FOS drives podocyte injury and renal fibrosis in diabetic kidney disease via direct Smad3 binding

Abstract

Introduction

Epithelial-mesenchymal transition (EMT) plays a crucial role in diabetic kidney disease (DKD), particularly in the context of podocyte EMT. However, the regulatory networks controlling EMT activation in podocytes remain incompletely mapped. The FOS proto-oncogene (FOS) has emerged as a key player in EMT, yet its functional significance in DKD-podocytes remains poorly defined.

Methods

The predictive validity of FOS was confirmed through least absolute shrinkage and selection operator (LASSO) regression and multivariable logistic analysis in clinical data. A streptozotocin-induced DKD rat model was established, and podocyte cells (MPC5, HPC) treated with high glucose served as an in vitro model. The functional effect of FOS was explored using genetic interventions in vivo and in vitro.

Results

FOS expression is increased in patients with DKD, correlating with deterioration in renal function. LASSO and logistic analyses identified FOS as a robust predictor of DKD progression. FOS deficiency in vivo attenuated renal dysfunction and fibrosis. In podocytes, high glucose-primed FOS induced EMT and collagen accumulation, effects reversed by FOS knockdown. Crucially, FOS directly bound and activated Smad3, while Smad3 silencing abolished FOS-mediated injury.

Conclusion

This study demonstrates that the FOS/Smad3 axis is a critical contributor to coordinating EMT and fibrosis in diabetic renal podocytes, proposing a novel therapeutic target in DKD.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00018-026-06215-z.

Background

Diabetic kidney disease (DKD), the leading cause of end-stage renal disease (ESRD), is characterized by progressive fibrosis and irreversible multi-organ damage [1–3]. Despite current therapeutic strategies, including glucose regulation and renin-angiotensin-aldosterone system (RAAS) inhibitors [4, 5], which have been applied clinically for decades, the prevalence of DKD continues to increase [6]. Hence, this underscores the urgent need for developing specific treatment strategies. Fibrosis serves as a cardinal drive of poor clinical outcomes in DKD [7, 8]. The fibrosis transformation observed in DKD results from a complex interplay between cellular signaling pathways and extracellular matrix dynamics [7, 9]. Excessive deposition of fibrillar collagens disrupts normal renal architecture while generating matrix-derived bioactive peptides that perpetuate fibrotic signaling [10–12]. Among various fibrosis pathways, the TGF-β/Smad3 signaling has emerged as particularly significant [13, 14], orchestrating the transcription of fibrosis-related genes through Smad3 phosphorylation and nuclear translocation [15, 16]. However, its context-dependent regulation in DKD remains poorly understood. Critically, A key cellular process in DKD progression involves epithelial-mesenchymal transition (EMT) [17], where renal epithelial cells undergo phenotypic conversion to matrix-producing myofibroblasts [18]. Podocyte EMT, specifically, disrupts glomerular filtration barrier integrity [19, 20], increasing glomerular permeability to albumin [21], compared to hepatic or cardiac fibrosis [22, 23], suggesting organ-specific modifiers of the EMT-TGF-β axis. However, the regulatory networks controlling EMT activation in podocytes remain incompletely mapped.

The FOS proto-oncogene (FOS) operates as a nodal signaling integrator through AP-1 complex formation with JUN proteins [24, 25]. Beyond its established roles in cell proliferation and stress responses [26], recent multi-omics analyses implicate FOS in fibrosis progression across cardiac and hepatic pathologies [27, 28]. While FOS potently amplifies TGF-β signaling in extrarenal systems [29, 30], its potential role in coordinating renal cell plasticity through this crosstalk during DKD progression remains undefined. Notably, FOS directly activates EMT in multiple malignancies [31–33], suggesting conserved pathological mechanisms across organ systems. Emerging evidence positions FOS at the intersection of TGF-β signaling and EMT transcriptional networks, yet the molecular circuitry underlying this regulatory axis demands systematic investigation.

