Abstract
Introduction
Lupus nephritis (LN) is kidney inflammation that commonly occurs from systemic lupus erythematosus. NF-κB activity is increased in LN patients, leading to elevated circulating concentrations of immune modulators that contribute to LN pathophysiology. This study sought to investigate this phenomenon with the aim of discovering novel serum biomarkers for LN.
Methods
A multiplex antibody-based assay was performed with serum from 24 LN patients and 7 healthy controls (HCs) to determine if 48 NF-κB-regulated proteins were elevated in LN patients. Confirmation ELISAs were performed on stem cell factor (SCF), macrophage colony-stimulating factor (M-CSF), and interleukin-2 receptor alpha (IL-2Rα) subunit in a separate sample cohort of 27 LN patients and 10 HC. Follow-up ELISAs were performed on samples obtained from the same patients during LN remission to determine if these candidates were reliable predictors of disease activity. Comparisons of protein levels between LN patients and HC were performed using a series of 2-tailed Mann-Whitney tests. Paired samples were analyzed using a Wilcoxon matched pairs test. Two-tailed Spearman’s correlation analyses were used to compare serum protein concentrations with LN clinical parameters. All p values were adjusted for multiple comparisons.
Results
SCF, M-CSF, and IL-2Rα were significantly elevated in LN serum. Elevated serum SCF and M-CSF were significantly correlated with elevated urine protein to creatinine ratio, decreased estimated glomerular filtration rate, and elevated serum creatinine. Elevated serum IL-2Rα was significantly correlated with elevated serum creatinine. Serum SCF concentration was significantly decreased during LN remission in paired samples from individuals, but it was not a good predictor at the population level (AUC = 0.6265).
Conclusion
We identified the NF-κB-regulated proteins SCF, M-CSF, and IL-2Rα as candidate serum biomarkers for consideration in monitoring LN activity. Our findings also indicate the importance for follow-up mechanistic studies pertaining to these inflammatory mediators.
Keywords: Systemic lupus erythematosus, Proliferative lupus nephritis, Serum biomarkers, Autoimmune disease, Nuclear factor-kappa B
Introduction
Lupus nephritis (LN) causes significant morbidity and mortality for patients with systemic lupus erythematosus (SLE) [1]. Both innate and adaptive immune responses are dysregulated, resulting in inflammation, reduced complement levels, and immune complex deposition in kidneys [2, 3]. Approximately 90% of those affected by SLE are women, predominately those of African and Latin American ancestry [4]. SLE etiology remains undetermined, but evidence suggests that genetic, hormonal, and environmental factors contribute to the inflammatory phenotype [5]. Currently, care includes monitoring, immune suppression, immune modulation, B-cell-targeted therapies, and supportive therapy for chronic kidney disease and kidney failure [2].
Nuclear factor-kappa B (NF-κB) is a transcription factor that regulates a multitude of processes, including inflammation, cell survival, differentiation, and proliferation [6]. Prior studies have demonstrated that NF-κB activation contributes to LN pathogenesis [7–10]. Moreover, inhibition of NF-κB activity is associated with improved renal outcomes [11, 12]. The central aim of this project was to characterize serum immune phenotypes associated with NF-κB activation in LN. We hypothesize that if NF-κB activation contributes to proliferative LN pathogenesis, then some proteins under its regulation should be elevated in active LN patients’ serum. Moreover, if protein concentrations fluctuate with disease activity, then they may provide additional biomarkers for tracking disease progression and response to treatment.
Methods
Human Samples
The University of Louisville Internal Review Board and/or the Oklahoma Medical Research Foundation approved all patient and healthy individual sample collections and studies. All LN patients had biopsy-proven proliferative LN (class III, IV, III/V, or IV/V). LN samples were obtained from an outpatient nephrology clinic at the University of Louisville and the outpatient rheumatology clinic at Oklahoma Medical Research Foundation. All patients were experiencing active LN at the time of sample collection used for Bio-Plex and confirmation ELISAs. Active LN was defined by proteinuria with a urine protein/creatinine ratio (UPCR) exceeding 500 mg/g, and inactive LN was defined by UPCR measurements below 500 mg/g at the time of donation. “Paired samples” refer to those collected from the same patient during both active and inactive LN at the time of sample collection. The demographic and clinical data for the LN patient cohorts used in experiments for each figure are included in Table 1. The prescribed medications for these patients are listed in Table 2. Healthy controls (HCs) had no underlying health conditions, were not taking any medications, and were not anemic on the day of collection. No other data were available for the HC.
Table 1.
