Measuring B cell reset and its relation to durable response
Immunophenotypes, complementary biomarkers and clinical evidence
A returning population dominated by naïve B cells is evidence of altered reconstitution. Its ability to predict lasting remission remains incompletely established. The most directly predictive evidence in this review comes from longitudinal rituximab studies in systemic lupus erythematosus (SLE). CAR-T studies provide evidence that clinical benefit can persist after B cells return, with increasing molecular detail, but limited validation of prospective relapse predictors. Sources and access limits are in the reference register.
Out of scope. Treatment selection, dose selection and comparative safety recommendations. The specification proposes measurements and analyses for research.
1. Definitions and Interpretation
An immunophenotype describes cells using measurable markers, usually proteins on the cell surface detected by flow cytometry. A gate is the rule used to assign measured cells to a subset. CD denotes cluster of differentiation; plus and minus signs indicate marker expression, while hi and lo indicate relatively high and low expression under the laboratory’s gating procedure.
A B cell receptor (BCR) is the cell’s antigen-binding immunoglobulin. Naïve cells have not yet undergone the usual antigen-driven maturation into memory or antibody-secreting populations. Class switching changes the immunoglobulin class, such as from IgM to IgG. Somatic hypermutation changes the sequence of the antigen-binding region during maturation. Neither class switching nor mutation burden identifies the antigen recognized by a cell.
The term reset can refer to a changed cell composition, a changed antigen repertoire, or durable control of disease. The Junt perspective explicitly addresses clinical, cellular and molecular definitions. Those levels should be reported separately. In particular, calling the returning cells naïve does not demonstrate that they cannot recognize self-antigens.
A Measurement Hierarchy
| Measurement | Supported conclusion | Additional evidence needed |
|---|---|---|
| Blood B cells below detection | Blood depletion at this assay’s sensitivity | Tissue depletion and identity of surviving cells |
| Return with altered subset composition | Phenotypic reconstitution | Persistence and antigen specificity |
| Changed BCR sequences or reduced baseline-clone overlap | Repertoire change within the sampled population | Whether lost or new clones recognize disease antigens |
| Reduced self-reactivity in a functional assay | Altered function under that assay | Persistence in vivo and prospective clinical association |
| Remission after return without maintenance therapy | Observed clinical durability | A comparison establishing prediction or mechanism |
A normal total count can conceal an abnormal subset distribution. Conversely, a high naïve fraction can arise because memory cells remain scarce while the absolute naïve count is still low. Both counts and fractions are needed.
2. Immunophenotypes
The table is a proposed harmonized measurement panel. Gates must be qualified locally and reconciled with each paper’s definitions before comparing results. Begin with viable singlet leukocytes, exclude unwanted lineages and identify B-lineage cells using markers that remain interpretable after the treatment. Do not require CD20 expression to identify every antibody-secreting cell. Therapeutic binding or altered target expression can also complicate CD19 or CD20 detection; record antibody clones and the detection strategy.
| Population | Candidate gate or assay | Interpretation during return |
|---|---|---|
| Total B cells | CD19-based enumeration with an independently qualified alternative when needed | Establishes abundance and the denominator for subset fractions |
| Transitional B cells | CD19+ CD24hi CD38hi, usually IgD+ CD27−; distinguish from plasmablasts | Immature returning cells; abundance alone does not establish regulatory function |
| Mature naïve B cells | IgD+ CD27− after separating transitional cells | Broad measure of reconstitution; add activation markers if disease biology requires |
| Unswitched memory | IgD+ CD27+, with IgM when available | Memory retained without the usual switched-isotype phenotype |
| Switched memory | IgD− CD27+, excluding plasmablasts | Recovery of an antigen-experienced compartment; surface IgG/IgA can refine classification |
| Double-negative cells | IgD− CD27− | A heterogeneous compartment requiring further subdivision |
| DN2-like cells | Within double-negative cells: CXCR5− CD11c+, often CD21lo/−; T-bet can be measured intracellularly | An activated, disease-associated candidate subset in lupus |
| Plasmablasts | CD27hi CD38hi, often CD20− with variable CD19 | Circulating antibody-secreting cells; record absolute counts at adequate assay sensitivity |
| Tissue plasma cells | Tissue-based panel including CD38, CD138 and BCMA, with CD19 status | A separate compartment; absence of blood plasmablasts does not establish its removal |
The IgD/CD27 framework and refinement of double-negative cells are supported by Jenks et al., who linked DN2 cells to an extrafollicular pathway toward antibody-secreting cells in SLE. This also explains why CD27 negativity alone is insufficient to label a population naïve. CXCR5 is a chemokine receptor associated with follicular homing; T-bet is a transcription factor. Their interpretation depends on the full gate. DN2, age-associated B cells and all CD11c+ cells should not be collapsed into one interchangeable category.