We revealed that FOS acts as a potential diagnostic biomarker and therapeutic target through clinical and experimental data. Circulating FOS levels show strong associations with renal function decline, including estimated glomerular filtration rate (eGFR) and urinary albumin excretion rate (UAER), while modulation of FOS expression significantly impacts disease progression. Mechanistically, FOS engages in direct physical interaction with Smad3, forming a functional unit that amplifies TGF-β signaling and drives fibrosis transformation. In DKD, the marked upregulation of FOS expression acts both as a consequence and a driver of diabetic renal injury, creating a vicious cycle that perpetuates disease progression.

The FOS/Smad3 axis represents a promising target for intervention, with experimental evidence suggesting that disrupting this interaction can ameliorate renal fibrosis while preserving essential renal functions. Future therapeutic development should focus on small molecules capable of selectively inhibiting this pathogenic signaling node. Such targeted approaches may finally provide effective treatment options for this currently irreversible complication of DKD.

Methods

Establishment of DKD-specific molecular signatures

The transcriptome profiling data were obtained from the NCBI Gene Expression Omnibus (GEO) database under accession numbers GSE96804 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE96804) and GSE99339 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE99339). The combined dataset included 55 DKD samples and 20 normal control samples. We applied the combat algorithm to address technical variability between the two studies. The success of batch correction was validated through principal component analysis (PCA) and hierarchical clustering. PCA was performed using the prcomp function in R. Hierarchical clustering was conducted using the top 1000 most variable genes with euclidean distance and complete linkage method. Identification of differentially expressed genes (DEGs) was implemented with the limma R [34] package (|logFC|>1 and p value < 0.05).

The process of the weighted gene co-expression networks analysis (WGCNA) [35] encompassed the following critical steps: Firstly, the outliers or batch effects were removed using the abline function. Secondly, the correlation matrix can be transformed into an adjacency matrix based on the β value. Building on this foundation, a topological overlap matrix (TOM) was constructed to capture gene co-expression modules effectively. Hierarchical clustering and dynamic tree cutting were employed to group genes with similar expression patterns into modules. Finally, module members (MMs) and gene significance (GS) were calculated, and highly associated modules were selected for further analysis.

LASSO [36] regression and random forest (RF) [37] algorithms were used to capture genes more relevant to DKD. Ultimately, we established the ROC curve and calculated the area under the ROC curve (AUC) value.

To explore the biological significance and potential upstream regulators of the identified DEGs, we performed transcriptional regulator enrichment analysis using the ChEA3 web platform (https://maayanlab.cloud/chea3/). The list of DEGs (with an adjusted |logFC|≥0.5) was analyzed for up-regulated genes. We used the ENCODE ChIP-seq library, which infers transcription factor targets from experimentally determined ChIP-seq data, to identify candidate master regulators. Transcription factors within the top 20 mean rank scores were considered significant.

Study participants and clinical measurements

This cross-sectional study recruited 183 participants, including 88 DKD patients and 95 healthy adults (CON) from the Fourth Affiliated Hospital of Harbin Medical University (Ethics Approval No.: 2022-WZYSLLSC-06).

Inclusion criteria: (1) age 20–75 years; (2) diagnostic criteria of 2-DM; (3) microalbuminuria (30–300 mg/24 h) or macroalbuminuria (> 300 mg/24 h); (4) blood pressure ≤ 140/90 mmHg.

Exclusion criteria: (1) primary kidney disease with a definite diagnosis; (2) non-diabetic proteinuria etiologies; (3) recent (< 1 month) acute complications of DM or urinary infections; (4) major comorbidities involving cardiovascular/cerebrovascular disorders, hepatic/renal dysfunction, hematopoietic abnormalities, or malignancies; (5) psychiatric or neurological impairment; (6) pregnancy or lactation status; (7) menstruation phase.

A 24-hour urine collection was performed for the determination of urinary microalbumin using BNII Specific Protein Analyzer (SIEMENS AG, Germany) and calculation of the UAER. Serum parameters, including serum creatinine (Scr), blood urea nitrogen (BUN), uric acid (UA) and lipid profiles, were using an Automatic Biochemical Analyzer (Beckman Coulter, USA) following kit instructions. HbA1c was analyzed using a dedicated HbA1c analyzer (Bio-Rad, USA). The eGFR was calculated from Scr based on the CKD-EPI equation.

