Discovery of non-invasive biomarkers for accurate detection of membranous nephropathy

Abstract

Membranous nephropathy (MN) is the most common cause of nephrotic syndrome in adults. Currently, MN diagnosis relies on either the presence of serum anti-phospholipase A2 receptor antibodies (αPLA2R Abs) or kidney biopsy. However, approximately 30% of patients with MN exhibit negative αPLA2R Ab, and the invasive nature of kidney biopsy limits its use in certain patients. This study aims to identify non-invasive biomarkers for predicting progression of MN and to compare the urinary metabolic profiles of patients with MN at different αPLA2R Ab statuses using urine metabolomics. Untargeted metabolomics was performed on urine samples that were retrospectively collected from biopsy-confirmed patients with MN (n = 73) and healthy controls (HC) (n = 79) using liquid chromatography-tandem mass spectrometry. Among the 156 identified endogenous metabolites, 35 primary urinary metabolite biomarkers were associated with MN. Diagnostic and predictive models for patients with MN and high-risk patients were established based on these biomarkers, respectively. Metabolic pathway analysis indicated that oxidative stress pathways may play a crucial role in the pathophysiology of MN. Additionally, two urinary metabolites (dodecanedioic acid [DDDA] and 5’-deoxy-5’-(methylthio)adenosine [MTA]) exhibited a significant negative association with αPLA2R Ab titer. In this study, we proved metabolomics is valuable for accurate diagnosis and early detection of MN by identifying novel urinary metabolites. Additionally, it may offer insights for future research and the development of novel therapeutic targets.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-18744-x.

Introduction

Membranous nephropathy (MN) is a glomerular disease characterized by the deposition of immune complexes on the glomerular basement membrane (GBM). This causes inflammation, thickening, and GBM damage, resulting in proteinuria and nephrotic syndrome¹. MN is the most common cause of nephrotic syndrome in adults². Compared to other glomerulonephritis types, proteinuria and edema may develop gradually and become evident at advanced stages, thereby delaying early diagnosis and treatment. This results in chronic kidney disease (CKD) and end-stage kidney disease (ESKD) where, patient mortality rates are significantly high, necessitating dialysis or kidney transplantation for survival^3–6. Due to the individual heterogeneity of kidney disease, early diagnosis and appropriate treatment are crucial for high-risk patients with MN who experience rapid renal function decline to reduce the risk of disease progression and enhance clinical outcomes⁷.

MN can be classified as primary or secondary based on its etiology and pathology, with primary MN accounting for approximately 70–80% of all cases⁸. Phospholipase A2 receptor (PLA2R) is the most common autoantigen in primary MN that plays a crucial role in regulating the structure and function of podocytes in GBM⁹. Anti-PLA2R antibodies (αPLA2R Abs) are observed in 70–80% of patients with primary MN, with their titers associated with disease remission and relapse^10–12. While αPLA2R Ab is highly specific to primary MN, not all patients with MN have detectable levels of these antibodies^11,12. Kidney biopsy is the gold standard for MN diagnosis. However, owing to its invasive nature, it has significant risks, including bleeding and infection, thereby rendering it unsuitable for certain patients^13–15. Therefore, there is a need to develop non-invasive and accurate diagnostic techniques for MN.

Metabolomics is a technique used to analyze small molecular compounds in biological samples to understand an organism’s physiological and pathological states¹⁶. Kidney diseases, specifically diabetic kidney disease, have been extensively studied using metabolomics, using diverse analytical methods and sample types¹⁷. However, MN-associated metabolomic research is limited, often involving a small study population and using gas chromatography-mass spectrometry or nuclear magnetic resonance for sample analysis^18–22. Additionally, there is a lack of comparative metabolomic profiling of patients with MN based on αPLA2R Ab status. Addressing this gap may significantly enhance our understanding of the specific metabolic variations among various MN subtypes, which is crucial for the precise diagnosis and development of personalized treatment strategies.

