Therapeutic B cell depletion and regeneration in rheumatoid arthritis: Emerging patterns and paradigms


  • Gregg J. Silverman

    Corresponding author
    1. University of California, San Diego
    • University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0663
    Search for more papers by this author
    • Dr. Silverman has received honoraria (less than $10,000) from the Genentech Speakers' Bureau.


In recent controlled clinical trials (1–3), B cell–targeted therapy using rituximab, a chimeric anti-CD20 monoclonal antibody, was shown to be effective in patients with rheumatoid arthritis (RA). However, even though use of these anti-CD20 antibody infusions has become increasingly commonplace, especially in hematology/oncology practices, this FDA-approved therapeutic modality is still unfamiliar to most rheumatologists. Although these treatments predictably induce the loss of most detectable circulating B lymphocytes, the greater implications of B cell depletion therapy for host immune defenses are only now slowly being revealed. In this issue of Arthritis & Rheumatism, the report by Roll et al (4) provides further contributions to our understanding of the immunobiologic sequelae of these treatments, as leading research groups have begun to characterize the central immunobiologic changes in the B cell compartment that are induced by anti-CD20 treatment.

Fundamental questions concerning B cell depletion and reconstitution

In general, the demographic characteristics of patients in recent small mechanistic studies (4, 5) and patients in larger controlled trials (1–3) are similar to those of patients involved in other recent trials of biologic agents for the treatment of seropositive classic RA. The aforementioned patients received either the rituximab regimen that was first developed for patients with non-Hodgkin's lymphoma (NHL) (i.e., 375 mg/m2) or the regimen that is now more commonly used for patients with rheumatic disease (i.e., 2 infusions of 1,000 mg given 2 weeks apart); treatment differences included the inclusion or exclusion of immunomodulatory or antiinflammatory agents and corticosteroids.

These previous studies have shown that in patients with RA, treatment with rituximab infusions commonly results in the selective loss of virtually all (>97%) B-lineage blood cells; several studies have demonstrated that inadequate blood B cell depletion is uncommon among seropositive patients with RA. Breedveld et al reported that the concentration of peripheral circulating B cells became highly depleted within days following the administration of rituximab (6). For the great majority of patients, the levels of circulating B cells remained greatly depressed, with few (or no) detectable CD19+ cells for at least 4–5 months, before increases were first detected. Although the time range was broad, in some patients the start of B cell return was evident at 16 weeks and coincided with a mean low-level serum rituximab concentration of 7.5 μg/ml, while in other patients the concentration of B cells in blood remained depleted for the entire 48-week followup period (6).

In the phase IIa controlled study described by Breedveld et al, the patterns of peripheral B cell return did not appear to be appreciably different between responders (who attained improvement of at least 20% according to the American College of Rheumatology criteria [ACR20] at 24 weeks) (7) and nonresponders, or between responders who maintained their response at 48 weeks and those who did not (6). In approximately half of patients with RA, clinical disease activity flared within 2–3 months after the first evidence of B cell “regeneration,” while in the remaining patients the clinical benefits persisted for 5–17 months after the return of blood B cells (5, 8, 9).

In the blinded, controlled, phase IIb DANCER (Dose-Ranging Assessment International Clinical Evaluation of Rituximab in RA) study, 65% of rituximab-treated patients achieved at least an ACR20 clinical response at week 24, and an ACR20 response was achieved by 73–75% of patients who received rituximab plus methotrexate (MTX) (3). Similarly, among patients in the phase III REFLEX (Randomized Evaluation of Long-Term Efficacy of Rituximab in RA) study, who had an inadequate response to anti–tumor necrosis factor α (anti-TNFα) therapy, ∼50% attained an ACR20 response or better (2). In that study, most clinical responses were demonstrated by week 8, and additional responses were achieved for 4 months after infusion (2). For most patients in the DANCER and REFLEX studies, the clinical benefits of rituximab persisted at least as long as peripheral B cell levels remained greatly depressed. In a study reported by Emery et al (10), 141 patients with RA who received a second or third course of rituximab had biologic and clinical response patterns similar to those associated with the first course of treatment, with the same attractive safety profile as that observed in previous trials (10). Taken together, these findings emphasize that the biologic changes responsible for effective blood B cell depletion are required for the induction of a clinical response. However, the larger trials also showed that one-third or more of the patients with RA who displayed a biologic response (i.e., blood B cell depletion) to rituximab infusions still did not gain appreciable clinical benefits.

The use of rituximab plus weekly MTX is more likely to induce longer-duration blood B cell depletion and more durable clinical responses, compared with use of rituximab alone (1). Rituximab co-treatment with MTX or cyclophosphamide was associated with a higher frequency of ACR50 and ACR70 clinical responses (1). Findings in a recent small nonblinded trial suggested that clinical responses may also be more durable after rituximab co-treatment with cyclophosphamide (11), but this was not demonstrated in a larger, controlled study (1). In any case, these findings suggest that the addition of certain second agents can affect the duration of the clinical response to anti-CD20 treatment. Although currently these enhanced benefits are poorly understood, we speculate that they may derive from a reduction in disease-associated inflammation that diminishes associated B cell pro-survival influences, as discussed below.

