Good and bad memories following rituximab therapy

Treatment with rituximab, a chimeric anti-CD20 monoclonal antibody, produces rapid depletion of >95% of CD20+ B cells from the circulation. Such a crushing blow to the B cell compartment rightfully creates a certain amount of anxiety about its potential impact on immune function. This shockwave is gradually mitigated as the peripheral blood B cell compartment begins to recover ∼6–9 months later with the emergence of a predominantly immature naive B cell population (1). Our understanding about the kinetics of depletion and reconstitution of B cells following rituximab therapy has come almost exclusively from detailed phenotypic analysis of circulating B cell subsets. Since the circulation contains only 2% of total B cells, this limited view fails to take into account the corresponding changes in the bone marrow, spleen, and lymph nodes where most of the B cells reside and generate immune responses. Therefore, valuable insights into the effects of rituximab therapy may be gained by complementing analysis of circulating B cell subsets with assessments of immune function.

It is for this reason that 2 articles in this issue of Arthritis & Rheumatism will be of interest to aficionados and practitioners of B cell–targeted therapy. Together, they highlight in patients with rheumatoid arthritis (RA) the effects of rituximab-mediated B cell depletion therapy on humoral and cellular immune responses. The first of these articles, by Bingham et al (2), describes a randomized, controlled study examining the effects of rituximab therapy on the antibody responses to 3 different immunogens, including tetanus toxoid (TT), a 23-valent pneumococcal polysaccharide vaccine (Pneumovax), and keyhole limpet hemocyanin (KLH), as well as its effects on the delayed-type hypersensitivity (DTH) reaction to Candida albicans. In this study, vaccine responses were evaluated between weeks 24 and 36 after rituximab therapy, coinciding with the initial recovery of the B cell compartment.

The second of these articles, by van Assen et al (3), compares serologic responses to influenza vaccine (Influvac 2007–2008; Solvay Pharmaceuticals, Weesp, The Netherlands) in rituximab-treated patients with RA, RA patients taking methotrexate (MTX) alone, and healthy controls. The rituximab-treated group was separated into early (4–8 weeks, n = 11) and late (6–10 months, n = 12) subgroups according to the timing of influenza vaccination following drug administration. Taken together, these studies showed that rituximab-mediated CD20+ B cell depletion produced the following effects on immune function: 1) no significant change in the rate of IgG TT seroconversion, 2) reduction in 23-valent pneumococcal polysaccharide vaccine–induced IgG antibody responses, 3) decrease in primary IgG anti-KLH response, 4) no appreciable change in DTH reactions to C albicans, and 5) impairment of IgG response to influenza vaccination. Do these results tell us anything about the effects of rituximab therapy on immunologic memory?

Immunologic memory is a fundamental outcome of successful vaccination. Generally, conventional vaccines are designed to expand antigen- (or pathogen)-specific CD4+ T helper and CD8+ cytotoxic cells and to induce antigen-specific B cell maturation and antibody production. The immune system recognizes and responds to antigens in the same way if it encounters a microbial pathogen or a self antigen. Two of the perturbing antigens in the study by Bingham et al (2), TT and KLH, are termed T cell–dependent antigens because they require T cell help to generate an antibody response. B cells encountering antigen move to the T cell zones of secondary lymphoid tissues where they interact with primed T cells. Next, they form germinal centers within follicles or expand in the extrafollicular regions and differentiate into short-lived plasmablasts. Within germinal centers, follicular B cells proliferate, mutate their antigen receptors, class switch, and differentiate into memory B cells and long-lived plasma cells.

B cell memory comes in two forms: 1) long-lived antibody-secreting plasma cells that do not express CD20, and 2) long-lived memory CD20+ B cells that respond quickly to secondary antigenic challenge (4). T helper cells are crucial for the generation of memory B cells and long-lived plasma cells by providing signals through interactions between CD40 ligand and CD40 as well as other costimulatory pathways. Memory B cells express isotype-switched high-affinity surface Ig, are long-lived, and produce large amounts of antibodies upon restimulation with antigen. Plasma cells also may be long-lived and secrete high-affinity antibodies for extended periods of time. Long-lived plasma cells reside in the bone marrow in a specialized “survival niche” where they are stimulated by prosurvival factors such as interleukin-5 (IL-5), IL-6, ligands for CD44, tumor necrosis factor, and the ligands for BCMA (4).

Several studies have shown that circulating antibody is maintained following treatment with rituximab. Levels of serum IgG and IgA are relatively unaffected by CD20+ B cell depletion, as shown by Bingham et al (2) as well as others (5, 6). However, circulating IgM and IgG levels may fall below the lower limit of normal in a minority of patients with subsequent rituximab courses (5). A single course of rituximab therapy leads to modest drops in serum IgM levels, suggesting that the pool of IgM antibody–secreting plasma cells may depend to some extent on a steady inflow of mature B cells. Serum titers of antibodies to TT and pneumococcal polysaccharides, as well as levels of natural antibodies, are maintained despite rituximab therapy (7, 8). In humans, therefore, circulating antimicrobial antibodies as well as the bulk of other Ig appear to arise from long-lived plasma cells which do not depend on memory B cells to replenish their pool. This interpretation is consistent with studies in mice where maintenance of the plasma cell pool is independent of memory B cells (9, 10).

