Urinary biomarkers for active Lupus Nephritis that have survived independent validation across cohorts

Abstract

Most reported biomarkers for lupus nephritis (LN) have not been independently validated across cohorts. Moreover, many of the documented biomarker candidates have been reported to be elevated in LN compared to healthy controls. In contrast, biomarkers that distinguish patients with active LN (ALN) from inactive systemic lupus erythematosus (iSLE) hold significant clinical utility. Hence, this focused review attempts to assemble urine protein biomarkers for LN that (a) have been independently validated across two or cohorts (using ELISA and/or comprehensive proteomics), and (b) exhibit good diagnostic potential for distinguishing ALN from iSLE. PubMed and OVID were screened for studies assessing the diagnostic value of urinary biomarkers in subjects with ALN as compared to iSLE. Forty peer-reviewed articles were evaluated, encompassing urine biomarker data from 3,411 distinct patients. Of the 32 candidate biomarkers identified, fourteen were repeatedly reported/tested in four or more papers each, namely ALCAM, CCL2 (MCP1), CD163, HAVCR1 (KIM-1), HPGDS, ICAM-1 (CD54), ICAM-2 (CD102), IGFBP-2, LCN2, NCAM-1 (CD56), SELE (E-Selectin), SELL (L-Selectin), TNFSF12 (TWEAK), and VCAM-1, with most exhibiting ROC AUC values ≥ 0.80 across multiple studies when discriminating ALN and iSLE patients. The 32 reproducibly elevated biomarkers for active LN mapped to nine functional categories. The urinary proteins reported here promise to serve as a liquid biopsy for ALN. Besides representing potential candidates for diagnostic, monitoring, predictive, and prognostic biomarkers in LN, they also provide a window into potential intra-renal processes that may be responsible for driving LN. Ongoing advances in proteomics, which offer wider proteome coverage at increased sensitivity, are likely to further reshape our perspective of urinary biomarkers for LN in the coming years.

Lay Summary

This review shortlists reliable urine protein biomarkers for identifying active lupus nephritis (LN), a complication of systemic lupus erythematosus (SLE). With a focus on independently validated biomarkers distinguishing active LN (ALN) from inactive (iSLE), the authors screened 40 peer-reviewed articles, covering data from 3,411 SLE patients. 32 reproducibly elevated biomarkers for ALN were identified, mapping to nine functional categories. Of note, urine ALCAM, CCL2, CD163, HAVCR1, HPGDS, ICAM-1, ICAM-2, IGFBP-2, LCN2, NCAM-1, SELE, SELL, TNFSF12, and VCAM-1, consistently exhibit promising diagnostic potential, with ROC AUC values ≥ 0.80 across four or more studies. These non-invasive urine biomarkers may help identify initial kidney involvement in a patient diagnosed with SLE.

Introduction

Lupus is a chronic autoimmune inflammatory disease that affects multiple systems in the body with a broad spectrum of clinical manifestations causing edema and tissue damage in any part of the body, including the joints, skin, kidneys, heart, lungs, and brain [1]. Systemic lupus erythematosus (SLE) affects an estimated 1.5 million Americans and 5 million people worldwide of which a disproportionate amount are women of child-bearing age and people of color [2].

Lupus nephritis is a serious complication of SLE that specifically affects the kidneys and is significantly associated with an increase in morbidity and mortality which can lead to end-stage renal disease [3]. According to the National Institute of Diabetes and Digestive and Kidney Diseases, LN affects up to 60% of people with lupus within the first five years which can then advance chronic kidney disease or progress to end-stage renal disease [4, 5].

There is a growing body of evidence investigating the use of urinary biomarkers for early diagnosis and monitoring of active lupus nephritis. Up to 70% of the proteins and peptides in urine are of renal origin and are typically normalized based on the concentration of creatinine to account for patient hydration status [6]. Because of the proximity to disease pathology several proteins such as adhesion molecules, autoantibodies, chemokines, complement proteins, and cytokines, have been recognized as possible biomarkers of disease activity in cross-sectional studies of LN patients. Urine is easily obtained non-invasively, allowing for sequential sampling which would aid in the validation of biomarkers in longitudinal studies and clinical trials. Identifying highly sensitive and specific LN biomarkers to monitor and predict disease exacerbation could aid in the advancement of precision medicine for LN patients.

The advent of OMICs, using technologies such as multiplexed immune assays, transcriptomics, proteomics, and mass cytometry, is giving a new boost to this field, allowing the discovery of promising biomarkers. Of relevance to this review, recent comprehensive proteomic screens examining 1000 or more proteins in the urine, using either an antibody-based or an aptamer-based targeted proteomic platform, have given rise to several novel biomarker candidates for LN [7–10].

