cell survival. However, BAFF is not the only molecule needed for the survival or depletion of autoreactive B cells and in a competitive environment not all autoreactive B cells are equally dependent on the availability of BAFF.

A potential consequence of this general model in which BAFF functions as a rheostat for autoreactive B cell survival is that increases in serum BAFF are predicted to increase both the breadth of the pre-immune B cell repertoire and the relative self-reactivity of naïve B cells, resulting in a trade-off between autoimmune risk and enhanced pathogen responses. Consistent with this idea, a genetic study of elevated rates of autoimmunity in the Sardinia population identified a novel insertion–deletion variant in the 3′ UTR of human TNFSF13B (encoding BAFF) ¹⁵. Variant TNFSF13B (BAFF-var) exhibited reduced microRNA degradation, resulting in 1.5–2 fold increase in circulating BAFF levels, B cell expansion and an elevated risk of SLE and multiple sclerosis (MS). Although a detailed analysis of the naïve B cell repertoire in Sardinian subjects has not yet been performed, murine modeling predicts that BAFF-var carriers likely express a B cell repertoire enriched for low-affinity autoreactive B cell specificities. Strikingly, this autoimmune risk variant showed evidence of positive selection within the Sardinian population, an observation the authors attributed to the selective pressure of malaria, a pathogen endemic in the region until the 1950s ⁶⁷. In keeping with this hypothesis, BAFF-Tg mice are protected from lethal malaria infection ⁶⁸ and human B cells expressing the intrinsically autoreactive VH4–34 heavy chain family are expanded in the memory compartment of malaria patients ⁶⁹, suggesting a link between relaxed B cell developmental selection and robust pathogen responses.

3.1. BAFF-R and TACI signals integrate to promote autoimmunity in BAFF transgenic mice

Given human data linking increased serum BAFF with the pathogenesis of SLE, the immune mechanisms whereby BAFF modulates B cell tolerance has been of considerable interest. Despite central and peripheral tolerance mechanisms, a substantial proportion of naïve B cells express autoreactive BCRs even at physiologic BAFF levels ⁷⁰. Thus, BAFF-driven expansion of low-affinity autoreactive B cells is unlikely the sole contributor to humoral autoimmunity observed in high BAFF settings. Rather, signals downstream of BAFF-family receptors integrate with antigen receptor and innate signals to drive breaks in B cell tolerance at later stages of B cell development.

Murine models of transgenic BAFF overexpression (BAFF-Tg) exhibit spontaneous humoral autoimmunity characterized by class-switched autoantibodies and plasmablast expansion in the absence of CD4⁺ T cell help ⁷¹. As detailed above, BAFF-R signals drive B cell expansion and the survival of low-affinity autoreactive B cells, suggesting that BAFF-R is the predominant receptor responsible for B cell tolerance breaks. In contrast, separate groups, including our own, used independently-generated BAFF-Tg and Taci^−/− murine strains to uncover a critical role for TACI signals in BAFF-Tg autoimmunity ^{50, 51, 72, 73}. These data were particular surprising given that TACI was initially proposed as a negative regulator of BAFF signaling ^{38, 74}, and highlight the complexity of BAFF family biology.

Based on these findings, we sought to identify the B cell subset that is the target of TACI-dependent activation in BAFF-Tg mice. Since the splenic marginal zone (MZ) is both enriched for autoreactive specificities and the number of MZ B cells is increased in BAFF-Tg mice ³, MZ B cells were initially predicted to promote BAFF-driven SLE ¹⁰. Consistent with this idea, MZ-like cells have been found in inflamed salivary glands of BAFF-Tg mice ⁷⁵. However, BAFF-Tg mice deficient in lymphotoxin-β (resulting in disordered splenic architecture and loss of MZ B cells) as well as genetically or surgically splenectomized BAFF-Tg mice (resulting in loss of both MZ B cells and peritoneal B1a B cells) develop immune-complex glomerulonephritis ^{76, 77}, findings suggesting that neither MZ B cells nor B1a B cells are required for BAFF-mediated lupus nephritis.

To address these controversies, we examined the effect of increased BAFF on surface TACI expression by B cell developmental subsets. Surprisingly, whereas TACI is typically limited to mature B cells at physiological BAFF levels, we observed a striking expansion of TACI-expressing transitional B cells in BAFF-Tg mice. We confirmed that this TACI⁺ transitional subset derived from bone fide transitional cells, using a fate-mapping approach. In settings of BAFF excess, TACI expression marked a subset of transitional B cells exhibiting cellular proliferation, activation marker expression, and a BCR repertoire enriched for autoreactivity. In keeping with a direct contribution of this subset to BAFF-driven autoimmunity, TACI⁺ transitional B cells expressed activation-induced cytidine deaminase (AID) and Blimp1 required for somatic hypermutation, class-switch recombination, and plasma cell differentiation, respectively ⁷³. Finally, sorted TACI⁺ transitional B cells spontaneously produced class-switched autoantibodies ex vivo, whereas no antibody was generated by BAFF-Tg follicular mature (FM) and MZ B cells. Importantly, whereas our data strongly suggest that transitional B cells are a prominent source of pathogenic autoantibodies in BAFF-Tg mice, these findings do not preclude additional contributions of alternate B cell subsets to BAFF-driven autoimmunity. Indeed, a recent study identified a role for peritoneal B1b B cells in the autoantibody repertoire of BAFF-Tg mice ⁷⁸, data in keeping with prominent TACI expression by peritoneal B1 B cells.

We next examined the additional B cell signals driving transitional B cell TACI expression and BAFF-driven breaks in tolerance. Integration of BCR and TLR signals enhances B cell activation, and autoimmunity in BAFF-Tg mice is abrogated in the absence of the TLR signaling adaptor, MyD88 ⁷¹. In addition, as described above, prior studies have shown that BCR and TLR activation ex vivo increases sTACI levels on mature B cells ⁴⁶, and that TACI also associates intracellularly with MyD88 in endosomes. In keeping with these data, deletion of the BCR signaling adaptor Bruton’s tyrosine kinase (BTK) or MyD88 similarly protected BAFF-Tg mice from progressive glomerulonephritis. However, these signaling pathways exerted distinct impacts on surface TACI levels and autoantibody production. Whereas loss of BCR signals abrogated transitional TACI expression and blocked autoantibody production across immunoglobulin isotypes, surface TACI was maintained in Myd88^−/−.BAFF-Tg mice which manifested a specific reduction in class-switched, but not IgM, autoantibodies ⁷².

