Soluble urine activated leukocyte cell adhesion molecule is a strong predictor of lupus nephritis

Abstract

Objectives

To evaluate urinary activated leucocyte cell adhesion molecule (ALCAM) and CD6 as predictors of LN progression or disease resolution across a 1-year study.

Methods

Serum and urine samples from biopsy proven LN subjects (n = 122) were prospectively collected over the course of a year at 3- or 6-month intervals (weeks 0, 12, 26 and 52) across multiple study sites and assessed for soluble ALCAM and CD6 levels. Urine creatinine from the same urine sample was used to normalize the levels of urinary ALCAM and urinary CD6. Measured levels of serum and urine ALCAM and CD6 were then analysed against disease metrics cross-sectionally and longitudinally.

Results

Cross-sectional analysis at baseline revealed that urinary ALCAM significantly correlated with urine protein creatinine ratio, renal SLEDAI, and the Physician Global Assessment (PGA), and negatively correlated with serum C3 and C4. Receiver operating characteristic curve analysis demonstrated that urinary ALCAM is a predictor of LN with an area under the curve (AUC) of 0.97, compared with urinary CD6 with an AUC of 0.71. Importantly, the change in urinary ALCAM over a 3-month period distinguished between non-responders and responders at week 52.

Conclusion

Urinary ALCAM is reflective of changes in LN and may be predictive of response status.

Rheumatology key messages.

• Urinary ALCAM is elevated in subjects with LN and correlates with commonly used disease metrics, as well as the renal activity index.

• The elevation of urinary ALCAM is selective and not due to non-specific proteinuria.

• Changes in urinary ALCAM over time, but not CD6, may be predictive of response status.

Introduction

LN is a serious complication of SLE mediated by the deposition of immune complexes in the kidney that can lead to impaired kidney function and kidney failure [1]. Currently, diagnosis of LN includes blood and urine tests, with kidney biopsies being the definitive test for determining LN class and treatment [2]. Identifying and validating non-invasive biomarkers may enable treatment at an earlier stage, delaying the progression to kidney failure [3]. Mechanistic studies utilizing two established mouse models of SLE/LN have demonstrated that the CD6–ALCAM pathway promotes LN through T cell-mediated responses [4]. Expression of CD6 (on T cells) and ALCAM (on antigen-presenting cells and tubular cells) is upregulated in the kidneys in LN models, and blockade of the pathway significantly reduces disease severity, indicating an important role for CD6–ALCAM interactions in the pathogenesis of LN [4, 5].

In humans, soluble urinary ALCAM (uALCAM) is significantly elevated in patients with LN [6–8]. Previous studies have established the potential of uALCAM to distinguish between patients with and without renal involvement [8, 9]. However, longitudinal studies to assess the utility of urine proteins as LN biomarkers are limited. The Accelerating Medicines Partnership (AMP) Lupus Network is a large consortium working to uncover the biological pathways that play a role in disease. To accomplish this, AMP investigators prospectively collected urine, serum and kidney biopsies from patients with active LN across a 1-year study [10]. This carefully standardized biorepository allows for the longitudinal tracking of soluble ALCAM and CD6 at weeks 0 (V0), 12 (V1), 26 (V2) and 52 (V3) that can be correlated with measures of disease activity. While urine protein creatinine ratio (UPCR) is a pharmacodynamic biomarker of LN, its utility to inform real-time clinical decisions is limited [11, 12]. Hence, identification of biomarkers that change faster than UPCR to provide decision-making information is critical to improving patient outcomes. Here, we evaluate and provide evidence for using uALCAM as a non-invasive biomarker that can be used to not only monitor LN and distinguish between patient populations, but potentially predict response to treatment.

Methods

Serum and urine sample collection

Samples were acquired from subjects that qualified for the AMP phase 2 study of SLE/LN [10]. Serum and urine samples were obtained from patients with biopsy proven LN and healthy controls across multiple sites. Follow-up longitudinal samples at weeks 0 (V0), 12 (V1), 26 (V2), and 52 (V3) were available for a subset of patients. All subjects provided written informed consent.

Experimental details concerning soluble ALCAM and CD6 detection, Quantibody array, urine creatinine measurement, and the statistical methods employed in the analysis can be found in the Supplementary Data S1, available at Rheumatology online.

