A discrete cluster of urinary biomarkers discriminates between active systemic lupus erythematosus patients with and without glomerulonephritis

Abstract

Background: Management of lupus nephritis (LN) would be greatly aided by the discovery of biomarkers that accurately reflect changes in disease activity. Here, we used a proteomics approach to identify potential urinary biomarkers associated with LN.

Methods: Urine was obtained from 60 LN patients with paired renal biopsies, 25 active non-LN SLE patients, and 24 healthy controls. Using Luminex, 128 analytes were quantified and normalized to urinary creatinine levels. Data were analyzed by linear modeling and non-parametric statistics, with corrections for multiple comparisons. A second cohort of 33 active LN, 16 active non-LN, and 30 remission LN SLE patients was used to validate the results.

Results: Forty-four analytes were identified that were significantly increased in active LN as compared to active non-LN. This included a number of unique proteins (e.g., TIMP-1, PAI-1, PF4, vWF, and IL-15) as well as known candidate LN biomarkers (e.g., adiponectin, sVCAM-1, and IL-6), that differed markedly (>4-fold) between active LN and non-LN, all of which were confirmed in the validation cohort and normalized in remission LN patients. These proteins demonstrated an enhanced ability to discriminate between active LN and non-LN patients over several previously reported biomarkers. Ten proteins were found to significantly correlate with the activity score on renal biopsy, eight of which strongly discriminated between active proliferative and non-proliferative/chronic renal lesions.

Conclusions: A number of promising urinary biomarkers that correlate with the presence of active renal disease and/or renal biopsy changes were identified and appear to outperform many of the existing proposed biomarkers.

Keywords: Glomerulonephritis; Renal biopsy; Systemic lupus erythematosus; Urinary biomarkers.

Fig. 1

Urinary protein differences between patient groups. a Urinary protein levels for 85 systemic lupus erythematosus (SLE) patients (60 active LN, 25 active non-LN) and 24 controls in the discovery cohort. Shown are hierarchical clustering results for the 128 analytes tested as measured by Luminex. Signal intensities were adjusted using SD scaling, with green indicating overexpression. Hierarchical clustering was performed on the samples (rows) and proteins (columns) using divisive analysis. b Log₂ fold-change of normalized protein abundance for comparisons between SLE patients and healthy controls (left column) and between active SLE patients with and without LN (right column). The size of the circles indicates the fold-change for the comparison, with orange circles indicating overexpression and blue circles indicating underexpression in all SLE patients (left column) or patients with LN (right column). The background in each row/column indicates the statistical significance of the comparison as determined by multivariate linear modeling followed by FDR correction

Fig. 2

Comparison of the most highly discriminative urinary analytes to several previously proposed biomarkers. Scatter plots showing the normalized concentration of selected urinary proteins in healthy controls (HC; n = 24, open circles) and SLE patients with active non-LN (ANLN; n = 25, closed circles) or active LN (ALN; n = 60, closed triangles). Urinary concentrations were corrected for creatinine to normalize for osmolality. Units for all graphs are pg/μmol, except PF4 and TWEAK which are in ng/μmol. a Urinary analytes that demonstrated >3 log₂ fold-difference between ANLN and ALN. b Previously proposed urinary biomarkers. For all graphs each symbol represents the determination from a single individual, with the mean value for each group indicated by a horizontal line. Fold-differences and statistical comparisons for the analytes are shown in Table 2. c Correlation between the urinary analytes that demonstrated >3 log₂ fold-difference between ANLN and ALN and that showed the strongest association with albuminuria, estimated glomerular filtration rate (GFR), or Systemic Lupus Erythematosus Disease Activity Index-2000 (SLEDAI). Lines indicate linear regression curves. Spearman’s correlation coefficients are shown

Fig. 3

Urinary proteins that discriminate between proliferative and non-proliferative/chronic lesions on renal biopsy. Scatter plots showing the normalized concentration of selected urinary proteins in SLE patients with proliferative LN (Prol; Class III and IV (A) or (A/C), n = 38, closed triangles) and non-proliferative or chronic lesions (NPC; Class I, II, V, or VI, and Class III or IV (C), n = 22, open triangles) on renal biopsy as compared to active SLE patients without LN (NLN; n = 25, closed circles). Units for all analytes are pg/μmol, except vWF and PF4 which are in ng/μmol. a Plots for the eight urinary proteins that discriminated between proliferative and NPC lesions on renal biopsy. In the two columns on the right of each plot the NPC results have been subdivided into chronic class III/IV (III/IV(C)) and other non-proliferative/chronic lesions (NP). b Plots for three representative urinary proteins that did not discriminate between proliferative and non-proliferative or chronic lesions on renal biopsy, demonstrating their ability to discriminate between non-proliferative/chronic lesions and active non-LN. For all graphs, each symbol represents the determination from a single individual, with the mean value for each group indicated by a horizontal line. Significance levels were determined by Mann-Whitney non-parametric testing and have been corrected for multiple testing. Only the differences between proliferative and non-proliferative/chronic lesions, and non-proliferative/chronic lesions and active non-LN for each analyte are shown. In graphs in panel (a), significant differences between active proliferative and chronic class III/IV lesions are also indicated, and represent uncorrected p values. *p < 0.05, **p <0.005, NS not significant

Fig. 4

Normalization of urinary protein levels in LN patients in remission. Scatter plots showing the normalized concentration of selected urinary proteins in SLE patients from the validation cohort with active non-LN (ANLN; n = 16, closed circles), active LN (ALN; n = 33, closed triangles) or LN in remission (RLN; n = 30, open triangles). Units for all graphs are pg/μmol, except for PF4 which is in ng/μmol. Each symbol represents the determination from a single individual, with the mean value for each group indicated by a horizontal line. Statistical comparisons for the analytes are shown in Table 2