Of the Myd88-dependent TLRs, the endosomal receptors TLR7 and TLR9 have been implicated in the production of RNA- or DNA-associated autoantibody specificities, respectively. Although original studies reported high-titer anti-double stranded DNA (dsDNA) autoantibodies in BAFF over-expression models ^{3, 71}, we noted a marked enrichment for RNA-associated autoantibodies in BAFF-Tg mice, with relatively low-titer anti-dsDNA autoantibodies observed using the stringent Crithidia luciliae assay for native DNA reactivity. Consistent with these findings, TLR7 deletion abrogated class-switched autoantibodies in BAFF-Tg mice. Moreover, TACI deletion resulted in a relatively specific reduction in BAFF-driven RNA-associated autoantibodies, including against RNP and Sm antigens. Strikingly, loss of these specificities correlated with a specific decrease in endocapillary, but not mesangial, immune complex deposits within inflamed glomeruli of aged BAFF-Tg mice ⁵¹.

Importantly, whereas TACI expression is critical for BAFF-driven autoimmunity, BAFF-R signals exerted complementary roles in facilitating breaks in B cells tolerance. Specifically, the expansion of cycling TACI⁺ transitional B cells and production of class-switched autoantibodies was absent in BAFF-R^−/−.BAFF-Tg mice ⁷². Although the mechanisms whereby BAFF-R and TACI signals integrate to break B cell tolerance have not been completely defined, we hypothesize that BAFF-R signals both facilitate the survival of low-affinity autoreactive transitional B cells and promote the expansion of autoreactive clones that have engaged self-antigen. Notably, this model requires that BAFF-R signals impact B cell biology earlier in development as described above. Alternatively, BAFF-R and TACI ligation could enhance B cell TLR mediated signals ⁴⁹ and thereby facilitate class-switched autoantibody production.

In summary, our data demonstrate that multiple signaling pathways, including BCR, TLR, TACI and BAFF-R signals, integrate to promote class-switched autoantibodies in high BAFF settings. It remains to be determined whether modest BAFF excess such as that observed in Sardinian individuals or in the setting of SLE are sufficient to break tolerance in the transitional compartment and drive T cell independent autoimmunity.

3.2. Distinct contribution of EF and GC pathways to lupus pathogenesis

During a protective or pathogenic humoral response, antibody-secreting plasma cells can be generated by either extrafollicular (EF) or germinal center (GC)-dependent B cell activation pathways. After infectious challenge, EF B cell activation produces an initial wave of relatively low-affinity protective antibody, which is followed by high-affinity, durable antibodies derived from GCs. Functionally, EF-derived plasma cells are believed to represent a short-lived population residing in the splenic red pulp or lymph node medullary cords, while GC-derived plasma cells are maintained long-term in bone marrow survival niches ⁷⁹. Importantly, both EF and GC pathways have been implicated as sources for pathogenic autoantibodies in humoral autoimmune diseases, in particular SLE. Specifically, EF B cell activation underlies the generation of somatically mutated plasma cells in the MRL.Fas^lpr80 and BAFF-Tg lupus-prone stains ⁷¹. In addition, immunophenotyping and V_H-family clonal analysis has uncovered an in vivo developmental link between expanded active naïve B cells (aNAV; defined as CD19⁺IgD⁺CD27⁻CXCR5⁻), double-negative 2 (DN2; CD19⁺IgD⁻CD27⁻CXCR5⁻) B cells and circulating plasma cells in lupus patients ^{81, 82}; findings implicating EF B cell activation as important contributor to pathogenic autoantibodies in human SLE. Mechanisms for regulation of autoimmunity generated via this EF pathway remain poorly understood.

In addition to EF B cell activation, dysregulated GCs represent an alternate source for class-switched autoantibodies in SLE. Murine studies have shown that the entry of non-anergic autoreactive B cells into the GC is usually tightly regulated, perhaps by the availability of T cell help ⁸³. In addition, those B cells that acquire autoreactivity as a result of somatic mutation are censored before they reach the memory compartment ^84–87. Despite these data, evidence from both animal and human studies implicates dysregulated GCs as an important site for B cell tolerance breaks. For example, the development of spontaneous GCs is frequently observed in murine lupus models and human subjects can develop ectopic GCs within inflamed target organs ^88–90. Moreover, a subset of autoantibody specificities can persist despite anti-CD20 B cell depletion, implicating GC-derived CD20^neg long-lived plasma cells in their genesis. Finally, SLE is characterized by a prolonged period of pre-clinical autoimmunity during which patients sequentially accumulate autoantibody specificities in a manner consistent with epitope spreading within an ongoing dysregulated GC response ⁹¹.

Based on these observations, the contribution of BAFF signals to autoreactive GC formation is particularly relevant. Notably, the GC is a relatively BAFF-poor environment. While BAFF-R occupation by BAFF occurs in the GC light zone, B cells in the dark zone downregulate BAFF-R, most likely as a result of receptor cleavage ²⁵. Despite these data, Baff^−/− mice exhibit a defect in the formation of a mature follicular dendritic cell (FDC) network and are characterized by premature GC termination ^{92, 93}. In addition, evidence suggests that T follicular helper (Tfh) cells regulate light zone affinity selection of GC B cells by producing BAFF. Although T cell-derived BAFF was not required for GC formation, loss of T cell-intrinsic BAFF production reduced the emergence of somatically mutated high-affinity B cell clones following candidate antigen immunization ⁹⁴. Notably, TACI is profoundly downregulated on GC B cells in an IL-21 dependent manner. By limiting the sequestration of free BAFF, decreased TACI expression may thereby allow reduced concentrations of locally-generated BAFF to modulate GC selection via interaction with BAFF-R ⁹⁴. These data are compatible with the general model wherein higher affinity B cells present more antigen to Tfh cells and thereby receive enhanced survival signals, including BAFF. These findings suggest that increased serum BAFF and therapeutic BAFF inhibition could reciprocally modulate GC selection via either enhancing selection or deletion of autoreactive clones.