Although memory B cells uniformly express CD20 on their cell surface, it seems unlikely that they are all depleted following rituximab therapy. The recall T cell–dependent antibody response to TT was apparently left unscathed despite a period of profound CD20+ B cell depletion (2). This finding may mean that memory B cells escape the depleting effects of rituximab and are poised during the recovery phase to rapidly expand and differentiate into antibody-producing cells (2). Nascent immunity is also possible given that an anti-TT IgG response may have developed during the 4-week time period between vaccination and serologic evaluation. What would happen to the anti-TT response if it was evaluated during the time of maximal B cell depletion? We know from studies in mice that anti-CD20 immunotherapy severely impairs secondary IgM and IgG responses during the time when circulating, spleen, and lymph node B cells are reduced by >95% (9). Mouse germinal center B cells express CD20 at high levels and are highly sensitive to depletion by anti-CD20 monoclonal antibodies (9). In humans and nonhuman primates, previous studies have shown that recall antibody responses are dramatically reduced when the immunizations are administered within a few weeks after rituximab therapy (11–14).

Immediately after rituximab therapy, the extant shortage of B cells may hamper T cell help. Cognate T cell help depends on B cells to present antigen on class II major histocompatibility complex (MHC) molecules, after which T cells are stimulated to secrete cytokines and activate B cells. In mice, depletion of CD20+ B cells has been shown to impair CD4+ T cell activation and clonal expansion in response to immunization with protein antigens (15). These results imply that T cell help for antibody responses would be severely compromised when CD20+ B cells are maximally depleted from the secondary lymphoid tissue. Later, with partial restoration of the B cell compartment, T cell help may be sufficient to generate a memory humoral response.

Bingham et al (2) found that the reconstituting peripheral blood B cell compartment following rituximab therapy was not favorable for generating an optimal 23-valent pneumococcal polysaccharide vaccine–induced antibody response, which is classically derived from T cell–independent mechanisms. In their study, a positive response against a pneumococcal serotype was defined as a 2-fold increase or an increase of >1 μg/ml from prevaccination levels. Compared with the group receiving MTX alone, the group receiving rituximab plus MTX had a lower proportion of patients with positive responses across the different serotypes and a lower overall postvaccination level of serum antipneumococcal IgG antibodies (2). It is important to note that normal subjects may not mount a 2-fold increase in antibody titer to all of the serotypes present in the vaccine (16). The possible role of MTX in suppressing this response is also of interest. A previous study found that RA patients taking MTX had lower rates of ≥2-fold IgG antibody responses to pneumococcal vaccination than did healthy controls (17). Thus, MTX therapy and CD20+ B cell depletion likely conspire to inhibit this T cell–independent vaccine response.

T cell antigen-independent antibody responses are divided into types I and II. Type I antigens (e.g., lipopolysaccharide) are polyclonal activators of B cells, while type II antigens, such as pneumococcal polysaccharide, elicit antibody responses by multivalent crosslinking of the B cell antigen receptor in the absence of class II MHC–mediated T cell help. Our understanding about the role of B cell subsets in T cell antigen-independent antibody response comes mainly from studies in mice of B-1 and marginal-zone B lymphocytes. Mouse B-1 cells have been subdivided into B-1a (CD5+) and B-1b (CD5−) cells. Haas et al (18) have shown that B-1 and marginal-zone B lymphocytes team up in mice to combat Streptococcus pneumoniae infection. In that study, mouse B-1a cells were shown to be primarily responsible for the production of natural polyreactive antibodies, the first line of defense against infection and a bridge between innate and adaptive immunity. In contrast, mouse B-1b cells were deemed to be the main players in generating an IgM response to immunization with pneumococcal polysaccharide serotype 3. In those experiments, marginal-zone B cells contributed little to the IgM antibody responses to pneumococcal polysaccharide serotype 3, but they were capable of generating an IgM response against another bacterial antigen.

The caveat is that mouse B-1 lymphocytes have no known human equivalents, and that marginal-zone B lymphocytes in humans are different from those in mice. The human marginal zone is located in the spleen, where it is surrounded by a large perifollicular area (for review, see ref.19). The human marginal zone differs from that of the mouse in its structure (e.g., lack of a marginal sinus) and cellular content (e.g., absence of metallophilic macrophages). The human marginal-zone B cell region does not develop during the first years of life. As a consequence, infants at this age show poor responsiveness to T cell–independent antigens and increased susceptibility to infections with encapsulated bacteria. Unlike marginal-zone B cells in mice, marginal-zone B cells in humans display antigen receptors with somatic mutations, express a memory phenotype, and circulate in the peripheral blood (19). The extent to which marginal-zone B cells contribute to T cell–independent antigen responses in humans remains a subject of intense debate.