LN clinical parameters organized by experimental design
| Bio-Plex (Fig. 1) | ||||
|---|---|---|---|---|
| active LN (n = 24) | average (SD) | median | maximum | minimum |
| Age at visit, years | 36.33 (11.38) | 34 | 56 | 22 |
| UPCR, mg/g | 2,573.88 (2,835.79) | 1,645 | 14,611 | 735 |
| eGFR, mL/min/1.73 m2 | 62.77 (35.23) | 62 | 137 | 15 |
| Serum creatinine, mg/dL | 1.72 (1.06) | 1.41 | 4.17 | 0.71 |
| dsDNA antibody titer, IU/mL | 120.79 (184.39) | 37 | 739 | 1 |
| C3, mg/dL | 94.47 (34.12) | 90 | 194 | 53 |
| C4, mg/dL | 23.33 (10.59) | 20 | 48 | 10 |
| Confirmation ELISAs (Fig. 2) | ||||
|---|---|---|---|---|
| active LN (n = 27) | average (SD) | median | maximum | minimum |
| Age at visit, years | 34.52 (10.22) | 31 | 63 | 22 |
| UPCR, mg/g | 2,707.94 (3,837.58) | 1,387 | 20,058 | 568 |
| eGFR, mL/min/1.73 m2 | 73.78 (32.63) | 68 | 133 | 21 |
| Serum creatinine, mg/dL | 1.29 (0.59) | 1.18 | 2.93 | 0.54 |
| dsDNA antibody titer, IU/mL | 240.39 (840.08) | 51 | 4,079 | 3 |
| C3, mg/dL | 92.12 (35.12) | 92 | 181 | 25 |
| C4, mg/dL | 22.87 (10.88) | 20 | 52 | 10 |
| Replicated samples (Figs. 1-2) | ||||
|---|---|---|---|---|
| active LN (n = 4) | average (SD) | median | maximum | minimum |
| Age at visit, years | 35 (10.65) | 35 | 48 | 22 |
| UPCR, mg/g | 3,181.25 (2,342.32) | 2,609.5 | 6,355 | 1,151 |
| eGFR, mL/min/1.73 m2 | 74 (34.46) | 74 | 108 | 40 |
| Serum creatinine, mg/dL | 1.29 (0.55) | 1.17 | 1.99 | 0.81 |
| dsDNA antibody titer, IU/mL | 75.5 (102.54) | 32 | 227 | 11 |
| C3, mg/dL | 103.25 (30.3) | 100.5 | 136 | 76 |
| C4, mg/dL | 25.5 (10.34) | 23.5 | 38 | 17 |
| Correlations (Fig. 3) | |||||
|---|---|---|---|---|---|
| active LN (n = 27), inactive LN (n = 18) | active LN: average (SD) | inactive LN: average (SD) | global median | global maximum | global minimum |
| Age at visit, years | 34.52 (10.22) | 35 (12.53) | 31 | 66 | 20 |
| UPCR, mg/g | 2,707.94 (3,837.58) | 205.61 (79.01) | 881 | 20,058 | 67 |
| eGFR, mL/min/1.73 m2 | 73.78 (32.63) | 81.56 (29.6) | 73 | 133 | 21 |
| Serum creatinine, mg/dL | 1.29 (0.59) | 1.15 (0.36) | 1 | 3 | 1 |
| dsDNA antibody titer, IU/mL | 240.39 (840.08) | 403.53 (1,453.75) | 31 | 6,035 | 2 |
| C3, mg/dL | 93.21 (35.44) | 100.72 (27.4) | 99 | 181 | 25 |
| C4, mg/dL | 23.14 (11.06) | 22.78 (9.93) | 21 | 52 | 6 |
| Paired samples (Fig. 4) | ||||
|---|---|---|---|---|
| active LN (n = 18) | average (SD) | median | maximum | minimum |
| Age at visit, years | 34.61 (11.18) | 31 | 63 | 22 |
| UPCR, mg/g | 1,401.11 (736.82) | 1,182.50 | 3,507 | 568 |
| eGFR, mL/min/1.73 m2 | 82.11 (30.89) | 85 | 133 | 43 |
| Serum creatinine, mg/dL | 1.10 (0.4) | 1.02 | 1.83 | 0.54 |
| dsDNA antibody titer, IU/mL | 50.67 (53.21) | 27 | 188 | 3 |
| C3, mg/dL | 91.94 (27.2) | 95.50 | 136 | 25 |
| C4, mg/dL | 22.2 (8.23) | 21 | 38 | 10 |
| Paired samples (Fig. 4) | ||||
|---|---|---|---|---|
| inactive LN (n = 18) | average (SD) | median | maximum | minimum |
| Age at visit, years | 35 (12.53) | 34 | 66 | 20 |
| UPCR, mg/g | 205.61 (79.01) | 196 | 375 | 67 |
| eGFR, mL/min/1.73 m2 | 81.56 (29.6) | 83.5 | 125 | 36 |
| Serum creatinine, mg/dL | 1.15 (0.36) | 1.06 | 1.77 | 0.71 |
| dsDNA antibody titer, IU/mL | 403.53 (1,453.75) | 15 | 6,035 | 2 |
| C3, mg/dL | 100.72 (27.4) | 102.5 | 174 | 43 |
| C4, mg/dL | 22.78 (9.93) | 22.5 | 44 | 6 |
Table 2.
Prescribed medications of LN patients organized by experimental design
| Bio-Plex (Fig. 1) | |
|---|---|
| medication | active LN (n = 24) |
| Prednisone | 16 (66.67%) |
| Hydroxychloroquine | 19 (79.17%) |
| Mycophenolate | 20 (83.33%) |
| Methotrexate | 0 |
| Azathioprine | 2 (8.33%) |
| Cyclophosphamide | 2 (8.33%) |
| Rituximab | 1 (4.17%) |
| Belimumab | 2 (8.33%) |
| Tacrolimus | 1 (4.17%) |
| Confirmation ELISAs (Fig. 2) | |
|---|---|
| medication | active LN (n = 27) |
| Prednisone | 16 (59.26%) |
| Hydroxychloroquine | 24 (88.89%) |
| Mycophenolate | 20 (74.07%) |
| Methotrexate | 1 (3.70%) |
| Azathioprine | 4 (14.81%) |
| Cyclophosphamide | 3 (11.11%) |
| Rituximab | 0 |
| Belimumab | 0 |
| Tacrolimus | 3 (11.11%) |
| Replicated samples (Figs. 1–2) | |
|---|---|
| medication | active LN (n = 4) |
| Prednisone | 2 (50%) |
| Hydroxychloroquine | 3 (75%) |
| Mycophenolate | 1 (25%) |
| Methotrexate | 0 |
| Azathioprine | 2 (50%) |
| Cyclophosphamide | 2 (50%) |
| Rituximab | 0 |
| Belimumab | 0 |
| Tacrolimus | 0 |
| Correlations (Figs. 3–5) | ||
|---|---|---|
| medication | active LN (n = 27) | inactive LN (n = 18) |
| Prednisone | 16 (59.26%) | 6 (33.33%) |
| Hydroxychloroquine | 24 (88.89%) | 18 (100%) |
| Mycophenolate | 20 (74.07%) | 13 (72.22%) |
| Methotrexate | 1 (3.70%) | 0 |
| Azathioprine | 4 (14.81%) | 3 (16.67%) |
| Cyclophosphamide | 3 (11.11%) | 0 |
| Rituximab | 0 | 1 (5.56%) |
| Belimumab | 0 | 2 (11.11%) |
| Tacrolimus | 3 (11.11%) | 4 (22.22%) |
| Paired samples (Fig. 6) | ||
|---|---|---|
| medication | active LN (n = 18) | inactive LN (n = 18) |
| Prednisone | 9 (50%) | 3 (16.67%) |
| Hydroxychloroquine | 17 (94.44%) | 18 (100%) |
| Mycophenolate | 12 (66.67%) | 13 (72.22%) |
| Methotrexate | 0 | 0 |
| Azathioprine | 4 (22.22%) | 3 (16.67%) |
| Cyclophosphamide | 2 (11.11%) | 0 |
| Rituximab | 0 | 1 (5.56%) |
| Belimumab | 2 (11.11%) | 2 (11.11%) |