Importantly, recent studies suggest that, as in mice, the drive towards either the GC or EF pathway in SLE may be patient dependent, resulting in differences in the accumulation of autoreactive class switched memory cells vs. plasma cells in individual lupus patients ⁹⁵. Cumulatively, these murine and human data indicate that significant complexity underlies lupus pathogenesis with both EF and GC-dependent B cell activation pathways contributing to pathogenic autoantibody production in SLE. More importantly, this complexity may explain contradictory data from human clinical trials. Since the various BAFF-family cytokines and receptors have distinct roles in different B cell activation pathways, therapeutic targeting of a specific pathway may result in modest or inconsistent benefit based on the patient population studied.

3.3. Roles for individual BAFF family cytokines and receptors in lupus pathogenesis

The study of murine SLE strains rendered genetically deficient in specific BAFF family cytokines and receptors has informed our understanding of lupus pathogenesis. However, the at times contradictory nature of these data based on the specific lupus-prone strain studied has complicated the interpretation of human clinical trials of BAFF blockade in SLE. As detailed above, the potential generation of lupus-associated autoantibodies via distinct B cell activation pathways may account in part for these conflicting results. Viewed through this lens, individual BAFF cytokines and/or receptors could facilitate disease pathogenesis in individual patients and may explain the heterogeneity in clinical responses to BAFF inhibition.

The interpretation of murine genetic models is further complicated by fluctuations in serum BAFF and APRIL levels conferred by changes in the availability of surface receptors and the ability of shed forms of TACI and BCMA to act as decoy receptors. Thus, deficiency of either BAFF-R or TACI increases the circulating levels of their respective ligands and genetic deletion of either TACI or BCMA results in the loss of the regulatory effect of the soluble decoy receptors. Despite these caveats, we will briefly summarize the available data regarding the impact of genetic targeting of both BAFF-family cytokines and the three known receptors in murine SLE, with the goal of informing the interpretation of human clinical trials of BAFF blockade (Table 1).

Table 1:

The effect of deficiency of BAFF family members in murine lupus models

Deficiency	Mouse strain	Result	Reference
BAFF-R	MRL.Faslpr1	Profound B cell depletion No reduction in autoreactive plasma cells or effect on disease progression	102
	NZM2328 2		101
TACI	MRL.Faslpr	Decreases autoantibodies Improves outcome	51
	Nba2.Yaa 3		99
	BAFF Tg		50, 51, 73
	NZM2328	No effect on autoantibodies, plasma cells or mortality	101
BCMA	Nba2.Yaa	Increases plasma cells, enhances autoantibodies and exacerbates disease	103
	NZM2328	Increased spleen size No effect on autoantibodies, plasma cells or mortality	101
APRIL	NZM2328	No effect on autoantibodies, plasma cells or mortality	100
	Nba2.Yaa	Decreased autoantibodies and nephritis	99
BAFF	NZM2328	Delay in autoantibodies and failure to progress from glomerulonephritis to ESRD	97
BAFF/APRIL	NZM2328	Same as BAFF deficiency	98
BCMA/TACI	Nba2.Yaa	Same as TACI deficiency	99
BAFF-R/TACI	NZM2328	Protective	98
BAFF-R/BCMA	NZM2328	Protective	98
TACI/BCMA	NZM2328	Increases plasma cells, enhances autoantibodies and exacerbates disease	98

¹Loss of Fas, expansion of double negative T cells, EF B cell expansion and short-lived plasma cells

²Hyperactivity of B cells, enhanced GC formation and long-lived plasma cells

³Nba2 locus confers B cell hyperreactivity. Yaa translocation confers an extra copy of TLR7

BAFF:

BAFF deletion significantly attenuates autoimmunity in NZM2328 lupus-prone mice, as manifested by reduced glomerular pathology and delayed proteinuria even if the mice are challenged with Type I IFN ⁹⁶. Despite a profound reduction in circulating mature B cells, aged Baff^−/−.NZM2328 mice still developed class-switched IgG⁺ anti-nuclear antibodies. However, glomerular immune deposits in Baff^−/−.NZM2328 mice were predominantly IgG1 subclass, rather than the complement-fixing IgG2a in Baff-sufficient NZM2328 animals. Consistent with these findings, BAFF deletion prevented glomerular C3 complement deposition and limited progressive nephritis in aged Baff^−/−.NZM2328 mice ⁹⁷. APRIL may partially compensate for lack of BAFF in this model, a conclusion supported by loss of serum autoantibodies in dual-deficient Baff^−/−.April^−/−.NZM2328 mice ⁹⁸. The cellular mechanisms underlying BAFF-independent autoantibody generation have not yet been defined, however GCs are reduced, but not absent, in Baff-deficient NZM2328 mice. In addition, peritoneal B1 and splenic transitional B cells are relatively preserved, implicating these B cell subpopulations as potential autoantibody sources.

APRIL:

The impact of APRIL deletion is strain dependent, with attenuation of autoantibody production and lupus nephritis in Nba2.Yaa mice ⁹⁹, but no impact on disease development in NZM2328 mice ¹⁰⁰. The immune mechanisms accounting for this discrepancy have not been addressed but may relate the differential requirements for TACI in these respective models (see below).

BAFF-R:

Given the critical role for BAFF in B cell homeostasis including the requirement for BAFF-R signals in B cell maturation beyond the transitional stage, BAFF-R deletion was predicted to inhibit lupus development in murine models. Indeed, as described above, BAFF-R deficiency ameliorated autoantibody production in BAFF-Tg mice. However, lack of BAFF-R signals did not prevent humoral autoimmunity in either MRL.Fas^lpr or NZM2328 mice despite profound B cell depletion ^{101, 102}. These murine lupus models differ in several important immunologic characteristics. In the MRL.Fas^lpr strain, FAS deficiency promotes predominantly EF B cell activation resulting in the generation of short-lived TACI⁺ plasma cells. In contrast, NZM2328 mice develop spontaneous GCs generating long-lived plasma cells. These data suggest that increased stringency of naïve B cell selection in BAFF-R^−/− mice is not sufficient to prevent breaks in B cell tolerance in either the EF or GC pathway. One caveat is that increased available BAFF in the setting of BAFF-R deletion may induce supraphysiologic TACI signals in these models. In keeping with this hypothesis, dual BAFF-R and TACI deletion prevented disease in NZM2328 mice ⁹⁸.