Factors that may interfere with anti-CD20–mediated B cell depletion

Although the above-described findings indicate that most patients with RA are responsive to the B cell depletional activity of rituximab, the efficacy of this anti-CD20 reagent is not always equivalent in patients with other autoimmune diseases, and these differences are believed to reflect variations in immune set points that affect the efficiency of B cell depletion (12). For example, the expression of Fcγ receptor (FcγR) by leukocytes provides a signaling pathway that can act as a central interface between the adaptive and innate immune systems, enabling specific IgG-containing immune complexes to recruit proinflammatory mononuclear cells into immune responses. FcγRIII also regulates antibody-dependent cell cytotoxicity, which has been reported to play an essential role in rituximab-induced B cell depletion. Studies in patients with NHL have shown that clinical nonresponsiveness to rituximab can correlate with inheritance of the low-affinity (i.e., F176) FcγRIII allele (13). Similarly, in a small nonblinded study, many of the patients with SLE who were resistant to rituximab-mediated B cell depletion were found to have inherited this same low-affinity allele (14). The cause of inefficient blood B cell depletion was not clearly defined in the remainder of patients, but in some cases low circulating levels of complement were suspected (14), as discussed below.

Certain FcγR genes can themselves represent susceptibility factors for the development of autoimmune disease. In fact, inheritance of the low-affinity FcγRIIIa allele is reported to be a strong risk factor for the development of SLE and nephritis across many ethnic populations, which explains the overrepresentation of such genetically predisposed individuals among patients with SLE (15, 16). Although it remains controversial whether certain FcγRIII alleles predispose to RA, a recent study demonstrated no association with (and, even less frequently, inheritance of) this low-affinity FcγR allele, and disease susceptibility may instead correlate with inheritance of the high-affinity allele (17).

The pathway of complement-dependent cell lysis has also been implicated in anti-CD20–mediated B lymphocyte killing in some patients (for review, see ref. 18). It therefore remains likely that in some patients with SLE, depressed circulating complement levels, either due to inherited or acquired deficiency states, may also contribute to the lack of response to anti-CD20 therapy. In contrast, in patients with uncomplicated RA, the levels of complement in the circulation are generally normal, although there can be significant complement consumption in affected rheumatoid joints (19). Hence, it remains possible that localized complement deficiency could reduce the efficacy of B cell depletion in rheumatoid joints. Therefore, there are many reasons why the immune responsiveness of patients with RA to anti-CD20 infusions is often different from that of patients with SLE (20). Moreover, it may be predicted that for many (or most) patients with RA, the deletion of synovial B cells may be much less efficient after anti-CD20 treatments compared with the deletion of blood B cells.

Which B cells remain in blood after anti-CD20 infusions?

Even in healthy individuals, the bloodstream contains a mixture of types of B-lineage cells, because diverse B-lineage cells pass through the circulation as they come from, and travel to, different anatomic sites. In fact, among patients with RA, absolute B cell levels vary greatly, and a subset of patients is reported to have chronic, often severe B lymphopenia, which may be more common among patients expressing the HLA–DR4/shared epitope and those with greater acute-phase responses (21).

While we currently know little about the effects of anti-CD20 therapy within lymphoid tissues (and sites of disease), as described above, rituximab treatment commonly induces >98% depletion of measurable CD19-bearing B-lineage cells in the blood of patients with RA. Flow cytometric analyses have shown that most (∼80%) of these remaining B cells display phenotypic markers identifying them as memory B cells (i.e., CD27+), representing a 2-fold enrichment compared with that in healthy individuals. CD27+ memory B cells are also reported to persist in the blood of patients with SLE after rituximab treatment (22; for review, see ref. 23). These findings therefore support the notion that although many (or most) mature B cells are deleted following anti-CD20 infusions, current anti-CD20 treatment regimens do not completely erase immune memory related to the immunopathogenesis of RA (discussed further below).

Surveys of patients with RA have shown that at the induced nadirs of their blood B cell levels, residual circulating plasma cell precursors (i.e., CD19+,CD38+++) remained, which is consistent with the notion that plasma cells, which do not bear CD20, are not substantially affected by rituximab. Because plasma cells are major sources of circulating immunoglobulins, these findings are believed to explain why total IgM, IgA, and IgG levels in most patients show, at most, minor decreases but generally stay within normal limits. However, treatment often decreases the level of rheumatoid factor (RF) autoantibody (9), which may indicate that some RF-bearing B cells involved in the pathologic process are affected by these treatments. In a few cases, RF levels have become undetectable after treatment, but this effect has been seen in individuals with low pretreatment levels of RF. The level of IgG anti–cyclic citrullinated peptide can also be decreased, although reductions are often more modest (19).