The study protocol was approved by the institutional review boards and ethics committees of participating sites in adherence with the Declaration of Helsinki, including the Institutional Review Board of the Albert Einstein College of Medicine (no. 2014-4079).

Results

Urinary ALCAM and CD6 levels are elevated in LN cases

Urine samples from 269 subjects with LN and 70 healthy controls, and serum samples from 252 subjects with LN and 61 healthy controls were obtained from individuals that qualified for phase 2 of the AMP SLE/LN study at the first visit (V0). Levels of uALCAM and urinary CD6 (uCD6) were normalized to urine creatinine from the same urine sample. Detailed demographic and clinical data are provided in Table 1. Analysis of urine and serum levels revealed that both uALCAM (P < 0.0001) and uCD6 (P < 0.0001) were significantly elevated in LN subjects over healthy controls (Fig. 1A and D). A significant difference was also detected in serum ALCAM (sALCAM) levels between LN and control subjects, while serum CD6 (sCD6) levels were not significantly different (Fig. 1A and D). Hence, the increase in uALCAM and uCD6 levels is indicative of an increase in renal activity of the CD6–ALCAM pathway in subjects with active LN.

Table 1.

Baseline demographics for subjects included in analyses

	Healthy controls a  (n = 70)	LN cases a  (n = 269)	P-value (healthy controls vs LN cases)	Subjects with at least three visits b  (n = 122)	P-value (LN cases vs subjects with at least three visits)
Sex, n (%)
Male	30 (42.86)	35 (13.01)	<0.0001	16 (13.11)	1.00
Female	40 (57.14)	232 (86.25)	<0.0001	106 (86.89)	1.00
Unknown	0 (0.00)	2 (0.74)	1.0	0 (0.00)	1.00
Age, mean (s.d.), years	46.64 (12.22) (n = 70)	37.03 (11.98) (n = 267)	<0.0001	36.21 (11.03) (n = 122)	0.52
Race, n (%)
Asian	1 (1.43)	40 (14.87)	0.0008	19 (15.57)	0.88
Black/African American	4 (5.71)	117 (43.49)	<0.0001	57 (46.72)	0.58
White	64 (91.43)	83 (30.86)	<0.0001	36 (29.51)	0.81
Other	1 (1.43)	32 (11.90)	0.0057	11 (9.02)	0.49
Ethnicity, n (%)
Hispanic or Latino	0 (0.00)	79 (29.37)	<0.0001	35 (28.69)	1.00
Not Hispanic	66 (94.29)	186 (69.14)	<0.0001	87 (71.31)	0.72
Unknown	4 (5.71)	4 (1.49)	0.060	0 (0.00)	0.31
ISN LN class, n (%)
No LN, I, II	NA	33 (12.27)	NA	2 (1.64)	0.0002
III	NA	46 (17.10)	NA	24 (19.67)	0.57
IV	NA	42 (15.61)	NA	22 (18.03)	0.56
Mixed (III+V, IV+V)	NA	58 (21.56)	NA	36 (29.51)	0.098
V	NA	61 (22.68)	NA	37 (30.33)	0.13
VI	NA	11 (4.09)	NA	1 (0.82)	0.11
Unknown	NA	18 (6.69)	NA	0 (0.00)	0.0013
dsDNA, n (%)
Positive	NA	163 (60.59)	NA	82 (67.21)	0.22
Negative	NA	79 (29.37)	NA	37 (30.33)	0.91
Unknown	NA	27 (10.04)	NA	3 (2.46)	0.0074
Serum C3c, mean (S.D.), mg/dl	NA	80.28 (35.82) (n = 255)	NA	73.23 (35.38) (n = 120)	0.075
Serum C4c, mean (S.D.), mg/dl	NA	15.83 (9.88) (n = 253)	NA	14.26 (9.19) (n = 120)	0.14
UPCRc, mean (S.D.), ratio	NA	2.48 (1.20) (n = 249)	NA	2.31 (1.85) (n = 114)	0.42
SELENA-SLEDAIc, mean (S.D.)	NA	10.92 (6.080) (n = 269)	NA	12.23 (5.96) (n = 122)	0.048
Renal SLEDAIc, mean (S.D.)	NA	6.65 (3.63) (n = 269)	NA	7.21 (3.44) (n = 122)	0.15
Medicationsd, n (%)
None	NA	38 (14.13)	NA	9 (7.38)	0.065
Azathioprine	NA	29 (10.78)	NA	16 (13.11)	0.50
Calcineurin inhibitor	NA	15 (5.58)	NA	9 (7.38)	0.50
Cyclophosphamide	NA	5 (1.86)	NA	1 (0.82)	0.67
Mycophenolate	NA	129 (47.96)	NA	66 (54.10)	0.28
Steroid	NA	183 (68.03)	NA	89 (72.95)	0.35
Other	NA	10 (3.72)	NA	3 (2.46)	0.76