TACI:

In addition to critical roles for TACI in BAFF-Tg autoimmunity ^{50, 51}, TACI deletion limited disease progression in both the MRL.Fas^lpr and Nba2.Yaa ⁹⁹ lupus models. Notably, in each of these strains, autoantibody secreting plasma cells are believed to be generated predominantly by an EF activation pathway ⁹⁵. In contrast, TACI deletion exerted no impact on serum autoantibodies, plasma cell numbers and lifespan in NZM2328 lupus-prone mice, a model characterized by spontaneous GC formation and long-lived plasma cell expansion ¹⁰¹. Importantly, these data are consistent with the known requirement for TACI signals in T cell-independent, but not T-cell dependent, antibody responses. Most importantly, these conflicting animal models suggest that the therapeutic benefit of TACI blockade may be limited to a subset of human lupus patients in whom disease is driven by a TACI-dependent B cell activation program.

BCMA:

BCMA, like TACI, contributes to plasma cell maintenance, but surprisingly deficiency of BCMA increases the number of plasma cells, enhances autoantibody production and exacerbates disease in several lupus models ^{44, 103}. Several mechanisms have been proposed to account for these unexpected data. First, BCMA has been proposed to act in a T cell-intrinsic manner to restrain autoreactive T follicular helper cell (T_FH) expansion, resulting in expanded GCs and increased autoantibodies in BCMA-null lupus models ⁴⁴. Second, BCMA deletion may impact lupus pathogenesis via loss of a soluble decoy protein. Consistent with this hypothesis, dual TACI and BCMA deficiency in NZM2328 mice results in prominent B cell expansion, increased plasma cells, high-titer autoantibodies and exacerbated disease ⁹⁸, an outcome attributed to increased serum BAFF in the setting of combined decoy receptor deletion. These data are also consistent with a model in which plasma cell survival in inflammatory settings can be independent of both BCMA and TACI. The protective effect of double BAFF-R/BCMA deficiency in NZM2328 mice ⁹⁸, leaving TACI as the only functional receptor, is difficult to explain given the limited impact of BAFF-R deletion alone and the exacerbating effect of BCMA deficiency.

4. Pharmacologic targeting of BAFF family cytokines and receptors

It is clear from the above studies that the BAFF cytokine family plays an important role in the pathogenesis of lupus, but that heterogeneity in responses to BAFF inhibition should be expected based on the variety of B cell developmental pathways that contribute to lupus pathogenesis. In addition, an important distinction between the genetic models described above and treatment of human SLE is that targeted therapies are provided to patients with established disease. Thus, in addition to roles in driving breaks in tolerance, an important therapeutic consideration is the requirement for BAFF-family cytokines in the long-term maintenance of autoreactive memory B cells and plasma cells.

Therapeutic approaches to BAFF family blockade include the use of antibodies against specific cytokines or receptors, or the construction of receptor fusion proteins that target BAFF, or both BAFF and APRIL. In this regard, BAFF-R-Ig blocks BAFF binding to BAFF-R and TACI, whereas TACI-Ig blocks the interaction of BAFF and APRIL with all three receptors. Clinically, there has been much debate about the utility of targeting BAFF vs. targeting both BAFF and APRIL, targeting APRIL alone or using newer approaches that spare the regulatory decoy proteins of TACI and BCMA (Figure 1).

Among the agents currently in development for the treatment of human SLE, anti-BAFF and variants of BAFF-R that act as pure BAFF antagonists all have a similar mechanism of action in preventing BAFF interactions with its receptors. Of these, the human anti-BAFF antibody belimumab has been the most effective in human clinical trials and is the only agent with FDA approval for the treatment of SLE ^{8, 9, 104}. In addition, dual targeting of BAFF and APRIL using TACI-Ig (atacicept) completed phase 2 studies ¹⁰⁵. Recent studies have shown that belimumab binds to and inhibits the activity of BAFF trimers but not multimers whereas atacicept can bind to both trimers and multimers ¹². Atacicept also antagonizes the activity of APRIL and of BAFF-APRIL heterotrimers. Both belimumab and atacicept bind to and antagonize the activity of membrane bound BAFF ¹⁰⁶ but do not mediate complement or antibody dependent cytotoxicity in vitro.

5. Insights into the impact of BAFF family inhibition from murine models

To inform the design and interpretation of human clinical trials, our group has examined the role of BAFF and its receptors in B cell selection and disease progression in three different mouse models of SLE. Importantly, the various therapeutic approaches resulted in distinct impacts on each model tested, highlighting the complexity of lupus pathogenesis and predicting similar heterogeneity of clinical response in human studies. To briefly summarize the murine models evaluated, NZB/W F1 mice are a classic murine lupus model in which T cell-driven immune responses result in the GC-driven production of somatically mutated high affinity IgG anti-DNA antibodies that deposit in the kidneys and initiate a proliferative glomerulonephritis with abundant interstitial inflammatory infiltrates reminiscent of human Class IV lupus nephritis. Similarly, NZM2410 mice are an inbred strain derived from NZB and NZW lineage, which also generate T cell-dependent anti-DNA autoantibodies, although renal disease in the model is primarily glomerulosclerosis with limited interstitial inflammation. Finally, NZW/BXSB male mice express an additional copy of TLR7 via the Yaa translocation, produce both GC and EF derived autoantibodies and develop severe glomerulonephritis and interstitial renal disease. Autoantibodies in these mice are predominantly directed towards RNA-associated and phospholipid autoantigens ^107–112.

