Objective To identify systemic lupus erythematosus (SLE) patient endotypes according to B cell immunophenotyping and serological profile and assess their response to belimumab.

Patients and Methods We analysed data from 796 patients with SLE from the phase III BLISS-SC clinical trial of belimumab. Using an unsupervised machine learning algorithm, patients were stratified into distinct clusters based on B cell immunophenotype and autoantibody profile. Analysis of variance was used to compare characteristics across clusters. Cox regression was used to show the effect of belimumab on the attainment of sustained lupus low disease activity state (LLDAS) or DORIS remission across clusters.

Results Cluster 1 (n=193) was characterised by significantly higher proportions [mean (SD)] of CD19⁺CD24b⁺CD27⁺regulatory B cells [35.8%(12.5%)], CD19⁺CD20⁺CD27⁺bulk memory B cells [32.0% (9.9%)], CD19⁺CD20⁺CD69⁺activated B cells [0.2%(0.3%)], CD19⁺CD20^-CD138⁺ long-lived plasma cells [0.6% (0.9%)], and CD19⁺CD38b⁺CD27b⁺ SLE-associated plasma cells [6.6%(7.0%)]. Cluster 2 (n=358) was predominantly characterised by higher proportions of CD19⁺CD24b⁺CD38b⁺CD27^- transitional B cells [6.5% (9.2%)], and CD19⁺CD20⁺CD27^- naïve B cells [85.5%(7.2%)], and lower proportions of CD19⁺CD20^-CD138⁺ peripheral long-lived plasma cells [0.2%(0.3%)] and CD19⁺CD38b+CD27b+ SLE-associated plasma cells [1.6%(6.1%)]. Cluster 3 was primarily characterised by a higher proportion of CD19⁺CD20⁺CD138⁺ short-lived plasma cells [0.1%(0.1%)] and was the most serologically active with respect to low C3 and C4 levels and anti-dsDNA positivity compared to clusters 1 and 2. Cluster 2 was dominated by musculoskeletal manifestations (85.8%) when compared with cluster 1 (75.1%; p<0.005) and cluster 3 (69.0%, p<0.001) and mucocutaneous manifestations when compared with cluster 3 (90.8% versus 83.7%; p<0.001). Cluster 3 was characterised by the highest baseline SLEDAI-2K scores, a greater proportion of patients with renal involvement, greater use of mycophenolate, and a higher median prednisone dose. Use of belimumab in cluster 2 was associated with a >2-fold increased probability of sustained LLDAS [HR 2.12 IC 95% (1.1–4.0) p <0.005] and >3-fold increased probability of sustained DORIS remission [HR 3.45 IC 95% (1.2–9.9) p<0.005], while no such benefit from belimumab was seen in clusters 1 and 3.

Conclusion Three distinct SLE endotypes were identified through an unbiased approach based on B cell immunophenotyping and autoantibody profiles, showing differential benefit from belimumab therapy.