Which autoimmune diseases are B-cell or plasma-cell driven?
Evidence review and provisional disease ranking — 22 September 2026
1. Current assessment
B-cell targeting has demonstrated clinical benefit in several autoimmune diseases, but a diagnosis alone rarely identifies the responsible B-lineage population in an individual. Plasma cells are differentiated B-lineage cells; “B-cell driven” and “plasma-cell driven” describe potentially overlapping mechanisms. B cells can also sustain disease by presenting antigen to T cells and producing inflammatory signals. Abeles et al., 2024.
For the three main indications in this project, seropositive rheumatoid arthritis (RA) and active systemic lupus erythematosus (SLE) have stronger clinical intervention evidence than systemic sclerosis (SSc). SLE now has positive anti-CD20 trials in both nephritis and nonrenal disease. None of these findings establishes that all manifestations, or all patients, will respond. REFLEX, REGENCY, ALLEGORY, DESIRES.
Rituximab response supports the involvement of a CD20-accessible immune pathway. Nonresponse has several explanations: inadequate depletion, spared plasma cells, a different inflammatory pathway, damage that cannot reverse, or an endpoint and timing that miss the benefit. Current reviews do not provide a validated universal test assigning patients to a B-cell, plasma-cell or non-B-cell category. Stockfelt et al., 2025, Wincup et al., 2026.
Scope: a focused research synthesis for indication and biomarker planning. It is not an individual treatment recommendation or a systematic review. Sources and access limits are recorded in References.
2. What the targets reach
A plasmablast is a proliferating antibody-secreting cell. A plasma cell is a differentiated antibody-secreting cell; some survive for years in tissue niches. Long survival cannot be inferred from anatomical location alone. Fooksman et al., 2024.
| Target | Main biological reach | Consequence for interpreting response |
|---|---|---|
| CD20 | Most mature naïve and memory B cells; generally absent from terminal plasma cells | Can interrupt B-cell functions and replacement of short-lived antibody-secreting cells while sparing established plasma cells |
| CD19 | A broader B-lineage range, including many plasmablasts and some plasma cells | Can reach populations spared by CD20 targeting, but cannot be assumed to remove CD19-negative plasma cells |
| BCMA: B-cell maturation antigen | Many plasmablasts and plasma cells, including populations beyond CD19/CD20 reach | Tests a more direct antibody-secreting-cell hypothesis; conventional B-cell precursors can remain |
These are biological ranges, not guarantees of in vivo depletion. Human marrow contains a CD19-negative population with long-lived antibody production, but CD19-positive plasma cells can also persist for years. CD19 expression therefore does not divide plasma cells into short-lived and long-lived populations. Expression varies with differentiation and tissue. Halliley et al., 2015, KDIGO conference report, 2026, Lee et al., 2021.
The target does not specify the modality. Rituximab and obinutuzumab both target CD20 but differ in their biological properties. Inebilizumab is a CD19 antibody; a CD19 chimeric antigen receptor T-cell (CAR-T) product adds engineered effector cells and often lymphodepleting chemotherapy. BCMA-directed T-cell engagers and transient messenger-RNA CAR-T products also differ in persistence, dosing and conditioning. An apparent benefit after switching modalities does not isolate the contribution of the new antigen target. Schett et al., 2024, Vu et al., 2026.
A useful working classification asks whether disease depends on continual replacement of antibody-secreting cells, persistent tissue immune activity, or established plasma-cell populations. Yuan and Xu propose a related framework for comparing targets and modalities. These categories organize hypotheses; they have not been validated as patient-assignment rules. Yuan and Xu, 2026.