What are the features of peripheral blood reconstitution?

Available reports are in agreement that at the time of B cell recovery following anti-CD20 treatment, the predominant blood B-lineage population is CD38high,IgD+,CD10+,CD24high. These cells do not display the CD27 memory B cell marker, and their antibody gene rearrangements lack extensive somatic hypermutation. Based on a differentiation schema from studies of tonsil from healthy individuals, this early wave of regenerating B cells was previously characterized as naive germinal center founder populations. However, more recent studies have identified these as transitional B cells (24, 25), which are antigenically naive recent bone marrow emigrants that derive from immature B cells. While the period of blood B cell depletion can be associated with elevated serum levels of BAFF, a TNF family B cell prosurvival factor, BAFF levels later decrease when B cell levels begin to normalize (11). Although bone marrow analyses have not directly confirmed this notion, the early period of B cell recovery following rituximab treatment likely reflects the reestablishment of B cell neogenesis in bone marrow. These newly generated B cell precursors, which are released into the bloodstream, must then enter peripheral lymphoid tissue for further maturation and differentiation (Figure 1). In recent studies, circulating transitional B cells were also shown to be increased in neonates and in patients with common variable immunodeficiency, those with severe human immunodeficiency virus disease, or in patients recovering from hematopoietic stem cell transplantation (26, 27).

Figure 1.

Distinct stages of B cell maturation as identified by differential expression of patterns of surface phenotype markers. In adults, B cell progenitors and precursors are produced in the bone marrow, leading to the differentiation of transitional B cells that are released into the bloodstream before they enter peripheral lymphoid organs. In patients with rheumatoid arthritis, transitional B cells, aberrant related B cell precursors, recirculating follicular B cells, and plasma cells may enter inflamed joints in response to chemokine gradients along high endothelial venules. MHC II = class II major histocompatibility complex.

A small population of plasma cells with the phenotype of CD38high,sIgD−,CD27high B cells has been described during the early regeneration period; pretreatment levels of these cells were rapidly attained. These cells had been undetectable during the period of induced B cell depletion, despite the fact that they did not express CD20. The demonstration of these plasma cells during the recovery phase appeared to identify the likely source of the hypermutated antibody genes characterized in an earlier survey of post-rituximab B cell regeneration (28). Although such expansions have not been documented in other reports, this could reflect differences in sampling time points, technical difficulties in the detection of transient cellular changes, or differences between individual patients evaluated. Taken together, all studies to date agree that most early-returning B cells are naive and are likely newly generated in the bone marrow. In contrast, in many patients with RA, the overall representation of circulating memory B cells may be slow to recover after treatment; the level of CD27+ B cells was reported to remain reduced to less than 50% of pretreatment values for more than 2 years posttreatment (4). Among the few reports of patients with RA who were studied after re-treatment with rituximab, one described progressive decreases in the levels of memory B cells after each sequential cycle of re-treatment with rituximab (4). Significantly, patients who displayed higher posttreatment levels of CD27+ memory B cells were reported to be more likely to experience rapid clinical relapse soon after B cell recovery (5), although the general relevance of these observations will need to be confirmed in larger controlled studies.

What are the long-term sequelae of B cell depletion treatment?

Although rituximab does effectively delete the vast majority of circulating B cells, surveys of >900 patients with RA who were enrolled in controlled trials have not demonstrated an increased rate of serious infection. Moreover, mycobacterial and opportunistic infections in treated patients with RA are distinctly uncommon. The type of adverse events in patients with RA are similar to those in patients with NHL (29); more than 700,000 full courses of rituximab therapy have been administered over the past 9 years since FDA approval for NHL was granted. Therefore, although it is feasible that other adverse outcomes will be recognized later, the available extensive data indicate that rituximab has a highly attractive safety profile that likely matches or exceeds that of current TNF inhibitors.

RA is a disease of ectopic infiltrative neolymphogenesis

Based on evolving perspectives on the pathogenesis of RA, for many patients with this chronic inflammatory disease, RA can be viewed as a disease of ectopic lymphoid tissue generation. The key lesion involves synovial accumulations of infiltrating leukocytes, with a predominance of B cells, at an anatomic site that is bereft of these cells in healthy individuals.

Although the healthy synovium contains only a veil-like layering of fibroblast-like synoviocytes that is generally 2–3 cells thick, the rheumatoid synovial pannus can form into organized frond-like growths of defined architectural organization. Hence, the central pathologic feature of RA is a chronic cellular accumulation, which in many cases resembles the features of a lymph node, at a site that should have no such accumulation.