Clopper–Pearson exact method used to calculate P-values between healthy controls vs LN cases at baseline and between LN cases at baseline vs LN subjects with at least three visits for categorical variables and t-test for age as a continuous variable.

^aSubjects with baseline value for urinary ALCAM.

^bSubjects with baseline and at least two follow-up values for urinary ALCAM.

^cC3 and C4: complement components; UPCR: urine protein creatinine ratio; SLEDAI: SLE Disease Activity Index.

^dSubjects receiving more than one medication are included under each medication being received and are counted more than once under the medications category; a total of 269 subjects is used as the denominator to calculate percentage of total. Values highlighted in bold indicate statistical significance (P < 0.05). ALCAM: activated leucocyte cell adhesion molecule; LN: lupus nephritis; NA: not applicable.

Figure 1.

uALCAM is elevated in LN. (A and D) uALCAM (labelled as ALCAM), sALCAM and uCD6 (labelled as CD6) are elevated in LN. (B) Correlation of uALCAM and sALCAM at V0. (C) uALCAM decreases while sALCAM remains relatively stable. (E) uALCAM is elevated in mixed LN (n = 58) relative to early-stage disease (n = 33) and proliferative LN (n = 88). (F) No differences were observed with uCD6. (G) UPCR is elevated in mixed LN (n = 57) and membranous LN (n = 60) relative to early-stage disease (n = 30). (H–J) Correlation plots between uALCAM, uCD6 and UPCR with the activity and chronicity indices. (K and L) Percentage of cells positive for ADAM17 (K) and ALCAM (L) between controls and LN. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ADAM17: a disintegrin and metalloproteinase 17; ALCAM: activated leucocyte cell adhesion molecule; LN: lupus nephritis; s: serum; u: urinary; UPCR: urine protein creatinine ratio

Since LN subjects had increased levels of sALCAM, it is important to consider whether the high uALCAM levels are simply a reflection of the serum levels. To address this possibility, we performed a correlation analysis between uALCAM vs sALCAM at the first visit and indeed found a correlation (Fig. 1B). Nevertheless, when examined longitudinally, uALCAM decreased over time whereas sALCAM levels remained relatively stable (Fig. 1C). Moreover, while there is a 1.3-fold increase in sALCAM in LN subjects compared with controls, uALCAM levels increased 18.8-fold (Fig. 1A). The discrepant fold changes between LN subjects and controls for sALCAM and uALCAM is consistent with a renal-specific mechanism for the LN-associated increase in uALCAM. We also found a significant correlation between uCD6 vs sCD6 with similar longitudinal tendencies (Supplementary Fig. S1, available at Rheumatology online). However, when comparing the trends, the relationship between serum and urine CD6 over time appears to be tighter than that between serum and urine ALCAM. Finally, if changes in the CD6–ALCAM pathway proteins in the urine were solely dependent on serum levels, both ALCAM and CD6 would be expected to correlate with LN, although this was not the case. Altogether, while the changes in uALCAM may be influenced by serum levels to a degree, it is likely the changes in LN severity are predominantly responsible for driving the levels of uALCAM.

Urinary ALCAM is significantly elevated in patients with mixed LN

We next evaluated the associations between soluble levels of ALCAM and CD6 and the International Society of Nephrology (ISN) histopathological classifications of the kidney biopsy tissue. Analysis of uALCAM revealed that subjects with mixed LN (class III+V or IV+V) exhibited significantly higher levels of the protein in the urine compared with subjects without renal involvement or with early-stage disease (no LN, class I or class II) (P < 0.001) or proliferative LN (class III or class IV) (P < 0.01) (Fig. 1E). In contrast, uCD6 was unable to distinguish between the different classes of LN (Fig. 1F), suggesting that while there is an overall increase in activity of the CD6–ALCAM pathway in LN (Fig. 1A and D), uALCAM is most elevated in mixed LN.