Published CAR-T panels sometimes use a broader naïve gate, CD20+ CD27−, and an immature gate different from CD24hi CD38hi. Preserve those original labels when extracting results. The proposed panel above deliberately provides more resolution; it cannot retroactively resolve populations in older data.
Wang et al. found that transitional cells from a subset of patients with lupus produced autoantibodies following Toll-like receptor 7 stimulation. Immaturity therefore cannot be taken as evidence of self-tolerance.
Regulatory B cells require a functional qualification. A transitional gate can identify a population for testing, but a claim of restored immune regulation should specify an assay, such as cytokine production under defined stimulation or suppression of another cell response. This is an assay proposal, not an established predictor from the clinical cohorts below.
3. Complementary Biomarkers
| Measurement | Question it can answer | Interpretation constraint |
|---|---|---|
| Disease-specific autoantibodies | Does a disease-associated humoral response decline or recur? | Dependence on antibody specificity and disease; seronegativity is not a universal remission criterion |
| Complement C3/C4, urine and organ measures in lupus | Does immune activity and organ involvement improve? | Clinical response measures can overlap with the outcome being predicted |
| Total IgG/IgA/IgM and protective antibody titres | Is humoral immune capacity preserved or recovering? | Record immunoglobulin replacement; retained antibodies do not prove ability to make new responses |
| BCR sequencing | Do pretreatment clones persist, and how do diversity, mutation and isotype change? | Match cell numbers and sequencing depth; an undetected clone may be unsampled |
| Antigen-specific B cell assays or recombinant antibodies from single cells | Are returning cells autoreactive to the relevant antigens? | Greater mechanistic specificity, but assay availability and antigen selection limit coverage |
| B cell activating factor (BAFF), interferon-related transcripts and T cell phenotypes | Does the environment that supports B cell activation change? | Exploratory context; specify assay and avoid calling an association causal |
| Drug concentration, CAR abundance and tissue measurements | Could continuing treatment activity or residual tissue disease explain the result? | Blood measurements may not reflect all relevant compartments |
The BCMA-targeted myasthenia study discussed below supports studying a wider immune response than the circulating B cell count. In contrast, a full multi-omics panel in every patient is not necessary to test a prespecified flow-cytometry predictor. Bank samples for later mechanistic work and keep the primary analysis small.
4. Evidence After Anti-CD20 Depletion
The rows summarize published observations; they are not pooled estimates. Reported numbers are literature extractions, not calculations from patient data.
| Study and population | Measurement and timing | Clinical association | Interpretation |
|---|---|---|---|
| Anolik 2007, 15 rituximab-treated patients with SLE; mean follow-up 41 months | Blood and tonsil phenotyping; delayed CD27+ memory recovery | A subset with prolonged responses and autoantibody normalization had delayed memory recovery for years | Longitudinal association; no prospectively validated reset cut-off. Abstract-level extraction |
| Vital 2011, 39 patients with SLE | Highly sensitive flow cytometry; memory and plasmablast return after week 26 | Faster return in patients with earlier relapse; incomplete depletion also associated with nonresponse | Supports separate depletion and reconstitution questions. Observational; abstract-level extraction |
| Lazarus 2012, 61 patients with refractory SLE; phenotyping in a subgroup | B cell counts, IgD/CD27 subsets and anti-double-stranded DNA antibodies | Relapse was delayed beyond return; the count and phenotype at relapse differed with anti-DNA levels | A single total-count threshold loses disease heterogeneity. Phenotype at relapse is not automatically a pre-relapse predictor |
| Md Yusof 2017, prospective rituximab cohort; relapse validation subset of 25 | Plasmablast count at month 6, before subsequent relapse | >0.0008 × 10⁹/L predicted relapse by 12 months with 73% sensitivity, 90% specificity and area under the curve 0.86 | Most explicit predictive rule here; small validation subset within the same research programme; not validated for CAR-T or other diseases |
The last threshold equals 0.8 cells/µL. Sensitivity describes the fraction of earlier relapses identified; specificity describes the fraction of later relapsers correctly classified. The reported confidence intervals were wide (45–92% and 56–99%, respectively). These results justify testing plasmablast counts prospectively, while leaving transportability and clinical utility open. The article’s Plasmablast repopulation as a biomarker of relapse section is the source of these values.