Predictive modeling and interpretability analysis for DKD

To identify predictive factors for DKD, univariate analysis was performed using a significance threshold of P < 0.05. Meanwhile, the correlations between FOS and clinical parameters were determined. LASSO regression was subsequently applied for dimensionality reduction of candidate variables. Variables selected through multivariate analysis were used to construct a clinical nomogram. Calibration curves demonstrated satisfactory agreement between nomogram predictions and actual observations. Decision curve analysis (DCA) was employed to assess the clinical utility of the predictive model, while ROC analysis quantified discriminative ability through the AUC. Precision-recall (PR) curves were additionally employed for comprehensive model evaluation under class imbalance conditions. To enhance interpretability, SHAP analysis was implemented to quantify the feature.

Rats models

A total of 40 male Sprague-Dawley rats, aged 6–8 weeks and weighing 180 ± 10 g (Changsheng, Liaoning), were included. All rats were kept in specific pathogen-free facilities and housed with 12 h cycles of light/darkness and a standard chow diet at 25 °C. Animal experiments were approved by the Fourth Affiliated Hospital of Harbin Medical University (Ethics Approval No.: 2022-WZYSLLSC-06).

STZ-induced diabetic rat model

The 30 SD rats were fed a high-fat diet for 4 weeks, then administered STZ, 35 mg/kg by intraperitoneal injection after a 12 h fasting period. The third and seventh days after injection, the blood glucose levels of the rats were measured. Diabetes was confirmed by a fasting blood glucose level ≥ 16.7mmol/L at least twice. All rats were euthanized at 18 weeks. Body weight and blood glucose levels were monitored biweekly. Blood, urine and partial kidney tissue samples are stored at −80℃, and the other part of the kidney tissues were fixed in 4% paraformaldehyde for pathological analysis.

Silence FOS diabetic rat

To achieve FOS knockdown, a recombinant lentivirus (GeneChem, Shanghai, China) expressing shRNA specifically targeting the rat c-FOS transcript (sequence: 5’-ATCCGAAGGGAAAGGAATAAGCTCGAGCTTATTCCTTTCCCTTCGGAT-3’) was injected through the tail vein. Rats in the control group were injected with a non-targeting lentivirus vector. The rats were treated with FOS silencing vectors every 4 weeks until 18 weeks.

Measurements of renal function

The urinary microalbumin concentration and urine protein were measured using an ELISA kit (Elabscience, Wuhan). Scr and BUN levels were measured using an ELISA kit (Jining, Shanghai), respectively, according to the manufacturer’s instructions.

Kidney histologic and morphometric analyses

Kidney samples were fixed in 10% formalin, embedded in paraffin, and sectioned to 2 μm thickness. The morphology and structure of the kidney tissue were visualized by Hematoxylin and Eosin (HE) staining. Periodic acid-Schiff staining (PAS) was employed to localize glycogen and assess the impact of diabetes on renal glucose metabolism; black arrowheads indicate the glomerulosclerotic pathological characteristics (including mesangial matrix hyperplasia, capillary loop collapse/occlusion, glomerular capsule wall thickening, and segmental/global sclerosis) to be quantified. Masson Trichrome staining (MASSON) was utilized for detecting collagen fibers in tissues, allowing for the assessment of the degree of renal fibrosis.

Glomerular damage was assessed on PAS-stained kidney sections. For each animal, at least 50 consecutive glomeruli in the renal cortex were examined in a blinded manner. Each glomerulus was semi-quantitatively scored based on the percentage of the glomerular area exhibiting mesangial matrix expansion, capillary lumen loss, and/or adhesion to Bowman’s capsule: 0 (no sclerosis), 1 (mild, < 25%), 2 (moderate, 25–50%), 3 (severe, 50–75%), and 4 (global sclerosis, > 75% or collapsed). The glomerulosclerosis index (GSI) for each sample was calculated using the formula: GSI = [(1 × N₁) + (2 × N₂) + (3 × N₃) + (4 × N₄)]/(N₀ + N₁ + N₂ + N₃ + N₄), where Nₓ represents the number of glomeruli assigned a score of x.

