Elevated SLC1A5 links to inflamed endothelial cells and proteinuria in membranous nephropathy patients

Transcriptome data acquisition

Glomerular microarray data from patients with MN and healthy transplanted kidneys were obtained from GEO database. The expression matrix was downloaded from the GSE104948 dataset, including samples from European Renal cDNA Bank subjects and living donors (Grayson et al., 2018). A total of 21 MN patient samples and 21 healthy donor samples were selected for subsequent analyses. Using GEO-provided annotation files, the expression matrix underwent Probe ID-Gene Symbol conversion. Additionally, single-cell RNA sequencing (scRNA-seq) data of kidney samples from MN patients (n =6) and healthy control subjects (n =2) were obtained from GSE171458 (Xu et al., 2021).

Identification of differentially expressed genes

The microarray transcriptome data were normalized and analyzed for differential expression using the limma R package (version 3.56.2). P-values were adjusted via Benjamini–Hochberg correction to generate adjusted p-values (p.adj). Genes with absolute log2-fold change (—log2FC—) > 1 and p.adj < 0.05 were identified as differentially expressed genes (DEGs). Differential analysis results were visualized using the ggplot2 and ComplexHeatmap of R packages.

Gene enrichment analysis

DEGs from microarray data underwent enrichment analysis using the KOBAS web tool (Xie et al., 2011). For scRNA-seq pseudotime and Olink proteomic analyses, gene lists were processed with the R package clusterProfiler (version 4.8.3) (Wu et al., 2021). Enrichment terms were derived from Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene Ontology (GO), Reactome, and WikiPathways (WP) databases (Li et al., 2025; Xu et al., 2024). R package fgsea (version 1.26.0) was employed for gene set enrichment analysis (GSEA) on scRNA-seq data. The gene set pertaining to ferroptosis drivers was acquired from the FerrDb V2 database (Zhou et al., 2023).

Immune infiltration analysis

The CIBERSORT deconvolution algorithm was applied to the normalized glomerular expression matrix to estimate the proportions of various immune cells in tissues of both healthy and MN individuals (Newman et al., 2015). Specifically, LM22 feature matrix and 1,000 permutations were utilized for prediction, with quantile normalization performed. Box plots were constructed to display the proportions of each immune cell type in MN and healthy groups for the retained samples, and their inter-group differences were compared using Wilcoxon test. Spearman correlation coefficient was used to evaluate the relationship between expression of specific cytokines and immune infiltration.

scRNA-seq analysis

Bidirectional mendelian randomization (MR)

To investigate the causal relationship between module gene expression and MN, we conducted a bidirectional Mendelian Randomization (MR) analysis using the TwoSampleMR package (version 0.5.7). eQTL data were obtained from the IEU OpenGWAS database and NephQTL (Gillies et al., 2018). We performed clumping to select suitable instrumental variables, excluding SNPs in linkage disequilibrium (r² > 0.001) or lacking genome-wide significance (p ≥ 5 × 10⁻⁸). Instrument strength was assessed using the F-statistic, with F > 10 indicating robust associations. MN outcome data were sourced from the IEU OpenGWAS database, integrating datasets from diverse populations, including ebi-a-GCST010005 (2,150 cases and 5,829 controls of European ancestry) and ebi-a-GCST010004 (1,632 cases and 3,209 controls of East Asian ancestry) for discovery and validation. We applied the inverse-variance weighted (IVW) method for causal assessment, conducted heterogeneity testing to confirm IVW applicability, and assessed pleiotropy with the MR-Egger intercept test. Finally, reverse MR analysis was performed to explore reverse causality between gene expression and MN.

Serum SLC1A5 protein expression was detected by enzyme-linked immunosorbent assay (ELISA)

Serum samples from 50 MN patients and 31 healthy controls were collected from the First Affiliated Hospital of Soochow University. Serum was diluted 5-fold for analysis. Optical density (OD) was measured at 450 nm using a microplate reader, following the manufacturer’s instructions (Jingmei). SLC1A5 concentrations were quantified via a standard curve, with a detection limit of 1.0–160 pg/mL. The 50 MN patients were divided into high and low urinary protein groups, based on whether 24-hour urinary protein exceeded 3.5 g/day. Patients were further classified into nephrotic syndrome (urinary protein >3.5 g/day and serum albumin <30 g/L) and non-nephrotic syndrome groups. This study was retrospective, and the waived consent was given to and approved by the Medical Ethics Committee of the First Affiliated Hospital of Soochow University (2024-394).