Subjects with mixed LN had significantly higher UPCR compared with subjects with early-stage disease (P < 0.001) and subjects with proliferative LN (P < 0.05). Similar to uALCAM, UPCR was significantly different between subjects with mixed LN and proliferative LN (P < 0.05) (Fig. 1G). When the same analysis was performed with serum ALCAM and CD6 levels, no significant differences amongst the subjects with different ISN classifications were observed (data not shown). Hence, the strongest differences amongst the biopsy classes were observed with uALCAM and not uCD6, suggesting that uALCAM is selectively elevated in LN subjects and could help distinguish between LN classes.

To further explore the association of ALCAM with renal pathology, we performed additional analyses examining the correlation of uALCAM with the activity and chronicity indices. There was a significant positive correlation between uALCAM (at the first visit) and the activity index, while there was a significant negative correlation between uALCAM and the chronicity index (Fig. 1H). We performed the same analyses with uCD6 but did not observe a significant relationship between uCD6 and the activity index although a negative correlation with the chronicity index was present (Fig. 1I). Lastly, we performed correlation analyses between UPCR and the activity and chronicity indices but these did not reach statistical significance (Fig. 1J).

Increased expression of ALCAM and ADAM17 is detected in kidney cells of LN subjects

To identify the source of the elevated levels of uALCAM in LN subjects, single-cell RNA sequencing data were obtained from kidney biopsies at baseline. Expression of ALCAM and ADAM17, the protease that cleaves ALCAM [13], was evaluated on the different types of kidney cells and compared between controls and LN subjects (Fig. 1K and L). The percentage of cells positive for ADAM17 was significantly elevated in the glomeruli (P < 0.05) and loop of Henle (P < 0.0001) of LN subjects. The percentage of cells positive for ALCAM in the distal tubule of LN subjects was significantly higher compared with controls (P < 0.0001). In contrast, the percentage of cells positive for ALCAM in the glomeruli, loop of Henle and proximal tubule was not significantly different. The increase in cells positive for ADAM17 in resident kidney cells suggests that the increase in soluble ALCAM detected in the urine of LN subjects is a result of ALCAM cleavage and not non-specific proteinuria, which is a general loss of the glomerular filtration barrier due to kidney damage.

Urinary ALCAM trends longitudinally with regression in LN

LN subjects with at least three visits were included in a longitudinal analysis in which uALCAM and uCD6 levels were followed for the same subject and compared against available measures used to monitor LN. Both uALCAM and uCD6 showed a decreasing trend as UPCR and renal SLEDAI decreased (Fig. 2A and B). Furthermore, there was an inverse relationship between uALCAM and C3 and C4 as well as between uCD6 and serum C3 and C4 (Fig. 2C and D), suggesting that over time the activity of the CD6–ALCAM pathway trends downwards with therapy-related improvements in nephritis and disease activity.

Figure 2.

Urinary ALCAM and CD6 trends longitudinally with LN. Subjects with a urine baseline sample and at least two follow-up samples were included in a longitudinal analysis where urinary ALCAM and CD6 were analysed against measures of LN including (A) UPCR, (B) renal SLEDAI, (C) C3, and (D) C4 complement proteins. Both urinary ALCAM and CD6 trends with UPCR and renal SLEDAI while there is an inverse relationship between urinary ALCAM and CD6 with C3 and C4. The number of available subjects is indicated above each visit. Datapoints denote mean, and bars denote standard error. ALCAM: activated leucocyte cell adhesion molecule; LN: lupus nephritis; UPCR: urine protein creatinine ratio

Urinary ALCAM correlates with measures of LN

Next, we assessed the correlations between uALCAM and uCD6 with measures of disease activity including UPCR and renal SLEDAI at V0. uALCAM significantly correlated with UPCR [r(257) = 0.58, P < 0.0001] and renal SLEDAI [r(249) = 0.25, P < 0.0001] (Fig. 3A and B). In contrast, no significant correlations were detected between uCD6 and UPCR (Fig. 3C and D). The significant correlations between uALCAM, and not uCD6, with standard measures of LN at V0 further demonstrate the association between uALCAM and LN, suggesting that monitoring changes in uALCAM may be informative about the changes in disease.

