BAFF appeared in the literature when it was cloned in 1999 (1–4). At that time, the pharmaceutical/biotech world had entered the era of DNA database mining, and a race for identifying the full sequence of new factors with potential as therapeutic targets was in progress. Considering the established clinical success of TNF inhibition, discovery of new genes belonging to the TNF family was an obvious focus of the data-mining effort. In 1999, four publications on BAFF were released within 2 months of each other, giving the molecule four different names, BLyS, BAFF, TALL-1, and THANK (1, 2–4). Of these four, two were of particular interest, including a collaboration between Biogen (Cambridge, Massachusetts) (now BiogenIdec) and the University of Lausanne in Switzerland (2), and the other from HGS (Rockville, Maryland) (1) which concluded with the prophecy: “As such, BLyS, its receptor, or related antagonists may find medical utility in the treatment of B cell disorders associated with autoimmunity, neoplasia, or immunodeficiency syndromes” (1). These early articles were correct about the importance of BAFF in B cell biology, but the exact details of BAFF's function in health, and its potential role in diseases was still lacking. The above articles proposed BAFF as an important co-stimulator of B cells, but it was not until later that year (1999) that Biogen scientists suggested that BAFF provided a critical B cell survival signal (rather than a proliferation signal), and that excess BAFF had a role in driving an autoimmune condition in mice which closely resembled SLE in humans (5). This last report thus became the first proof of concept that BAFF was indeed a therapeutic target, and that inhibitors of BAFF may have a use as a treatment for SLE (and potentially other autoimmune conditions). What was still unclear was how BAFF was driving autoimmunity. Another 10 years of research would be needed to obtain a much clearer picture as to how BAFF really works in health and autoimmunity.
BAFF is produced by macrophages, neutrophils, and monocytes and is required for the survival of mature B cells (4, 6). Soluble BAFF adopts two forms, one typical of a TNF-family ligand, a homotrimer, and the other a capsid-like structure of twenty trimers (a 60-mer) (7). In 2000, research showed that BAFF was a survival factor in vitro for maturing peripheral B cells in mice, in particular splenic transitional B cells (8). This study proved highly predictive of the phenotype of mice lacking BAFF. As published in the following year, BAFF deficiency in mice led to loss of mature B cells from both the follicular and marginal zone B cell subsets as a result of a developmental block in the periphery at the early transitional (T1) B-cell stage (9, 10).
BAFF is able to bind to three receptors: transmembrane activator and calcium modulator and cyclophilin ligand interactor (TACI), B-cell maturation antigen (BCMA), and BAFF receptor (BAFF-R) (11). However, BAFF-R is the only receptor that receives signals uniquely from BAFF, whereas another homologous TNF-family member, APRIL, also signals through BCMA and TACI (12). The mAb Belimumab is able to bind to soluble, but not membrane-bound BAFF (13), and as such the only mechanism of action is a reduction of signaling through the BAFF receptors by “mopping up” soluble BAFF in circulation. As Belimumab does not bind to APRIL, not all signaling through BAFF receptors is blocked (see Figure 1). Also, Belimumab has been designed to preclude antibody-dependent cell cytotoxicity, and the complement activation cascade, thus avoiding direct killing of B cells.
BAFF-R is the receptor that signals BAFF-mediated B cell survival. This was shown in A/WySnJ mice, which have a disruption in the BAFF-R locus by a dramatic drop in B cell numbers, and in BAFF-R-deficient mice (14, 15): both strains have a phenotype very similar to that of BAFF-deficient mice (9). BAFF-R is expressed on most peripheral B cells and its expression augments as B cells mature (15).
In contrast, BCMA expression is restricted to plasma cells. Investigation of its function in mice has yielded conflicting data. BCMA−/− mice, when compared with controls, have similar serum immunoglobulin levels, normal T-cell independent immune responses to nitrophenyl-Ficoll (NP-Ficoll), and normal serum titres in response to immunisation with a T-cell-dependent antigen, nitrophenyl-chicken gamma globulin (NP-CGG), out to day 45 (16). However, administration of TACI-Ig (blocking both BAFF and APRIL signaling) 6 weeks after an immunization, when the antigen-specific cells should have all been activated and the response effectively resolved, resulted in the number of antigen-specific antibody secreting cells dropping, suggesting that BAFF signaling does play an important role in cell survival of antibody-producing plasma cells in the bone marrow (17). As BCMA is the only receptor for BAFF expressed on long-lived bone marrow plasma cells, treatment with TACI-Ig in that experiment was likely to be acting by blocking BCMA signaling. Current views are that BCMA may be important for the survival of some, but not all plasma cells, and that both BAFF and APRIL may contribute to this effect. Blocking either BAFF, or APRIL, independently does not result in differences in numbers of bone marrow plasma cells, but blocking both reduces numbers dramatically (18).