An estimated 60% of rheumatoid synovial biopsy specimens contain lymphocytes, which have been reported to accumulate into infiltrates with 3 predominant histopathologic patterns: lymphocytic infiltrates with interdigitating dendritic cells (DCs) and variable representations of local B cell accumulations; disorganized groupings of infiltrating B and T cells in more substantial numbers, associated with interdigitating DCs; and T and B lymphocyte aggregates that are associated with follicular DC (FDC) networks (30–32). In the latter pattern, which is observed in fewer than one-third of patients, the synovial cellular infiltrates include distinct B cell follicle-like structures that appear to be organized into close spatial relationship to clusters of T cells (33), and these aggregates may be directly adjacent to areas of cartilage and bone destruction. It was recently reported that inflamed RA synovial tissue can also be associated with breaches in the cortex of the adjacent bone, resulting in the formation of B cell–rich aggregates in the underlying bone marrow at sites of increased new bone formation (34). It is currently unknown whether disease progression is associated with a linear, predictable, sequential change in these histologic patterns, and/or whether a specific pattern can reflect disease severity or can be used as a predictor of prognosis. However, as discussed further below, observations in SCID mouse–human synovial tissue chimeras suggested that the response to certain biologic therapeutic agents can differ based on the dominant histologic pattern in the rheumatoid synovial infiltrates.

Although organized lymphoid infiltrates are common in the diseased joints of patients with RA, these synovial ectopic foci are, by themselves, not entirely specific and diagnostic for RA. The affected joints of patients with ankylosing spondylitis have also been shown, at times, to harbor germinal center–like aggregates (35). Even the synovia of osteoarthritic joints can occasionally contain infiltrates of activated B cells that display clonally related antibody gene sequences (36, 37). These findings suggest that mechanisms common to many chronic inflammatory processes may be responsible for the synovial accumulation of B cells in ectopic lymphoid tissue; however, the accumulation of plasma cells does appear to be specific to RA. In RA, these plasma cells may be organized in concentric rings around the large clusters of T cells and B cells or as perivascular clusters (30). Moreover, in vitro studies have confirmed the special capacity of activated rheumatoid synoviocytes to produce factors that support the terminal differentiation and survival of plasma cells (38). These findings support the hypothesis that there are immunopathogenetic pathways intrinsic to RA that are intimately linked to the observed disease-associated abnormalities in the maturation of B-lineage cells (39).

Stable ectopic lymphocytic infiltrates have also been described in several other chronic diseases; examples include salivary infiltrates in Sjögren's syndrome, thyroid infiltrates in autoimmune thyroiditis, and thymic infiltrates in myasthenia gravis, among others. There is also increasing interest in understanding the potential roles of infiltrating B cells in lupus nephritis. Cassese et al reported that disease progression in a murine lupus disease model was associated with the accumulation in the kidney of large numbers of infiltrating lymphocytes, which included autoantibody-secreting plasma cells (40). However, the currently available data on lymphocytes that infiltrate the kidney during clinical lupus are very limited (41).

Pathways responsible for the development of organized lymphoid tissue

Advances in our understanding of the pathways responsible for the guided migration of lymphocytes to peripheral lymphoid tissues and the subsequent maintenance of their survival have provided a new perspective on the origins of the lymphoid infiltrates of rheumatoid synovium. During RA pathogenesis, a distortion of physiologic cellular trafficking mechanisms no doubt occurs, as earlier-arriving cells release chemokine factors that become the beacons that attract responsive sets of mononuclear cells in the circulation. Lymphocytes that pass through the circulation express distinct and specific chemokine receptors on their membrane surfaces as a consequence of their level of maturation and phenotypic subset, which enables them to be actively guided to peripheral sites as they follow the gradients of chemoattractant chemokines.

The organization of peripheral lymphoid tissue is dependent on lymphotoxin, a TNF family member, which exists in 2 forms: a membrane-bound heterotrimer comprising 2 β-chains and 1 α-chain (lymphotoxin α1β2) that is a ligand for the lymphotoxin β receptor, and a soluble homotrimer (lymphotoxin α3) that is a ligand for both TNF receptor I (TNFRI; p55) and TNFRII (p75) (42). In physiologic peripheral lymphoid tissue, B cells constitutively express lymphotoxin α1β2, which can engage lymphotoxin β receptor on stromal cells and cells of myeloid lineage, resulting in the induction of the chemokine, CXCL13 (also termed B lymphocyte chemoattractant or B cell–attracting chemokine). B cells are attracted toward the CXCL13 produced by these stromal cells through their membrane expression of the chemokine receptor, CXCR5. In response, these facilitated interactions with stromal cells induce the display of more lymphotoxin α/β. In parallel, T cells that express the receptor CCR7 can recognize the chemokine ligands, CCL21 and CCL19, which are produced by other stromal cells. By these pathways, the cells that express these lymphocyte-targeted chemokines influence the trafficking B cells and T cells to become segregated into distinct lymphocytic accumulations that are key features of peripheral lymphoid tissues.