Statistics

The aforementioned bioinformatics analyses were conducted using R 4.3.1 (R Core Team, 2023). Specific testing methods and p-value correction approaches were detailed in their respective methodology sections. The Wilcoxon rank-sum test was used to analyze the differences between two groups, and correlations were evaluated via Spearman’s test. A p-value of less than 0.05 was considered statistically significant.

Key interactions between macrophages and endothelial cells in MN

In Fig. 1, we depict the key interactions between macrophages and endothelial cells in MN. Differential gene expression analysis revealed significant transcriptional variations in the glomeruli of MN patients compared to healthy donors. Subsequent single-cell localization studies indicated that these differentially expressed genes are predominantly enriched in macrophages and endothelial cells. Trajectory analysis and cell communication assessments identified pro-inflammatory macrophage and inflammatory endothelial cell subclusters, elucidating their critical interactions as central drivers of MN pathogenesis. Additionally, weighted gene co-expression network analysis (WGCNA) facilitated the identification of gene modules that are essential for each subcluster. Notably, Mendelian randomization (MR) analysis provided compelling evidence for a causal relationship between the module gene SLC1A5 and MN. Clinical validation further confirmed the significance of serum SLC1A5 protein expression, highlighting its potential as a biomarker for MN diagnosis and disease progression monitoring.

Bulk-transcriptome analysis of glomeruli revealed macrophage subtype differences between MN and control groups

The pathological changes of MN mainly affect the glomeruli, indicating that gene expression profiling in this region may elucidate its pathogenesis (Sealfon et al., 2022). Hence, gene expression analyses were performed on glomeruli from 21 MN patients and 21 healthy donors, identifying 95 differentially expressed genes (DEGs, —log2FC— > 1, p.adj < 0.05, Table S1). Of these, 39 genes were up-regulated, while 56 were down-regulated (Figs. 2A–2B).

Subsequently, enrichment analysis was conducted to characterize the potential functions of differentially expressed genes. The results showed that up-regulated genes were primarily involved in immune-related pathways, including “Inflammatory response”, “Cytokine-cytokine receptor interaction”, “Chemokine signaling pathway” and “Neutrophil degranulation”. Conversely, down-regulated genes were significantly enriched in metabolic pathways, such as “Oxidation–reduction process”, “Metabolism of lipids”, “Glycolysis/Gluconeogenesis” and “PPAR signaling pathway” (Figs. 2B–2C).

The CIBERSORT algorithm was employed to estimate the proportion of various immune cell types in glomerulus and investigate changes in the immune microenvironment of MN glomerulus. The results revealed a significant increase in monocyte infiltration, as well as a notable rise in M1 macrophages (p < 0.05) and activated NK cells (p < 0.01), accompanied by a significant decrease in M2 macrophages (p < 0.001, Figs. 1D and 1E). Additionally, significant reductions were observed in naïve B-cells, neutrophils, and activated dendritic cells, suggesting that the immune milieu of MN glomerulus is uniquely altered, with particularly increased macrophage infiltration and disrupted polarization.

Expression patterns of up-regulated DEGs from transcriptome analysis in MN renal tissues were examined at the single-cell level

To elucidate the role of up-regulated DEGs in the MN immune microenvironment identified by transcriptome analysis, scRNA-Seq data from renal tissues of MN patients and healthy controls were reanalyzed. Following quality control and the reduction of batch effects using the R package harmony (Figs. S1A, S1B), a total of 30,671 cells were obtained. Ten distinct cell populations were initially identified through automated annotation with ingleR and prior knowledge, including proximal tubular cells, mesangial cells, podocytes, Loop of Henle cells, distal tubular cells, intercalated cells, principal cells, endothelial cells, fibroblasts/pericytes, and myeloid cells (Fig. 3A). The markers for these cell populations are displayed in Figs. 3B and 3C.

Subsequently, the expression patterns of the 39 up-regulated DEGs identified by transcriptome analysis were explored at the single-cell level in MN renal tissues. These genes were found to be primarily localized in endothelial cells, fibroblasts/pericytes, myeloid cells, and podocytes (Fig. S1C). Given the low cell counts of fibroblasts/pericytes (383 cells) and podocytes (51 cells), the analysis focused on gene expression differences in myeloid and endothelial cells. Among 1,405 myeloid cells, 939 were identified as monocytes/macrophages, while other cell types, such as mixed lymphocytes and other myeloid cells, were excluded from further analysis due to insufficient numbers.