TACI is the most intriguing of the three receptors and appears to have a dual function. Loss of TACI in mice leads to B-cell hyperplasia, autoimmunity and lymphomas (19, 20). The TACI−/− mice have an expanded B cell compartment, particularly affecting MZ and B1 cell subpopulations, suggesting a role for TACI in B cell homeostasis and immune tolerance. On the other hand, loss of TACI in mice leads to defective B cell responses to T-independent antigens (21, 22). Proliferative signals through TACI do not occur efficiently in response to BAFF 3-mers (the most prevalent form of BAFF in the circulation), but do to BAFF 60-mers and membrane-bound BAFF (23).
Complementing the studies in knockout mice, investigations of the phenotypes associated with lack of each of the receptors for BAFF and blocking BAFF signaling in vivo was accomplished soon after the first description of BAFF. These studies used soluble forms of the receptors engineered as fusion proteins with the Fc portion of an antibody, thus creating receptor-Ig decoy reagents (6). Biogen Idec used a soluble form of the BAFF receptor, BCMA-Ig, to block BAFF activity. Results showed a dramatic decrease in the number of B cells in vivo, further suggesting an important role for the BAFF system in B cell survival (6, 24). Mice transgenic for TACI-Ig were also created, and displayed blocks in B cell development similar to those observed in the BAFF-deficient mice (10, 24).
These studies were later extended to use of these blocking reagents in autoimmune-prone strains of mice to see if this could ameliorate disease. Studies in the SLE-like NZBxNZW F1 mouse showed that blocking BAFF in vivo using a BAFF-R-Fc resulted in significant improvement in the health of the mice, as judged by decreased proteinuria, decreased levels of anti-dsDNA, improved glomerular histology and increased survival (25). Importantly mice were treated with the BAFF-R-Fc three times a week for 5 weeks, but the decrease in proteinuria, and anti-dsDNA was seen out to 26 weeks, at which point the mice had B cell numbers that were within the normal range, indicating that the BAFF-R-Fc was depleting specifically some cells that were critical for disease progression and that protective effects of treatment extended beyond the point at which treatment was ceased. These studies demonstrated the potential of using antibodies to block the BAFF receptor in mice with established lupus, and encouraged pharmaceutical companies to develop BAFF blocking reagents for human use. It was just 4 years later, in 2003, that the development of Belimumab (also known as Lymphostat-B and Benlysta), a humanized IgG1 mAb that blocked human BAFF in vitro and BAFF in monkeys was reported (13). Belimumab inhibits the binding of soluble BAFF to BAFF receptors, but does not bind to membrane-bound BAFF, and has recently reached its primary endpoint in two multicenter Phase III trials comprising more than 1,600 patients (26).
The early studies, as well as highlighting the importance of BAFF as a survival factor for B cells, showed the adverse consequences of excessive BAFF signaling, with elevated BAFF as a driver of autoimmunity. The relevance of BAFF to lupus is highlighted by the pathology that develops in BAFF transgenic mice. This includes B cell hyperplasia, hypergammaglobulinaemia (with elevated antinuclear and antihistone autoantibodies), enlargement of spleen and lymph nodes, proteinuria and lupus-like features including complement deposition in the kidney and destruction of glomeruli (5). In addition, aged BAFF-transgenic (BAFF-Tg) mice developed symptoms reminiscent of Sjögren's syndrome, such as enlarged salivary glands, reduced saliva production and destruction of acinar cells (27). Interestingly, levels of circulating BAFF are elevated in as many as 61% of patients with SLE, and BAFF levels have been found to correlate with levels of anti-dsDNA (28, 29). Whilst the BAFF transgenic mice are a useful model it must be noted that the amount of BAFF produced in these animals is supraphysiological, and the elevation of BAFF in human patients is more modest, although serum BAFF varies greatly from individual to individual.