Lymphotoxin, which is a major determinant of the tropisms that underlie the organization of normal physiologic lymphoid tissue, also likely contributes to the architectural organization within disease-associated inflammatory tissue infiltrates. For instance, activated T cells in the joint produce lymphotoxin α1β2 (43), which may act as a downstream effector that is required for the development of primary B cell follicles in these synovial infiltrates. In a recent study, the expression of lymphotoxin α1β2 and expression of CXCL13 transcripts were found to be independent variables that correlated with the presence of germinal center–like lymphoid aggregates in rheumatoid synovial biopsy specimens (44). Moreover, treatment with a decoy receptor that blocks lymphotoxin was shown to ameliorate synovitis in murine collagen-induced arthritis (CIA), a relevant and accepted model with many of the features of RA (45).

During their passage through the venules of affected synovial tissues in patients with RA, circulating lymphocytes may preferentially interact with endothelial surfaces bearing a sulfotransferase termed GlcNAc6ST-2 (also known as HEC-GlcNAc6ST, GST-3, LSST, or CHST4). Both TNFα and lymphotoxin α/β can induce the expression of GlcNAc6ST-2 in cultured human umbilical vein endothelial cells (46). These observations suggest that the cytokine/chemokine pathways associated with the pathogenesis of RA also contribute to the induction of specific vascular “addressins” expressed on synovial vessels. These vascular signals direct circulating myeloid and stromal cells to enter articular sites of inflammation and tissue damage for their accumulation into these pathologic infiltrates.

Although largely overlooked in recent literature, etanercept, a soluble TNFR:Fc hybrid molecule that was the first TNF-targeted therapeutic agent approved for human use, also has binding activity for the lymphotoxin α3 homotrimer. In a preliminary study involving a small number of patients with RA, Anolik et al (47) showed that etanercept treatment significantly reduced memory B cell levels without affecting levels of naive mature B cells in the blood, although these changes were seen after treatment with other disease-modifying antirheumatic drugs. Treatment with etanercept also greatly reduced the levels of memory B cells, FDCs, and germinal center structures in tonsil biopsy specimens. These findings suggest that clinical benefits from etanercept could derive from blockade of the organizing influence of lymphotoxin, although this apparent blockade in the recruitment of B cells into the pathogenesis of RA could also reflect the influence of TNF on factors involved in chemotaxis and survival.

Survival factors for B cells create niches and sanctuaries

The survival of B lymphocytes is believed to require active interactions with specific cell-surface and soluble factors that are locally expressed, to create supportive local environments termed niches. Such survival signals are believed to extend the life span of these lymphocytes at local sites in lymphoid tissues, and similar processes are likely to also convey resistance of B cells to therapeutic CD20-targeted interactions. Studies in human CD20–transgenic mice and in nonhuman primates have shown that certain types of mature B cells are not efficiently depleted by rituximab, even though these B cells bear substantial levels of CD20 (48–50). Similar results were observed with anti–murine CD20 treatments (51, 52). For example, although circulating blood B cells and follicular B cells are rapidly and efficiently depleted by anti-CD20 treatment, marginal zone B cells, B1 cells, and germinal center B cells in Peyer's patches of gut-associated lymphoid tissue are resistant (51–53). Hence, there is concern that survival factors produced as part of the proinflammatory milieu may provide sanctuaries for B cells involved in pathogenesis that protect against rituximab-induced depletion.

Rheumatoid synovial tissues are rich sources of CXCL12 (also termed stromal cell–derived factor 1 [SDF-1]) (54). SDF-1 was first described as a chemotactic and survival factor for early B cell precursors (pro-B, pre-B, and immature B cells), without substantial effects on mature B cells. However, plasma cells, which represent end-differentiated antibody-producing B-lineage cells, are also highly responsive to SDF-1 (Figure 2). When stimulated with TNF and interferon-γ (IFNγ), RA fibroblast-like synoviocytes produce SDF-1 as well as CXCL13, which likely contribute to the accumulation of B-lineage cells in ectopic lymphoid aggregates in rheumatoid synovium (44, 55, 56). In fact, the prominence of plasma cell accumulations in rheumatoid tissues in part likely reflects the migration of these cells toward gradients of SDF-1, CXCL9 (monokine induced by IFNγ), CXCL10 (IFNγ-inducible protein 10), and CXCL11 (IFN-inducible T cell chemoattractant) (57). SDF-1, as well as interleukin-5 (IL-5), IL-6, TNFα, and ligands for CD44, can also prolong the longevity of murine plasma cells (58). Contacts mediated through B cell membrane–associated α4 integrins provide additional pro-survival influences for marginal zone B cells and plasma cells, and perhaps other B-lineage cells (52). These factors are believed to support the survival of plasma cells that are often prominent in rheumatoid synovial infiltrates and that play key roles in the immune complex–mediated pathologic pathways responsible for joint destruction (for review, see ref. 59).

Figure 2.