3. Disease ranking for expected clinical benefit
Rank 1 is the highest tier. Conditions within a tier are tied. The ranking reflects confidence that a specified intervention can improve a specified outcome, using randomized evidence, reproducibility and population definition. It is the project’s interpretation of the cited studies. It does not estimate which disease has the largest response rate or compare unlike trial endpoints.
| Rank | Indication or subgroup | Evidence supporting B-lineage targeting | Main qualification |
|---|---|---|---|
| 1 — established controlled benefit | Pemphigus, especially antibody-defined pemphigus vulgaris | Rituximab plus short-course prednisone improved complete remission off therapy in Ritux 3 | Benefit from interrupting B-cell supply does not require direct killing of every plasma cell. Ritux 3 |
| 1 | Aquaporin-4-antibody-positive neuromyelitis optica spectrum disorder (AQP4-positive NMOSD) | CD19 antibody inebilizumab reduced attacks in N-MOmentum | Preventing new attacks differs from repairing established neurological injury. Do not transfer to seronegative disease. N-MOmentum |
| 1 | Antineutrophil cytoplasmic antibody (ANCA)-associated granulomatosis with polyangiitis or microscopic polyangiitis | Rituximab was noninferior to cyclophosphamide for induction in RAVE | Specify subtype and organ activity; this is not a claim about all vasculitis. RAVE |
| 1 | Active immunoglobulin G4-related disease (IgG4-RD) | Inebilizumab reduced flares in the phase 3 MITIGATE trial | B-cell intervention evidence is stronger than a claim that serum IgG4 alone causes all injury. MITIGATE |
| 1 | Antibody-positive generalized myasthenia gravis (MG) | Positive phase 3 CD19-antibody evidence in MINT | Separate acetylcholine-receptor (AChR) and muscle-specific kinase (MuSK) antibodies. Rituximab and BCMA evidence differ by subtype and trial. MINT |
| 1 | Relapsing multiple sclerosis (MS) | Anti-CD20 therapy reduced relapse activity in two OPERA trials | Does not establish plasma-cell dominance or equivalent benefit for established noninflammatory progression. OPERA |
| 1 | Active seropositive RA | Controlled rituximab benefit; rheumatoid factor and anticitrullinated protein antibodies enrich response | Several other inflammatory pathways can remain active. REFLEX, RA consensus |
| 1 | Active SLE, with renal and nonrenal populations assessed separately | Positive obinutuzumab trials: REGENCY and ALLEGORY | Earlier rituximab trials were negative; anti-CD20 antibodies and trial populations are not interchangeable. REGENCY, ALLEGORY |
| 1 | Primary membranous nephropathy at risk of progression | Rituximab maintained proteinuria remission better than cyclosporine at 24 months in MENTOR | Antibody activity and clinical recovery can occur on different timescales; distinguish primary disease from secondary causes. MENTOR, KDIGO |
| 2 — benefit supported, narrower or less mature evidence | SSc with measurable active skin disease | DESIRES supports a skin-score effect of rituximab | A small trial and a defined endpoint; fibrosis, vasculopathy and organ-specific disease require separate assessment. DESIRES |
| 2 | Active systemic Sjögren’s disease | New phase 3 ianalumab results support B-cell-pathway intervention | Evidence located here is a conference abstract; the target is BAFF receptor, outside CD19/CD20/BCMA. Rituximab symptom trials were negative. NEPTUNUS, TRACTISS |
| 2 | Immune thrombocytopenia (ITP) | Rituximab can produce responses, but long-term treatment-free durability is limited | Antibodies, cellular immunity and platelet production contribute differently across patients. ITP review, ITP follow-up |
| 3 — selected-subgroup promise; comparative effect uncertain | Selected inflammatory myopathies | Observational and subgroup signals, with emerging cell-therapy reports | The RIM trial did not meet its randomized primary comparison. Do not pool all myositis or infer efficacy from its overall improvement proportion. RIM, myositis review |
Low expectation of reversing established damage is a separate judgment from a low expectation of suppressing future immune injury. Persistent dryness after gland loss, contractures, scarred organs or fixed neurological deficits should not be scored as direct evidence against a B-cell mechanism.
Ranking by antibody causality
The clearest antibody-effector examples in this comparison are pemphigus, AQP4-positive NMOSD and antibody-defined MG. ANCA-associated vasculitis also has substantial antibody-mechanism evidence. These diseases are useful starting points for an antibody-secreting-cell hypothesis, although clinical response to upstream B-cell depletion can still be strong. Abeles et al., 2024, Lee et al., 2021, MG review.