A comprehensive overview of urine biomarkers in LN and biomarkers from other body fluids has recently been reported [11]. Unfortunately, many potential biomarkers reported in the literature have not been independently verified by other groups or orthogonal assay platforms. Moreover, many of the documented biomarker candidates have been reported to be elevated in LN compared to healthy controls. In contrast, biomarkers that distinguish patients with active LN (ALN) from inactive SLE patients without active renal disease (iSLE) may hold higher clinical utility. Hence, the objective of this focused review attempts to assemble urine protein biomarkers for LN that (a) have been independently validated across two or more cohorts, and (b) exhibit diagnostic potential for distinguishing ALN from iSLE, with the following diagnostic metrics: Fold Change (FC) ≥2, at p<0.05, and/or Fold Change ≥2 with a ROC Area Under the Curve (ROC AUC) ≥0.60, and (c) at least one of the assay platforms used for validation is the ELISA.

Methods

A systematic literature review was conducted by two individuals independently (SV, CM) to identify articles that reported findings from clinical laboratory studies that analyzed the diagnostic role of urinary biomarkers in patients with ALN. Relevant studies were identified using the PubMed electronic database and the following search terms ‘lupus nephritis,’ ‘biomarkers,’ ‘urine samples,’ and ‘human.’ Supplementary searches in OVID and Web of Science using the same search terms yielded only duplicates of the PubMed results. The final list of included urine proteins was also used as search terms, together with “lupus nephritis” and “urine” in Google Scholar, with an updated search date of August 9, 2024, in order to identify any remaining studies that met the inclusion criteria. The detailed search procedure is presented in the PRISMA flow chart in Figure 1. The search terms yielded 232 articles that were manually reviewed to exclude reviews, microRNAs, studies based on qualitative design, and case reports. As an additional inclusion criterion, each urine protein biomarker included was manually confirmed to meet at least one of the following statistical parameters, when comparing active LN to inactive SLE: Fold Change (FC) ≥2, at p<0.05, and/or Fold Change (FC) ≥2 with a ROC Area Under the Curve (ROC AUC) ≥0.60. Finally, each included urinary biomarker had to meet the criterion of independent validation, meaning that the selected urinary protein had been independently validated in two or more independent cohorts, consistently demonstrating a difference between ALN and iSLE. Finally, we mandated that at least one of the assay platforms used for validation should be an ELISA, since not all urine proteomic hits readily clear validation [7–9]. All of these criteria reduced the number of peer-reviewed articles to 40. The criteria adopted for this systematic review have not been published as a protocol, nor was a protocol registered. This study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [12] as detailed in Figure 1 and supported by PRISMA checklist (Supplementary Material S2). Quadas 2 was used to assess the quality of the diagnostic accuracy of the studies included in this review (Supplemental Figure 3).

Figure 1:

PRISMA flow diagram for systematic review of LN urine biomarkers based on searches of public databases. The assessment of eligibility was performed by S.V. and C.M., and uncertainties were discussed and resolved by consensus [12].

Most reviewed manuscripts used the systemic lupus erythematosus disease activity index (SLEDAI) which is a cumulative and weighted scoring system using 24 clinical and laboratory indices and renal SLEDAI (rSLEDAI) which refers to the sum of the scores of the four renal criteria in SLEDAI to define ALN and iSLE, although some used proteinuria cutoffs, as detailed in Figure 2. The vast majority of the studies classified patients with biopsy-proven LN with rSLEDAI ≥4 (or uPCR >0.5g/24hr) as having ALN (Figure 2A). There was more variability in how iSLE was defined. However, in most studies, subjects were classified as having iSLE if their rSLEDAI was zero, and global SLEDAI was less than or equal to 4, as detailed in Figure 2B.

Figure 2. Criteria used for defining ALN and iSLE in the included studies.

Detailed are the criteria used for defining ALN (A) and iSLE (B)in the studies tabulated In Table 1 (7–9,13–49). Note: 1: All but 2 studies also mandated biopsy evidence of LN for defining ALN. In addition, one study included the presence of an active urinary sediment as an additional criterion for ALN. 2: Most studies that included proteinuria as a criterion used 0.5g/day as the cutoff, while some used 0.5mg/mg as the cut-off. 3: Two studies did not clearly specify the criteria for defining ALN. 4: Most studies that used proteinuria as a cutoff for defining iSLE used <0.5g/day as the cut-off, while one used <0.25g/day, and another used <1 g/day as the cut-off.

The literature pertaining to each included urine protein was scrutinized in detail using Google Scholar in order to understand the functional role of the protein. Based on an understanding of each protein’s role, the included proteins were classified into nine different functional groups as detailed in Figure 3.

Figure 3. Functional Categories of Urine Biomarkers for ALN.

The shortlisted biomarkers belong to at least nine different functional categories, each of which has been well researched in the lupus literature. The 32 urinary biomarkers distinguishing ALN from iSLE independently validated across two or more independent cohorts are indicated in bold. Shown in italics are the urinary biomarkers that did not meet the criteria set for inclusion but were otherwise significantly elevated in ALN compared to healthy urine, by one independent research group or by only one orthogonal analytical method for biomarker detection.