| Tacrolimus | 5 (27.78%) | 4 (22.22%) |
Bio-Plex Pro Human Immune Modulator Screening Panel
Serum was collected from 7 HC individuals and 24 LN patients by centrifugation of whole blood at a relative centrifugal force of 800 for 10 min at 4°C. Samples were measured using the Bio-Plex Pro assay (Hercules, CA, USA) following the manufacturer’s instructions. Briefly, serum samples were incubated with magnetic beads that were conjugated to capture antibodies for the panel listed below. After a series of washes, the beads were subjected to a biotinylated detection antibody. Streptavidin-phycoerythrin was then added to the beads to form the detection complex. The fluorescent signal created by the detection complex was read using a Bio-Plex 200 Reader. The following NF-κB-regulated proteins were targeted: basic fibroblast growth factor (FGF), cutaneous T-cell-attracting chemokine (CTACK), eotaxin, granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), growth-regulated oncogene-alpha (GRO-α), hepatocyte growth factor (HGF), interferon-alpha two (IFN-α2), interferon-gamma (IFN-γ), interferon-gamma-inducible protein 10 (IP-10), interleukin-one alpha (IL-1α), interleukin-one receptor alpha (IL-1rα), interleukin-one beta (IL-1β), interleukin-two (IL-2), interleukin-two receptor alpha (IL-2Rα), interleukin-three (IL-3), interleukin-four (IL-4), interleukin-five (IL-5), interleukin-six (IL-6), interleukin-seven (IL-7), interleukin-eight (IL-8), interleukin-nine (IL-9), interleukin-ten (IL-10), interleukin-twelve p40 (IL-12p40), interleukin-twelve p70 (IL-12p70), interleukin-thirteen (IL-13), interleukin-fifteen (IL-15), interleukin-sixteen (IL-16), interleukin-seventeen alpha (IL-17), interleukin-eighteen (IL-18), leukemia-inhibitory factor (LIF), macrophage colony-stimulating factor (M-CSF), macrophage inflammatory protein-one alpha (MIP-1α), macrophage inflammatory protein-one beta (MIP-1β), macrophage migration inhibitory factor (MIF), monocyte chemoattractant protein 1 (MCP-1), monocyte chemoattractant protein 3 (MCP-3), monokine induced by interferon gamma (MIG), nerve growth factor beta (β-NGF), platelet-derived growth factor-BB (PDGF-BB), regulated upon activation, normal T-cell expressed and secreted (RANTES), stem cell factor (SCF), stem cell growth factor beta (SCGF-β), stromal cell-derived factor-1 alpha (SDF-1α), tumor necrosis factor alpha (TNF-α), tumor necrosis factor alpha-related apoptosis inducing ligand (TRAIL), tumor necrosis factor-beta (TNF-β), and vascular endothelial growth factor (VEGF).
Confirmation ELISAs
Serum was collected from 10 additional HC individuals and 27 proliferative LN patients by centrifugation of whole blood at a relative centrifugal force of 800 for 10 min at 4°C. LN patients were selected from the University of Louisville and the Oklahoma Medical Research Foundation biorepositories according to their activity status (UPCR ≥500 mg/g). The majority (n = 23) of samples measured in validation experiments were obtained from separate patients and/or collection dates than those used in the Bio-Plex array experiment. A small number (n = 4) of samples were remeasured from the Bio-Plex assay, based on limited sample availability and in the interest of confirming our preliminary data.
ELISA on Paired Active and Inactive LN Samples
Paired active disease and remission samples were analyzed when available. Some patients never entered clinical remission or the remission samples were not available in our biorepository. Eighteen of the 27 LN patients were included in the paired analyses. Inactive proliferative LN was determined by patients’ UPCR <500 mg/g. When an individual’s electronic medical record contained multiple dates with UPCR <500 mg/g, we chose the sample with the lowest double-stranded DNA (dsDNA) antibody titer and the highest complement levels.
Statistical Analyses
Quantitative and statistical analyses were performed using GraphPad Prism 10.1.1 and R 4.3.2. Comparisons of protein levels between LN patients and HC were performed using a series of 2-tailed Mann-Whitney tests. p values obtained from these calculations were adjusted for multiple comparisons using the Bonferroni correction method. Paired active and inactive LN samples were analyzed using a Wilcoxon matched pairs test with post hoc Bonferroni correction for multiple comparisons. Two-tailed Spearman’s correlation analyses were used to compare serum protein concentrations with clinical and socioeconomic characteristics from active LN patients, including UPCR, dsDNA antibody titer, C3, C4, serum creatinine, and estimated glomerular filtration rate (eGFR). The obtained p values were adjusted for false discovery rate using the Benjamini-Hochberg method. Logistic regression analyses from the data acquired by ELISA were used to model each protein independently and all combinations to determine which protein(s) provide the highest degree of binary predictability, indicated by low Akaike’s information criterion (AIC) values. Comparisons were made between LN patients with active disease compared to HCs (Table 3) and paired samples from LN patients during active disease and LN remission (Table 4). Optimal protein cutoff concentrations were determined using the cutpointr package in R.
Table 3.