Figure 3.

Urinary ALCAM correlates with disease parameters and is associated with UPCR. (A–D) uALCAM correlated with UPCR (P < 0.0001) (A) and renal SLEDAI (P < 0.0001) (B); uCD6 did not correlate with UPCR (P = 0.36) (C) nor renal SLEDAI (P = 0.054) (D) as determined by Spearman’s rank correlation. (E) Receiver operating characteristic (ROC) curve shows that uALCAM (left) is a better predictor of active LN compared with uCD6 (right). (F) A mixed model for repeated measures was used to observe the patterns of change in UPCR and their associations with fixed effects. The fit of the model revealed that uALCAM (P < 0.0001) at baseline significantly associated with UPCR. Values highlighted in bold indicate statistical significance (P < 0.05). ALCAM: activated leucocyte cell adhesion molecule; DF: degrees of freedom; LN: lupus nephritis; UPCR: urine protein creatinine ratio; pr: proteinuria

Urinary ALCAM is a predictor of active LN

The sensitivity and specificity of uALCAM and CD6 at baseline to distinguish between healthy controls and LN subjects was assessed using ROC curves. uALCAM is a more specific and sensitive predictor of LN compared with uCD6 with an area under the curve of 0.97 compared with 0.71 (Fig. 3E). This is in line with previous publications demonstrating the sensitivity and specificity of uALCAM cross-sectionally across SLE/LN patient cohorts [6, 7]. Hence, while an increase in uCD6 was observed at baseline in LN subjects (Fig. 1D), further analyses revealed here that uCD6 is less sensitive at distinguishing between healthy controls and LN subjects compared with uALCAM.

To further evaluate the associations between uALCAM and metrics used to monitor LN, a mixed model for repeated measures analysis was used to observe the patterns of change in UPCR from baseline over time, with UPCR as the dependent variable [14, 15]. The fit of the model revealed that in addition to the significant effect of post-V0 visits (which was most likely driven by treatment), C3 (P = 0.024), C4 (P = 0.025), Physician Global Assessment (PGA) (P = 0.0007) and baseline uALCAM (P < 0.0001) were significantly associated with UPCR (Fig. 3F). This demonstrates that while uALCAM is significantly associated with UPCR the elevated levels of uCD6 are not, suggesting that uALCAM is specific to changes in disease.

Urinary ALCAM correlates with lupus indicators

Proteinuria is routinely used to follow LN progression or response to treatment. Hence, the correlation of uALCAM with other LN metrics was compared with that of UPCR. First, both UPCR [r(191) = 0.34, P < 0.0001] and uALCAM [r(198) = 0.33, P < 0.0001] significantly correlated with the PGA (Fig. 4A) at baseline. However, UPCR did not correlate with either C3 (Fig. 4B) nor C4 (Fig. 4C). In contrast, uALCAM negatively correlated with both C3 [r(255) = −0.22, P < 0.001] (Fig. 4B) and C4 [r(253) = −0.14, P < 0.05] (Fig. 4C). This suggests that as uALCAM decreases there is less active disease as measured by the decrease in consumption of both serum C3 and C4 complement proteins, a change that was not detected by alterations in the UPCR. Lastly, no significant correlations were detected between uALCAM and anti-dsDNA antibody titres or between UPCR and anti-dsDNA antibody titres (Fig. 4D). Since urine creatinine levels can fluctuate as a response to LN treatment (including corticosteroids), especially in patients with low estimated glomerular filtration rate (eGFR), a separate analysis was conducted after removing subjects with low eGFR (<30 ml/min/1.73 m²) to evaluate whether normalizing uALCAM to urine creatinine is driving the effects observed with uALCAM. There were minimal changes to the correlations between uALCAM and PGA, C3 and C4 after removing the patients with low eGFR (data not shown), suggesting that the changes observed with uALCAM are a result of changes in the soluble levels of the protein and not due to fluctuations in urine creatinine.

Figure 4.