Results

Data from 40 published studies [7–9, 13–49] were evaluated based on the criteria defined in Methods. Insights from recently reviewed articles [10, 11] were also included in this analysis. Only urinary biomarkers that have been independently validated to distinguish ALN from iSLE patients in two or more cohorts were included. This yielded a total of 32 distinctive urinary proteins, as listed in Table 1, which met at least one of the following cutoff parameters, FC ≥2 and p<0.05, OR ROC AUC ≥0.60 and p<0.05, comparing ALN to iSLE, across two or more cohorts. Interestingly, the shortlisted urine proteins have all been independently validated not only across different cohorts, but also by independent research groups and assay platforms.

Table 1:

Independently validated urine proteins discriminating active LN from inactive SLE at p<0.05, FC > 2, ROC AUC >0.60

Protein	Alias	Proteomic Screen 1	Proteomic Screen 2	ELISA-based reports
ALCAM	CD166	FC:8752.7 [7]	FC:6.6; AUC=0.84 [8]	FC:3.0; AUC=0.74 [8]; AUC=0.74 [13]; AUC=0.64–0.90 [14]; AUC=0.83 [15]
Angiostatin		AUC=0.83 [40]		AUC=0.85 [19]
BCL2A1	BFL-1	FC:11.9; AUC=0.93 [8]		FC:3.3; AUC=0.74 [8]
CAST	Calpastatin	FC:6.7; AUC=0.95 [8]		FC:63.0; AUC=0.74 [8]
CCL2	MCP-1	FC:3.5e14; AUC=0.95 [7]	FC:9.1; AUC=0.79 [8]	FC:2.9; AUC=0.71 [8]; FC:4.9; AUC=0.78 [20]; AUC=0.87 [25]; AUC=0.70 [27];
			FC:7.7; AUC=0.89 [16]	AUC=0.76 [28]; AUC=0.81 [29]; AUC=1.00 [30]; AUC=0.73 [31]; AUC=0.81 [33];
			AUC=0.90[22]; AUC=0.70[23]	FC:3.3 [34]
CD163		FC:1.9e4 [7]	FC:5.5; AUC=0.77 [8]	AUC=0.98 [35]; AUC=0.81–0.96 [36]; AUC=0.76 [45]
CFP	Properdin	FC:10.2; AUC=0.91 [8]		FC:6.2; AUC=0.79 [8]
CP	Ceruloplasmin			AUC=0.68 [42]; AUC=0.77 [48]
CXCL 10	IP-10	FC:868.3 [7]		FC:3.2; AUC=0.68 [20]
CXCL 16		FC:3.5e6; AUC=0.91 [7]		FC:11.1; AUC=0.68 [7]
FABP4		FC:1.8e10; AUC=0.94 [7]		FC:16.6; AUC=0.82 [7]
FASLG	CD95L	FC:2.3e4 [7]	FC:2.8 [8]	FC:38.8; AUC=0.82 [7]
HAVCR1	KIM-1	FC:9.4e11 [7]	AUC=0.76 [23]	AUC=0.70 [27]; AUC=0.92 [37]
HPGDS	PGDS			AUC=0.86 [17]; AUC=0.93 [26]*; AUC=0.79 [41]; AUC=0.71 [42]*
ICAM-1	CD54	FC:1.5e10 [7]	FC:4.2; AUC=0.91 [8]	AUC=0.75 [17]
			FC:4.2; AUC=0.83 [16]
			AUC=0.97 [18]
ICAM-2	CD102	FC:2.36 [7]	FC:2.8 [8]	FC:7.6; AUC=0.83 [7]
			FC:8.1 [9]
IGFBP-2		FC:2e3; AUC=0.88 [7]	FC:2.8; AUC=0.88 [8]	FC:7.9; AUC=0.81 [7]
			FC:2.7; AUC=0.70 [9]
IL-1RT2	CD121b	FC: 1.0e4 [7]	FC:62.9 [9]	FC:3.2; AUC=0.70 [7]
LCN2	NGAL	FC:115.1 [7]	AUC=0.78 [23]	AUC=0.70 [27]; AUC=0.67 [31]; AUC=0.99 [37]; AUC=0.67 [44]; AUC=0.83 [46]
NCAM-1	CD56	FC:20 [7]	FC:4.0; AUC=0.89 [8]	FC:7.3; AUC=0.91 [47]*
			AUC=0.88 [18]
OPG	TNFRSF11B	FC:6.5e6; AUC=0.94 [7]		FC:7.9; AUC=0.81 [7]
ORM-1	AGP-1			AUC=0.91 [24]; AUC=0.81 [42]*
PF-4	CXCL 4	FC:46.0; AUC=0.98 [8]		FC:8.9; AUC=0.81 [8]; AUC=0.76 [19]
RBP4				AUC=0.80 [32]; FC:5.68 [49]
SELE	E-Selectin	FC:5.1e7 [7]	FC:8.3; AUC=0.84 [8]	FC:90.1; AUC=0.80 [7]; FC:2.6; AUC=0.73 [8]
SELL	L-Selectin	FC:3.3e12 [7]	FC:3.5; AUC=0.79 [8]	FC:27.2; AUC=0.85 [9]; FC:29.2; AUC=0.97 [47]*
TF	Transferrin			AUC=0.80 [42]*; AUC=0.82 [48]
TFPI		FC:6.5; AUC=1.00 [8]		AUC=0.77 [43]
TIMP-1		FC:7.5e4; AUC=0.99 [7]		FC:7.9; AUC=0.81 [7]
TNFSF12	TWEAK	FC:4.2e3; AUC=0.94 [7]	FC:2.0; AUC=0.79 [8]	AUC=0.82 [25]; AUC=1.00 [30]; FC:2.9 [34]; AUC=0.88 [38]; AUC=0.88 [39]
TNFSF13B	BAFF	FC:4.6e7 [7]		FC:12.0; AUC=0.80 [7]
VCAM-1	CD106	FC:1.1e11 [7]	FC:15.4; AUC=0.91 [8]	FC:2.5; AUC=0.85 [8]; AUC=0.77 [13]; AUC=0.80 [19]; AUC=0.84 [21]
			FC:6.5; AUC=0.91 [16]