Logistic regression models between active LN and HC for each protein combination
| Protein(s) in model: LN vs. HC | AIC value |
|---|---|
| SCF + IL-2Rα | 19.237 |
| SCF + M-CSF + IL-2Rα | 20.926 |
| SCF + M-CSF | 22.031 |
| SCF | 25.69 |
| IL-2Rα + M-CSF | 27.519 |
| M-CSF | 29.672 |
| IL-2Rα | 31.339 |
Models are organized from the best fit (top) to the worst (bottom) according to the AIC values.
LN, lupus nephritis; HC, healthy control; AIC, Akaike’s information criterion; SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; IL-2Rα, interleukin-2 receptor alpha.
Table 4.
Logistic regression models between active and inactive LN for each protein combination
| Protein(s) in model: active vs. inactive LN | AIC value |
|---|---|
| SCF | 52.073 |
| M-CSF | 53.61 |
| IL-2Rα | 53.759 |
| SCF + IL-2Rα | 54.046 |
| SCF + M-CSF | 54.073 |
| IL-2Rα + M-CSF | 55.574 |
| SCF + M-CSF + IL-2Rα | 56.036 |
Models are organized from the best fit (top) to the worst (bottom) according to the AIC values.
LN, lupus nephritis; HC, healthy control; AIC, Akaike’s information criterion; SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; IL-2Rα, interleukin-2 receptor alpha.
Results
SCF, M-CSF, and IL-2Rα were expressed more in LN serum than HC serum. A multiplex antibody-based array (Bio-Plex) was performed on 48 NF-κB-regulated proteins with serum from 24 active LN patients and 7 HCs. Twenty-five (25) of the 48 proteins measured from serum of patients with active LN and HC controls yielded results within the range of detection and contained enough data points to perform statistical analyses between LN and HC. Significantly different protein concentrations were found for SCF (Fig. 1a, p = 0.0004), M-CSF (Fig. 1b, p = 0.0004), and IL-2Rα (Fig. 1c, p = 0.0005).
Fig. 1.
Exploratory analysis of serum protein concentrations between active biopsy-proven lupus nephritis and HCs. SCF (a), M-CSF (b), and IL-2Rα (c) concentrations were measured in LN patients (n = 24) and HC (n = 7) using Bio-Plex. Adjusted p values <0.05 were considered statistically significant. HC, healthy control; LN, lupus nephritis; SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; IL-2Rα, interleukin-2 receptor alpha.
Confirmation ELISAs Support Bio-Plex Data
The results obtained during the initial Array-based screening were validated using a series of ELISAs, which confirmed that SCF, M-CSF, and IL-2Rα were expressed more in LN serum (n = 27) than HC serum (n = 10) (Fig. 2). These experiments provide a measure of reliability to the array data because the ELISAs use antibody against a single protein and included different sample time points from our University of Louisville LN cohort and additional samples from Oklahoma Medical Research Foundation LN cohort. Because we used samples from separate HCs and the data were contained within a narrow range, we can also assume that the variation in quantified protein expression is the result of the method used.
Fig. 2.
Confirmatory analysis of serum protein concentrations between active biopsy-proven lupus nephritis and HCs. SCF (a), M-CSF (b), and IL-2Rα (c) concentrations were measured in LN patients (n = 27) and HC (n = 10) using individual ELISAs for each protein. Adjusted p values <0.05 were considered statistically significant. HC, healthy control; LN, lupus nephritis; SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; IL-2Rα, interleukin-2 receptor alpha.
Serum SCF, M-CSF, and IL-2Rα Concentrations Correlated with Clinical LN Parameters
Spearman’s correlation analyses (Figs. 3–5; Table 5) were used to evaluate if serum SCF, M-CSF, and IL-2Rα concentrations from active and inactive LN patients were correlated with clinical LN parameters, including UPCR, eGFR, serum creatinine, dsDNA antibody titers, C3, and C4. After adjusting for multiple comparisons, we found significant correlations with regards to UPCR (SCF r = 0.56, p = 0.0001; M-CSF r = 0.38, p = 0.034), eGFR (SCF r = −0.80, p < 0.0001; M-CSF r = −0.35, p = 0.034), and serum creatinine (SCF r = 0.79, p < 0.0001; M-CSF r = 0.37, p = 0.034; IL-2Rα r = 0.41, p = 0.028). It was not surprising that SCF and M-CSF concentrations were correlated with eGFR and serum creatinine because the two parameters are inversely related. IL-2Rα was correlated with eGFR prior to adjusting for multiple comparisons, but this relationship was lost after the post hoc analysis was conducted. Given that we found significant correlations between serum protein concentrations and multiple clinical parameters, we then investigated if samples obtained from the same patient during times of active and inactive LN were different.
Fig. 3.
Correlation analyses between serum SCF concentrations and LN clinical parameters. Spearman’s rank coefficient (r) values were obtained between SCF and UPCR (a), eGFR (b), serum creatinine (c), dsDNA antibody titer (d), C3 (e), and C4 (f) in LN patients (n = 27 active LN, n = 18 inactive LN). Adjusted p values <0.05 were considered statistically significant. SCF, stem cell factor; UPCR, urine protein to creatinine ratio; eGFR, estimated glomerular filtration rate; dsDNA, double-stranded DNA; C3, complement component 3; C4, complement component 4.
Fig. 5.
Correlation analyses between serum IL-2Rα concentrations and LN clinical parameters. Spearman’s rank coefficient (r) values were obtained between IL-2Rα and UPCR (a), eGFR (b), serum creatinine (c), dsDNA antibody titer (d), C3 (e), and C4 (f) in LN patients (n = 27 active LN, n = 18, inactive LN). Adjusted p values <0.05 were considered statistically significant. IL-2Rα, interleukin-2 receptor alpha; UPCR, urine protein to creatinine ratio; eGFR, estimated glomerular filtration rate; dsDNA, double-stranded DNA; C3, complement component 3; C4, complement component 4.
Table 5.