Urinary ALCAM correlates with LN indicators. The relationship between UPCR with the Physician Global Assessment (PGA), C3 and C4 complement proteins, and dsDNA was assessed and compared with that of urinary ALCAM. (A) Both UPCR and urinary ALCAM significantly correlated with PGA. (B and C) No correlations between UPCR and C3 or UPCR and C4 were detected while urinary ALCAM negatively correlated with both C3 and C4 complement proteins. (D) Neither UPCR nor urinary ALCAM correlated with anti-dsDNA antibody levels. Correlations determined by Spearman’s rank correlation. Values highlighted in bold indicate statistical significance (P < 0.05). ALCAM: activated leucocyte cell adhesion molecule; LN: lupus nephritis; UPCR: urine protein creatinine ratio

Change in urinary ALCAM is indicative of response status

To evaluate if uALCAM is predictive of response to treatment, the change in the urine proteins for the same subject was determined and related to their response status at the end of the study (week 52). In addition, the change in UPCR between the indicated time points was determined and used as a comparator. Subject response status, i.e. no response (NR), partial response (PR) or complete response (CR), was defined at week 52 as follows: CR = UPCR <0.5, normal serum creatinine (≤1.3 mg/dl) or, if abnormal, ≤125% of baseline, and prednisone ≤ 10 mg/day; PR = UPCR >50% reduction from baseline, normal serum creatinine or, if abnormal, ≤125% of baseline, and prednisone ≤15 mg/day; NR = not meeting the previous criteria [16]. The change in uALCAM and UPCR over a 3-month period was determined between weeks 0 and 12 (Fig. 5A, C and D) and between weeks 12 and 26 (Fig. 5B, E and F). A decrease in both uALCAM and UPCR over the 1-year study could distinguish between NR vs PR, and between NR vs CR (Supplementary Fig. S2, available at Rheumatology online). A 3-month change in uALCAM between weeks 0 and 12 could distinguish between NR and PR (P < 0.01) and between NR and CR (P < 0.05) (Fig. 5C) while a change in UPCR over the same period was only able to distinguish between NR and CR (P < 0.05) (Fig. 5D). A 3-month change in uALCAM between weeks 12 and 26 was able to distinguish between NR and PR (P < 0.001) (Fig. 5E) while a change in UPCR over the same period did not predict response status at week 52 (Fig. 5F). The same analyses were performed for uCD6 and no significant differences in uCD6 amongst response statuses were detected (data not shown). These data suggest that changes in uALCAM, and not uCD6 or UPCR, may be predictive of response to treatment in LN.

Figure 5.

Reductions in urinary ALCAM is predictive of response status. (A and B) Schematic illustration of the 3-month intervals evaluated. (C) A change in urinary ALCAM between weeks 0 and 12 distinguished NR from PR (P < 0.01) and CR (P < 0.05). (D) A change in UPCR between weeks 0 and 12 distinguished NR from CR (P < 0.05). (E) A change in urinary ALCAM between weeks 12 and 26 distinguished NR from PR (P < 0.001). (F) A change in UPCR between weeks 12 and 26 did not distinguish amongst response statuses. One-way ANOVA with multiple comparisons. (G) CR experienced a decrease in urinary ALCAM over the course of the 1-year study. *P < 0.05, **P < 0.01, ***P < 0.001. ALCAM: activated leucocyte cell adhesion molecule; CR: complete response; LN: lupus nephritis; NR: no response; PR: partial response; UPCR: urine protein creatinine ratio

Finally, to evaluate the selectively of uALCAM as a biomarker and confirm that increased uALCAM in LN is not due to damage to the glomerular filtration barrier, which would allow soluble proteins to pass into the urine in a non-specific manner, the change in abundance of uALCAM was compared with the change of selected proteins based on response status at week 52. ALCAM and CD6 are similar in size (105 kDa) but while an early and consistent decrease in ALCAM abundance is detected in CR (Fig. 5G), the same trend is not observed with CD6 (Supplementary Fig. S3, available at Rheumatology online). Additionally, the abundance of E-cadherin (110 kDa) and CD7 (40 kDa) was assessed [17–19]. The protein abundance of E-cadherin and CD7 remained constant for the duration of the study regardless of the subjects’ response status (Supplementary Fig. S3, available at Rheumatology online). Therefore, uALCAM is most likely reflective of active kidney inflammation in LN, rather than being due to an unselective loss through the glomerular filtration barrier (which would likewise affect similar sized or smaller proteins), and this highlights the selectivity and utility of uALCAM as a biomarker that can be used to monitor LN regression/progression.