So what have the trials with the BAFF-blocking antibody, Belimumab taught us? Notably, the phase II clinical trial with Belimumab did not meet the primary endpoint. However, this was likely due to flawed patient selection since as many as 28% of recruited patients were sero-negative for lupus-type autoantibodies. When the data obtained were subsequently partitioned, with segregation into sero-negative and sero-positive patients, the sero-positive patients who received Belimumab, did meet the primary endpoints (30).
Both of the phase III clinical trials were assessed at 52 weeks of treatment, with the second trial further analysed at 76 weeks; the two trials were named BLISS-52 and BLISS-76, respectively. Both trials recruited only SLE patients that were “serologically active” (titres of antinuclear antibody of at least 1:80, or concentrations of anti-dsDNA antibodies of at least 30 IU/mL) at entry, a decision made on the basis of examination of data released from the phase II trials. Both trials used an “SLE-responder index” (SRI) to assess patients (31), a new procedure developed by clinicians across several medical centres combined with academic staff at several universities and in collaboration with personnel at HGS. The SRI had not been previously used in any other clinical trial for the treatment of SLE; this careful planning may explain, in part, why this trial did attain the primary endpoints. The SRI defines a disease response that is a reduction from baseline of at least four points on the “Safety of Estrogens in Lupus Erythematosus National Assessment-SLE Disease Activity Index” (SELENA-SLEDAI), with no clinically significant worsening of the British Isles Lupus Assessment Group (BILAG) index (no new BILAG A organ domain score indicating severe lupus disease activity and no more than one new BILAG B organ domain score, indicating moderate disease activity), and no worsening of the Physician's Global Assessment (indicated by an increase of 0.30 points above baseline) (31). This effectively means that the SRI is based on improvement of disease activity, without deterioration of the symptoms or wellbeing of the individual, and without development of new symptoms in previously unaffected organs. Use of the SRI metric meant that patients recruited into the trial were required to have a SELENA-SLEDAI score of greater than 6 when recruited, representing moderate to severe SLE. The BLISS-76 trial showed that at week 52 of treatment patients who were treated with Belimumab 10mg/kg (given at day 0, 14, 28, and every 28 days thereafter for the duration of the study) plus standard care, experienced a significantly higher SRI rate of 43.2% of individuals, as compared to patients treated with standard care plus placebo (intravenous saline injection), where the SRI improvement rate was 33.5%, a statistically significant, albeit modest, result (P = 0.02). The current “standard of care” for lupus includes corticosteroids, antimalarial agents or immunosuppressives (32). The standard of care in these studies allowed changes in the doses of drugs for the first six months, which differs from many other trials. The BLISS-52 trial reported similar results, at week 52 Belimumab 10mg/kg plus standard of care the SRI response rate was 57.6% compared to 43.6% for standard of care plus placebo, a statistically significant result of P = 0.0006 (30, 33, 37). Although the SRI improvement rate is modest, it must also be noted that the proportion of patients with a greater than 50% reduction in prednisone dose was also higher in the Belimumab treatment group than the placebo group (from week 24 until the conclusion of the study). This is an important point as sustained high doses of corticosteroids is a main cause of organ damage and morbidity in patients with SLE (34, 35). It is important to note that the BLISS-76 trial no longer met the primary outcome at 76 weeks, underlining the difficulties in achieving significance in lupus trials, and also highlighting that the differences between the treatment and placebo groups were modest. A number of factors unrelated to treatment may have contributed to this outcome, in particular patients withdrawing from the trial.