Raw and Benjamini-Hochberg-adjusted p values from Spearman’s correlation analyses with serum protein concentrations collected within 6 months of renal biopsy (n = 7)
| Clinical parameter | Raw p values | Adjusted p values |
|---|---|---|
| SCF | ||
| UPCR | 0.00329926250 | 0.009897787 |
| eGFR | 0.00187700420 | 0.009897787 |
| Serum creatinine | 0.00565095400 | 0.011301908 |
| dsDNA ab titer | >0.9999999999 | 1.0 |
| C3 | 0.83822641090 | 1.0 |
| C4 | 0.19821428570 | 0.297321429 |
| IL-2Rα | ||
| UPCR | 0.18256774090 | 0.20490079 |
| eGFR | 0.20490079365 | 0.20490079 |
| Serum creatinine | 0.12290764791 | 0.18436147 |
| dsDNA ab titer | 0.04583333333 | 0.1375 |
| C3 | 0.01492945326 | 0.08957672 |
| C4 | 0.11626984127 | 0.18436147 |
| M-CSF | ||
| UPCR | 0.06561177249 | 0.7568452 |
| eGFR | 0.08516945246 | 0.1898461 |
| Serum creatinine | 0.09492303992 | 0.1898461 |
| dsDNA ab titer | 0.61909722222 | 0.1898461 |
| C3 | 0.68208884480 | 0.7568452 |
| C4 | 0.75684523810 | 0.7568452 |
Statistically significant correlations are highlighted in orange. SCF stands as the most reliable predictor of proteinuria and renal function because the correlations found in this subanalysis correspond with analyses performed on samples averaging 2.9 years from biopsy (range: 0–13.7 years).
Fig. 4.
Correlation analyses between serum M-CSF concentrations and LN clinical parameters. Spearman’s rank coefficient (r) values were obtained between M-CSF and UPCR (a), eGFR (b), serum creatinine (c), dsDNA antibody titer (d), C3 (e), and C4 (f) in LN patients (n = 27 active LN, n = 18 inactive LN). Adjusted p values <0.05 were considered statistically significant. M-CSF, macrophage colony-stimulating factor; UPCR, urine protein to creatinine ratio; eGFR, estimated glomerular filtration rate; dsDNA, double-stranded DNA; C3, complement component 3; C4, complement component 4.
SCF Concentrations Are Significantly Lower during Remission than Active LN
To evaluate whether SCF, M-CSF, and IL-2Rα decrease in remission, paired active disease and remission samples from the same patients were analyzed. SCF was the only protein that was significantly decreased with LN remission (p = 0.014) (Fig. 6). However, due to an overlap in the distributions of protein concentration during active and inactive disease, the receiver operating characteristic curve did not provide a reliable optimal cutoff value between disease states (AUC = 0.6265).
Fig. 6.
Serum protein concentrations obtained from the same individuals during active and inactive LN. SCF (a), M-CSF (b), and IL-2Rα (c) concentrations were measured using ELISA (n = 18). ROC curves with the AUC values were generated for SCF (d), M-CSF (e), and IL-2Rα (f) concentrations between disease states. Adjusted p values <0.05 were considered statistically significant. LN, lupus nephritis; SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; IL-2Rα, interleukin-2 receptor alpha; ROC, receiver operating characteristic; AUC, area under curve.
Best Predictive Model Uses Serum SCF and IL-2Rα Concentrations to Differentiate People with LN from Healthy Individuals
Candidate proteins were modeled independently and in all possible combinations to determine the potential to serve as clinical markers of LN. Two binary datasets were used, those from the confirmation ELISAs comparing active LN patient samples to HC and those from the paired active-inactive LN samples. Logistic regression was used to model each dataset, with the optimal model having the lowest AIC value. The best predictive model discriminated LN from HC using SCF and IL-2Rα concentrations, and SCF concentration was independently the best predictive model for discerning active and inactive LN (Tables 3, 4).
STRING Analysis Indicates Convergence on the PI3K/Akt Signaling Pathway
To investigate the mechanism(s) underlying our findings, STRING protein network analysis was used to determine convergent biological processes amongst SCF, M-CSF, and IL-2Rα [13]. All three proteins are involved in leukocyte homeostasis, proliferation, and activation as well as positively regulating mononuclear cell proliferation, leukocyte differentiation, and cell adhesion (Fig. 7a). They are also all involved in hematopoiesis and the PI3K-Akt signaling pathway (Fig. 7b).
Fig. 7.
Biological process enrichment (a) and KEGG pathways enrichment (b) obtained by STRING protein network analysis. SCF, M-CSF, and IL-2Rα were all involved in these biological processes, predominately pertaining to leukocyte homeostasis, proliferation, and differentiation. SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; IL-2Rα, interleukin-2 receptor alpha; FDR, false discovery rate.
When we performed the STRING analysis with all twelve significantly elevated proteins from the Bio-Plex, upregulation of proteins involved in the PI3K-Akt pathway remained in the top pathways/biological processes, including 5 of the 12 proteins (Fig. 8). The five proteins involved in PI3K-Akt signaling were IL-2Ra, M-CSF, SCF, HGF, and G-CSF.
Fig. 8.
Biological process enrichment (a) and KEGG pathways enrichment (b) obtained by STRING protein network analysis. SCF, M-CSF, IL-2Rα, HGF, and G-CSF were involved in these biological processes, predominately pertaining to leukocyte homeostasis, proliferation, and differentiation. SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; IL-2Rα, interleukin-2 receptor alpha; HGF, hepatocyte growth factor; G-CSF, granulocyte colony-stimulating factor; FDR, false discovery rate.
Discussion
This dual-center study provides evidence that serum levels of three NF-κB-regulated proteins, SCF, M-CSF, and IL-2Rα are elevated in active proliferative LN patients, relative to healthy individuals. The LN patients whose samples were analyzed in this report adequately represent those most commonly diagnosed with LN (Black, Indigenous, Asian, and Latina women of childbearing age, see Table 1). A significant correlation exists between each of these three serum proteins and clinical parameters of renal function and proteinuria. All three proteins were correlated with elevated serum creatinine concentration. Elevated SCF and M-CSF were correlated with increased UPCR and decreased eGFR. Serum levels of SCF also significantly decreased with LN remission.