Discussion

Currently, it is common practice to perform a kidney biopsy for the diagnosis of LN and to determine the severity of disease. However, this process is burdensome to the patient, has associated morbidity that can be significant, and the biopsy may be performed after fibrotic change and irreversible renal damage has occurred [3]. Hence, there is an unmet need for less invasive biomarkers that can not only monitor the course of disease but predict response to treatment. Previous studies have identified uALCAM as a urine biomarker that has shown high diagnostic accuracy for renal pathology in active LN patients [6–8] and more recently has been used to develop a novel technology for home monitoring of LN [20].

In this large multicentre cohort of LN subjects, we demonstrate that both uALCAM and uCD6 are significantly elevated in LN cases, providing additional evidence for the aberrantly high activity of the CD6–ALCAM pathway in LN. Moreover, the positive correlation between uALCAM and the activity index (the latter which reflects active inflammatory and possibly reversible lesions) not found for proteinuria, as well as the negative correlation with the chronicity index, further indicates that uALCAM is elevated with more active LN and that this biomarker may indicate currently active and potentially treatable disease.

The variations in the demographics between the LN cases and healthy controls (Table 1) reflect the characteristic epidemiology of SLE and LN, with over-representation of younger females from ethnic minorities. As for the minor differences between the total LN cohort and those LN patients with at least three visits (Table 1), these are likely driven by the presence of more severe disease in the latter (as reflected in the biopsy class and SLEDAI scores) requiring closer follow-up over longer periods of time.

The increased activity of the CD6–ALCAM pathway in LN is further supported by the longitudinal trends between both uALCAM and uCD6 and measures of LN including UPCR, renal SLEDAI, C3 and C4. While both proteins are elevated in LN cases, uALCAM is more sensitive and specific to distinguish between healthy controls and LN subjects compared with uCD6. Longitudinal modelling shows that uALCAM, and not uCD6, is associated with UPCR across the 1-year study. Cross-sectional analysis also revealed that uALCAM correlated with UPCR and the renal SLEDAI whereas uCD6 did not. This suggests that while both CD6 and ALCAM are elevated in the urine of LN subjects, only uALCAM is associated with disease severity and could be used as a urine biomarker to monitor LN disease and response to treatment. Furthermore, the elevation of ADAM17 in kidney cells of LN subjects provides evidence for the selective cleavage of ALCAM into LN urine.

ALCAM is a transmembrane glycoprotein that is increased in epithelial cells isolated from renal tissues of LN subjects [4]. In contrast, CD6 is primarily expressed on T cells and a subset of NK cells [21, 22]. Hence, the longitudinal trends observed with both uALCAM and uCD6 may be attributed to the sources of these soluble proteins and not non-specific proteinuria. The increase in both uALCAM and uCD6 suggests an increase in the activity of the CD6–ALCAM pathway, which has been successfully targeted to ameliorate LN disease in non-human models [4]. ALCAM can be actively shed through proteolytic cleavage by ADAM17/TACE from the cell surface [13]. Furthermore, activation of ADAM17-dependent TNF and eGFR signalling has been shown to mediate irreversible kidney damage [23]. This suggests that greater activity of ADAM17 and potentially an increase in soluble uALCAM may be indicative of the degree of kidney tissue damage. While additional mechanistic studies will need to be performed to identify the exact source(s) and form(s) of soluble ALCAM detected in the urine, evaluation of the transcript levels of ALCAM and ADAM17 suggests that expression of both is elevated in subjects with LN in select renal tissues. Hence, the increase in soluble uALCAM can be a result of two non-exclusive mechanisms, through the upregulation of ALCAM and/or ADAM17 on renal cells and/or immune-mediated damage of renal cells expressing high levels of ALCAM.

Renal infiltration by immune cells induces inflammation that can result in tissue damage before clinically measurable impact in renal function is discernible [24, 25]. Since CD6 is primarily expressed by immune cells, the increase in soluble uCD6 in LN patients may be attributed to the number of infiltrating immune cells and shedding of the receptor from the cell surface under inflammatory conditions [26]. However, further studies are warranted to correlate the level of uCD6 with T cell activity in LN.