In this study, we aimed to identify non-invasive urine biomarkers for primary MN using untargeted metabolomics with liquid chromatography-tandem mass spectrometry (LC-MS/MS). Additionally, we compared the urinary profiles of MN subtypes based on αPLA2R Ab status to offer novel insights for clinical research.

Results

Clinical characteristics of study participants

Table 1 summarizes the demographic and clinical characteristics of the study population. Clinical characteristics are presented as the mean ± standard deviation (SD) or median with interquartile range (IQR). Compared with the HC group, patients with MN with a median age of 57 years (IQR, 49–63) were older, with a higher percentage of males (65.8%). Additionally, laboratory findings associated with proteinuria and nephrotic syndrome were evident, including lower serum albumin levels of 3.0 g/dL (SD, 0.7), higher median urine protein-to-creatinine ratio of 4.6 mg/mg (IQR, 2.9–8.2), and higher median urine albumin-to-creatinine ratio of 3.1 mg/mg (IQR, 2.1–5.8). Moreover, 37 patients (50.7%) were positive for αPLA2R Ab, with a median αPLA2R Ab titer of 20.2 (IQR, 0.0–79.2).

Table 1.

Demographic and clinical characteristics of the study cohort at baseline.

Variables	MN (n = 73)	HC (n = 79)	Total (n = 152)	P
Age, median (IQR), years	57.0 (49.0, 63.0)	53.0 (45.0, 59.0)	55.0 (47.0, 61.0)	0.008
Sex, n (%), male	48 (65.8%)	38 (48.1%)	86 (56.6%)	0.042
Body mass index, median (IQR), kg/m2	23.9 (21.8, 25.5)	24.0 (22.2, 26.4)	24.0 (21.9, 26.0)	0.548
Hemoglobin, mean ± SD, g/dL	12.9 ± 1.7	13.5 ± 1.5	13.2 ± 1.6	0.04
Blood urea nitrogen, median (IQR), mg/dL	14.0 (12.0, 17.0)	13.0 (11.0, 16.0)	13.0 (11.0, 17.0)	0.227
Creatinine, median (IQR), mg/dL	0.8 (0.7, 0.9)	0.8 (0.6, 0.9)	0.8 (0.7, 0.9)	0.413
eGFR, median (IQR), mL/min/1.73 m2	98.7 (82.9, 124.2)	99.0 (91.2, 108.2)	98.8 (89.7, 113.2)	0.511
Phosphate, median (IQR), mg/dL	3.5 (3.2, 3.8)	3.5 (3.1, 3.8)	3.5 (3.2, 3.8)	0.805
Potassium, median (IQR), mmol/L	4.2 (4.0, 4.5)	4.2 (4.0, 4.4)	4.2 (4.0, 4.4)	0.338
TCO2, mean ± SD, mmol/L	28.2 ± 3.0	26.5 ± 3.0	27.4 ± 3.1	0.002
Albumin, mean ± SD, g/dL	3.0 ± 0.7	3.8 ± 0.3	3.4 ± 0.7	< 0.001
Calcium, mean ± SD, mg/dL	8.4 ± 0.5	8.3 ± 0.5	8.4 ± 0.5	0.568
Uric acid, median (IQR), mg/dL	6.0 (5.0, 7.4)	4.8 (4.1, 5.8)	5.4 (4.5, 6.7)	< 0.001
HbA1C, median (IQR), %	5.8 (5.5, 6.4)	5.3 (5.1, 5.6)	5.5 (5.2, 5.9)	< 0.001
LDL-C, median (IQR), mg/dL	140.0 (106.0, 189.0)	124.0 (103.0, 137.5)	126.5 (104.5, 155.0)	0.012
hsCRP, median (IQR), mg/L	0.0 (0.0, 0.1)	4.8 (0.4, 8.3)	0.2 (0.0, 5.1)	< 0.001
PCR, median (IQR), mg/mg	4.6 (2.9, 8.2)	0.0 (0.0, 0.1)	1.1 (0.0, 4.9)	< 0.001
ACR, median (IQR), mg/mg	3.1 (2.1, 5.8)	0.0 (0.0, 0.0)	0.0 (0.0, 2.4)	< 0.001
Comorbidities, n (%)
Diabetes mellitus	14 (19.2%)	0 (0.0%)	14 (10.1%)	0.001
Hypertension	29 (39.7%)	0 (0.0%)	29 (20.9%)	0.001
αPLA2R Ab positivity, n (%)	37 (50.7%)	-	-	-
αPLA2R Ab titer, median (IQR), RU/mL	20.2 (0.0, 79.2)	-	-	-