SLE and ITP have substantial antibody involvement with heterogeneous additional mechanisms. RA and MS illustrate clinically useful B-cell targeting that need not remove all circulating antibodies. In IgG4-RD, activated B cells interact with T cells and fibrotic tissue; the disease name is not proof that total IgG4 is the pathogenic agent. SSc and Sjögren’s also combine immune activity with tissue processes that can persist after immune intervention. Stockfelt, ITP review, IgG4 review, SSc review, TRACTISS.
No defensible disease-wide ordering of “plasma-cell dependence” follows from this evidence. Autoantibody pathogenicity, cellular source and replaceability of that source are separate questions.
A negative CD20 result despite antibody involvement
IgA nephropathy provides a useful counterexample. A 34-patient randomized trial found no significant improvement in proteinuria or kidney function with rituximab over one year, despite blood B-cell depletion. Galactose-deficient IgA1 and antibodies against it persisted. That supports investigating the source of the relevant antibodies; it does not demonstrate that BCMA treatment will work. Evidence for other B-lineage pathways in IgA nephropathy should be assessed separately from this negative rituximab result. Lafayette et al., 2017, KDIGO, 2026.
4. RA, SLE and SSc in more detail
Rheumatoid arthritis
Rituximab works in a substantial subset of active RA. In REFLEX, patients with inadequate response to tumour necrosis factor inhibitors received rituximab or placebo with methotrexate. At week 24, 51% versus 18% met the American College of Rheumatology 20% improvement criterion. That endpoint is improvement, not remission or proof that the disease has a single driver. REFLEX.
Rheumatoid factor and anticitrullinated protein antibody positivity are useful enrichment features, but neither guarantees response nor establishes a long-lived plasma-cell mechanism. Synovial biology adds information: in R4RA, tocilizumab outperformed rituximab in the subgroup classified as B-cell poor by RNA sequencing. The primary histology-based comparison was not significant. Later STRAP molecular work identified response-associated signatures, including B-cell modules shared across responses to different drugs. These observations support further validation of tissue-based selection, not a universal binary biopsy test. RA consensus, R4RA, STRAP.
Patient-level interpretation: combine serology with objective synovitis, prior treatment effects and evidence of depletion. Persistent pain with little inflammation is a different endpoint problem from persisting swollen joints. The hypothesis that a rituximab nonresponder needs CD19 or BCMA targeting remains unproven without more specific evidence; the reviewed sources do not establish a comparative advantage of either target for that subgroup.
Systemic lupus erythematosus
B-lineage involvement is strong, but the relevant compartment varies. Immune-complex disease, antibody-secreting cells, B-cell antigen presentation and innate immune pathways can coexist. The negative rituximab EXPLORER and LUNAR trials remain part of the evidence; residual cells, immunogenicity and repopulation are among the explanations considered in mechanistic reviews. They are not proof that every negative trial is caused by shallow depletion. Stockfelt et al., Wincup et al..
Newer randomized results change the clinical assessment. In active lupus nephritis, REGENCY found complete renal response at week 76 in 46.4% with obinutuzumab plus standard therapy versus 33.1% with standard therapy alone. ALLEGORY, in active SLE without proliferative or membranous nephritis, reported an SLE Responder Index-4 response at week 52 in 76.7% versus 53.5%. SRI-4 requires improved disease activity without specified worsening in other assessments; it does not mean remission. Neither trial directly compared obinutuzumab with rituximab. REGENCY, ALLEGORY.
Patient-level interpretation: identify active organ disease, damage, anti-double-stranded-DNA antibodies, complement, plasmablasts and prior depletion/repopulation patterns. Serology may enrich or track an immune-active phenotype without selecting the best target. Persistent pathogenic antibody activity despite documented depletion would strengthen a plasma-cell hypothesis, but persistence of an antibody titre alone is insufficient. Small CD19 CAR-T and dual CD19/BCMA studies support further trials; dual targeting cannot reveal which target was necessary. Wincup, Müller, Feng.