Variation in responsiveness to survival factors during B-lineage cell maturation and differentiation. Most observations have been made in murine models, but human B cells may differ from mouse B cells, and certain subtleties may currently be undefined. For example, although interleukin-7 (IL-7) is essential for B cell maturation in mice, it is not essential in humans (83), although human pre-B cells express IL-7 receptors and respond to IL-7 in the presence of stromal cells (84). The role of α4 integrin in human B cell survival remains to be confirmed. General principles of the patterns of responsiveness to pro-survival factors are represented, but the details for certain factors may not be well defined. SDF-1 = stromal cell–derived factor 1.

As part of RA pathogenesis, locally produced TNF no doubt leaks into the bloodstream or enters the bone marrow by local diffusion, where it may mediate the early release of B cell precursors that may potentially bypass clonal selection mechanisms required to maintain immune tolerance (60) (Figure 3). The triggering of mast cells and the release of humoral factors during immune complex–mediated pathology induce the release from the bone marrow of neutrophils, which follow their own chemoattractants to these sites of synovial injury. In addition to directly exacerbating local tissue injury through the release of proteases and related factors, these neutrophils can be major sources of the TNF family B cell pro-survival factors, BAFF (also called B lymphocyte stimulator [BLyS; trademark of Human Genome Sciences, Rockville, MD], TALL-1, THANK, and zTNF-4) and APRIL (a proliferation-inducing ligand) (61, 62). Notably, IFNγ and TNF priming of synoviocytes can also induce high levels of BAFF expression (63) (Figure 3).

Figure 3.

Diagram showing that leukocyte trafficking, cytokine production, and local enhanced expression of B cell survival factors are integrally linked to the dynamic processes responsible for the pathogenesis of rheumatoid arthritis (RA). This simplified schematic details several pathways by which the local release of proinflammatory factors, such as interferon-γ (IFNγ) and tumor necrosis factor α (TNFα), leads to enhanced production of survival factors including BAFF, APRIL, and stromal cell–derived factor 1 (SDF-1), although others are certainly also expressed. Examples are shown in which local TNF release into the circulation induces the release of myeloid cells that later enter the joints by following chemokine gradients. These cells may then express B cell survival factors that amplify the accumulation of synovial lymphoctyes. Cytokine stimulation of synoviocytes stimulates their expression of SDF-1, BAFF, and APRIL. These are some of the relevant pathways, but no doubt there are others that also contribute to the self-perpetuating pathogenesis of RA. IL-1 = interleukin-1; pDC = plasmacytoid dendritic cell.

Levels of BAFF and APRIL are elevated in patients with RA, with significantly higher levels in synovial fluid than in serum (64, 65), confirming that there is high local production at these sites of inflammation. The cell surface receptors for these factors, TACI (TNFRSF13B, CD267), BAFF-R (BR-3, TNFRSF13C, CD268), and BCMA (TNFRSF17, CD269), which are differentially expressed by B-lineage cells based on the stage of differentiation, provide the signals required for survival and maturation (66) (Figure 2). These factors have therefore been proposed to protect synovial B-lineage cells from apoptosis in inflamed rheumatoid joints and contribute to the properties of specialized synovial nurse-like cells peculiar to RA synovium, which have been hypothesized to mediate both the homing and survival of B cells in these tissues (55). Notably, serum levels of the B cell survival factor, BAFF, were reported to rise markedly after rituximab treatment and remained elevated at least 1–2 months, during the period of depletion from blood (11). Although declines in these BAFF levels predated the return of blood B cell levels, there was no direct temporal relationship to the time of recurrence of disease activity (11).

In the murine model of autoimmune inflammatory polyarthritis, CIA, elevated serum levels of BAFF correlated with increased levels of anti-collagen antibodies during disease progression (67). Notably, dendritic cells in affected joints appeared to be the major source of BAFF during the acute phase, while macrophages were the major sources during the chronic phase. The blockade of these B cell survival factors can greatly ameliorate disease activity. Treatment with the soluble decoy receptor, TACI-Ig, which can inhibit both APRIL and BAFF, has been shown to prevent the progression of CIA and lower disease severity scores (68, 69). These treatments were also efficacious when given after the onset of inflammation and the generation of anti-collagen antibodies. Similar benefits were seen after treatment with BCMA-Fc (67).

Documenting a potential role in RA pathogenesis, a recent analysis of RA joint biopsy specimens (70) demonstrated TACI-bearing T cells in RA synovial samples with prominent local B cell accumulations and in RA synovial samples with disorganized groupings of infiltrating B and T cells. However, TACI-bearing T cells were not found in samples with infiltrates that appeared to be organized, with the features of germinal centers. Significantly, TACI-Ig treatment of human synovium–SCID mouse chimeras that were generated by implanting pieces of synovial tissue with the B-cell follicle pattern or the diffuse infiltrate histology type led to increased levels of the key proinflammatory cytokine, IFNγ. In contrast, treatment with TACI-Ig of chimeras generated by implanting synovial samples containing germinal center–like structures resulted in the destruction of these organized structures and greatly diminished levels of IFNγ and immunoglobulin transcripts (70). These findings were interpreted as evidence that the targeting of APRIL and BAFF could have either proinflammatory and/or antiinflammatory activities in patients with different histopathologically defined forms of RA, and that these factors can regulate T cell functions as well as B cell functions.