MN: membranous nephropathy; HC: healthy control; IQR: interquartile range; SD: standard deviation; eGFR: estimated glomerular filtration rate; TCO₂: total carbon dioxide; HbA1C: hemoglobin A1c; LDL-C: low-density lipoprotein cholesterol; hsCRP: high-sensitivity C-reactive protein; PCR: random urine protein-to-creatinine ratio; ACR: random urine albumin-to-creatinine ratio; αPLA2R Ab: anti-phospholipase A2 receptor antibody.

Various urinary metabolomic profiles between MN and HC

A comprehensive screening of the metabolomic database successfully identified 156 endogenous metabolites (Supplementary Table S1). PCA and unsupervised clustering heatmaps based on total metabolites did not exhibit clear differences between MN and HC groups (Fig. 1a, Supplementary Fig S1). However, a distinct separation was observed in OPLS-DA, indicating specific metabolic signatures that differentiated the two groups (Fig. 1b). Following the Student t-test, 35 metabolites with variable importance in projection (VIP) scores in OPLS-DA > 1 and P value < 0.05 were recognized as potential biomarkers for MN (Fig. 1c, Supplementary Fig S2). Based on OPLS-DA feature selection, metabolites with VIP > 1 were used to construct a logistic regression model for distinguishing MN from HC. The model demonstrated excellent diagnostic performance under five-fold cross-validation, yielding an average area under the curve (AUC) of 0.953 (Fig. 1d, Supplementary Table S3). Unsupervised clustering heatmaps based on these metabolites illustrated a good separation between MN and HC, indicating their crucial role in metabolic disruption of the disease (Fig. 1e). Further analysis of the metabolic pathways associated with these metabolites revealed significant alterations in amino acid metabolism, including taurine, phenylalanine, tyrosine, tryptophan, and histidine in MN compared to that in HC (Fig. 1f, Supplementary Table S3).

Fig. 1.

Multivariate metabolomics analysis of urine samples from study subjects. (a) PCA of HC and patients with MN, with QC samples or without QC samples. (b) OPLS-DA of HC and patients with MN. Bubble plot (c), logistic regression models with five-fold cross-validation (d), clustering heatmap (e), and pathway analysis (f) of selected significant urine metabolites.

Prediction of kidney deterioration risk in patients with MN

To investigate the prognostic relevance of the selected metabolites, the Cox proportional hazards model was applied to assess their association with disease progression. This time-to-event analysis accounts for both follow-up duration, and identified five metabolites—2-hydroxybutyric acid, taurine, 21-hydroxy-5β-pregnane-3,11,20-trione, pregnanolone sulfate, and glycoursodeoxycholic acid—as significant predictors of progression risk (Fig. 2a–b). These findings suggest that the identified metabolites may serve as potential prognostic biomarkers for membranous nephropathy. To further evaluate their predictive utility, these five metabolites were used to construct a logistic regression model aimed at classifying patients according to progression risk. Model performance was assessed using five-fold cross-validation (Fig. 2c, Supplementary Table S3). This model achieved an average AUC of 0.849, confirming its substantial predictive accuracy.

Fig. 2.

Selection and modeling of urine metabolites for CKD progression. (a) Cox proportional hazards model assessing the impact of selected urine metabolites on CKD outcomes. (b) Grouped violin plot of the five urine metabolites selected based on their significance in the Cox model. (c) Logistic regression model with five-fold cross-validation to predict high-risk MN patients.