Human memory B cells may be subdivided into IgM memory (IgM+IgD+CD27+) and switched memory (IgM−IgD−CD27+) subsets. It has been suggested that circulating IgM+IgD+CD27+ memory B cells represent marginal-zone B cells that are generated independently of germinal center reactions and undergo somatic hypermutation during the development of the preimmune repertoire (19, 20). Alternatively, this subset may develop from a germinal center reaction to a T cell–independent antigen with a lower rate of somatic mutations than a T cell–dependent response (19, 20). IgM+IgD+CD27+ B cells with somatic hypermutations are present during human fetal development and develop in Rag2^−/−γc^−/− mice independent of T cells (21). They are also present in patients with hyper-IgM syndrome who lack the cognate T cell help as a result of CD40 or CD40 ligand deficiency (22).

Remarkably, adoptively transferred human IgM+IgD+CD27+ memory B cells have been shown in a humanized SCID/SCID mouse model to be capable of reconstituting both an IgM and an IgG antipneumococcal polysaccharide response (22). In this model, switched memory B cells were able to reconstitute only the IgG response. The reduction in IgG antipneumococcal responses following rituximab therapy may derive from an abrogation of IgM B cell memory or switched memory. To this end, IgD+CD27+ B cells are severely depleted during the early recovery phase after rituximab therapy (23), suggesting that deficits in IgM memory may be the culprit. Perhaps a simpler explanation is that a paucity of mature B cells may have impaired de novo responses to the 23-valent pneumococcal polysaccharide vaccine.

Influvac for the season studied by van Assen et al (3) contained purified hemagglutinin and neuramidase from A/Wisconsin/67/2005 (H3N2)–like strain (A/H3N2 strain), A/Solomon Islands/3/2006 (H1N1)–like strain (A/H1N1 strain), and B/Malaysia/2506/2004-like strain. The cell surface antigens hemagglutinin and neuramidase are the immunologically relevant proteins for the induction of protective antibody responses (24). Influenza viruses come in 3 types: A, B, and C (types A and B are the most common causes of human disease). Influenza virus undergoes continuous antigenic change to evade the host immune response. In doubly infected cells, a virus can reassemble its genome with new RNA segments from an avian strain or other circulating subtype to create a new subtype, a process termed antigenic shift. Influenza virus also undergoes antigenic drift by accumulating mutations in its cell surface glycoproteins. Circulating antihemagglutinin antibodies have been detected in the serum for decades after natural influenza virus infection. The vaccine-induced antibody response typically reaches its peak at 4–6 weeks and is maintained for >6 months (25). Since high preimmunization titers are inversely correlated with seroconversion rate, immunologic memory from natural infection or prior vaccination shapes the most current vaccine responses. Therefore, intergroup differences in priming history (e.g., natural infection, vaccine administration) can produce significant heterogeneity in baseline serology and vaccine-induced antibody response.

The rituximab-treated patients clearly did not measure up to the MTX treatment group and the healthy controls in their ability to generate an antibody response to the influenza vaccine (3). Notably, the rituximab treatment group failed to show a significant increase in the geometric mean titer (GMT) of antihemagglutinin antibodies. Postvaccination GMTs for the A/H3N2 and A/H1N1 strains were somewhat higher in the late rituximab subgroup than at baseline, indicating partial recovery of the vaccine response by this stage. Notably, a ≥4-fold rise in postvaccination titers was achieved in only 3 of 12 patients (25%) in the late rituximab subgroup and none in the early rituximab subgroup. While intergroup differences in priming history may have biased the results, the negative impact of rituximab therapy on the antibody response to influenza vaccination appears to be striking in this study, especially 4–8 weeks after drug administration, and is consistent with a previous report (26).

While these studies add to our knowledge about the effects of rituximab therapy on immune function, they fall short in offering clues about the causal mechanisms. Such questions are difficult to tackle in humans for a variety of reasons. However, studies in the future should include analysis of IgM and IgA isotypes as well as IgG. For example, we may find that T cell antigen-independent antibody responses to pneumococcal polysaccharides preferentially induce IgM antibodies, and it is not known if IgM and IgG responses are differentially affected by CD20+ B cell depletion. More detailed phenotypic characterization of peripheral blood B cell subsets may also be informative for correlation of the magnitude of IgM and IgG antigen-specific responses with the presence or absence of circulating B cell subsets. A challenge with any human research is the heterogeneity of a study population and the consequent variability in immune responses. It is likely that the study subjects in both studies varied considerably in their priming history. Quantitative assessments of antigen-specific T and B cells using tetramer technology may be needed at baseline and followup to distinguish between primary and secondary responses to a vaccine challenge. Regardless, further studies like those described herein (2, 3) are needed to dissect the complex effects of rituximab therapy on pathologic and protective immune responses to develop strategies that balance effective B cell depletion with the risk of infections (27).

AUTHOR CONTRIBUTIONS

Dr. St.Clair was involved in drafting the article or revising it critically for important intellectual content, and Dr. St.Clair approved the final version to be published.

Acknowledgements

The author would like to thank David DiLillo and Thomas F. Tedder for their critical review of the manuscript and helpful comments.