Abbreviations used: AGP-1: Alpha-1-acid glycoprotein 1; ALCAM: activated leukocyte cell adhesion molecule; BFL-1: apoptosis-regulatory protein of the BCL2 family; CD163: Scavenger receptor cysteine-rich type 1 protein M130; ICAM-1: Intercellular Adhesion Molecule 1 (CD54); IP-10: Interferon gamma-induced protein 10; KIM-1; Kidney Injury Molecule-1; L-PGDS; Lipocalin-Type Prostaglandin D Synthase; MCP-1: monocyte chemoattractant protein-1; NCAM-1: neural cell adhesion molecule 1; NGAL: neutrophil gelatinase-associated lipocalin; PDGS: Prostaglandin D Synthase; PF4: Platelet factor 4; RBP4: (Retinol Binding Protein 4; TFPI: tissue factor pathway inhibitor; TNF R1: Tumor necrosis factor receptor 1; TWEAK: TNF-like weak inducer of apoptosis; VCAM-1: Vascular cell adhesion protein 1

Aptamer-based screen

Enzyme-linked immunosorbent assay (ELISA)

Array-based

Proximity Extension assay (PEA)

Electrochemiluminescence

The included studies had different study designs and varying sample sizes. There were 33 studies that used a cross-sectional approach, one study was longitudinal, one was prospective and cross-sectional, three were longitudinal and cross-sectional, and two studies used a prospective, longitudinal, and cross-sectional approach. For longitudinal studies, data from baseline were used.

The 40 studies encompassed a total of 3,411 distinct patients. These numbers represent independent subjects, with no double-counting, based on detailed scrutiny of the included manuscripts. As detailed in Figure 2, most studies classified patients with biopsy-proven LN with rSLEDAI ≥4 (or uPCR >0.5g/24hr) as having ALN, while subjects were classified as iSLE if their rSLEDAI was zero, and global SLEDAI was less than or equal to 4, though some variation was noted.

To identify urinary biomarkers, five different techniques were used across the 40 different studies, namely antibody array-based screen, aptamer-based proteomics, proteomic extension assay (PEA), and electrochemiluminescence (ECL), and enzyme-linked immunosorbent assay (ELISA), as shown color-coded in Table 1. Whereas the first 2 columns in Table 1 have employed comprehensive proteomic screening platforms, the final column lists ELISA studies focusing on single analytes.

In total, across the 40 studies, there were 809 unique urinary protein biomarkers were reported comparing ALN to iSLE patients but only 32 of those met the stipulated inclusion criteria. Of the 32 urine biomarkers identified, 14 were repeatedly reported in four or more papers each, namely ALCAM, CCL2 (MCP1), CD163, HAVCR1 (KIM-1), HPGDS, ICAM-1 (CD54), ICAM-2 (CD102), IGFBP-2, LCN2, NCAM-1 (CD56), SELE (E-Selectin), SELL (L-Selectin), TNFSF12 (TWEAK), and VCAM-1 (bolded in Table 1), most with ROC AUC values ≥ 0.80, indicating excellent discrimination between ALN and iSLE.

Besides their obvious potential in disease diagnostics and monitoring, these validated biomarkers also point to potential functional pathways that are likely to be important in the pathogenesis of LN. Classifying biomarkers into biological processes allows researchers to identify potential targets for therapeutic intervention and understand the potential multi-factorial role biomarkers may play in autoimmune diseases. As depicted in Figure 3, the shortlisted biomarkers belong to at least nine different functional categories, each of which has been well-researched in the lupus literature. The 32 urinary biomarkers distinguishing ALN from iSLE independently validated across two cohorts are indicated in bold. Shown in italics are the urinary biomarkers that did not meet the criteria set for inclusion but were otherwise significantly elevated in ALN compared to healthy urine in one or more independent cohorts.