Metabolomic analysis revealed differences between αPLA2R Ab-positive and -negative MN

Possibly due to the similarity of the overall urine metabolome, PCA did not reveal distinct differences between the αPLA2R Ab-positive and -negative MN subtypes (Fig. 3a). Seventeen metabolites with P < 0.05 were identified as potential biomarkers associated with αPLA2R Abs (Supplementary Fig. S3). Among these, three urinary metabolites exhibited higher levels in the αPLA2R Ab-positive group, whereas the other 14 were predominantly observed in the αPLA2R Ab-negative group (Fig. 3b). Unsupervised clustering heatmaps using these metabolites demonstrated partial separation between the two groups (Fig. 3c). Further pathway analysis revealed alterations in amino acid-related pathways, which were similar to those observed between the MN and HC groups (Fig. 3d, Supplementary Table S2). Specifically, amino acid-related pathways, such as methionine, cysteine, and histidine metabolism exhibited notable alterations.

Fig. 3.

Multivariate metabolomics analysis of urine samples from two MN subtypes defined by αPLA2R Ab statuses. (a) PCA of patients with different αPLA2R Ab statuses. Volcano plot (b), clustering heatmap (c), and pathway analysis (d) of selected significant urine metabolites. Ab-P: αPLA2R Ab-positive; Ab-N: αPLA2R Ab-negative.

Spearman’s rank correlation was conducted between 17 urinary metabolite levels and αPLA2R Ab titers (Fig. 4a). Ten metabolites presented statistically significant correlations (P < 0.05) with αPLA2R Ab titers, with two exhibiting positive correlations and eight exhibiting negative correlations, although all the correlation coefficients were below 0.4. Notably, dodecanedioic acid (DDDA) and 5’-deoxy-5’-(methylthio)adenosine (MTA), both with correlation coefficients > 0.3, were selected to develop a logistic regression classification model. The model based on DDDA levels achieved an average AUC of 0.868, highlighting its potential as a significant discriminative metabolite for distinguishing between the αPLA2R Ab-positive and -negative groups (Fig. 4b, Supplementary Table S3).

Fig. 4.

Analysis of αPLA2R Ab titer-related urine metabolites. (a) Correlation analysis between αPLA2R Ab titer and urine metabolites abundance. (b) Logistic regression model by two selected urine metabolites used to classify MN subtypes based on their antibody positivity. DDDA: Dodecanedioic acid; MTA: 5’-deoxy-5’-(methylthio)adenosine.

Discussion

In this study, we identified significant urine metabolites associated with MN through untargeted metabolomic analysis and successfully established models for MN diagnosis and prediction of disease progression. Additionally, we compared the urine metabolomic profiles of patients with MN and various αPLA2R Ab statuses, highlighting their potential clinical applications.

A total of 35 urine metabolites associated with amino acid metabolism and oxidative stress pathways, including taurine, phenylalanine, tyrosine, tryptophan, and histidine, were specific to patients with MN. Taurine is a sulfur-containing amino acid with antioxidant and anti-inflammatory properties²³. In certain animal and human clinical studies, taurine supplementation can protect renal function by reducing oxidative stress and inflammatory responses^24–26. Our results demonstrated that urine taurine levels were significantly higher in patients with MN than in HC. An increase in taurine levels was positively correlated with disease progression, likely owing to increased production in response to continued oxidative stress and inflammation in patients with MN, specifically those with disease progression and chronic inflammatory status. Phenylalanine is the precursor of protein synthesis, and increased phenylalanine levels may be associated with enhanced oxidative stress and inflammatory states²⁷. Tyrosine, a derivative of phenylalanine, is involved in the synthesis of dopamine and norepinephrine, affecting the function of the nervous and immune systems²⁸. The enhancement of phenylalanine metabolism may result in an increase in tyrosine synthesis, affecting cell function and signal transduction²⁹. Evidence indicates that dysregulated phenylalanine and tyrosine levels may be associated with renal and cardiovascular disease^30,31. Tryptophan is an essential amino acid from dietary sources that plays a crucial role in regulating immune response and intestinal microbial balance³². Increased levels of its metabolites activate the aryl hydrocarbon receptor that facilitates renal fibrosis and exacerbates CKD progression^33,34. Unlike other amino acids, reduced levels of urinary histidine and its metabolites were observed in patients with MN. Histidine exerts an antioxidant effect in the body by converting it into histamine and other metabolites that protect embryonic kidney cells from oxidative stress damage^35,36. Reduced histidine levels may result in abnormal immune responses and increase the risk of infection and inflammation³⁷.