Renal interstitial fibrosis was evaluated on MASSON-stained sections. For each kidney, a minimum of 20 randomly selected, non-overlapping cortical fields (excluding glomeruli and large vessels) were examined at 400× magnification. The extent of fibrosis in the cortical interstitium per field was scored as: 0 (none, < 5%), 1 (mild, 5–25%), 2 (moderate, 25–50%), and 3 (severe, > 50%). The interstitial fibrosis score (IFS) for each animal was calculated as the mean score of all assessed fields. All scoring was performed by an observer blinded to the experimental groups.

Immunohistochemistry (IHC)

For antigen retrieval, tissue sections were heated in a microwave oven for 15–20 min in 10 mM sodium citrate buffer (pH 6.0). Subsequently, the slides were allowed to cool naturally to room temperature. Diluted normal goat serum was added to sections and incubated at room temperature for 30 min. Then, the diluted primary antibody FOS was added at 4 °C overnight. HRP-labeled goat-anti-rabbit secondary antibody was added and incubated at 37 °C for 30 min, followed by color development with diaminobenzidine and counterstaining with hematoxylin.

FOS expression was quantified from IHC-stained kidney sections using digital image analysis with ImageJ software. The percentage of FOS-positive nuclei among total nuclei was determined in at least 20 glomeruli per animal. Positive nuclei were identified based on distinct brown DAB staining within the nucleus. Data are presented as the mean percentage of positive nuclei per group.

TUNELstaining

Apoptotic cells in paraffin-embedded kidney sections were detected using a TUNEL apoptosis detection kit. Sections (4 μm) were deparaffinized, rehydrated, and treated with proteinase K (20 µg/mL, 20 min, 37 °C). After PBS washing, sections were incubated with TUNEL reaction mixture (60 min, 37 °C in dark), then counterstained with DAPI. Apoptotic cells (green nuclear fluorescence) were imaged under a fluorescence microscope. The apoptotic index was calculated as the percentage of TUNEL-positive nuclei relative to total DAPI-stained nuclei in five random cortical/medullary fields per sample using ImageJ.

In vitro models and genetic engineering

Cell culture and treatment

MPC5 murine and HPC human podocyte lines (Meisen Biotechnology, Zhejiang) were maintained in RPMI-1640 supplemented with 10% FBS under standard culture conditions (37 °C, 5% CO₂). Podocytes were then exposed to normal glucose (NC, 5.5mM glucose) or high glucose (HG, 35mM glucose) medium for 48h to simulate diabetic conditions.

RNA interference and overexpression

Gene-specific siRNAs targeting FOS and Smad3 (Hanyin Biotech, Shanghai) and overexpression plasmids for c-FOS and p-Smad3 (Jikai Gene, Shanghai) were transfected using Lipofectamine 2000 according to the manufacturer’s protocol. Negative control siRNA and empty vector served as experimental controls. The siRNA sequences: Human FOS: 5’–CCGAGCCCUUUGAUGACUUTT–3’, Mouse Fos: 5’–GCAGAUCUGUCCGUCUCUATT–3’, Human SMAD3: 5’–GCGUGAAUCCCUACCACUATT–3’, Mouse Smad3: 5’–CCCCAGCACACAAUAACUUTT–3’. The overexpression plasmids were custom-built on a pcDNA3.1(+) backbone, each driven by a CMV promoter. Detailed plasmid maps are provided in Fig. S1.

Functional Assays

CCK-8 assays: 1 × 10⁴ cells were seeded into 96-well plates and cultured. Then, CCK-8 reagents were added at 24 h, 48 h, and 72 h. After 2 h, the absorbance value was measured at 450 nm.

Scratch wound healing assays: Confluent MPC5 and HPC monolayers were wounded using sterile 200µL pipette tips. Cells were then incubated in serum-free medium. Images of each wound were captured at 0 h and at 48 h post-wounding using an inverted microscope. Wound closure was quantified using ImageJ software. The wound area at each time point was measured, and the percentage of wound closure was calculated as: (Area₀ₕ - Area₄₈ₕ)/Area₀ₕ × 100%. For each condition, at least 1 independent fields were analyzed per biological replicate and the experiment was independently repeated 3 times.