5.1. Is there a difference between targeting BAFF or APRIL alone compared with dual targeting?

We first asked whether targeting BAFF or both BAFF and APRIL differentially impacted autoimmune responses and longevity in NZB/W F1 mice. Whereas delivery of BAFF-R-Ig or TACI-Ig to pre-nephritic mice using an adenoviral vector delayed disease onset by ~6 months, neither agent was sufficient to reverse established disease. However, the addition of a brief two-week course of costimulatory antagonist CTLA4-Ig synergized with both BAFF-R-Ig and TACI-Ig to limit renal damage, proteinuria, and enhance survival ^{111, 113}. Notably, a related study confirmed TACI-Ig-mediated protection from progressive nephritis, but showed no effect on disease with BAFF-R-Ig treatment ¹¹⁴, findings that contrast with the robust impacts of even short-term BAFF-R-Ig in our hands. Potential explanations for this discrepancy include differences in dose or protein structure of the BAFF-R-Ig used in each study. Mechanistically, both BAFF-R-Ig and TACI-Ig resulted in a marked reduction in T2, marginal zone and follicular B cells, resulting in a secondary decrease in the accumulation of splenic class switched B cells, memory T cells and dendritic cells. In contrast, the number of splenic B1 B cells was unchanged. Moreover, despite prolonged B cell depletion, spontaneous GCs still formed, resulting in the generation of high-affinity somatically-mutated anti-DNA autoantibodies shortly after treatment discontinuation.

Plasma cell survival in mice depends on signals via BCMA or TACI that can be delivered by either BAFF or APRIL ligation. Notably, in the NZB/W model, IgM and IgG plasma cells exhibited distinct requirements for BAFF family survival signals. Short course TACI-Ig treatment markedly reduced serum IgM levels for several months, whereas BAFF-R-Ig exerted minimal impacts on IgM titers ¹¹³. These findings suggested that BAFF and APRIL together are required for the terminal differentiation and maintenance of IgM plasma cells. In contrast, IgG plasma cells are more resistant to the effects of TACI-Ig. Treatment with BAFF-R-Ig minimally affected serum IgG. In keeping with compensatory roles for BAFF and APRIL in plasma cell survival, a specific antibody to APRIL only modestly decreased autoantibody titers in aged NZB/W mice without affecting overall survival ^{16, 115}. Consistent with these results, long-term TACI-Ig treatment reduced splenic and bone marrow plasma cells in NZB/W mice, resulting in reduced anti-dsDNA titers and protection from proteinuria ¹¹⁴.

5.2. How does strain influence the response to BAFF/APRIL inhibition?

We next determined whether other lupus strains would be similarly responsive to administration of adenoviral vectors expressing BAFF-R-Ig or TACI-Ig. NZM2410 mice are characterized by the expansion of splenic plasma cells secreting IgG1 autoantibodies. This strain was highly responsive to short-term BAFF-R-Ig and TACI-Ig, even in animals treated after the onset of proteinuria. Similar to NZB/W mice, short-term exposure to TACI-Ig profoundly depleted serum IgM but induced only a temporary decrease in total IgG1 which was followed by the rapid return of class-switched autoantibodies ¹¹⁰.

NZM2410 mice manifest rapid progression of glomerular damage with predominant glomerulosclerosis and minimal interstitial inflammation. Despite glomerular IgG deposition in both BAFF-R-Ig and TACI-Ig treated animals, BAFF family blockade significantly attenuated renal damage and protected against proteinuria. Surprisingly, BAFF-R-Ig treated animals developed significant interstitial infiltrates, but lacked the tubular and endothelial damage associated with ESRD in this model suggesting that an additional checkpoint in the progression of renal damage is regulated by BAFF, perhaps via direct modulation of intra-renal myeloid cells ^{110, 116}.

In the MRL.Fas^lpr model, class-switched autoantibodies are produced by predominantly short-lived plasma cells generated via a T cell-dependent EF pathway. As predicted, TACI-Ig treatment resulted in a rapid decrease in both IgM and IgG autoantibodies and prolonged survival, without impacting T cell activation or renal interstitial infiltrates. Therefore, in MRL.Fas^lpr mice, plasma cell expansion requires continued BAFF and APRIL signals to maintain IgM and IgG plasma cells ¹¹⁷.

Finally, we studied the impact of BAFF inhibition in NZW/BXSB mice in which the Yaa translocation promotes TLR7-dependent anti-RNA and anti-cardiolipin autoantibody generation. Notably, both BAFF-R-Ig and TACI-Ig treatment limited proteinuria and attenuated the cardiac and renal damage characteristic of this model. However, protection from these disease manifestations occurred in the absence of any effect of serum autoantibody titers, glomerular IgG deposition or immune thrombocytopenia. Continued antibody-mediated platelet destruction despite BAFF/APRIL inhibition indicates no treatment-induced loss of autoantibody pathogenicity. Rather, BAFF inhibition likely attenuated the inflammatory response to autoantibody deposition in target organs ¹¹².

5.3. What is the effect of Type I IFN on response to BAFF/APRIL inhibition?

SLE is characterized by a Type I interferon (IFN) signature. Since type 1 IFN is known to induce feed-forward BAFF production by myeloid dendritic cells, we asked whether BAFF inhibition could protect NZB/W mice from type 1 IFN-accelerated autoimmunity. Continuous TACI-Ig initiated at the time of IFN treatment delayed disease onset and prolonged lifespan of IFN-accelerated NZB/W mice but lacked efficacy if TACI-Ig was delayed until after autoantibody formation. In this IFN-accelerated model, TACI-Ig still induced profound IgM depletion but exerted no impact on IgG autoantibodies, findings consistent with abundant GCs and IgG⁺ plasma cells in treated animals ¹¹⁸. Together, these data show that Type 1 IFN modulates the splenic microenvironment to support IgG plasma cell survival in the absence of BAFF/APRIL signals, perhaps by promoting B cell TLR7 expression or increasing expression of cytokines known to support B cell differentiation and survival. Moreover, these findings indicated that differences in the ability of BAFF/APRIL to support IgM vs. IgG plasma cells is not due either to differences in the microenvironment or intrinsic differences between short-lived and long-lived plasma cells, since IgG plasma cells in the IFN-induced model persist mainly in the spleen as a short-lived population. Despite conserved glomerular IgG deposition, renal inflammation was attenuated in the TACI-Ig treated mice. These findings correlated with a reduction in inflammatory mediators produced by renal myeloid cells, including TNF, CXCL13 and CCL5 ¹¹⁹. Thus, TACI-Ig either decreased the relative pathogenicity of deposited autoantibodies, or BAFF/APRIL inhibition exerted B cell-extrinsic effects on renal inflammation.