Previous studies of urinary metabolome in MN have provided valuable insights into disease-related metabolic alterations^20–22. However, the key metabolic pathways reported in these studies differ partially from those identified in our analysis. A shared finding across studies is the consistent dysregulation of aromatic amino acid metabolism, particularly involving phenylalanine, tyrosine, and tryptophan pathways, supporting their relevance in MN pathophysiology. Taherkhani et al. employed GC-MS and NMR, which are limited in sensitivity and chemical coverage. GC-MS predominantly captures volatile and derivatizable metabolites, while NMR detects only high-abundance compounds. In contrast, our LC-MS/MS platform offers broader and deeper coverage, especially for lipid species, bile acids, and sulfur-containing metabolites. This technical advantage enabled the identification of additional pathway alterations. Another key distinction lies in cohort design. Unlike Qu et al., who compared MN with IgAN without healthy controls, our inclusion of healthy individuals allowed us to identify disease-specific metabolic deviations rather than inter-subtype differences. As a result, we identified several novel pathway perturbations not highlighted in previous reports, including bile acid biosynthesis, taurine metabolism, and histidine metabolism. These findings may reflect additional biological mechanisms, such as dysregulation of the gut-liver axis, osmolyte imbalance, oxidative stress, and inflammation, which may be more prominent in MN than previously recognized. Together, these differences highlight the importance of both analytical methodology and cohort design in shaping metabolomic interpretations. Our results not only support previous findings but also extend the current understanding of MN-associated metabolic perturbations, offering new insights into disease pathophysiology.

Seventeen significant urinary metabolites associated with αPLA2R Ab positivity were identified, indicating various metabolic pathophysiologies between αPLA2R Ab-positive and -negative patients with MN. The primary metabolic pathways distinguishing the two groups include caffeine, pyruvate, and methionine metabolism. Variations in caffeine metabolism may affect renal function and drug metabolism ability³⁸. Dysregulation in pyruvate metabolism may indicate shifts in cellular metabolic states, potentially associated with oxidative stress and energy metabolism³⁹. Increasing evidence indicates that pyruvate has multiple metabolic effects that directly affect the development of acute kidney injury⁴⁰. The cysteine ​​and methionine metabolic pathways play a significant role in the antioxidant and methylation reactions, and significant alterations in these pathways may result in higher oxidative stress levels and antioxidant requirements^41,42. Based on these results, it can be inferred that the metabolic differences observed between αPLA2R Ab-positive and -negative patients with MN reflect distinct underlying pathophysiological mechanisms. Specifically, the αPLA2R Ab-negative group may exhibit higher oxidative stress and energy demands, whereas the αPLA2R Ab-positive group may be more sensitive to immune responses and specific pathological processes associated with antibody presence. Although these findings indicate potential metabolic pathways that differentiate the two groups, further research is required to understand how these metabolic alterations affect disease progression, treatment responses, and clinical outcomes. This understanding may potentially guide personalized therapeutic approaches for patients with MN based on their αPLA2R Ab status.