Belimumab would be expected to have the greatest effect on signals downstream of the BAFF-R, as it reduces signaling via BAFF, but not via APRIL. Human studies have shown BAFF-R is expressed on mature B cells and memory B cells, and on CD138+ plasmablasts (early plasma cells) within the tonsil, but not CD138+ plasma cells within the bone marrow (mature plasma cells) (36). The BLISS-76 trial provided some interesting clues as to which B cells were influenced by the mAb treatment. Statistical results from the phase III trial showed significant reductions in CD20+ cells at week 24, 52, and 76, including reductions in naïve B cells (CD20+CD27−) and activated B cells (CD20+CD69+) and reductions in plasmablasts (CD20+CD138+) (37, 38). At the highest dose of Belimumab used in the trial (10 mg/kg), reductions were also seen in CD20−/CD27bright short-lived plasma cells and CD20+/CD138+ plasma cells, although, as these measurements were taken from the blood, they almost certainly represent a reduction in the generation of plasma cells due to the reduction in B cells and plasmablasts, rather than a specific ablation of long-lived plasma cells (37). Supporting this idea, Belimumab treatment did not substantially affect the ability of SLE patients to maintain a protective secondary immune response to pneumococcal, tetanus and influenza vaccines, suggesting that both the long-lived plasma cell compartment and the memory B cell compartment were intact in the SLE patients treated with Belimumab, although the number of memory B cells in the blood was not tested directly. In addition, there is some controversy as to whether pneumococcal, tetanus and influenza titres reflect just long-lived plasma cell survival or both long-lived plasma cell and memory B cell survival (38–40).
Although Belimumab treatment did not influence protective pneumococcal, tetanus and influenza titres, it did decrease the production of a range of auto-antibodies including anti-dsDNA (median change from baseline measured at week 52: 37.6% 10 mg/kg Belimumab vs. 12.3% placebo, BLISS-52 trial), anti-Sm (56.3 vs. 29.6% placebo), anti-ribosomal P (55.6 vs. 5% placebo), anticardiolipin IgM (32.1 vs. 14.8% placebo), and anticardiolipin IgG (27.9 vs. 21.9%) (30, 41, 42). This suggests that the autoantibody secreting plasma cells have different survival features from the nonautoreactive plasma cells, and that the nonautoreactive plasma cells are resistant to Belimumab treatment, whereas the autoreactive plasma cells are responsive. The phase II trial with Belimumab showed that IgM and IgE median serum concentrations in the Belimumab 10 mg/kg treatment group decreased by 29% and 34% from baseline, respectively (P < 0.0001), where the placebo group showed <5% change from baseline. The IgG and IgA levels were reduced 10 and 14%, respectively (the Phase III data were similar with a 16% and 16% decrease, respectively, compared to baseline in Belimumab treatment group observed, and 4% and 3% decrease in placebo group) (30, 38). The IgM serum concentration is likely to be affected by the loss of B cells, as IgM-secreting plasma cells are thought to be predominantly short-lived. BAFF signaling is known to play a role in Ig class switching to IgE; IgE is turned over very rapidly with a half-life of only 2.7 days, and as 93% of the intravascular pool of IgE is catabolized per day, a decrease in IgE titres is likely to reflect a decrease in plasma cell numbers (43); however, since little is known about the longevity of IgE plasma cells it is not possible to tell if these effects relate to the production of plasma cells from B cells, or the survival of long-lived plasma cells. The small differences in the IgG and IgA levels suggest that Belimumab is not having an effect on the survival of the long-lived plasma cells that secrete these subclasses, but will affect the generation of new plasma cells of these isotypes. Although Rituximab treatment has led to more robust decreases in autoantibody titres it has not been as effective as Belimumab in SLE clinical trials. Therefore, perhaps the specific decreases in autoantibody titres and reduction in activated B cell numbers following treatment with Belimumab, while retaining other subsets of B cells and nonautoreactive plasma cells might explain why this therapy has been more effective.
Belimumab is the first new SLE treatment in over 40 years that holds promise for the treatment of SLE patients. The question is, what has been different this time? The BAFF-transgenic mouse model offered a number of interesting new clues. BAFF is important for determining the threshold for tolerance in the B cell compartment. High levels of BAFF allow the survival of B cells that would ordinarily be deleted in the periphery (44). It is not the high-affinity self-reactive B-cell clones that escape deletion, but rather the relatively low-affinity B cells that can enter the B-cell follicle and undergo proliferation. The survival of these low-affinity self-reactive B cells is not a result of any catastrophic breakdown of B cell tolerance in the bone marrow or the periphery, but rather reflects a loss of the fine-tuning of the immune system in inducing deletion, or anergy, of weakly self-reactive B cells in the periphery (44). The absence of any effect ensuing from an excess of BAFF production on B cell tolerance mechanisms in the bone marrow does not entirely surprise, as developing B cells in the bone marrow do not express any of the receptors for BAFF (44).