Innate Immunity

Identified urinary biomarker proteins that participate in initiating and/or amplifying the innate immune responses include BCL2A1 (BFL-1), CAST (Calpastatin), CD163, CFP (Properdin), FABP4, and TNFSF12 (TWEAK). Two of these proteins have been extensively validated across 4–5 cohorts each, using different assay platforms, by multiple research groups, resulting in ROC AUC values for distinguishing ALN from iSLE as high as 0.98 to 1.0, namely CD163 and TNFSF12 (TWEAK).

Urine CD163 also predicted renal response to treatment, correlated well with renal pathology activity index (AI), SLEDAI, rSLEDAI, uPCR, C3/C4, anti-dsDNA [26, 35, 45] and discriminated proliferative from non-proliferative LN [36]. Because CD163 is a surface marker expressed by M2 macrophages that can be released from the plasma membrane by metalloproteinases in response to inflammatory stimuli, urine CD163 is a useful non-invasive marker of M2 macrophage infiltration/activation within nephritic kidneys [50], that can potentially be used as a marker of treatment response in LN.

Similarly, urine TNFSF12 (TWEAK) also exhibited positive correlation with SLEDAI, rSLEDAI, renal pathology AI scores, uPCR, anti-dsDNA, and a negative correlation with serum C3/C4 and has also been shown to predict renal flares and treatment response [51, 52]. TNFSF12 interacts with Fn14 to activate multiple signaling pathways in target cells and enhance mesangial and epithelial cell proliferation which is pathogenic in LN [53]. Indeed, subduing this pathway reduces proteinuria and renal IgG deposition, cytokine/chemokine production and macrophage infiltration in murine LN. Higher levels of TNFSF12 were found in patients with more severe renal pathology and have been shown to promote inflammation and fibrosis in the kidneys [25]. Not surprisingly, neutralization of TNFSF12 as a potential therapeutic modality in LN is also of great interest [54, 55].

Urine BCL2A1, CAST (Calpastatin) and CFP (Properdin) are 3 additional innate immune molecules, whose discriminatory potential for ALN outperformed that of C3/C4 and anti-dsDNA, in terms of assay sensitivity, PPV, and/or NPV [8]. Most of these urine proteins correlated well with proteinuria, PGA, SLEDAI, rSLEDAI and/or anti-DNA [8]. BCL2A1 protects cells from apoptosis induced by various stimuli such as cytokines and oxidative stress and is upregulated by inflammatory signals, including Epstein-Barr Virus LMP1 protein, which itself is associated with LN [56]. Calpastatin tightly regulates calpain, a mediator of acute inflammation, NFkB activation, and glomerulonephritis (GN) [57]. On the other hand, Properdin functions as a positive regulator of the alternative pathway of complement activation and has been shown to facilitate the pathogenesis of LN [58, 59].

We also documented additional innate immune molecules that did not meet the criteria set for inclusion in Table 1 but were otherwise significantly elevated in active LN urine compared to healthy urine [8, 9]. These include CD14, CD36, and MCSF-R as italicized in Figure 3, all of which are key surface molecules on macrophages [60]. To sum, most of the innate immune molecules uncovered as potential biomarkers for ALN appear to play a pathogenic role in LN.

Adaptive Immunity

Several proteins validated as urine biomarkers for ALN impact adaptive immunity, namely FASLG (CD95L), HAVCR1 (KIM-1), and TNFSF13B (BAFF). Urine KIM-1 also correlates with serum Creatine (sCr), rSLEDAI, renal pathology AI, and discriminates proliferative from membranous LN [23, 27]. It is a type-1 transmembrane glycoprotein that boosts T-cell and B-cell activation/survival, and prevents autoimmune responses by acting as a co-stimulatory molecule on immune cell surfaces [61]. The other 3 proteins in this functional category have not been compared to conventional diagnostic yardsticks, but play essential pathogenic roles.

FASLG or CD95L is expressed on activated T-cells and B-cells that are earmarked for activation-induced cell death (when engaged by FAS). Its vital role in keeping the immune system in check is underscored by studies that show that mice lacking FAS or FASLG develop lymphoproliferative autoimmunity [62]. Importantly, FASLG is expressed on renal tubulo-interstitial CD8/NK T-cells in LN kidneys, most likely representing activated T-cells programmed for cell death [7]. On the other hand, TNFSF13B/BAFF, a TNF-like cytokine produced by myeloid cells, supports the survival and differentiation of B cells. BAFF is elevated in LN serum, associated with anti-dsDNA antibodies, disease flares, renal activity and damage (99–103). Moreover, BAFF expression is elevated in glomerular macrophages and mesangial cells in LN patients (99). Indeed, Belimumab, a human antibody targeting BAFF, is an FDA-approved drug for lupus (103).