Additionally, DDDA (r = − 0.310, P = 0.008) and MTA (r = − 0.324, P = 0.005) exhibited significant correlations with the αPLA2R Ab titer. Specifically, the prediction model based on DDDA levels exhibited an AUC of 0.868, highlighting its potential as a significant discriminative metabolite for distinguishing between αPLA2R Ab-positive and -negative subgroups. DDDA is a medium-chain dicarboxylic acid that provides high-density energy through fatty acid oxidation⁴³. DDDA levels were negatively correlated with the αPLA2R Ab titer, with lower levels detected in patients with αPLA2R Ab positivity. This may reflect that the excessive consumption of DDDA results in a higher energy supply of the immune system, causing the hyperactivation of antibody response^44,45. MTA, a naturally occurring sulfur-containing nucleoside derived from S-adenosylmethionine primarily through the polyamine biosynthesis pathway⁴⁶has immunosuppressive effects and can reprogram Toll-like receptor (TLR) activation pathways through adenosine receptors to suppress inflammation⁴⁷. Low MTA levels may indicate enhanced conversion of adenine, which may result in increased inflammation of the kidneys⁴⁸.

Despite these findings, this study has several limitations. First, the study population was Korean, which may limit the generalizability of our findings to other ethnic groups. Second, variations observed in the urinary metabolomic profiles of patients with MN compared to HC may not be completely attributable to the onset of MN because certain patients had other medical histories or prior medication treatments. Third, while we confirmed the identities of 20 key metabolites using authentic standards, most compounds remain at MSI Level 2, annotated based on accurate mass and MS/MS spectral matching. Although this is a common limitation in untargeted metabolomics studies, future work incorporating a broader range of reference standards will help further strengthen the biological interpretation of urinary metabolic alterations in MN. Finally, owing to the limited number of participants, cross-validation was used, rather than an independent validation cohort, to confirm the selected biomarkers. Therefore, the role of these biomarkers in MN pathology requires further assessment.

In conclusion, this study identified significant urinary metabolite biomarkers associated with primary MN using urine metabolomic analysis. The significantly altered metabolites and associated metabolic pathways indicated that oxidative stress metabolic pathways may play a crucial role in the pathophysiology of MN.

Methods

Participants and sample collection

A total of 73 biopsy-proven primary patients with MN and 77 healthy control (HC) adults were enrolled at the Seoul National University Hospital between April 2021 and December 2022. Patients with MN were categorized into two groups based on the presence of αPLA2R Abs: αPLA2R Ab positive and αPLA2R Ab negative. An enzyme-linked immunosorbent assay was conducted for αPLA2R Ab detection, and αPLA2R Ab positivity was defined as a titer of ≥ 20 RU/mL. The baseline urine samples were collected from each participant and stored at − 80 °C. The study was approved by the Institutional Review Board of Seoul National University Hospital (No. H-2104-120-1214) and performed in accordance with the Declaration of Helsinki and the institutional guidelines. Informed consent was obtained from all subjects.

Sample preparation and quality control (QC)

Frozen urine samples were thawed at 4 °C before use. QC samples were prepared from small aliquots of all the samples. To extract metabolites, 50 µL of each sample and QC were added to 200 µL of cold (− 20 °C) methanol: ethanol mixture (1:1) in a fresh tube. After vortexing for 5 min, the samples were centrifuged (14,000 rpm, 10 min, 4 °C) to remove the protein pellet. The supernatant was transferred to two injection vials (100 µL each) for analysis.

Untargeted metabolomics

LC-MS/MS analysis was performed on a Vanquish Flex ultra-high-performance liquid chromatography system equipped with an Orbitrap Exploris 120 mass spectrometer (Thermo Fisher Scientific). Chromatographic separation was performed using an ACQUITY UPLC HSS T3 Column (2.1 × 100 mm, 1.7 μm; Waters). The mobile phase consisted of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile: methanol, 7:3). The gradient elution started with 5% B for 1 min, linearly increased to 80% at 12 min for 0.5 min and returned to 5% at 13.5 min for 15 min. The flow rate of mobile phase was 0.4 mL/min. The column oven temperature was maintained at 40 °C. The auto-sampler temperature was 4 °C, with an injection volume of 4 µL. The sample sequence was randomized, and QC samples were injected once after 10 sample runs.