Pseudotemporal trajectory highlights pro-inflammatory macrophages and endothelial cells in NM progression

Monocle was used to create pseudotemporal trajectories for monocytes/macrophages and endothelial cells, tracking changes in cell states. For monocytes/macrophages, the trajectory revealed nine distinct states, forming two branches (Fig. 3D). Among them, State 6 exhibited the lowest differentiation potential, while State 9 was identified as the cluster with the highest differentiation potential (Figs. S1D, S1E). Both states were associated with patients presenting with significant proteinuria, although no such notable changes were observed between individuals with and without MN (Fig. 3E).

The Branching Expression Analysis Model (BEAM) identified genes undergoing significant changes during different cell state transitions. In the trajectory leading to branch 1 (State 6), inflammatory signaling pathways and proinflammatory genes (e.g., IL1B and CXCL8) were upregulated. Conversely, in the trajectory leading to branch 2, pathways related to the negative regulation of immune system and markers of M2 macrophages (e.g., CD163 and MRC1) were upregulated (Fig. 3F). Therefore, State 6 and State 9 were classified as proinflammatory macrophages and undifferentiated monocytes, respectively. These findings suggest that the immune environment within MN kidney is characterized by robust monocyte recruitment and differentiation into proinflammatory macrophages, consistent with transcriptome-based immune infiltration analyses.

Notably, the increased expression of VEGFA and significant enrichment of VEGFA-VEGFR2 signaling in proinflammatory macrophages suggest their potential involvement in the interaction between immune cells and glomerular capillaries.

For endothelial cells, monocle identified five states along a trajectory (Fig. 3G). Differential abundance analysis revealed a significant increase in State4 and decrease in State3 within MN patient kidneys (Fig. 3H). Cytotrace designated State4 as the cluster with the highest differentiation potential, while State3 exhibited the lowest (Figs. S1F, S1G). Although Cytotrace positioned State4 early in the differentiation sequence, due to its close association with disease progression and significant up-regulation of cytokines such as CXCL2 and CCL2 during the transition of cells towards State4 (Fig. 3I), it was considered as the endpoint of the pseudotime trajectory and classified as inflammatory endothelial cells (IECs). Furthermore, enrichment analysis indicated significant up-regulation of biological process such as “TNF signaling pathway”, “IL-17 signaling pathway”, “NF-kappa B signaling pathway” and “Autophagy” in IECs. Simultaneously, pathways including “Oxidative phosphorylation”, “ATP biosynthetic process”, “Signaling by VEGF” and “Glomerular filtration” were significantly down-regulated (Fig. 3J).

Cellular communication between pro-inflammatory macrophages and endothelial cells during MN progression

The strength of cell signaling pathways between distinct macrophage and endothelial cell (EC) subpopulations, particularly the cellular communication between pro-inflammatory macrophages and inflammatory endothelial cells (IECs), was characterized using CellChat. Compared to other macrophage subtypes, proinflammatory macrophages exhibited fewer autocrine ligand–receptor pairs but enhanced interaction with IECs. Conversely, IECs showed more autocrine ligand–receptor pairs than normal endothelial cells (Fig. 4A). At the signaling pathway level, the CCL signaling pathway was primarily secreted by proinflammatory macrophages and targeted anti-inflammatory macrophages, while also partially mediating autocrine effects in IECs. The CXCL signaling pathway predominantly facilitated interactions between various macrophage subtypes and IECs, whereas the VEGF signaling pathway was exclusively mediated by proinflammatory macrophages (Fig. 4B). At the molecular level, CXCL8 released by proinflammatory macrophages was shown to target IECs, while VEGFA acted on both endothelial cell types. Notably, normal endothelial cells exhibited higher sensitivity to VEGFA secreted by proinflammatory macrophages, possibly due to lower VEGFR expression in IECs. Additionally, autocrine effects in IECs were largely driven by CXCL2 and CCL2 (Fig. 4C). To further investigate the association between these cytokines and immune cells, Spearman correlation coefficients were calculated between cytokines expressions and immune cell proportions estimated by the CIBERSORT algorithm from RNA-seq data.

CXCL8 and CCL2 displayed significant positive correlations with monocyte infiltration, while CXCL8 was negatively correlated with M2 macrophages. Monocytes were positively correlated with M1 macrophages and negatively correlated with M2 macrophages (Fig. 4D). These findings suggest that CCL2 and CXCL8 may play a role in recruiting monocyte to the kidney and promoting their differentiation into M1 pro-inflammatory macrophages.