Transwell assays: MPC5/HPC cells (2 × 10⁵ cells/insert) were seeded in serum-free upper chambers. Lower chambers contained 600µL RPMI-1640 with 15% FBS as a chemoattractant. After 48 h of migration, membranes were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. Migrated cells were imaged using an inverted microscope. Quantification was performed by measuring the crystal violet-positive area per field of view using ImageJ software from five random fields per membrane. The entire experiment was independently repeated 3 times.

Flow cytometry

Podocyte apoptosis was assessed using Annexin V-FITC/PI double staining flow cytometry. HPC and MPC5 cells were treated with 35mM HG or 5.5mM glucose (control) for 48 h, harvested, and stained with Annexin V-FITC and PI. Analysis was performed on a BD FACSCanto™ II flow cytometer within 1 h after staining. Data were analyzed with FlowJo software, and the percentage of apoptotic cells (Annexin V-FITC positive and PI negative) was calculated.

Quantitative real-time PCR (RT-qPCR)

Total RNA was extracted from cells or whole blood using by Trizol reagent. RNA was reverse transcribed to cDNA using the cDNA transcription kit (Tiangen). RT-qPCR was performed using SYBR Green SuperMix. The data were normalized to β-actin and analyzed using the 2^ΔΔCt method.

The primer sequences were presented as: Human FOS: Forward: 5‘-CAAGCGGAGACAGACCAACT-3’, Reverse: 5‘-GTGAGCTGCCAGGATGAACT-3’, Human β-actin: Forward: 5‘- ACCGCGAGAAGATGACCCAG − 3’, Reverse: 5‘- GGATAGCACAGCCTGGATAGCAA − 3’.

Western blot

Lysate of tissues or cells was centrifuged for 20 min (4 °C, 13000 rpm), and the protein concentrations were determined with the BCA protein analysis kit. Thirty micrograms of protein was separated by 10% SDS–PAGE and subsequently transferred to NC membranes, and incubated overnight at 4 °C with primary antibodies, including phosphorylated FOS (5348T, Cell signaling technology), FOS (ab208942, Abcam), E-Cadherin (20874-1-AP, Proteintech), N-Cadherin (66219-1-IG, Proteintech), Vimentin (10366-1-AP, Proteintech), CollagenI (66761-1-Ig, Proteintech), CollagenIII (22734-1-AP, Proteintech), α-SMA (14395-1-AP, Proteintech), Podocin (20384-1-AP, Proteintech), TGF-β (81746-2-RR, Proteintech), p-smad3 (ab63403, Abcam), Smad3 (ab208182, Abcam), and β-actin (81115-1-RR, Proteintech). Membranes were incubated with secondary antibodies for 1 h at room temperature and were exposed to ECL solution.

Co-immunoprecipitation (Co-IP)

Two independent Co-IP assays were performed in podocytes to validate the interaction between FOS and p-Smad3.

Co-IP of HA-FOS with endogenous p-Smad3.

Podocytes were transfected with a HA-tagged FOS (HA-FOS) plasmid. After 48 h, cells were lysed. Cleared lysates were incubated with Anti-HA tag agnetic beads overnight at 4 °C. Beads were washed, and bound proteins were eluted. Samples (Input and IP) were analyzed by western blott using antibodies against HA (ab9110, Abcam) to confirm transfection and IP efficiency and p-Smad3 (9520T, Cell signaling technology) to detect co-precipitated endogenous p-Smad3.

Co-IP of FLAG-p-Smad3 with endogenous FOS.

Podocytes were transfected with a FLAG-tagged p-Smad3 (FLAG-p-Smad3) plasmid. After 48 h, cell lysates were prepared and incubated with Anti-FLAG M2 magnetic beads. Subsequent steps were as in Co-IP of HA-FOS with endogenous p-Smad3. Western blot was performed using antibodies against FLAG(ab205606, Abcam) and FOS (A0236, ABclonal) to detect co-precipitated endogenous FOS.

