Glomerular CD68+ macrophages infiltration at initial biopsy predicts response to standard immunosuppression in proliferative lupus nephritis

Abstract

Objective

Predictive models of kidney response to standard immunosuppression are needed in proliferative lupus nephritis (LN). We tested the kidney macrophage infiltration at initial biopsy.

Methods

The prospective study was performed in 247 patients with newly diagnosed proliferative LN in 2 independent cohorts. Infiltrates of macrophages and lymphocytes in initial biopsies were identified using single-cell RNA sequencing and immunostaining analysis. The outcome was kidney response to standard immunosuppression at 1 year, defined clinically and histologically. Kidney infiltrates were investigated for association with kidney response. Models that combined kidney infiltrates and clinical parameters for predicting kidney response were developed and validated using machine learning algorithms.

Results

In Derivation cohort, glomerular infiltration of CD68⁺ macrophages at initial biopsy was associated with 1-year clinical response. Subjects in the highest tertile of glomerular CD68⁺ macrophage infiltrate (versus the lowest) had a 7.92-fold increase in probability of clinical response. An intelligent model incorporating infiltration score of glomerular CD68⁺ macrophage into clinical measures (area under the curve [AUC] 0.82) outperformed traditional clinical measure-based model (AUC 0.76) in predicting clinical response (P = 0.01). This intelligent model performed well in an independent Validation cohort. Furthermore, in 10 patients undergoing repeat kidney biopsy after 1 year of standard immunosuppression, our intelligent model effectively predicted histological response.

Conclusion

Intensity of glomerular CD68⁺ macrophage infiltration at initial biopsy predicted 1-year kidney response to standard therapy in proliferative LN. The intelligent model, which combines glomerular CD68⁺ macrophage infiltrates with clinical data at biopsy, could help discriminate responders from non-responders, enabling personalized therapy.