Proteinuria is commonly used to follow LN as it is a measure of kidney damage and severity of disease. To demonstrate that uALCAM is selectively secreted into the urine, we showed that while uALCAM correlated with UPCR and renal SLEDAI, uCD6, which is similar in size to uALCAM, did not. Furthermore, while uALCAM negatively correlated with serum C3 and C4, UPCR did not. This suggests that uALCAM may be used to provide valuable insight into the degree of complement activation and consumption, and may provide information about aspects of disease that UPCR does not reflect.

One limitation of this study was that the participants were receiving multiple lines of treatment that were left to the discretion of the treating physician. Follow-up analyses will be required to investigate the response of uALCAM to different medication regiments. However, these data demonstrate that regardless of treatment, a decrease in uALCAM can be predictive of response status 6–9 months later. In contrast, a decrease in UPCR between weeks 0 and 12 could only differentiate between non- and complete responders; no differences were detected between non- and partial responders where uALCAM was able to differentiate these subsets. Moreover, the change in UPCR between weeks 12 and 26 did not differentiate between response statuses at week 52, while the change in uALCAM could differentiate between non- and partial responders. This suggests that the change in UPCR may require more time to be predictive of response status whereas uALCAM changes quickly and can be used as a measure to predict if patients are responding to treatment sooner. This demonstrates that monitoring uALCAM may inform clinicians about a patients’ response to treatment early on so that alternative options can be evaluated. While comparing uALCAM levels in patients with LN with other renal disorders might be of interest and could be addressed in a future study, our aim was to show that uALCAM is a biomarker in LN and demonstrate its utility in monitoring disease. The overall goal of the AMP program was to identify biomarkers and drug targets solely for SLE (and rheumatoid arthritis), and samples from renal disease patients of a non-lupus aetiology was not part of this carefully designed multicentre study. Nevertheless, our recent studies indicate that uALCAM is not a biomarker of diabetic nephropathy or other diseases of the urinary tract ([27] and unpublished observations).

In conclusion, uALCAM is a biomarker of LN that could be indicative of disease progression or regression. Another urine protein under investigation is CD163, a surface protein expressed on monocytes and macrophages, which has shown both correlative and predictive value [28–30]. Nevertheless, while individual urine biomarkers such as CD163 and CD206 had promising performance characteristics in LN [30], at the present time it is not likely any single biomarker would provide sufficient information for critical follow-up and treatment decisions. Future studies are necessary to evaluate whether uALCAM in combination with other urinary biomarkers can be used to develop a multiplexed panel that can better predict histology, the optimal drug for personalized treatment (efficacy and duration of response), the course of disease and the ultimate prognosis. While previous studies have identified uALCAM as a biomarker across subject cohorts [8, 31], this is the first study to show that early changes in uALCAM could be used as a tool to predict response to treatment. Hence, the identification of non-invasive and predictive biomarkers may allow for adjustments to interventions earlier on in the course of disease, potentially contributing to improving patient outcomes.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

The soluble ALCAM and CD6 datasets generated and analysed during the current study are available from the corresponding author upon reasonable request. The clinical datasets used with the current study are available from the NIH.

Contribution statement

D.C. and C.T.N contributed to data collection, data analysis, and wrote the paper. N.S. contributed to data collection. J.G. and J.J. contributed to data collection and study design. J.P.B., A.F., M.P. and C.M. contributed to data collection, data analysis, study design, and review of the paper. J.A., S.C. and M.F. contributed to data collection and data analysis. C.P. contributed to data collection, data analysis, study design, and wrote the paper.

Funding

Funding was provided through grants from the National Institutes of Health (UH2-AR067676, UH2-AR067677, UH2-AR067679, UH2-AR067681, UH2-AR067685, UH2- AR067688, UH2-AR067689, UH2-AR067690, UH2-AR067691, UH2-AR067694, and UM2- AR067678).

Disclosure statement: D.C., J.A., S.C., M.F. and C.T.N. receive income and stock from Equillium; J.P.B. served as a consultant to Equillium; C.M. and C.P. served as consultants and received research support from Equillium. The remaining authors have declared no conflicts of interest.