Systemic sclerosis
B cells contribute, within a disease that also involves vascular injury and fibrotic tissue responses. SSc should be classified by organ, current activity and damage, not by an assumption that autoantibody positivity means plasma-cell-dominant fibrosis. A recent review describes interactions among B cells, macrophages and fibroblasts, together with other mechanisms maintaining fibrosis. Fibrotic tissue remodelling review.
DESIRES randomized 56 patients and found a week-24 difference of −8.44 points in modified Rodnan skin-score change with rituximab versus placebo (95% confidence interval −11.00 to −5.88). Lower scores reflect less skin thickening. The trial supports an effect on skin disease; it does not establish reversal of every organ manifestation. Its exploratory response predictors require external validation. DESIRES, DESIRES predictor analysis.
Patient-level interpretation: document the organ outcome one expects to change and evidence of ongoing activity. Disease-specific antibodies help define clinical subgroups, but the evidence checked here does not establish an antibody-based rule for choosing CD20, CD19 or BCMA. Cell-therapy series are encouraging; small uncontrolled studies with conditioning cannot supply a patient-specific response probability. Müller.
5. How far CD19 and BCMA evidence goes
The following comparison summarizes the evidence reviewed here. An early-study entry means that comparative benefit is unresolved, not that the target is biologically irrelevant.
| Population | CD20 | CD19 | BCMA |
|---|---|---|---|
| Active RA, especially seropositive | Established rituximab benefit; selection remains imperfect. REFLEX | A mechanistic rationale for broader depletion; no comparative advantage established in the sources reviewed. Yuan and Xu | Early mixed-disease reports do not establish a RA treatment effect. Bucci |
| Active SLE | Positive obinutuzumab trials; earlier rituximab trials negative. REGENCY, ALLEGORY | Encouraging uncontrolled CAR-T results; conditioning complicates attribution. Müller | Early reports, including dual CD19/BCMA; the necessary target is unresolved. Feng |
| SSc | Small controlled skin-score trial. DESIRES | Uncontrolled CAR-T series; organ-specific effects need confirmation. Müller | Patient selection and comparative efficacy remain unresolved. Yuan and Xu |
| Generalized MG | Rituximab evidence varies with antibody subtype, disease duration and trial design. Cochrane review | Positive phase 3 in AChR- or MuSK-positive disease. MINT | Small randomized signal with analysis-set limitations; MuSK-positive disease excluded. Descartes-08 |
CD19 antibody therapy has randomized evidence in IgG4-RD, antibody-positive MG and predominantly AQP4-positive NMOSD. In MITIGATE, treated flares occurred in 7/68 participants receiving inebilizumab and 40/67 receiving placebo during 52 weeks, with identical glucocorticoid tapers. In MINT, the week-26 adjusted between-group difference in the MG Activities of Daily Living score was −1.9 points (95% confidence interval −2.9 to −1.0), favouring inebilizumab. These are disease-specific results, not evidence that CD19 always exceeds CD20. MITIGATE, MINT, N-MOmentum.
BCMA has early randomized support in MG, with important design limits. Descartes-08 produced the reported primary month-3 response in 10/15 versus 3/11 patients in the reported primary efficacy population. The full report describes 36 randomized patients but an academic-site efficacy subset of 26, and a primary-endpoint change before unblinding. MuSK-positive patients were excluded. This is a signal for confirmation, not proof of superiority over CD19/CD20 or of a universal plasma-cell classifier. Vu et al., 2026.
For refractory rheumatic diseases, a ten-patient teclistamab report provides an uncontrolled BCMA-directed treatment signal. The separate dual CD19/BCMA lupus study cannot isolate the BCMA contribution. These results justify studying antibody-secreting cells but do not establish a routine rescue sequence after rituximab failure. Bucci et al., 2025, Feng et al., 2025.
