Identification, Confirmation, and Replication of Novel Urinary MicroRNA Biomarkers in Lupus Nephritis and Diabetic Nephropathy

Abstract

Background: The prevalence of chronic kidney disease (CKD) is increasing, leading to significant morbidity and mortality. Kidney biopsy remains the gold standard for diagnosing the underlying etiology of CKD, but the procedure carries complication risks. The aim of this study was to identify novel noninvasive biomarkers correlating with kidney function and histopathology in biopsy-proven CKD patients.

Methods: We profiled 2402 urinary microRNAs (miRNAs) to identify and confirm differentially expressed miRNAs associated with kidney function and histopathology in patients with diabetic nephropathy (n = 58) or lupus nephritis (n = 89), important etiologies of CKD, compared with healthy controls (n = 93 and 119, respectively). Top performing miRNAs were then measured in 2 independent multi-institutional cohorts of patients with diabetes mellitus with (n = 74) or without nephropathy (n = 71) and systemic lupus erythematosus with (n = 86) or without (n = 37) nephritis.

Results: In patients with diabetic nephropathy, miR-2861, miR-1915-3p, and miR-4532 were down-regulated (>10-fold, P < 0.0001) and were associated with estimated glomerular filtration rate (P < 0.01) and interstitial fibrosis/tubular atrophy (P < 0.05). The c-statistics for miR-2861, miR-1915-3p, and miR-4532 were 0.91, 0.86, and 0.85, respectively. In lupus nephritis patients, miR-3201 and miR-1273e were down-regulated (>3-fold, P < 0.0001) and associated with endocapillary glomerular inflammation (P < 0.01), with c-statistics of 0.97 and 0.91, respectively.

Conclusions: We have identified novel miRNAs that correlate with histopathological lesions and functional markers of kidney damage to facilitate sensitive, specific, and noninvasive detection of diabetic nephropathy and lupus nephritis.

Fig. 1. Schematic overview of the strategy for discovery of urinary miRNAs as biomarkers for CKD

Fig. 2. Selection of the best performing urinary miRNAs for DN in the BKBx cohort

miRNAs differentially present in DN (n = 99) compared with healthy controls (HE = 119) (t-test, FDR P < 0.05) (A). Box plot represents the median ±25th and 75th percentiles, and whiskers are determined by the 5th and 95th percentiles (**P < 0.01; ***P < 0.001). Spearman rank correlation analysis between miRNAs and renal function parameters: proteinuria (Prot), SCr and eGFR, as well as Spearman correlation coefficients (r_s) are presented in the boxes (B). Associations between miRNAs and renal histopathological parameters; P values are presented in the boxes (Wilcoxon rank-sum test for 2 independent observations, and Kruskal–Wallis followed by Dunn test for multiple comparisons) (C). Note: for panels B and C, only miRNAs that had significant correlation or association are shown.

Fig. 3. Selection of the best performing urinary miRNAs for LN in the BKBx cohort

miRNAs differentially present in LN (n = 99) compared with healthy controls (HE = 119) (t-test, FDR P < 0.05) (A). Box plot represents the median ±25th and 75th percentiles, and whiskers are determined by the 5th and 95th percentiles (*P < 0.05; **P < 0.01). Spearman rank correlation analysis between miRNAs and renal function parameters: proteinuria (Prot), SCr and eGFR, as well as Spearman correlation coefficients (r_s) are presented in the boxes (B). Associations between miRNAs and renal histopathological parameters; P values are presented in the boxes (Wilcoxon rank-sum test for 2 independent observations, and Kruskal–Wallis followed by Dunn test for multiple comparisons) (C). Note: for panels B and C, only miRNAs that had significant correlation or association are shown.

Fig. 4. Biomarker replication in the UW DN and OSU LN cohorts

miRNAs differentially present in DN (n = 74) compared with the control groups: healthy (HE = 30) and diabetes mellitus (DM = 71) (A). miRNAs differentially present in LN (n = 86) compared with the control groups: healthy (HE = 86) and systemic lupus erythematosus (SLE = 37) (B). Box plot represents the median±25th and 75th percentiles, and whiskers are determined by the 5th and 95th percentiles (*P< 0.05, **P < 0.01, ***P < 0.001; ANOVA, Tukey multiple comparison). Spearman rank correlation analysis between miRNAs and renal function parameters in the DN and LN groups, respectively. Spearman correlation coefficients (r_s) are presented in the boxes (C and D).

Fig. 5. Association of individual miRNAs with DN and LN in the replication cohorts

Forest plot of individual miRNAs and AUC-ROC of replicating miRNAs associated with outcome of DN vs HE in the UW cohort with unadjusted ORs and 95% CIs (A). Forest plot of individual miRNAs and AUC-ROC of replicating miRNAs associated with outcome of DN vs DM in the UW cohort with unadjusted ORs and 95% CIs. miR-6747-3p and miR-4536-3p replicated more weakly in this cohort compared withDNvs HE in the UW cohort (B). Forest plot of individual miRNAs and AUC-ROC of replicating miRNAs associated with outcome of LN vs HE in the OSU cohort with unadjusted ORs and 95% CIs (C). Forest plot of individual miRNAs and AUC-ROC of replicating miRNAs associated with outcome of LN vs SLE in the OSU cohort with unadjusted ORs and 95% CIs (D). miR-204-5p and miR-30c-5p replicated more weakly in this cohort compared with LN vs HE in the OSU cohort.