When we performed correlation analyses on samples collected within 6 months of a biopsy (n = 7), we found significant correlations between SCF × UPCR, eGFR, and serum creatinine concentrations (Table 5). We also found significant correlations with IL-2Ra × dsDNA antibody titer and C3 concentrations. After correcting for multiple comparisons, correlations relative to SCF remained significant. This subanalysis suggests that serum SCF is a promising biomarker for monitoring LN activity because the clinical correlations to proteinuria and renal function remain consistent whether patients are within 6 months of biopsy or have not been biopsied for several years.
Considering that clinical markers of proteinuria (UPCR) and renal function (eGFR and serum creatinine) were correlated with serum protein concentrations, we investigated how these relationships could affect the predictability of our models. We could not include proteinuria (UPCR) as an explanatory variable for our model because we chose patients with values above and below a threshold value (500 mg/g) to determine whether a patient was experiencing active disease or clinical remission. Adding this parameter to our models would artificially inflate their predictability. Due to the fact that eGFR includes serum creatinine as a component of its calculation, it would be redundant to include both variables. The addition of eGFR to each logistic regression model negligibly improved the predictive ability of serum SCF levels (online suppl. data; for all online suppl. material, see https://doi.org/10.1159/000547044).
SCF (KIT ligand, steel factor, mast cell growth factor) is a circulating growth factor that binds to KIT, a receptor tyrosine kinase [14], and elevated serum concentration have been associated with LN activity [15]. The mean values that we obtained from LN serum are similar to those reported by others [15]. The values reported for healthy individuals were similar to those that we found [15]. Prior studies found correlations between serum SCF concentration and elevated dsDNA ab titer, decreased C3, and decreased C4, while our study is the first to investigate correlations with UPCR, eGFR, and serum creatinine. [15]. Another report found no difference in serum SCF concentrations between SLE patients and HCs, likely due to the comparatively high concentrations measured in healthy individuals [14]. Lupus-prone mouse models have also demonstrated that enhanced renal SCF production occurs during kidney inflammation [16].
Interleukin-2 receptor alpha (IL-2Rα; CD25; Tac antigen) becomes soluble when it is cleaved from the membrane of activated mononuclear cells [17], and elevated serum IL-2Rα concentrations are implicated and as predictor of LN in SLE patients [17–19]. Moreover, elevated IL-2Rα in serum [19] and urine [20, 21] has been associated with LN activity [22]. Compared to reported studies, we found much higher serum concentrations in HC, active LN patients, and inactive LN patients. While we did find that serum IL-2Rα concentration was higher in active LN than HC, we did not find a significant decrease between active and inactive disease states. An important difference between studies is that we measured serum proteins during active and inactive disease as paired samples within individuals, while others categorized an individual’s sample as either active or inactive [22]. We found significant correlations between IL-2Rα and serum creatinine concentrations. While others report significant correlations with C3, C4, anti-dsDNA antibodies, we did not find these relationships in our cohort [22].
Macrophage colony-stimulating factor-1 (M-CSF; CSF1) is a cytokine that regulates mononuclear phagocyte survival, proliferation, and differentiation [23] and elevated M-CSF concentration has also been previously recognized as a urinary [24, 25] and serum [25, 26] biomarker in LN. Others have found that M-CSF concentrations in serum and urine decrease with LN remission, suggesting that this is a useful marker of LN activity [25]. Moreover, both serum and urinary M-CSF concentrations elevate prior to the development of proteinuria in LN flares, suggesting that the circulating cytokine may influence the renal environment to promote inflammation that leads to glomerular damage [25]. Others have also demonstrated that monocytes from SLE patients with renal involvement express more of the receptor for M-CSF (CD115) than SLE patients without renal involvement, although they did not find a difference between SLE patients with active and inactive disease [27]. Our investigation found much higher (approximately 50–100×) serum M-CSF concentrations in LN patients and healthy individuals than others have reported [25–27]. One differentiating factor between these reports and our findings are that we strictly chose LN patients with class III or class IV LN, while Menke et al. [15] and Wang et al. [26] included class II and pure class V LN patients. Zeisbrich et al. [27] do not mention the LN classes included in their investigation. Our and Menke et al.’s [15] measurements were obtained from ELISA, whereas Wang et al. [26] measured serum proteins using Luminex, a multiplex immunoassay that is less sensitive than ELISA [28]. Similarly, we found higher serum protein concentrations when measuring by ELISA than when using multiplex platforms.
Regarding the findings from STRING network analysis, the PI3K-Akt pathway occurs upstream of NF-κB and plays a role in its regulation through Akt-mediated phosphorylation of IKKα [29]. IKKα phosphorylation leads to p65 phosphorylation [29], a requirement for nuclear translocation and NF-κB activation [30]. Pharmacological PI3K inhibitors ameliorate murine lupus and reduce the degree of renal macrophage infiltration [31], circulating dsDNA [32], dsDNA antibody titers [33], glomerular immunoglobulin/complement deposition [32], and proteinuria [33], implicating PI3K as a potential therapeutic target for LN. Inhibition of PI3K-Akt pathway in murine models improves LN through elevated FOXO3 production [34].
We were surprised to not find significant differences in proteins that others groups have previously reported, such as IFN-γ and TNF-α [35]. IFN-γ and TNF-α were elevated in LN patients compared to HCs, but the difference was not significant (unadjusted p = 0.16 and 0.27, respectively). We speculate that these values were not significantly elevated in our cohort because of heterogeneity of our patient population. The healthy individuals’ protein concentrations fell within a much more narrow range than the lupus patients’ concentrations, of whom several had much higher values. One limitation of our preliminary multiplex assay was that nearly half of the proteins assessed were below the detectable range, so we were unable to quantify these values. All samples were prepared for analysis at a 1:200 dilution. If there was a difference in the concentrations of these proteins between healthy individuals and lupus patients, we would not be able to recognize it due to limitations of the protocol.