Several additional adaptive immune response proteins were documented that did not meet the criteria set for inclusion in Table 1 but were significantly elevated in active LN urine compared to healthy urine [8, 9, 30]. These include BCMA (TNFRSF13A, CD269), CD26 (DPPIV), CD30 (TNFRSF8), CD40L (CD154), and TNFSF9 (CD127/41BB) as italicized in Figure 3, all of which are involved in the regulation of B-cell and T-cell responses, including activation, survival, and cytokine/antibody production; interestingly, some of these predict clinical and histopathological response to therapy in LN [63].

Adhesion Molecules

Multiple urinary cellular adhesion molecules were identified in this study, including ALCAM (CD166), ICAM-1 (CD54), ICAM-2 (CD102), VCAM-1 (CD106), NCAM-1 (CD56), SELE (E-Selectin, CD62E), SELL (L-Selectin, CD62L), all of which may mediate the “sticky interactions” amongst the different immune cells that mediate LN and the endothelium [64]. Not only were these adhesion molecules highly discriminatory of ALN, outperforming conventional biomarkers such as anti-DNA and C3/C4, they also showed good correlation/association with SLEDAI/rSLEDAI, PGA, and/or proteinuria [8, 9, 14–16, 19, 47].

Urine ALCAM and VCAM have been extensively validated by multiple research groups across many cohorts, and both are reflective of renal pathology AI, as well as long-term renal outcome [8, 13–15]. ALCAM expressed primarily on myeloid cells and endothelial cells mediates engagement with its cognate ligand on T-cells, CD6, to mediate cell activation [14, 15]. Blockade of the ALCAM/CD6 pathway is therapeutically effective in murine models, and human subject testing has begun using an anti-CD6 antibody (Itolizumab) [14]. VCAM-1 expressed on endothelial and other cells mediate engagement to leukocytes expressing VLA-4, thus facilitating the pathogenesis of lupus nephritis [65]. Thus, both these urine biomarkers are also integral in the pathogenesis of the disease.

In fact, endothelial VCAM-1, ICAM-1, and ICAM-2 together play an intricate role in facilitating leukocyte trafficking and transmigration across endothelial cells during inflammation [66]. Not surprisingly, both ICAM-1 and ICAM-2 are also highly discriminatory of ALN (Table 1). Importantly, urine ICAM-1 predicts renal flares [17, 18], and also plays a pathogenic role in murine LN [67].

SELL is a leukocyte selectin upregulated on T cells and neutrophils, that accumulates in the kidneys, and exacerbates local inflammation [68]. In contrast, SELE is an endothelial selectin expressed in blood vessels, and facilitates leukocyte recruitment, which plays a significant role in LN progression [66, 69–71]. Both these selectins are not only highly discriminatory for ALN, but both are also correlated with a variety of clinical indices (including PGA, SLEDAI, rSLEDAI, and proteinuria), and both outperformed anti-dsDNA, and C3/C4 in their diagnostic potential. In addition, urine SELL also correlates with renal pathology AI, and predicts long-term disease activity [9, 47].

Finally, urine NCAM-1 correlated with SLEDAI and rSLEDAI, outperformed anti-dsDNA, and C3, and showed association with renal pathology AI and glomerulosclerosis [47]. NCAM-1 is a cell surface glycoprotein expressed mostly in interstitial cells but also occurs on podocytes and has been associated with interstitial fibrosis in lupus GN [72].

Chemokines

Among the urine chemokines implicated, CCL2 (MCP-1) is perhaps the most extensively investigated urinary biomarker in LN. Besides being shown to be discriminatory of ALN across >10 independent studies, it is correlated with disease activity and chronicity indices, reduced upon successful treatment, and predicted renal disease progression [8, 22, 25, 29, 30, 33]. MCP-1, a chemokine that attracts monocytes, is upregulated by several proinflammatory cytokines, and amplifies inflammation [30]. Indeed, the inhibition of MCP-1 ameliorates murine LN [73].

Among the other chemokines, urine CXCL10 (IP-10), CXCL16, and PF-4 (CXCL4) also correlated with rSLEDAI, and outperformed serum C3/C4 and anti-dsDNA antibodies in their diagnostic potential [7, 22]. IFN-driven CXCL10 (IP-10) upregulation may be involved in the transport of lymphocytes into afflicted organs in murine lupus models and SLE patients [74], while CXCL16 plays a role in attracting and activating T-lymphocytes during inflammation. Urine PF-4 (CXCL4) also correlated with SLEDAI and rSLEDAI, and surpassed proteinuria, C3/C4, and anti-dsDNA in its diagnostic potential [8]. PF-4 modulates angiogenesis, promotes expression of pro-fibrotic cytokines such as IL-4 and IL-13, and enhances the proliferation of T cells which can lead to kidney damage [75–81].