In a recent preliminary study, Goodyear et al (71) showed the potential relevance of these pathways to the efficacy of anti-CD20 therapy. While in vitro incubation of activated rheumatoid synoviocytes with TNF and IFNγ significantly increased BAFF production, these treatments also greatly diminished the efficiency of human B cell depletion by rituximab. However, when TACI-Ig was added to B cells co-cultured with activated rheumatoid synoviocytes, in studies designed to blockade BAFF and APRIL, the efficiency of human B cell depletion induced by rituximab treatment was greatly enhanced (71).

Clinical studies on BAFF blockade have so far progressed to completed phase 2 trials with the human recombinant anti-BAFF antibody, belimumab (LymphoStat-B) (72). Although these studies documented a desirable safety profile, the clinical efficacy of belimumab in patients with RA was much less than that derived from the current benchmark of biologic therapy, TNF inhibitors. In addition, no dose-response relationship was documented. A phase II trial in patients with SLE also demonstrated only limited efficacy (72). However, this outcome may be predictable, likely because belimumab does not recognize and block BAFF displayed on cell surfaces. Belimumab also cannot block APRIL or the soluble heterotrimers that can form from BAFF and APRIL. Therefore, based on the known redundant functional activities of APRIL and BAFF, which are often produced in the same cells, the blockade of BAFF alone may be predicted not to provide optimal benefits. In any case, due to the overall importance of these factors in B cell survival and the potential benefits that may derive from targeting B cells in RA synovium, it is certain that the topic of BAFF/APRIL blockade merits further investigation.


Recent trials of rituximab have shown that patients with RA derive clinical benefit from B cell depletion therapy, and we are now beginning to consider how current treatments impact the immune system in these patients. Akin to the above-described evidence of residual circulating B-lineage cells after rituximab treatment, recent reports in murine models demonstrated that anti-CD20 antibody treatment also did not completely delete all CD20-bearing murine B-lineage cells, even when blood B cell levels were efficiently depleted (51–53) (for review, see refs. 50 and73). Although currently very little data are available on the impact on B cells within human tissues, analyses in mice have confirmed that lymphoid tissue can provide sanctuaries that protect certain B-lineage cells from anti-CD20–mediated depletion. It is also relevant to appreciate that the human immune system is distinct in many ways from the murine immune system, and that the marginal zone of the human spleen also contains memory B cells. Integrin-mediated retention signals at this site may provide both tissue localization and pro-survival signals that enable some B cells to resist anti-CD20 depletion. In light of what we know about the expression of B cell survival factors and how inflammation can increase their expression, it is therefore predictable that the efficacy of current rituximab regimens may be even more limited in the disease-affected tissues of some patients.

From a practical perspective, there is a need to better understand the roles of memory B cells in the pathogenesis of RA, as well as how B cell–targeted treatments affect immunologic memory. Although current data indicate that levels of circulating CD27+ memory B cells may be diminished, no patient has yet had all immune memory within the B cell compartment “erased” by rituximab treatment, even after receiving multiple cycles of therapy. The evidence therefore indicates that the currently used anti-CD20 treatments do not return the patient to an earlier state of immunity akin to the “tabula rasa” (i.e., blank slate), in which all past cellular remnants of memory and previous (auto) immune responses have been excised.

Based on recent clinical successes, we can now begin to ponder whether it is feasible (or even desirable) to develop B cell–targeted therapeutic agents/regimens that delete all clonal remnants of pathologic autoreactive lymphocytes in the hope of reestablishing immune tolerance. At first glance, a therapeutic reversion of the immune system to an earlier pristine state holds appeal from a philosophical and theoretical perspective, as we may seek to completely erase the “bad” B cell memory, which retains the pathogenic imprinting of autoantigens that is at the core of immunopathogenesis. However, we must accept that there are at least 2 distinct faces to B cell immunologic memory in patients with RA. Such therapeutic interventions could carry an unacceptable level of risk if we also erase protective immune memory against infectious agents and perhaps adversely affect the capacity to control cancer (i.e., tumor immune surveillance).