In conclusion, this investigation provides additional evidence that circulating NF-κB-regulated proteins are elevated in LN patients. Of the three proteins discussed in this report, SCF was the best independent predictor of LN, while the combination of SCF and IL-2Rα provided the highest degree of predictability. Current LN clinical biomarkers such as serum creatinine and proteinuria are inadequate. Elevated serum creatinine is often a late finding, occurring only after the accrual of significant renal damage. Proteinuria is nonspecific and can be present in the absence of inflammation. Novel biomarkers are needed to provide additional measures of disease activity and also provide insight into potential pathways to target for future therapies.
Acknowledgments
We would like to thank all donors who provided samples to this study. Michael W. Daniels at the University of Louisville’s School of Public Health and Information Sciences provided insight on the statistical methods used.
Statement of Ethics
All studies were conducted with approval from the University of Louisville Institutional Review Board, approval no. 01.0536. Written informed consent was obtained from all participants prior to this study. The research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.
Conflict of Interest Statement
D.J.C. has received research funding from Alexion, Chinook (Novartis), and Travere; received speaker fees for Aurinia, Calliditas, GSK, and Roche; and received consulting fees for Aurinia, Alexion, Cabaletta, Calliditas, Chinook, GSK, Novartis, and Travere. D.J.C.’s conflicts of interest fall outside of this submitted work. Dr. Dawn Caster was a member of the journal’s Editorial Board at the time of submission. C.G.A. has received grant support from AstraZeneca and Bristol Myers Squibb; is on the advisory or review panel with AstraZeneca, Aurinia, Bristol Myers Squibb, Cabaletta, GSK, Health and Wellness Partners, Kezar, Synthekine, and UCB; and is speaker/honoraria with AstraZeneca and Aurinia. No other authors have conflicts of interest to disclose.
Funding Sources
This research was supported by NIH NIDDK grants R01 DK126777 and K08 DK102542. The Bio-Plex analysis was performed by the Functional Microbiomics Core (FMC) of the FMIP COBRE supported by the NIGMS award 5P20GM125504.
Author Contributions
The project was conceptualized by N.A.S., M.P.B., S.T., M.J.R., M.T.B., D.W.P., and D.J.C. Data were produced by N.A.S., M.P.B., R.R.L., K.N.O., and S.P.M. N.A.S. performed all statistical analyses. The manuscript was written by N.A.S. and reviewed/edited by all authors. C.G.A. provided serum samples from her biorepository. Each author made original and valuable intellectual contributions.
Funding Statement
This research was supported by NIH NIDDK grants R01 DK126777 and K08 DK102542. The Bio-Plex analysis was performed by the Functional Microbiomics Core (FMC) of the FMIP COBRE supported by the NIGMS award 5P20GM125504.
Data Availability Statement
The data that support the findings of this study are openly available through figshare at http://doi.org/10.6084/m9.figshare.28229120. Inquiries regarding the data or statistics used in this study may be directed to the corresponding author.
References
- 1. Teh CL, Phui VE, Ling GR, Ngu LS, Wan SA, Tan CHH. Causes and predictors of mortality in biopsy-proven lupus nephritis: the Sarawak experience. Clin Kidney J. 2018;11(1):56–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Kidney Disease Improving Global Outcomes KDIGO Lupus Nephritis Work Group . KDIGO 2024 clinical practice guideline for the management of LUPUS NEPHRITIS. Kidney Int. 2024;105(1S):S1–69. [DOI] [PubMed] [Google Scholar]
- 3. Siegel CH, Sammaritano LR. Systemic lupus erythematosus: a review. JAMA. 2024;331(17):1480–91. [DOI] [PubMed] [Google Scholar]
- 4. Izmirly PM, Parton H, Wang L, McCune WJ, Lim SS, Drenkard C, et al. Prevalence of systemic lupus erythematosus in the United States: estimates from a meta-analysis of the centers for disease control and prevention national lupus registries. Arthritis Rheumatol. 2021;73(6):991–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Pan L, Lu MP, Wang JH, Xu M, Yang SR. Immunological pathogenesis and treatment of systemic lupus erythematosus. World J Pediatr. 2020;16(1):19–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Oeckinghaus A, Ghosh S. The NF-kappaB family of transcription factors and its regulation. Cold Spring Harb Perspect Biol. 2009;1(4):a000034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Chalmers SA, Garcia SJ, Reynolds JA, Herlitz L, Putterman C. NF-kB signaling in myeloid cells mediates the pathogenesis of immune-mediated nephritis. J Autoimmun. 2019;98:33–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Chen J, Chen P, Song Y, Wei J, Wu F, Sun J, et al. STING upregulation mediates ferroptosis and inflammatory response in lupus nephritis by upregulating TBK1 and activating NF-κB signal pathway. J Biosci. 2024;49(1):9. [PubMed] [Google Scholar]
- 9. Zheng L, Sinniah R, Hsu SI. In situ glomerular expression of activated NF-kappaB in human lupus nephritis and other non-proliferative proteinuric glomerulopathy. Virchows Arch. 2006;448(2):172–83. [DOI] [PubMed] [Google Scholar]
- 10. Sun F, Teng J, Yu P, Li W, Chang J, Xu H. Involvement of TWEAK and the NF-κB signaling pathway in lupus nephritis. Exp Ther Med. 2018;15(3):2611–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Brightbill HD, Suto E, Blaquiere N, Ramamoorthi N, Sujatha-Bhaskar S, Gogol EB, et al. NF-κB inducing kinase is a therapeutic target for systemic lupus erythematosus. Nat Commun. 2018;9(1):179. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Cao HY, Li D, Wang YP, Lu HX, Sun J, Li HB. The protection of NF-κB inhibition on kidney injury of systemic lupus erythematosus mice may be correlated with lncRNA TUG1. Kaohsiung J Med Sci. 2020;36(5):354–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. STRING Network Analysis . CSF1, KITLG, and IL2RA. 2024 3 septebmer 2024]. Version 12.0:[Available from: https://string-db.org/cgi/network?taskId=bIRnw3fOQ2tv&sessionId=brvfIrimsdUP