Other chemokines that did not meet the criteria set for inclusion in Table 1 but were otherwise significantly elevated in LN urine compared to healthy urine [8, 9, 16, 82] include CCL11 (Eotaxin), MCP-4 (CCL13), MDC (CCL22), MIF, and MIP1B (CCL4) as italicized in Figure 3.

Hemostasis

Angiostatin, HPGDS (PGDS), and TFPI are three biomarkers identified as proteins that play a crucial role in regulating blood clotting and may contribute to inflammation and immune dysregulation. Urine Angiostatin also correlated with SLEDAI rSLEDAI, uPCR, SLICC renal activity score, and renal pathology chronicity index, outperforming serum anti-dsDNA and C3 [19, 40]. Angiostatin is a proteolytic fragment of plasminogen that inhibits the formation of new blood vessels and is protective in cancer growth through the blockade of angiogenesis [83].

Urine TFPI also correlated with rSLEDAI and SLICC RAS disease activity indices, and is an independent predictor of eGFR, surpassing the specificity and positive predictive value of anti-dsDNA and C3/C4 [43]. TFPI, secreted by human mesangial cells, podocytes, and proximal tubule cells, regulates the extrinsic pathway of blood coagulation and reduces fibrin deposition in the chronic stages of crescentic glomerulonephritis [84].

Other molecules that did not meet the inclusion criteria for Table 1 but were significantly elevated in LN urine compared to healthy urine [8, 9, 43] include CLEC-2, D-dimer, and Plasmin as italicized in Figure 3. Of note, the biomarker potential of urinary plasmin for ALN has been independently validated [43].

Iron Balance

CD163, CP (Ceruloplasmin), and TF (Transferrin) represent 4 urinary biomarkers for ALN, that also play inter-connected roles in iron balance. Whereas HP binds to free hemoglobin, allowing the removal of hemoglobin by macrophages thus preventing the loss of iron [85], TF, a protein coregulated by interferon-α, mediates iron delivery [86]. Both are modestly discriminatory for ALN (Table 1), and both correlated with clinical indices such as SLEDAI, and outperformed conventional diagnostic yardsticks [49]. Other molecules that did not meet the inclusion criteria for Table 1 but were significantly elevated in LN urine compared to healthy urine include Ferritin and Hemopexin [8, 9, 87], as italicized in Figure 3, both of which play an important role in iron homeostasis.

Lipocalin Family

Three of the identified urinary proteins, HPGDS, LCN2 (NGAL), and RBP4 are members of the Lipocalin family, involved in the binding and transporting of small hydrophobic molecules. Of these, the most extensively studied is urine LCN2 (Table 1). LCN2 may be an early predictor of LN activity, where elevated levels track progression to chronic kidney disease [88–91] and demonstrated the ability to predict renal response to induction therapy [92]. LCN2 has been shown to affect T-cell proliferation, cytokine production and migration, and can modulate immune inflammation through the balance of regulatory T-cells and effector T-cells [93]. It has been shown to be essential for lupus nephritis in preclinical models [67, 94]. Interestingly LCN2 has also emerged as a CSF biomarker for neuropsychiatric lupus [95].

Discussion

Research over the past several years has uncovered a large number of potential urine biomarkers that are elevated in ALN compared to the levels in healthy urine. However, biomarkers that are clinically more useful are markers that are elevated in ALN compared to inactive disease. This gap in knowledge has fueled the present systematic review. Given the observation that comprehensive proteomic hits of urine may not always survive platform validation, we restricted the search to biomarkers that had also been ELISA-validated. Finally, it would be important that the identified biomarker is informative in at least two independent LN cohorts. Such a focused literature search has uncovered 32 urinary proteins that reproducibly (and significantly) distinguish ALN from inactive SLE controls, across cohorts, above a 2-fold change level, or with ROC AUC values exceeding 0.60.

There are several potential contexts of use for these validated biomarkers. Not only are these urine proteins highly diagnostic of ALN, most also correlate with disease activity, as measured by SLEDAI, rSLEDAI, PGA or proteinuria, as detailed above. Thus, in a patient diagnosed with SLE but without renal involvement, serial monitoring of these biomarkers may help signal the very first occurrence of renal disease. Along the same vein, in a patient with LN, these proteins may be useful for monitoring oncoming renal flares. Indeed some of the short-listed proteins have already been documented to predict renal flares, including CD163, ICAM-1, and TWEAK. Unfortunately, most of the proteins in Table 1 have not been tested for this context of use.

Since urine is easily obtained, and urine testing readily lends itself to home-based monitoring using lateral flow test strips [96], frequent serial monitoring is feasible. However, more clarity is needed regarding which of the 32 proteins to use. A key task ahead for the field is a systematic head-to-head comparison of the biomarker candidates in Table 1, using longitudinal SLE cohorts where pre- and post-renal-involvement urine samples are available, and in well annotated LN cohorts where pre-flare and flare urine samples are available. These systematic studies should also address whether the biomarker candidates outperform current yardsticks in identifying ALN; in this context, only a couple of the proteins in Table 1 have been documented to outperform current laboratory tests such as anti-DNA and C3/C4, notably urine ALCAM, CD163, MCP-1, TWEAK, and TMP-1. Whether the identified biomarkers are additive with current clinical and serologic markers associated with LN activity also warrants investigation.