Benefits must be weighed against loss of protective immunity and modality-specific toxicity. Removing more B-lineage populations can also remove protective antibody production. Infection, reduced immunoglobulins, infusion reactions, and the conditioning and inflammatory toxicities of some cellular approaches belong in the comparison. Small series cannot establish comparative long-term safety. Junt, Müller, REGENCY.
6. What can identify a likely responder?
| Observation | What it can contribute | What it cannot establish |
|---|---|---|
| A disease-defining functional antibody, such as AQP4, MuSK or AChR | A more specific causal and clinical subgroup | The lifespan or target expression of the cells producing it |
| RA rheumatoid factor or anticitrullinated protein antibodies | Enrichment for rituximab benefit | Certainty of response or a plasma-cell indication |
| Autoantibody titre and complement over time in SLE | Disease-state and monitoring information in context | A universal treatment-selection threshold |
| Peripheral B-cell depletion | Evidence that treatment affected the circulating compartment | Depletion in every involved tissue |
| Synovial B-cell and plasma-cell profiles in RA | Candidate predictors of comparative response | A fully validated routine classifier across diseases |
| Persistent disease-associated antibody after depletion | A reason to investigate surviving antibody-secreting cells | Proof that the antibody explains current symptoms |
| Objective inflammation versus established damage | Whether the measured outcome plausibly changes with immune treatment | Which B-lineage target is best |
| Previous response, relapse and retreatment response | Within-patient evidence consistent with pathway dependence | An unconfounded experiment or guaranteed benefit from another modality |
The table combines clinical evidence with proposed interpretation; it is not a validated decision algorithm. The strongest examples here are antibody-defined MG/NMOSD populations and RA serologic enrichment. Tissue and molecular classifiers remain a research priority. MINT, N-MOmentum, RA consensus, R4RA, STRAP, Stockfelt.
A baseline association with response may be prognostic: patients with more reversible disease may improve on several treatments. A predictive marker must help distinguish benefit from the treatment under consideration relative to a comparator. A post-treatment depletion measurement is pharmacodynamic and cannot be used as though it were a pretreatment selection test.
7. Interpreting rituximab nonresponse
Use the following questions to organize evidence before assigning a mechanism. They are a research framework, not instructions to escalate treatment.
- Was the endpoint active disease? Separate inflammation from damage, unrelated symptoms and a measurement that cannot improve over the interval.
- Was treatment delivered and active? Examine exposure, treatment course, immunogenicity where measured and the depth and duration of depletion.
- Were the relevant tissues represented? Negative blood findings do not exclude tissue B cells. Use the biopsy framework.
- What remained? Document antibody-secreting cells, their target expression, disease-relevant antibody activity and alternative inflammatory pathways.
- Was there enough time, and what happened on recovery? Persistent disease, delayed improvement and relapse after reconstitution are different outcomes.
- What does the next intervention test? A deeper CD20 intervention, broader CD19 intervention and plasma-cell-directed BCMA intervention test different hypotheses; CAR-T conditioning can affect several of them simultaneously.
This framework follows the resistance mechanisms discussed in Stockfelt and Wincup. Repeated objective benefit temporally related to depletion increases the plausibility of a treatment-sensitive pathway. Clinical benefit without a fall in a measured antibody can reflect non-antibody B-cell functions, an incomplete antibody assay, or other treatment effects. Lee.
8. Where to read first
- Abeles et al., 2024: the broad comparison of B-cell functions and clinical interventions.
- Yuan and Xu, 2026: the closest cross-disease comparison of target coverage, disease mechanisms and development strategy.
- Stockfelt et al., 2025: the most directly useful review of why rituximab response and failure are hard to interpret.
- Wincup et al., 2026: recent SLE framing of target, patient selection and timing.
- Fooksman et al., 2024, with the 2026 KDIGO report: plasma-cell longevity and why CD19 status is not a complete measure of persistence.
- The 2026 SSc fibrosis review: how non-B-cell processes complicate the disease-level ranking.
- R4RA and STRAP: primary studies to read when moving from an indication-level argument to a patient-selection claim.