- 14. Kitoh T, Ishikawa H, Sawada S, Koshino K, Tokano Y, Hashimoto H, et al. Significance of stem cell factor and soluble KIT in patients with systemic lupus erythematosus. Clin Rheumatol. 1998;17(4):293–300. [DOI] [PubMed] [Google Scholar]
- 15. Zheng D. Correlation between serum SCF level and disease activity, pathological type and refractory degree of lupus nephritis. Acta Med Mediterranea. 2021;37. [Google Scholar]
- 16. Sung SJ, Fu SM. Interactions among glomerulus infiltrating macrophages and intrinsic cells via cytokines in chronic lupus glomerulonephritis. J Autoimmun. 2020;106:102331. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Zhang RJ, Zhang X, Chen J, Shao M, Yang Y, Balaubramaniam B, et al. Serum soluble CD25 as a risk factor of renal impairment in systemic lupus erythematosus - a prospective cohort study. Lupus. 2018;27(7):1100–6. [DOI] [PubMed] [Google Scholar]
- 18. El-Shafey EM, El-Nagar GF, El-Bendary AS, Sabry AA, Selim AGA. Serum soluble interleukin-2 receptor alpha in systemic lupus erythematosus. Iran J Kidney Dis. 2008;2(2):80–5. [PubMed] [Google Scholar]
- 19. Gomes RR, Khan S. Serum soluble interleukin 2 receptor alpha vs. Urinary parameter and serum immunological markers as a monitoring tool of treatment response of lupus nephritis and their relation to lupus nephritis class on renal biopsy. J Urol Nephrol. 2022;7(4):1–10. [Google Scholar]
- 20. Gupta R, Yadav A, Misra R, Aggarwal A. Urinary sCD25 as a biomarker of lupus nephritis disease activity. Lupus. 2015;24(3):273–9. [DOI] [PubMed] [Google Scholar]
- 21. Tsai CY, Wu TH, Sun KH, Lin WM, Yu CL. Increased excretion of soluble interleukin 2 receptors and free light chain immunoglobulins in the urine of patients with active lupus nephritis. Ann Rheum Dis. 1992;51(2):168–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Davas EM, Tsirogianni A, Kappou I, Karamitsos D, Economidou I, Dantis PC. Serum IL-6, TNFalpha, p55 srTNFalpha, p75srTNFalpha, srIL-2alpha levels and disease activity in systemic lupus erythematosus. Clin Rheumatol. 1999;18(1):17–22. [DOI] [PubMed] [Google Scholar]
- 23. UniProt. P09603- CSF1_HUMAN. [4 September 2024]. Available from: https://www.uniprot.org/uniprotkb/P09603/entry
- 24. Tian S, Li J, Wang L, Liu T, Liu H, Cheng G, et al. Urinary levels of RANTES and M-CSF are predictors of lupus nephritis flare. Inflamm Res. 2007;56(7):304–10. [DOI] [PubMed] [Google Scholar]
- 25. Menke J, Amann K, Cavagna L, Blettner M, Weinmann A, Schwarting A, et al. Colony-stimulating factor-1: a potential biomarker for lupus nephritis. J Am Soc Nephrol. 2015;26(2):379–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Wang R, Zhao H, Liu Y, Li Y, Cai J. Macrophage colony-stimulating factor could evaluate both disease activity and renal involvement in systemic lupus erythematosus. Ann Palliat Med. 2021;10(2):2098–107. [DOI] [PubMed] [Google Scholar]
- 27. Zeisbrich M, Rzepka R, Finzel S, Venhoff N, Voll R. Macrophage colony-stimulating factor receptor/CD115(+) non-classical monocytes are expanded in systemic lupus erythematosus and associated with lupus nephritis. Scand J Rheumatol. 2024;54(2):125–34. [DOI] [PubMed] [Google Scholar]
- 28. Tighe PJ, Ryder RR, Todd I, Fairclough LC. ELISA in the multiplex era: potentials and pitfalls. Proteomics Clin Appl. 2015;9(3–4):406–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Bai D, Ueno L, Vogt PK. Akt-mediated regulation of NFkappaB and the essentialness of NFkappaB for the oncogenicity of PI3K and Akt. Int J Cancer. 2009;125(12):2863–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Maguire O, O’Loughlin K, Minderman H. Simultaneous assessment of NF-kappaB/p65 phosphorylation and nuclear localization using imaging flow cytometry. J Immunol Methods. 2015;423:3–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Suárez-Fueyo A, Rojas JM, Cariaga AE, García E, Steiner BH, Barber DF, et al. Inhibition of PI3Kδ reduces kidney infiltration by macrophages and ameliorates systemic lupus in the mouse. J Immunol. 2014;193(2):544–54. [DOI] [PubMed] [Google Scholar]
- 32. Yamaguchi J, Isnard P, Robil N, de la Grange P, Hoguin C, Schmitt A, et al. PIK3CA inhibition in models of proliferative glomerulonephritis and lupus nephritis. J Clin Investig. 2024;134(15):e176402. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Winkler DG, Faia KL, DiNitto JP, Ali JA, White KF, Brophy EE, et al. PI3K-delta and PI3K-gamma inhibition by IPI-145 abrogates immune responses and suppresses activity in autoimmune and inflammatory disease models. Chem Biol. 2013;20(11):1364–74. [DOI] [PubMed] [Google Scholar]
- 34. Zhao C, Gu Y, Chen L, Su X. Upregulation of FoxO3a expression through PI3K/Akt pathway attenuates the progression of lupus nephritis in MRL/lpr mice. Int Immunopharmacol. 2020;89(Pt A):107027. [DOI] [PubMed] [Google Scholar]
- 35. Fava A, Rao DA, Mohan C, Zhang T, Rosenberg A, Fenaroli P, et al. Urine proteomics and renal single-cell transcriptomics implicate interleukin-16 in lupus nephritis. Arthritis Rheumatol. 2022;74(5):829–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data that support the findings of this study are openly available through figshare at http://doi.org/10.6084/m9.figshare.28229120. Inquiries regarding the data or statistics used in this study may be directed to the corresponding author.