A third potential context of use is to use the identified proteins to monitor treatment response in LN. Indeed, all proteins from Table 1 that have been tested in this context have demonstrated reduced urine levels in treatment responders compared to non-responders, including urine CD163, LCN2/NGAL, MCP-1, TWEAK and VCAM-1. Again, the remaining proteins in Table 1 should also be systematically assessed for this specific context of use.

Several of the proteins in Table 1 have been shown to reflect concurrent renal pathology, notably proliferative LN and/or renal pathology activity (as marked by the AI score), including urine ALCAM, CD163, and VCAM-1. How the other proteins may relate to renal pathology also needs to be investigated. Identifying urine proteins that can reliably reflect concurrent renal pathology AI or CI, may fill a key void in this field – the need for non-invasive monitoring of evolving renal pathology in LN. These urine biomarkers could potentially be useful for monitoring histological response to treatment, thus avoiding the need for repeat renal biopsies.

For all of the above potential contexts of use, systematic testing of the shortlisted candidates will not only help identify the biomarkers with the best diagnostic metrics (ROC AUC, sensitivity, specificity, NPV, PPV), but also allow the formulation of biomarker panels that offer improved diagnostic performance, compared to single biomarkers. Two examples of such panels using a couple of urine proteins drawn from Table 1 have been reported in LN [17, 23, 27, 42]. This effort should be expanded further incorporating all qualifying proteins from Table 1, in order to identify the optimal panels for each potential context of use in LN. Though several of the biomarkers listed in Table 1 were first reported more than a decade ago, they have not been translated into clinical tests yet, pointing to the barrier between basic research and clinical translation. A concerted effort by academia, industry, NIH and non-profit lupus organizations is warranted to systematically test the shortlisted biomarker candidates in large well documented SLE/LN cohorts with longitudinal urine collections.

Finally, the identified proteins also point to potential pathogenic mechanisms underlying LN, as summarized in Figure 3. Studies are warranted to investigate the expression of these proteins within specific intra-renal cells in LN. The relevance of the implicated functional pathways to disease also needs to be investigated using preclinical models, as has been reported for ALCAM [8, 13–15], CCL2 (MCP-1) [8, 22–25, 27–30, 32, 33, 82], ICAM-1 [8, 16–18], CFP (Properdin) [8], and TNFSF12 (TWEAK) [8, 25, 30, 38, 39]. Targeting selected functional pathways that are implicated as biomarkers represents a fruitful approach to discovering novel therapeutics for LN, although therapeutics effective in murine LN may not always translate well to human LN, for a variety of reasons, as exemplified by the failed trials targeting TWEAK [97].

A couple of limitations are acknowledged. First, since raw data was not available from the contributing primary studies, and since different assay platforms (including different ELISA manufacturers) were used across the different studies, we could not conduct a meta-analysis to assess the global estimates of the index tests, or to generate summary ROC curves for the examined biomarkers, or to examine the underlying heterogeneity across the different studies and reasons for the underlying differences – such as the use of different populations, different diagnostic platforms, potential biases, measurement errors, etc. For the same reason, we were not able to identify or recommend any specific diagnostic cut-offs or metrics (sensitivity, specificity, NPV, PPV) for any of the listed biomarkers. Second, the use of two-fold change and ROC AUC > 0.60 as cut-offs may have been too stringent, and relaxing these thresholds may yield additional candidates for evaluation. Third, although most of the included studies had used the same criteria for ALN, variation was noted in defining iSLE; this could potentially have led to the omission of some potential candidates from Table 1. Finally, since raw data was not available, the potential impact of confounders could not be assessed. This is indeed important given that ethnic variations in some of these biomarker levels have been documented [8], while some of them (e.g., CD163) can directly be induced by glucocorticoids [98].

Concluding Remarks

This review has screened thousands of urine biomarkers reported in the LN literature to identify candidates that may hold the highest clinical potential, based on independent validation across multiple cohorts. The 32 reproducibly elevated urinary proteins reported here, individually, or in combination, promise to serve as a liquid biopsy for LN. Besides representing potential candidates for diagnostic, monitoring, predictive, and prognostic biomarkers in LN, they also provide a window into potential intra-renal processes that may be responsible for driving LN. Ongoing advances in proteomics that offer even wider proteome coverage at increased sensitivity are likely to profoundly reshape our perspective of urinary biomarkers for LN over the coming years.

Supplementary Material

• S1. List of Abbreviations (pdf)

• S2. PRISMA 2020 Checklist (pdf)

• S3. QUADAS 2 Report (pdf)

Supplementary information is available at KI Report’s website.

Data Sharing Statement

Any data generated and any other material in this review are available from the corresponding author upon reasonable request.