5.4. Does BAFF inhibition alter B cell selection?

An important question is whether BAFF inhibition restores tolerance in the naïve, EF, or GC B cell repertoire in SLE. To address this question, we utilized NZB/W mice bearing a germline heavy chain (D42) that was derived from an anti-dsDNA hybridoma. Despite the potential for diverse light chain pairing with D42, the majority of IgG anti-DNA autoantibodies in 6-month-old NZB/W mice derive from D42 paired with the VκRF light chain, the original hybridoma heavy/light chain combination which confers high affinity dsDNA binding without additional somatic hypermutation.

To study autoreactive B cell selection and survival in competition with wild-type cells, we established mixed BM chimeras which were treated with TACI-Ig from 3 days post-transplant to facilitate B cell reconstitution in a BAFF/APRIL poor environment. Consistent with the models described above, high-affinity D42/VκRF B cells were deleted in the BM at the pre-B / immature B cell stage. In contrast, lower affinity D42 B cells were deleted between the T1 and T2 stages, a process which was enhanced by TACI-Ig treatment. Using single cell PCR analysis, we observed no preferential effect of TACI-Ig on the light chain repertoire of D42 cells. Thus, our data show that competition with non-autoreactive B cells is sufficient for deletion of high affinity D42/VκRF B cells, while TACI-Ig broadly impairs the survival of lower affinity D24 B cells. However, TACI-Ig did not prevent GC formation or selection of D42/VκRF B cells into the GC and plasma cell compartments, resulting in equivalent anti-DNA IgG autoantibodies in treated and untreated chimeras despite profound B cell depletion ¹²⁰.

Since D42/VκRF DNA reactivity does not require somatic mutation, we performed similar experiments using 3H9 heavy chain knock-in mice. In this model, the 3H9 heavy chain can pair with multiple light chains to encode autoreactive and non-autoreactive specificities, and somatic mutation improves the affinity of self-reactive 3H9 autoantibodies. Deletion of 3H9 B cells occurs predominantly at the pre-B cell stage in this strain, with 3H9 cells surviving BM deletion exhibiting reduced surface IgM expression. Using BAFF-R-Ig to inhibit BAFF in this study, we observed prolonged proteinuria-free survival in both NZB/W and NZW/BXSB 3H9 chimeras. However, BAFF-R-Ig had no effect on 3H9 B cell deletion, IgM downregulation or selection of the naïve B cell repertoire in either strain ¹⁰⁸.

Despite central and peripheral tolerance mechanisms, 3H9 B cells can be positively selected into the GC compartment. DNA and cardiolipin specificity of 3H9-derived autoantibodies is facilitated by pairing with Vκ5–43 and Vκ5–48 light chains that confer either low or moderate autoreactivity, respectively. For this reason, we asked whether BAFF-R-Ig alters the selection and mutational frequency of these heavy/light chain combinations. In the NZB/W model, the relative representation of 3H9/Vκ5–43 and 3H9/Vκ5–48 GC B cells was not impacted by BAFF-R-Ig treatment, but the lower affinity Vκ5–43 encoded light chains expressed fewer mutations and were underrepresented in the plasma cell compartment following BAFF inhibition. In contrast, BAFF-R-Ig treated NZW/BXSB mice exhibited a ~4 fold reduction in the higher affinity 3H9.Vκ5–48 clones in the GC and plasma cell compartment, compared with Vκ5–43-expressing B cells, with no difference in somatic mutation rates. Altered GC selection correlated with modest reductions in anti-cardiolipin titers in NZW/BXSB 3H9 chimeras, events perhaps explained by BAFF-dependent modulation of TLR7 signals in this Yaa-expressing strain ¹⁰⁸.

Together, these data confirm the variable impacts of BAFF inhibition on the selection of naïve and activated B cells in lupus models. High affinity autoreactive B cells are deleted in the BM without further regulation by BAFF inhibition. Although cells escaping BM deletion may be subject to regulation by BAFF at the transitional stage, this process does not apply equally to all autoreactive specificities. Moreover, in spite of tolerance mechanisms, a subset of high affinity self-reactive B cells within the naïve compartment can enter autoimmune GCs, undergo somatic hypermutation and differentiate into pathogenic plasma cells producing class-switched autoantibodies. BAFF inhibition has modest, strain-specific impacts on GC selection at this stage. Whereas only lower affinity GC B cells are regulated by BAFF in NZB/W mice, the selection of pathogenic anti-cardiolipin expressing B cells is modestly impacted by BAFF inhibition in NZW/BXSB mice. Thus, these data indicate that reduced serum BAFF can exert a context-dependent impact on B cell selection, somatic mutation and plasma cell differentiation within autoimmune GCs.

5.5. Summary of therapeutic studies in murine lupus models

Together, these studies provide several important insights into BAFF/APRIL inhibition in SLE that are relevant to the treatment of human patients. First, there is considerable heterogeneity between murine strains with respect to responsiveness to treatment with either selective BAFF or dual BAFF/APRIL blockade. These distinctions are likely applicable to the diverse immunologic underpinnings of human lupus and may account for variable treatment responses in lupus clinical trials. Second, BAFF/APRIL inhibition is more effective when initiated early in the disease course, with reduced effectiveness after immune complex deposition and the onset tissue damage. Third, increased Type I IFN may confer resistance to BAFF/APRIL inhibition. Fourth, there are differences between blocking BAFF alone vs. combined BAFF/APRIL inhibition. Most prominent amongst these is the sustained decrease in serum IgM induced by TACI-Ig, which may increase infectious complications following TACI-Ig treatment. In contrast, TACI-Ig treatment exhibited greater efficacy than BAFF-R-Ig in certain models, including in MRL.Fas^lpr mice. Fifth, there is a modest and strain specific effect of BAFF inhibition on GC B cell selection that may modulate the pathogenicity of the emerging autoantibodies. Sixth, plasma cells required both BAFF and APRIL for long-term survival but the effect of BAFF/APRIL inhibition on IgG plasma cells is variable between models. Finally, BAFF inhibition appears to confer a component of protection from renal damage that is independent of direct impacts on B cells.

