Serum sPD-1 and sTIM-3 as T-cell exhaustion biomarkers in systemic lupus erythematosus: The IL-6-associated exhaustion axis and the SLE Immune Exhaustion Index.

BACKGROUND

T-cell exhaustion marked by soluble PD-1 (sPD-1) and soluble TIM-3 (sTIM-3) has been implicated in systemic lupus erythematosus (SLE), yet the cytokine mechanisms driving checkpoint-receptor upregulation, the relationship with complement consumption, and the clinical utility of these markers remain incompletely characterised.

METHODS

In a case-control with cross-sectional study of 120 participants (40 active SLE, 40 quiescent SLE, 40 healthy controls), serum sPD-1, sTIM-3, IL-6, ferritin, complement proteins (C3, C4, C1q), anti-dsDNA, CRP, ESR, and urinary protein-to-creatinine ratio (UPCR) were measured. A 12-parameter Spearman correlation matrix, principal component analysis (PCA) of 10 standardised biomarkers, ROC curve analysis with bootstrap 95% CIs and DeLong pairwise comparisons, lupus-nephritis (LN) subgroup analysis, and four pre-specified novel analyses were performed: (i) formal bootstrap mediation analysis of the IL-6 → sPD-1/sTIM-3 → SLEDAI-2K pathway; (ii) the sTIM-3/sPD-1 imbalance ratio; (iii) an Exhaustion-to-Complement Index (ECI); and (iv) multivariable logistic regression for LN prediction. The SLE Immune Exhaustion Index (SIEI) was evaluated by ROC analysis.

RESULTS

Four primary biomarkers were significantly elevated in active versus quiescent SLE and healthy controls (Kruskal-Wallis H 87.2-99.0; all p < 0.001), with strong positive correlations with SLEDAI-2K (rho = 0.647-0.747). PCA identified a single dominant axis (PC1 = 78.9% variance) driven by exhaustion and inflammatory markers opposing complement proteins. Individual ROC AUCs were: sPD-1 0.932, sTIM-3 0.966, IL-6 0.944, and ferritin 0.953; the SIEI achieved 0.974, significantly exceeding sPD-1 alone (p = 0.002) but not the other individual markers. Bootstrap mediation analysis confirmed that sPD-1 and sTIM-3 each mediated approximately one-third of the IL-6 effect on SLEDAI-2K (indirect effects 0.276 and 0.303; bootstrap 95% CI excluding zero; Sobel p < 0.001). The ECI correlated strongly with SLEDAI-2K (rho = 0.86, p < 0.001; within-cohort AUC = 0.985). The UPCR plus sPD-1 plus IL-6 model significantly improved LN prediction over UPCR alone (AUC 0.869 vs 0.699; bootstrap p = 0.020), with sPD-1 carrying the dominant predictive contribution.

CONCLUSIONS

Serum sPD-1 and sTIM-3 are closely associated with the IL-6-related, complement-consuming inflammatory network of SLE, in a pattern consistent with but not proving a mechanistic link. The SIEI is the first named, weighted composite exhaustion score in SLE, with within-cohort discrimination that requires prospective external validation. These findings support T-cell exhaustion as a serologically accessible disease axis with direct implications for disease monitoring and lupus-nephritis risk stratification.