Statistical analysis

Data represent mean ± SD or mean ± SEM from ≥ 3 independent experiments. Between-group differences were assessed via Student’s t-test (two groups) or one-way ANOVA with post-hoc tests (multiple groups). All the analyses in this study were conducted using R software (R-4.3.2) and GraphPad Prism 9.0. Significance threshold was set at p < 0.05.

Results

Construction of DKD-related gene signatures

The mRNA expression microarray dataset GSE96804 and GSE99339 were downloaded and normalized.

As shown in Fig. S2, both PCA and hierarchical clustering were employed to assess sample variation before and after batch correction. After combat adjustment, PCA revealed distinct separation between CON and DKD groups along PC1, while PC2 captured within-group variation (Fig. S2A, B). This pattern was mirrored by hierarchical clustering, where samples robustly grouped by biological condition post-correction (Fig. S2C, D). The tighter clustering of CON samples in both analyses suggests greater heterogeneity within the DKD group, likely reflecting disease variability. Together, these results confirm successful batch correction and establish DKD status as the primary source of transcriptional variation.

Differential expression analysis conducted through the limma package revealed 187 DEGs, comprising 122 downregulated and 65 upregulated genes in DKD specimens. Visual representation through heatmap and volcano plot demonstrated distinct patterns between DKD and CON groups (Fig. S2 E, F).

Machine learning-based identification of hub genes: Scale-free topology network analysis identified an optimal soft thresholding power (β = 20, scale-free fit R²=0.9) for module construction. Dynamic hybrid tree cutting algorithm delineated 4 functionally distinct modules, among which the grey module exhibited the strongest disease association (module-trait correlation coefficient = 0.82) with DKD clinical parameters (Fig. S2, G). Feature selection through LASSO regression identified 16 DKD-associated candidates (Fig. S2 H, I), subsequently refined to 3 genes by the RF algorithm (Fig. S2 J, K). Intersection analysis identified FOS across all three algorithms (Fig. 1A).

Fig. 1.

The identification of related genes in DKD. (A) Venn plot exhibiting DEGs among WGCNA, LASSO, and RF. (B) Box plot for differential expression analysis of FOS. (C) ROC curves of FOS. (D) qRT–PCR showed FOS mRNA expression in blood samples from control and DKD patients. Error bars represent SEM. (E-H) Spearman correlation assessment revealed correlations between FOS expression levels and clinical parameters in DKD patient samples

To complement our mathematical analyses (WGCNA, LASSO, and RF), we performed a transcriptional regulator enrichment analysis using ChEA3. Analysis of upregulated DEGs (|logFC| ≥ 0.5) from the ENCODE ChIP-seq library predicted the transcription factor FOS as a key upstream master regulator (ranked 11th) (Fig. S2M). This finding strongly corroborates our prior results, suggesting FOS is not merely a correlative gene but a potential driver of the transcriptional network in DKD.

Transcriptional profiling of human kidney biopsies from the GSE96804 and GSE99339 dataset revealed FOS upregulation in DKD compared to controls (Fig. 1B). Diagnostic performance evaluation through ROC analysis demonstrated outstanding predictive capacity for FOS (AUC = 0.939) (Fig. 1C). Differential FOS mRNA expression was observed in blood of DKD patients compared to healthy controls (Fig. 1D).

Circulating FOS levels exhibit robust clinical correlations, displaying inverse associations with eGFR and positive relationships with Scr, BUN, UAER (Fig. 1E-H).

Clinically actionable predictors of DKD

In our 183-patient cohort, serum FOS levels correlated strongly with DKD progression. To bridge gaps between biomarker discovery and clinical deployment, we developed an interpretable machine learning-powered nomogram integrating routinely available clinical parameters (eGFR, Scr, BUN, UA, FOS, age, sex, and lipid profiles). Candidate variables were first selected through LASSO regression (Fig. 2B, C) and refined using multivariate logistic regression, forming the foundation for our predictive framework (Fig. 2D). Forest plot identified FOS, eGFR, Scr, BUN, UA, TC, and TG as independent risk factors for DKD (Fig. 2A), which were subsequently incorporated into the nomogram. This model serves as a clinical decision tool, offering visual quantification of individualized risk probabilities for DKD progression. The multivariable nomogram demonstrated robust predictive performance, validated through multi-dimensional evaluation. Calibration curves exhibited strong concordance between predicted and observed outcomes (Fig. 2F). ROC analysis achieved an AUC of 0.739.(Fig. 2E). DCA and PR curves confirmed superior net clinical benefit and risk stratification accuracy compared to alternative models (Fig. 2G, H). SHAP value analysis enhanced interpretability, revealing FOS and eGFR as dominant contributors to DKD risk predictions (Fig. 2I, J).

