Rituximab abrogates aquaporin-4–specific germinal center activity in patients with neuromyelitis optica spectrum disorders

Neuromyelitis optica spectrum disorders (NMOSDs) are caused by immunoglobulin G (IgG) autoantibodies directed against the water channel aquaporin-4 (AQP4). In NMOSDs, discrete clinical relapses lead to disability and are robustly prevented by the anti-CD20 therapeutic rituximab; however, its mechanism of action in autoantibody-mediated disorders remains poorly understood. We hypothesized that AQP4-IgG production in germinal centers (GCs) was a core feature of NMOSDs and could be terminated by rituximab. To investigate this directly, deep cervical lymph node (dCLN) aspirates (n = 36) and blood (n = 406) were studied in a total of 63 NMOSD patients. Clinical relapses were associated with AQP4-IgM generation or shifts in AQP4-IgG subclasses (odds ratio = 6.0; range of 3.3 to 10.8; P < 0.0001), features consistent with GC activity. From seven dCLN aspirates of patients not administered rituximab, AQP4-IgGs were detected alongside specific intranodal synthesis of AQP4-IgG. AQP4-reactive B cells were isolated from unmutated naive and mutated memory populations in both blood and dCLNs. After rituximab administration, fewer clinical relapses (annual relapse rate of 0.79 to 0; P < 0.001) were accompanied by marked reductions in both AQP4-IgG (fourfold; P = 0.004) and intranodal B cells (430-fold; P < 0.0001) from 11 dCLNs. Our findings implicate ongoing GC activity as a rituximab-sensitive driver of AQP4 antibody production. They may explain rituximab’s clinical efficacy in several autoantibody-mediated diseases and highlight the potential value of direct GC measurements across autoimmune conditions.

Immunoglobulin G (IgG) autoantibodies directed against the extracellular domain of the water channel aquaporin-4 (AQP4) are directly causative in patients with neuromyelitis optica spectrum disorders (NMOSDs) (1–4). AQP4-IgGs are predominantly of the IgG1 subclass, and their major proposed pathogenic mechanism is via complement-mediated damage to the AQP4-rich astrocyte end feet (5). In NMOSDs, patient disability is accrued through discrete clinical relapses, typically affecting the spinal cord and/or optic nerve (6, 7). However, the immunobiology underlying these attacks is poorly understood, and few serum biomarkers can accurately predict relapses (8).Traditionally, ongoing autoantibody production is considered to occur via two broadly discrete cellular pathways: continual germinal center (GC) activity versus long-lived plasma cells (LLPCs) (9). GCs are specialized microenvironments, typically located within secondary lymphoid organs, where antigen-reactive B cells diversify and mature their immunoglobulin genes via somatic hypermutation, with help from specialized lymphoid-resident T follicular helper (Tfh) cells (10). The process of somatic hypermutation is commonly observed alongside a DNA excision process known as class-switch recombination. Together, somatic hypermutation and class-switch recombination can generate high-affinity IgG responses. Autoantigen reactivity of the B cell receptor (BCR) may either arise de novo following somatic hypermutation in GCs or be originally encoded by antigen-reactive germline BCRs expressed by naive B cells (10, 11). Ongoing GC activity may be responsible for the prolonged presence of autoantibodies, such as AQP4-IgGs (9, 12). In an alternative model, LLPCs that successfully exit active GCs and acquire a bone marrow niche may autonomously persist for decades after an autoimmunizing event. These niched LLPCs are thought to secrete >90% of human serum IgG, including a variety of autoantibodies (13, 14).To date, a series of observations suggest that GC activity may play an important role in AQP4-IgG generation. First, close correlations between serum AQP4-IgG levels and AQP4-IgG secreted in vitro by circulating B cells suggest a limited role for LLPCs in AQP4-IgG generation (12, 15). Second, the detection of circulating AQP4-reactive naive B cells identifies a source of cells that could enter GCs and are reported to share clonal relationships with the hypermutated BCRs of intrathecal AQP4-reactive plasma cells (16, 17). Next, annualized relapse rates (ARR) in NMOSDs are robustly reduced by multiple immunotherapies likely to spare nonproliferative CD20⁻ LLPCs, including the anti-CD20 monoclonal antibody rituximab (RTX) (18–20); however, likely because RTX spares the LLPCs, it does not reduce serum AQP4-IgG levels, an observation that presents a potential clinical–serological paradox in a disease with proven pathogenic autoantibodies (21, 22).We hypothesized that the rapid clinical efficacy of RTX observed in patients with NMOSDs may be explained by its direct disruption of active GC reactions, impacting the most affinity matured, and hence pathogenic, B cells and antibodies. However, contradictory data from both human and mouse studies mean that it remains unclear whether RTX effectively depletes B cells within secondary lymphoid organs (23–25). Further, the putative role of GCs in NMOSDs has not been studied directly. In autoimmune diseases of the central nervous system (CNS), the lymphoid organs that drain meningeal lymphatics represent the most plausible anatomical site of active GCs, the deep cervical lymph nodes (dCLNs) (26).To address these concepts, we studied 63 patients with NMOSDs as a prototypical model of an autoantibody-mediated condition. From patients seen as part of routine clinical practice in two specialist NMO centers, we identified clinical relapses in association with proxy measures of an active GC response: class-switch recombination and de novo AQP4-IgM production. Next, to directly sample the secondary lymphoid organs most likely to generate a GC response to neuronal antigens, we aspirated dCLNs from NMOSD patients. These aspirates contained intranodal AQP4-specific B cells and evidence of local, intranodal AQP4-IgG synthesis, both of which were rapidly and efficiently abrogated by RTX over a timescale consistent with clinical remission. Our data present direct insights into the immunological drivers of NMOSD, highlight the effects of RTX in a model of human autoantibody-mediated illness, and provide a platform for the direct analyses of GCs in human autoimmunity.