Type I interferon gene expression signature as a marker to predict response to cyclophosphamide based treatment in proliferative lupus nephritis.

Objectives: To assess the longitudinal effect of cyclophosphamide (CYC) treatment on type-I interferon (IFN) signature in proliferative lupus nephritis (LN) and its role in predicting treatment response.

Methods: Fifty-four biopsy proven proliferative LN patients scheduled to receive high-dose (HD) or low-dose (LD) CYC were recruited and followed up for six months. At six months, patients were classified as clinical responders (CR) or non-responders (NR) to treatment, using the EULAR/EDTA criteria. An IFN-gene based score (IGS) was developed from the mean log-transformed gene expression of MX1, OAS1, IFIT1, OASL, IFIT4, LY6E, IRF7 at baseline, three and six months. Longitudinal changes of IGS within and between groups were assessed and ΔIGS, which is the difference in IGS between baseline and three months was calculated. Independent predictors of non-response were identified and an ROC analysis was performed to evaluate their utility to predict NR.

Results: There was a dynamic change in IGS within the HD, LD, CR, and NR groups. Compared to baseline, there was a significant decrease in IGS at three months in HD and LD groups (HD group: 2.01 to 1.14, p = .001; LD group = 2.01 to 0.81, p < .001), followed by a significant increase from three to six months in LD group (LD: 0.81 to 1.51, p = .03; HD: 1.14 to 1.54, p = .300). A decrease in IGS from baseline to three months was seen in both CR (2.13 to 0.79, p < .001) and NR groups (1.83 to 1.27, p = .046), and a significant increase from three to six months was observed only in the CR group (CR: 0.79 to 1.57, p = .006; NR: 1.27 to 1.46, p = 1). ΔIGS (baseline to three months) was higher in CR compared to NR group (-1.339 vs -0.563, p = .017). ROC analysis showed that the model comprising of 0.81 fold decrease in IGS from baseline to three months, endocapillary hypercellularity and interstitial inflammation on renal histopathology predicted non-response with a sensitivity of 83.3% and specificity of 71.4%.

Conclusion: In proliferative LN, treated with HD or LD-CYC, combined model comprising of decrease in IGS score by 0.81 fold from baseline to three months, along with important histopathological features such as endocapillary hypercellularity and interstitial inflammation had better predictive capability for non-response.