Discovery of urine biomarkers for lupus nephritis via quantitative and comparative proteome analysis

To the Editor: Lupus nephritis (LN) is the leading cause of morbidity and mortality in patients with systemic lupus erythematosus (SLE).1 Noninvasively obtainable biomarkers for LN could greatly improve early LN detection. In this study, we found that urine alpha-1-acid glycoprotein (ORM1) is a promising biomarker for early LN detection. Furthermore, with regard to histologic features, urine haemoglobin subunit delta (HBD) accurately discriminated proliferative LN from nonproliferative LN and correlated with activity index. Using sequential window acquisition of all theoretical mass spectra combined with liquid chromatography (SWATH LC–MS) platform, a novel mass spectrometrybased proteome analysis,2 we screened potential biomarker candidates in urine samples from 20 healthy controls (HCs) and 39 patients with SLE with or without newly diagnosed LN (n-LN). All patients with SLE met the 2012 Systemic Lupus International Collaborating Clinics classification criteria for SLE.3 Sample preparation and in-house urine proteome spectral library generation were performed as described previously.2 In our spectral library, 2323 proteins and 8371 peptides were identified using data-dependent acquisition analysis. A total of 1157 protein groups (581 ± 31, 506 ± 25 and 457 ± 22 in HCs, patients with SLE without nephritis, and patients with n-LN, respectively) were quantified from individual samples using an in-house spectral library. Among the 1157 protein groups, 143 protein groups (false discovery rate ≤.05, log2 fold change ±1) were increased in patients with SLE without nephritis than those in HCs (Figure 1A), and 67 protein groups were increased in patients with n-LN than those in patients with SLE without nephritis (Figure 1B). Among these protein groups, 23 protein groups were identified to be common between the two comparative analyses. Among these 23 urine proteins, five proteins (ORM1, antithrombin-III [SERPINC1], ceruloplasmin [CP], haemoglobin subunit beta [HBB] and HBD) were selected for enzyme-linked immunosorbent assay

(ELISA)-based validation. The characteristics between patients with SLE without nephritis (n = 22) and patients with n-LN (n = 17) included in the validation study were compared (Table 1). Importantly, three patients with n-LN (17.3%) had nonsignificant proteinuria (urine protein/creatinine ratio <500 mg/g). Therefore, the urine biomarkers identified herein could be useful in early LN detection, even before significant proteinuria develops. The ELISA results revealed that urine ORM1, SERPINC1, CP and HBD, normalised to urine creatinine, were significantly upregulated in patients with n-LN than in patients with SLE without nephritis ( Figure 1C-G). Similar results were observed with urine ORM1, SERPINC1, CP and HBD, not normalised to urine creatinine ( Figure S1).
The accuracy of the urine biomarkers and traditional serologic factors (C3, C4 and anti-dsDNA antibody) 4 for discriminating n-LN from SLE without nephritis was evaluated using receiver operating characteristic (ROC) analysis (Figure 2A-I). The area under the curves (AUCs) of urine ORM1, SERPINC1, CP and HBD were .914, .874, .896 and .757, respectively (Figure 2A-D). Urine ORM1, SER-PINC1 and CP had higher AUCs than the traditional serologic factors ( Figure 2E-G: C3, .759; C4, .689; and anti-dsDNA antibody, .715). Further, we evaluated whether combinations of variables enhanced discriminatory ability. Two composite parameters were assessed: combination of all four urine biomarkers (ORM1 + SERPINC1 + CP + HBD) and combination of factors selected using logistic regression analysis with stepwise forward selection (C3 + urine ORM1). ORM1 + SERPINC1 + CP + HBD had an AUC of .901 and C3 + urine ORM1 had an AUC of .920 ( Figure 2H,I). Although the AUC was the highest with C3 + urine ORM1, it was only slightly higher than that of urine ORM1 alone. This suggests that urine ORM1 accurately distinguishes patients with n-LN from patients with SLE without nephritis. ORM, a member of the acute phase protein family, activates monocytes, induces T-cell proliferation, and promotes the secretion of proinflammatory cytokines. 5 However, its biological role in LN is poorly    (Figure 2J-P). Urine HBD had a high accuracy (AUC = .964) in differentiating proliferative LN from nonproliferative LN ( Figure 2M). As differentiating proliferative LN from nonproliferative LN is important in the perspective of treatment strategy, 7 this discriminatory ability of urine HBD has clinical implication. We also assessed the correlation between urine biomarkers TA B L E 2 Correlation between potential urine biomarkers for LN and pathologic activity/chronicity index and activity/chronicity indices ( Table 2). Of the four urine biomarkers, only urine HBD was significantly positively correlated with the activity index (rho = .549, p = .024).
The correlation was stronger when urine HBD was used in combination with other urine proteins (ORM1 + SER-PINC1 + CP + HBD: rho = .727, p = .001). However, none of the urine biomarkers showed correlation with chronicity index. The activity index after treatment is an important prognostic factor of LN. 8 Clinical renal parameters do not accurately reflect the activity index, 9 and it is difficult to presume the activity index without performing renal biopsies. Nevertheless, renal biopsy is an invasive procedure and is often difficult to perform repeatedly. In such cases, urine HBD could be useful for speculating the activity index. HBD has been suggested to be a potential biomarker for early diabetic kidney disease in a recent proteomics study. 10 The possible mechanism underlying urine HBD in diabetic kidney disease pathogenesis is thought to be inflammation and oxidative stress. 10 However, its association with LN has not been known to date. We provide the first evidence that it could be used as a biomarker to reflect histologic information in patients with LN. In summary ( Figure 2Q), although limited by the small sample size and lack of replication in an independent cohort, we showed that urine ORM1 can accurately detect LN, even in the early disease where significant amount of proteinuria has yet developed. In addition, urine HBD had an excellent accuracy in differentiating proliferative LN from nonproliferative LN and correlated with activity index. Therefore, these urine biomarkers may provide important information, particularly when renal biopsies are unavailable. Although these findings are promising, further studies with more patients are warranted for a powerful conclusion.

E T H I C S A P P R O VA L A N D C O N S E N T T O PA R T I C I PAT E
The Institutional Review Board of Asan Medical Center in Seoul, South Korea, approved the study (IRB No. 2013-0405). Written informed consent was obtained from all patients.

C O N F L I C T O F I N T E R E S T
The authors declare that there is no conflict of interest.  Yong-Gil Kim 2,3