The clinical studies have shown that Belimumab does not completely deplete the B cell compartment, since the phase III BLISS-76 trial reported that some 44.3% of CD20+ B cells remained after treatment (from the analysis performed at week 76), this is not surprising as not all mature B cells are dependent on BAFF for survival as indicated by studies in BAFF-deficient animals. The degree of deletion of activated B cells (CD20+CD69+), was 49.1%, indicating that the Belimumab treatment was also effective at deleting activated B cells (although naïve cells were deleted most effectively with only 23.7% of naïve B cells remaining after treatment) and less effective at deleting memory B cells, as would be expected from the expression profile of the BAFF receptors (10, 45). These results were derived from B cell numbers in the blood, but studies in lupus-prone mice have shown that decreased B cell numbers were also seen in the spleen of mice treated with BAFF-R blocking antibodies (46). The inhibition of disease, despite retention of many B cells is an intriguing aspect of this therapy, and warrants further research.
Another intriguing question remains. HGS and GSK have shown that levels of autoantibodies, anti-dsDNA, anti-Sm, anti-ribosomal P, and anti-cardiolipin autoantibodies decreased during treatment, as well as activated B cells. Examining the total serum immunoglobulins and protective antibody titres to common vaccines in patients treated long-term will be very important to determine whether the treated patients will become more prone to infections over time, and will also help our understanding of how plasma cell homeostasis is achieved. Are these reductions in autoantibody titres sufficient to prevent reactivation of disease once patients cease Belimumab treatment, or will the inflammatory processes that have been set in motion in these patients require constant treatment for amelioration of disease? If Belimumab is protective for long periods following treatment, then the cost of treatment is brought down substantially, which certainly must be a concern with Belimumab's annual cost estimated to be $35,000 USD per patient (47, 48). This cost must be considered given that the Phase III trials showed that improvements in SRI in the Belimumab group over the current standard care group were only seen in a small number of patients (∼1 in 10, for the BLISS 52 trial. 57.6% responders in the Belimumab 10mg/kg group compared to 43.6% responders in the standard of care group) (30).
Following hard on the heels of HGS and GSK, ZymoGenetics have developed an alternative to Belimumab, Atacicept, another humanised mAb that targets both BAFF and APRIL. Atacicept is a soluble from of the extracellular domain of the TACI receptor, fused to a human IgG1 Fc domain. Thus Atacicept will target both B cells and also plasma cells by preventing signaling through all three BAFF receptors, BAFF-R, TACI, and BCMA (see Figure 1). It will be an interesting product in this context. Treatment of autoimmune-prone lupus mice with Atacicept resulted in delays in the development of proteinuria and increased survival (49). Initial human trials with Atacicept have demonstrated remarkable declines in total serum immunoglobulin levels: in patients treated with Atacicept every 2 weeks for 12 weeks, reductions were seen out to the endpoint at day 85, and median IgG values were reduced by 21%, IgA values by 37% and IgM values by 54% compared to baseline values (50). This is about double the reduction seen in Belimumab-treated patients, (IgG 10%, IgA 14%, IgM 29%, after Belimumab treatment iv at days 0, 14, 28 and every 28 days to 52 weeks along with standard of care, and then analysis at week 52). Presumably, as a result of the dramatic drop in serum immunoglobulins, the Phase II Atacicept trial (Atacicept in combination with mycophenolate) ran into problems with an increased rate of infections and was thus terminated (51). The problems associated with infections in the Atacicept trial highlight one of the dilemmas in treating patients with autoimmune disease associated with a range of autoantibodies. Is it necessary to deplete the long-lived plasma cells and risk increases in infections to obtain a favourable clinical outcome with respect to lupus symptoms? The Belimumab trial suggests that significant improvement for many patients can be achieved without targeting long-lived plasma cells. Experiments in the mouse showing that both BAFF-R-Ig and TACI-Ig are effective at relieving autoimmunity without eliminating all plasma cells support this possibility that targeting long-lived plasma cells is not required (25, 49). Only ongoing trials and long-term follow-up of patients will answer these questions, but with so many “anti-lupus” products coming forward, it is an encouraging time for therapists and their patients.