6. Therapeutic targeting of BAFF in humans

Given the variable effects of BAFF inhibition in mouse models it is likely that responses in humans will be heterogeneous and can only be evaluated in a clinical trial setting. It is therefore useful to compare the known effects of belimumab and atacicept in humans. Readers are referred to a recent comprehensive review that details the design and clinical outcomes of the human trials ¹⁰⁴.

6.1. Selective BAFF inhibition

Because of the mechanistic similarity of selective BAFF blocking agents, we will only consider the anti-BAFF monoclonal antibody belimumab here. Belimumab has consistently shown a benefit when added to standard of care therapy in moderately active SLE, particularly in those patients with higher disease activity, increased anti-dsDNA antibodies and low serum complement ¹²¹. Similar to mouse studies using BAFF-R-Ig, belimumab induces a profound reduction in naïve B cells and transitional B cells past the T1 stage, with lesser impacts on class switched memory cells, B1 cells and plasma cells ^{122, 123}.

Belimumab has an excellent long-term safety profile, and effects on serum total immunoglobulin titers are minimal. Although naïve B cells are depleted by ~95%, BAFF inhibition is associated with a temporary increase in memory B cells, perhaps because of expansion or peripheral mobilization following naïve B cell depletion. Nevertheless, memory B cells decreased substantially with prolonged belimumab treatment (by ~70% reduction after 7 years), likely because of reduced replacement from the naïve compartment. Similarly, circulating plasma cells are relatively resistant to short-term BAFF inhibition, but are reduced by ~70% in chronically-treated patients ^{122, 123}. This long-term reduction in memory B cells and plasma cells is associated with a 40–60% decrease in autoantibody titers, without increased infectious complications

Despite similar B cell depletion, not all treated subjects respond to belimumab. To identify predictors of belimumab response, PBMC gene expression was assessed in patients treated with the BAFF inhibitor tabalumab. Whereas ~75% of patients exhibited a Type 1 IFN signature at baseline, this did not change either with disease activity or in response to one year of treatment with tabalumab, and thus lacked efficacy as a clinical biomarker. In addition, serum BAFF levels could not be used to predict treatment response ¹²⁴.

While all three major belimumab clinical trials in general SLE met their primary endpoints, it is unclear whether this therapy will confer a similar benefit in lupus nephritis. Clinical trials in lupus nephritis are complex given the requirement for new therapies to be tested without withdrawal of standard immunosuppression, such as mycophenolic acid or cyclophosphamide. Therapeutic B cell depletion is accompanied by a homeostatic increase in circulating BAFF, which could result in two potential consequences. First, increased BAFF may alter the stringency of negative selection and allow the survival of low affinity autoreactive B cells, as described earlier. Second, BAFF may facilitate a niche for long-lived plasma cells in secondary lymphoid organs and thereby contribute to disease relapse ¹²⁵. For these reasons, the combination of B cell depletion with BAFF antagonists has been proposed as a useful strategy for severe lupus manifestations including lupus nephritis ¹²⁶. Unfortunately, initial data from the CALIBRATE trial, in which belimumab was added to a Rituximab, cyclophosphamide and prednisone induction regimen, did not show any clinical benefit at 48 weeks ¹²⁷. Nevertheless, several longer term studies have shown reduced SLE flare rates and slower progression of tissue damage in lupus patients taking belimumab ⁷. Thus, continued clinical experience with belimumab and other BAFF family inhibitors will be needed to define the therapeutic role of these agents in lupus nephritis.

6.2. Dual BAFF/APRIL inhibition

As described above, there are several reasons to predict that dual BAFF and APRIL inhibition may confer greater benefits to targeting BAFF alone. These include blocking BAFF/APRIL interactions with TACI, greater inhibition of signaling by BAFF multimers, and a more profound impact on plasma cell survival and other APRIL contributions to lupus pathogenesis. However, dual BAFF/APRIL inhibition may result in increased infectious complications because of reduced IgM responses and impacts in TACI-dependent innate signaling pathway mediated via endosomal TLRs. Clinical studies using the human TACI-Ig fusion protein atacicept are ongoing, but an initial study using two atacicept doses in SLE failed to meet the primary endpoint of reduction in SRI-4. However, several secondary endpoints were achieved, including an improvement in serological markers and reduced steroid requirement. As expected from murine models, serum IgM decreased by 70% whereas total IgG and autoantibody titers were only reduced by 30% ¹⁰⁵. These changes in immunoglobulin titers were not associated with increased infectious complications. Loss of protective immunity was evaluated in atacicept treated rheumatoid arthritis patients and varied from 0–25% depending on the antigen, reflecting the modest loss in total IgG ¹²⁸. Thus, there does not appear to be either a markedly increased infection risk or major therapeutic benefit of BAFF/APRIL blockade over BAFF inhibition alone. However, it is possible that a subset of individuals may respond better to a particular therapy.

6.3. Does BAFF inhibition alter B cell selection in humans?

Our mouse studies suggest considerable variability in the effects of BAFF inhibition on the regulation of B cell tolerance. To address this question in humans, we analyzed the B cell repertoire of lupus patients treated with belimumab for more than 7 years, a duration sufficient to allow complete turnover of the naïve B cell repertoire ¹²².

Naïve B cells:

Repertoire differences between transitional and mature naïve B cells reflect negative selection at the transitional checkpoint. Using next generation sequencing, we first asked whether belimumab treatment alters normal transitional selection. Surprisingly, we observed no impact of belimumab on VH family usage within the naïve repertoire compared with matched SLE controls. Furthermore, we observed no differences in pre- and post-belimumab naïve repertoire in the subset of patients were longitudinal samples were available. Prior studies have indicated that ~40% of naïve B cells from lupus patients are autoreactive, with a repertoire skewed towards VH4–34, 4–39 and 3–21 usage and more acidic CDR3 regions ⁷⁰. However, an in-depth analysis of multiple parameters of the B cell repertoire (including VDJ usage, D length, charge, and related biophysical characteristics) uncovered no additional differences between belimumab treated subjects and disease controls. Finally, VH4–34 is an intrinsically-autoreactive heavy chain gene that is censored in the GC and plasma cell compartments of healthy subjects, but enriched among lupus plasma cells, especially during disease flares ⁸⁷. Despite this putative role in lupus pathogenesis, we observed no preferential loss of VH4–34 in belimumab treated patients compared with SLE controls. Thus, despite profound depletion of the naïve compartment, BAFF inhibition resulted in no preferential loss or enrichment of specific VH genes among remaining naïve B cells ¹²².