5. Evidence After CAR-T Therapy
CAR-T denotes T cells engineered to express a chimeric antigen receptor. CD19-directed and BCMA-directed interventions target different B-lineage compartments and need separate interpretation.
| Study and population | Reconstitution or biomarker observation | Clinical follow-up | Interpretation |
|---|---|---|---|
| Mackensen 2022, five refractory SLE patients | Returning cells were naïve with non-class-switched receptors; mean return at 110 days | Drug-free remission continued after return; median follow-up 8 months | Demonstrates remission after reconstitution; no relapsing comparison group |
| Müller 2024, 15 patients: SLE, myositis and systemic sclerosis | Longitudinal subset analysis, including early and established reconstitution in six SLE patients | Expanded follow-up series | Supports characterization across time; the detailed phenotyping subset must not be treated as all 15 patients. Check overlap with the earlier SLE report before combining |
| CASTLE, Müller 2026, 24 patients across those three diseases | Paired analysis in 13 evaluable patients showed >90% naïve cells with very little memory/plasmablast representation | 22/24 met disease-specific efficacy endpoints; no relapse reported over median 13-month follow-up | Phenotypic and clinical observations coexist. With no relapse events, a relapse classifier cannot be validated |
| Li 2026, 11 patients with multirefractory autoimmune hemolytic anemia | Naïve predominance in remission; sequential molecular analyses implicated a relapse-associated B/T/plasma-cell niche | All initially achieved complete response; median follow-up 12.2 months | Adds relapse biology. Abstract does not establish a prospective classifier or provide enough detail here to extract relapse counts |
| Fedak 2026, exploratory analyses from a randomized BCMA-directed RNA CAR-T trial in myasthenia gravis | Plasma-cell, autoantibody-repertoire and other immune changes; only subtle broad B cell changes | Placebo comparison and responder analyses | Broader evidence for treatment-associated immune change; no universal naïve-cell reset rule follows |
| COMPARE 2026, six patients with refractory seropositive rheumatoid arthritis | Blood/tissue depletion and declines in autoantibodies | Follow-up 36–52 weeks; three patients reached DAS28-CRP remission and ACR70 response | Improvement and autoantibody change do not imply remission in every patient; no predictive reset threshold established by the abstract |
CASTLE also illustrates denominator discipline: its extended-data phenotype figure includes patients with only one available time point, whereas the paired reconstitution statement concerns a smaller subset. Drug withdrawal, clinical response and remission are distinct outcomes. Its sclerosis endpoint was absence of progression, which should not be relabelled remission.
In COMPARE, DAS28-CRP is a rheumatoid arthritis disease-activity score using 28 joints and C-reactive protein; ACR70 indicates at least 70% improvement under the American College of Rheumatology response criteria. These are clinical measures, not measures of B cell tolerance.
6. What Can Be Concluded About Long-Term Response?
There is evidence of association, with narrower evidence of prediction. The rituximab data motivate measuring returning memory cells and plasmablasts. The CAR-T data motivate testing whether a changed returning population permits clinical control after depletion ends. Neither observation supplies a validated cross-disease surrogate endpoint.
A surrogate endpoint would need to reliably capture treatment effects on clinical outcomes across appropriate studies. A patient-level correlation between a subset and relapse is insufficient. Likewise, treatment-induced change relative to placebo does not by itself show that the change mediates benefit.
Long-term must name a duration. Months of follow-up after treatment may contain much less follow-up after reconstitution. Report both clocks, the number still observed at each visit, and ongoing treatment. Persistent therapeutic activity, conditioning, prior drugs, tissue persistence and fixed organ damage can all complicate interpretation.
The available results favour a trajectory-based description: what returned, when it returned, whether the pattern persisted, and what happened afterward. A single high naïve percentage should remain a measured feature until a prespecified rule demonstrates predictive performance.
7. Reading Order
- Junt 2025 for the conceptual definitions; full-text access remains outstanding in this review.
- CASTLE 2026, especially the B cell results and Extended Data Figure 4, for a current phenotype example.
- Md Yusof 2017, the plasmablast repopulation section, for an actual prospective-time-order prediction question.
- Jenks 2018 for the biological reason to subdivide CD27-negative cells.
- Li 2026 for relapse mechanisms; obtain the full report before extracting a predictive rule.