Another new product that is entering the clinic is LY2127399, which binds both soluble and surface-bound BAFF, but not APRIL. LY2127399 has shown therapeutic efficacy in rheumatoid arthritis (RA) in a Phase II trial (52), and is currently in Phase III trials for SLE. Data from this trial could tell us more about a potential role for membrane-bound BAFF in RA and, if taken into lupus trials could reveal the importance of membrane-bound BAFF signaling when results are compared with those from Belimumab trials.
The Belimumab trials provide some important messages to pharmaceutical companies embarking on clinical trials. The primary endpoints in the phase III clinical trials may have been met only because HGS and GSK were prepared to examine the results from the phase II trial and learn from their experiences therewith. Developing the new SRI or SLE-responder index, which attempts to give a more global picture of the progress of SLE in patients, meant the primary endpoint was met, where several objectives were not significantly different.
Although these clinical trials have given us much information on the nature of the lymphocytes present in the blood, we still know very little about the efficacy of intra-venously delivered mAbs at various other tissue sites, such as the peritoneum in humans. In mice (and man) the peritoneal cavity represents an important reservoir of a subset of potentially self-reactive B-lymphocytes, the B1 B cells. Interestingly, in BAFF−/− mice, the B1 compartment within the peritoneum is largely unaffected by loss of BAFF signaling (9) suggesting that these B1 B cells may have access to other signals that promote their survival. Also, the peritoneum is an important site of B cells that produce IL-10 (53), and IL-10 produced by B cells is important in suppressing inflammation in relevant mouse models, with many studies conducted using models of experimental autoimmune encephalitis (EAE) (54, 55). Elevated serum levels of IL-10 have been found in SLE patients, and serum IL-10 levels correlate directly with disease severity (56), perhaps reflecting the body's attempts to control systemic inflammation. Whether or not Belimumab leaves the B10 (or interleukin-10 producing B cells) intact has not been determined, but this may come to be important for long-term remission.
An aspect of Belimumab treatment that has received little attention is the effect that blocking BAFF will have on T cells. It is known that BAFF can act as a co-stimulus with TCR ligation for T cell activation in both humans and mice, in much the same way as anti-CD28 acts as a T-cell co-stimulator (57, 58), and BAFF-R has been shown to be expressed on both CD4+ and CD8+ central memory and effector memory T cells (58). The co-stimulation that BAFF provided to T cells resulted in both proliferation and secretion of a wide range of cytokines, IL-2, TNF-alpha, IFN-gamma, IL-5, and IL-13 (57). Given these observations blocking BAFF activity in patients with high levels of BAFF, should have an inhibitory effect on T cells, as well as potently suppressing B cells and particularly auto-reactive B cells. In the many current mouse models for lupus (NZB.NZW F1 mice, Lyn−/− mice, Lpr mice, and others (59)) cooperative activity of B and T cell auto-reactive cells is required to cause disease. Thus, inhibition of auto-reactive T cells may well be an important aspect of treatment, and may in part explain why Belimumab and not Rituximab is effective in SLE. The BLISS-76 study enumerated T cells following Belimumab treatment, but there has been no published evidence to date showing whether Belimumab effects T cell activation (37).
However, work on BAFF-transgenic mice showed that B cells, even in the complete absence of T cells are capable of driving autoimmune disease. Indeed, the pathogenesis of lupus-like features in BAFF-transgenic mice is T cell independent, but does depend on toll-like receptor (TLR)-signaling, in that BAFF-transgenic mice reconstituted with MyD88−/− B cells (MyD88 being a common signaling element to many TLR) are protected from disease (60). This high dependency on MyD88 signaling demonstrates the important role that the innate immune system plays in balancing the immune system, based on its exquisite sensitivity to Toll signaling. Indeed, the use of anti-malarial drugs to treat lupus does have some efficacy due possibly in part to blocking some aspects of TLR signaling (61), and it is thought that an environmental infectious trigger can be one pathogenic event for lupus (62, 63). It is likely that there are some patients for whom the disease is independent of T cells (mirroring the BAFF-transgenic situation), and others for whom T cells play a major contributory role to pathogenesis (such as is observed in the Lyn−/− mice, Lpr mice, NZB.NZW F1 mice and others). Being able to identify which patients fall into each of these groups and tailoring treatment accordingly, could become relevant to therapy.