Activated B cells and plasma cells:

Examination of the repertoire of mutated CD27⁻ “activated” naïve B cells uncovered a loss of VH4–34 sequences between the naïve and activated, and the naïve and plasmablast stages, in belimumab treated subjects, but not lupus controls ¹²². Although the plasmablast compartment exhibited abundant clonal expansion in SLE, no differences were noted between patients with or without BAFF inhibition. These findings are provocative, since, together with mouse studies, they suggest an alternate mechanism for BAFF inhibition beyond impacts on B cell selection at the T1-T3 transition. One potential mechanism is anergy induction in autoreactive naïve B cells such that they are unlikely to be activated via either EF or GC pathways. This model is consistent studies from the Diamond group showing that anergic B cells are infrequent in lupus patients, but similar in number in belimumab treated subjects and healthy controls ¹²⁹. An alternate explanation could be modulated endosomal TLR/TACI interaction in activated B cells resulting in reduced survival in low BAFF settings.

Overall, our findings are consistent with animal studies showing a limited impact of BAFF inhibition on the naïve B cell repertoire in the presence of physiologic competition from non-autoreactive B cells. As in mice, a modest effect of belimumab on the antigen-activated autoreactive immunoglobulin gene repertoire is apparent, but it is unclear whether these events are responsible for declining autoantibody titers or whether loss of autoantibodies reflects chronic B cell depletion and the loss of plasma cells over time.

7. Other effects of BAFF inhibition on inflammation

Since BAFF receptors are expressed by T cells and myeloid cells it is possible that BAFF inhibition alters the activation of these cells in the inflammatory environment. In T cells, roles for BAFF are controversial since both activating and inhibitory BAFF-R effects have been reported ¹³⁰. BAFF supports the survival and differentiation of monocytes and enhances the activation of human myeloid DCs. TACI deficient macrophages skew towards M2 differentiation, and thus induced less inflammation with reduced infection clearance ^{131, 132}. In murine arthritis, BAFF-silenced synovial dendritic cells remained immature and were unable to support T helper type 17 (Th17) differentiation ¹³³.

BAFF inhibition may also impact inflammation by modulating B cell regulatory function. For example, an intriguing new study showed that TACI-dependent production of the regulatory cytokine IL-10 by IgA⁺ plasma cells protected against experimental autoimmune encephalomyelitis (EAE) in BAFF-Tg mice ¹³⁴. In a separate study, local APRIL production by infiltrating macrophages induced IL-10 production by astrocytes to limit multiple sclerosis (MS) lesions ¹³⁵. Together, these mechanisms may account for the unanticipated increase in disease flares in the atacicept MS clinical trials ¹³⁶, and suggest caution with dual BAFF/APRIL inhibition in other inflammatory diseases. Similarly, whereas global and B cell-specific BAFF-R deletion protected against atherosclerosis, BAFF neutralization increased and BAFF overexpression limited atheroma formation in mouse models ^{137, 138}. Mechanistically, BAFF-dependent TACI signals in atheroma M1 macrophages decreased the production of the pro-atherogenic chemokine CXCL10, resulting in accelerated atherosclerosis after global and myeloid-specific TACI deletion ¹³⁷. Since elevated serum BAFF levels are associated with increased atherosclerosis risk in human SLE ¹³⁹, the clinical relevance of these animal models to human lupus is unclear. However, these studies emphasize the need for long-term studies of cardiovascular event rates in belimumab treated patients. In this context, protective effect of BAFF inhibition on adiposity and insulin resistance may also be relevant ^{140, 141}.

Finally, multiple mouse studies have demonstrated decreased renal inflammation in BAFF-deficient and BAFF-R-Ig/TACI-Ig-treated lupus models in excess of impacts on serum autoantibody titers. In the NZB/W model, our studies have shown protection from progressive lupus nephritis that could not be explained by a reduction in the autoantibody levels or reduced glomerular immune complex formation. Rather, gene profiling studies in NZM2410 mice indicated that progression from proteinuria to ESRD correlated with podocyte loss, tubular and endothelial dysfunction, and metabolic derangements, which was limited in BAFF-R-Ig treated animals despite ongoing glomerular immune complex deposits and inflammatory infiltrates ¹⁴². How BAFF inhibition mediated this effect and whether a similar therapeutic benefit will be observed in human clinical trials has not yet been determined.

8. Concluding remarks

Much has been learned about BAFF, APRIL, and their receptors since their initial discoveries two decades ago, but the full therapeutic benefits of targeting this family have yet to be achieved. In addition to clinical successes of belimumab in lupus trials, the recent genetic association of variants impacting BAFF levels with SLE and MS risk confirms the importance of this cytokine in human autoimmunity. Recent data have increased our understanding of how BAFF interacts with both the BCR and TLR pathways to regulate B cell survival, activation and differentiation via EF- and GC-dependent pathways. These data expand BAFF’s role beyond its essential ability to support naïve B cell survival to a complex one integrating innate and adaptive B cell responses during humoral autoimmunity. In this model, increased serum BAFF levels associated with inflammatory states and following therapeutic B cell depletion may relax the stringency of B cell selection within both the naïve and activated B cell compartments. Selective BAFF or combined BAFF/APRIL inhibition appears safe in human studies, with chronic BAFF blockade decreasing lupus flare rates and progressive organ damage over time. However, the clinical response to BAFF inhibition is heterogenous in both murine models and human lupus patients, emphasizing the need for improved therapeutic approaches and mechanistic biomarkers that can identify patients likely to respond to specific therapies.