High-dimensional weighted gene co-expression network analysis of pro-inflammatory macrophages and endothelial cell-related key gene sets

Through transcriptome and single-cell analyses, pro-inflammatory cell subtypes and cellular communication closely related to MN pathogenesis and progression were identified. Subsequently, a set of genes potentially involved in regulating these cell subtypes was clustered using high-dimensional weighted gene co-expression network analysis (hdWGCNA). In monocytes/macrophages, 13 co-expression modules (Macro-M1 to Macro-M13) were identified and marked with distinct colors (Fig. 5A). Notably, Macro-M3 showed a significant positive correlation with pro-inflammatory macrophage phenotypes, while exhibiting negative correlations with undifferentiated monocytes and anti-inflammatory macrophages (Fig. 5B). Moreover, gene expression within this module was predominantly localized to State6 cells, corresponding to pro-inflammatory macrophages (Fig. 5C). In endothelial cells, six co-expression modules (Endo-M1 to Endo-M6) were identified (Fig. 5D). Among these, Endo-M2 and Endo-M4 exhibited significant positive correlations with the IEC phenotype, with gene expression concentrated in State4 cells, representing IECs (Figs. 5E and 5F). Figures 6A and 6C present the top 10 genes within each module. Functional enrichment analysis revealed common pathways in the key modules of pro-inflammatory macrophages and IECs, including ‘lipid and atherosclerosis’, ‘NF-kappa B signaling pathway’, ‘TNF signaling pathway’, and ‘apoptosis’. Furthermore, Macro-M3 was associated with ‘IL-17 signaling pathway’ and ‘antigen processing and presentation’, Endo-M2 with ‘NOD-like receptor signaling pathway’ and ‘HIF-1 signaling pathway’, while Endo-M4 was enriched in ‘other glycan degradation’, ‘ATP-dependent chromatin remodeling’, ‘sphingolipid metabolism’, and ‘peroxisome’ (Figs. 6B, 6D, and 6E).

Bidirectional MR identified SLC1A5 as a potential causative gene for MN

To determine the causal relationship between key module gene expression and MN occurrence, bidirectional MR was conducted using eQTL data and GWAS summary data from two distinct MN cohorts with varying pedigrees. In the European ancestry discovery cohort, potential associations with MN were identified for 17 genes (Fig. 7A). The directional trend of OR values for these candidate genes was observed to persist in the East Asian pedigree validation cohort. However, only SLC1A5 (OR = 1.41, p = 0.02 in the discovery cohort; OR = 1.36, p = 0.02 in the validation cohort) and TRIM4 (OR = 1.21, p = 0.02 in the discovery cohort; OR = 1.15, p = 0.04 in the validation cohort) maintained statistical significance (Fig. 7B). The evaluation of heterogeneity and pleiotropy for the two genes, using Q-value and MR-Egger intercept, showed no significance (p > 0.05). This supported the use of the IVW approach and suggested a direct causal association between the exposure and outcome. In both cohorts, reverse MR analysis confirmed that gene expression levels caused MN and ruled out reverse causation (Fig. S3).

Since SLC1A5 and TRIM4 were derived from the module associated with IECs, their expression in endothelial cells was examined. SLC1A5 expression was significantly concentrated in State4, representing IECs (Fig. 7C), while no such pattern was observed for TRIM4 (Fig. S4). As a key gene linked to ferroptosis, SLC1A5 was hypothesized to contribute to MN pathogenesis through endothelial cell ferroptosis. To investigate this, endothelial cells were stratified into high/low SLC1A5 expression groups, followed by GSEA. Endothelial cells with high SLC1A5 expression showed a notable upregulation of ferroptosis driver genes (Fig. 7D). Additionally, cytokine expression levels, including CCL2 and CXCL2, were elevated in SLC1A5-high-expressing endothelial cells (Fig. 7E).

SLC1A5 protein was highly expressed in the serum of MN patients and strongly associated with urinary protein levels

To evaluate the clinical relevance of SLC1A5 as a serum marker and its association with disease severity in MN, SLC1A5 protein expression was analyzed in 81 serum samples (31 healthy controls vs. 50 MN patients). The results revealed that SLC1A5 was significantly elevated in the serum of MN patients (p < 0.001) (Fig. 8A), and positively correlated with 24-hour total urinary protein (R = 0.418, p = 0.002) (Fig. 8B). Patients were divided into two groups based on urinary protein levels (>3.5 g/day vs. ≤3.5 g/day). In the high proteinuria group, SLC1A5 levels were significantly elevated in the serum (p < 0.001) (Fig. 8C). Further classification into nephrotic syndrome (NS) (urinary protein > 3.5 g/day + serum albumin <30 g/L) and non-NS groups revealed elevated SLC1A5 expression in the NS group (p = 0.047) (Fig. 8D).