Mass acquisition was performed in both positive and negative ionization modes in two stages. A mass scan was performed for both the samples and QC samples, whereas data-dependent mass fragmentation (ddMS2) was only performed for QC samples with absolute collision energies ranging from 10 to 50 eV. The mass resolution was set to 120,000 and 15,000 for the mass scan and ddMS2, respectively. The mass range of data acquisition was 60–1000 and 50–750 m/z for the mass scan and ddMS2, respectively. Four dependent scans were used for the ddMS2. Electrospray ionization source conditions were set as follows: sheath gas flow rate of 40 Arb, auxiliary gas flow rate of 10 Arb, sweep gas flow rate of 1 Arb, vaporizer temperature of 350 °C, and spraying voltage of 4.0 (positive) or − 3.0 (negative) kV.

Data processing

The data acquired from the mass spectrometer in positive and negative modes were processed separately. Mass scan and ddMS2 raw data were converted into mzXML and mgf formats, respectively, using the msConvert tool from ProteoWizard⁴⁹. Peak detection and alignment based on the mass scan data were conducted using the TidyMass R package⁵⁰. Features with > 20% missing values in QC samples or > 50% missing values in both the HC and MN groups were filtered to ensure data quality. Subsequently, the remaining missing values were inputted using the k-nearest neighbor algorithm. Probabilistic quotient normalization was used to adjust the dilution variability among samples in the urine metabolomic data. Metabolite annotation was performed by referencing three public MS2 databases (Human Metabolome Database, MassBank, and MassBank of North America) and two published in-house libraries supplied by TidyMass (version 1.0.8). The positive and negative mode data were combined, and endogenous metabolite data were extracted for subsequent analysis. To enhance identification confidence, several biologically relevant metabolites were confirmed by comparing their retention times and MS2 fragmentation patterns with those of authentic reference standards. These validated compounds are listed as MSI Level 1 in Supplementary Table S1.

Statistical analysis

The normalized metabolomic data were subjected to log transformation and Pareto Scaling before multivariate statistical analysis. Principal component analysis (PCA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) were performed to understand the relationships and variations among various sample groups in a multidimensional space. OPLS-DA models were constructed with automatic optimization of orthogonal components and evaluated using 7-fold cross-validation in the ropls package. Unsupervised hierarchical clustering heatmaps were created using Euclidean distance and Ward’s method to analyze and visualize the underlying patterns in the dataset. The Student’s t-test was performed to compare metabolites between different groups, with statistical significance set at P < 0.05. Pathway analysis was conducted using both the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Small Molecule Pathway Database (SMPDB)^51,52. Logistic regression models were constructed using five-fold cross-validation to ensure the robustness and reliability of our predictive capabilities. A Cox proportional hazards model was used to explore the relationship between urine metabolites and disease progression, defined as a reduction in estimated glomerular filtration rate (eGFR) of > 5 mL/min/1.73 m²/year. Additionally, the relationship between αPLA2R Ab titer and urine metabolites was calculated using Spearman rank correlation. All data analysis and visualization were conducted in the R environment (version 4.3.3).

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: Y.L., J.Kim, J.Kang, S.H., J.C.; Resources: J.Kim, S.H.; Data curation: Y.L., J.Kim; Formal analysis: Y.L.; Methodology: Y.L.; Software: Y.L.; Validation: Y.L.; Supervision: J.Kang, S.H., J.C.; Visualization: Y.L.; Writing - original draft: Y.L.; Writing - review & editing: J.Kim, J.Kang, S.H., J.C.; Funding acquisition: S.H., J.C.

Funding

This work was supported by a grant (Grant no: RS-2022-NR070765) from the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (MSIT). This work was supported by the National Institute of Health Research Project (project no. 2021-0317C93-00).

Data availability

Raw data is available at the Korea BioData Station (K-BDS, https://kbds.re.kr/) under the accession ID KAP241585.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

The study received approval from the institutional review boards of (No. H-2104-120-1214).