To better understand the impact of B cell deletion therapy on immunologic memory, we will need means to monitor these 2 facets of immunity. Hence we will need to develop quantitative assays that efficiently discriminate the lymphocyte clonal sets associated with the very different functional roles of immune defense from infection versus disease-associated autoreactivity. Such studies may also be complicated by potential problems with currently accepted markers for cells involved in immune memory. For instance, while studies of the human immune system have been greatly accelerated by the proven utility of CD27 to identify post–germinal center B cells that include memory B cells (74), recent reports suggest that some sets of memory B cells may not bear CD27, and CD27 may be less dependable in patients with autoimmune disease (75, 76). It also remains possible that under the nonphysiologic pressures that follow anti-CD20 treatments, memory B cells may undergo phenotypic shifts to lose CD27 expression. Therefore, future studies will also need to identify, validate, and apply additional phenotypic markers, like the recently reported ATP-binding cassette B1 transporter (77), to discriminate naive from antigen-experienced B cells and track bona fide memory cells even when they do not bear CD27.

Recent studies have shown that current rituximab regimens are associated with little increased susceptibility to infection in patients with RA, similar to what has been observed in patients with NHL. It is therefore relevant that the persistence of IgG responses to foreign antigens, which lasts for years (or decades) after the last known exposure, is believed to derive from a reservoir of long-lived plasma cells in the bone marrow. The low incidence of infection in trials of RA in which the current rituximab regimens were used has therefore been attributed to the limited effect of rituximab on plasma cells, which do not express CD20 and hence are not directly affected by these treatments. However, this may be an oversimplified and inaccurate model, because memory B cells and plasma cells are likely closely linked sets. Memory B cells have recently been reported to have a high rate of turnover (74, 77), and their homeostatic proliferation is required to maintain a continuous production of plasma blasts that then sustain serum antibody levels (78). Therefore, the development of regimens designed to totally “erase” B cell immune memory and/or plasma cells may be undesirable, because a thorough depletion of all memory B cells will certainly greatly increase the risk of infection, and this susceptibility could be quite prolonged.

Biologic and clinical responses vary greatly among patients with RA; in some patients, rituximab induces blood B cell depletion and clinical responses that last for years, while in other patients B cell depletion and clinical benefits last for only a few months. Current treatment regimens involve a clustered set of 2–4 infusions of this anti-CD20 antibody, which has an estimated in vivo half-life of ∼20 days; the pharmacokinetics of rituximab is not affected by co-treatment with MTX or cyclophosphamide (6). These anti-CD20 treatments are not believed to directly affect stem cells or early B cell progenitors. One might therefore predict that when in vivo posttreatment rituximab levels decrease below an effective level for biologic activity, new immature B cells should rapidly develop and continue their differentiation to emerge in the periphery in less than a week. In most patients, high circulating levels of this anti-CD20 antibody should last for only 4–6 months after the last infusion. Therefore, the central question of how blood B cell depletion can persist in some patients for 1 year or longer remains unanswered.

We speculate that the wide range of biologic and clinical responses to rituximab must result from mechanisms in addition to the simple pharmacokinetics of rituximab clearance from the circulation. Prolonged B cell depletion could result from the establishment of a tissue depot for rituximab, perhaps in bone marrow or in the iccosomes of FDCs in peripheral lymphoid tissue (79), where deposited rituximab may be encountered by B cells even though it is undetectable by serum assays. Alternatively, it is possible that, in some patients, these treatments may induce a negative feedback loop that adversely affects the nurse-like cells responsible for the survival niches for B cells and their progenitors in the bone marrow.

We now know that biologic and clinical responses are often greatly prolonged by weekly treatment with MTX (or perhaps other agents). We therefore postulate that this co-treatment agent may diminish the capacity to produce proinflammatory cells, such as neutrophils (61), macrophages, or dendritic cells (80), which can become dominant sources of B cell survival factors. Hence, some second-treatment agents likely provide benefits by reducing the availability in the bone marrow (and at other sites) of BAFF and APRIL and/or other survival factors for B cell precursors, such as IL-7. Optimizing these influences could further improve the response profile of B cell–targeted treatments.

In conclusion, although data on the biologic impact of rituximab infusions in patients with RA are still limited, we have already learned important lessons. These insights have identified fertile areas for future investigation that are relevant to better understanding the mechanism(s) of action of rituximab, as well as how dysregulation within the B cell compartment contributes to pathogenesis in general. Although several areas hold great promise, we need to go beyond blood surveys, to sample peripheral lymphoid tissue and quantitate residual antigen-specific memory B cells. Although a preliminary report has suggested that the regimen of rituximab plus MTX yielded significant improvement in serologic markers of both inflammation and bone turnover (81), expanded analyses of treatment-associated changes in a range of proteomic, transcript, and cell-based factors are needed. In particular, if validated in later studies, the recent evidence of response-dependent modulation of T cell activation in an open-label rituximab trial in lupus nephritis (82) may suggest an attractive surrogate end point biomarker for monitoring biologic and clinical responsiveness.

Advances in these topics should lead to improvements in the effectiveness of treatment regimens and the durability of treatment-induced clinical responses. The consequences of these efforts should also yield more favorable comparative medical economics of B cell depletion treatments for RA and many other diseases.


I appreciate the assistance of Andrea Cerutti (Weill Medical College of Cornell University) in refining the figures.