Fig. 2.

Development and validation of clinical prediction models for DKD (A) Forest plot demonstrated the association strength between clinical variables and DKD. (B, C) LASSO penalized regression analysis (λ=0.0183) with 10-fold cross-validation identified 7 characteristic predictors. (D) Clinically applicable nomogram integrated multivariable logistic regression coefficients for DKD risk stratification. (E-H) Comprehensive validation of predictive performance: (E) ROC curves. (F) Calibration curves demonstrated agreement between predicted vs observed probabilities. (G) DCA curves revealed the clinical net benefit threshold. (H) PR curves quantified predictive reliability. (I-K) SHAP value-based analysis delineated feature impact directions and global importance

FOS deletion ameliorated renal injury in DKD

To achieve renal-selective FOS knockdown, adeno-associated virus vectors encoding short hairpin RNAs targeting FOS (shFOS) were administered via tail vein injection. Robust FOS suppression compared to vector control in DKD rats was confirmed by immunohistochemistry and western blot 4 weeks post-injection (Fig. 3J, P). FOS-knockdown rats exhibited a lower kidney weight-to-body weight ratio (KW/BW) and renal function parameters than DKD controls, despite unchanged body weight and blood glucose levels (Fig. 3A, B). Therapeutically, shFOS-treated DKD rats exhibited marked renal functional improvements: urinary microalbumin, UACR, Scr, and BUN decreased (Fig. 3C-F). HE, PAS and MASSON evaluated histological changes of glomeruli and were quantified. This integrated approach revealed significant attenuation of glomerular basement membrane thickening, amelioration of glomerular hypertrophy and fibrosis (Fig. 3G-I). Furthermore, TUNEL staining demonstrated that FOS knockdown significantly reduced renal apoptosis, with apoptotic rates of 25.37% in DKD, 23.62% in DKD+Vector, and 11.79% in DKD+shFOS groups (Fig. 3K, N). Mechanistically, FOS knockdown reversed hyperglycemia-induced EMT, suppressing mesenchymal markers N-cadherin and Vimentin, while restoring E-cadherin expression (Fig. 3P). Attenuation of EMT related protein expression in DKD rat kidneys by FOS silencing supports a role for FOS as an orchestrator of EMT networks in DKD (Fig. 3P). These findings provide compelling evidence supporting FOS-targeted therapy for DKD.

Fig. 3.

shRNA-mediated FOS suppression demonstrated renoprotective effects in DKD rats. (A,B) Longitudinal monitoring of physiological parameters: (A) body weight and (B) blood glucose levels measured fortnightly. (C–F) Renal function parameters at study termination: (C) urinary microalbumin, (D) UACR, (E) Scr, and (F) BUN. (G–I) Histopathological evaluation of kidney tissues: (G) HE staining, (H) MASSON staining, and (I) PAS staining. Black arrowheads indicate pathological features quantified as glomerulosclerosis. Images acquired at 200× and 400× magnification. (J) Immunohistochemical staining of FOS expression in kidney tissues (200× and 800×). (K) TUNEL staining for apoptosis detection (200× and 400×). (L) Quantification of glomerulosclerosis index from PAS staining. (M) Quantification of interstitial fibrosis score from MASSON staining. (N) Apoptotic rate quantified as percentage of TUNEL‑positive nuclei in glomerular and tubular regions. (O) Percentage of FOS‑positive nuclei determined from IHC images. (P) Western blot analysis of FOS, p‑Smad3, Smad3, EMT‑related proteins, and fibrosis‑related proteins. Data in C–F, L–O are presented as mean ± SEM; n = 6 rats per group. Five images were analyzed per biological replicate, with three biological replicates assessed per condition. *P < 0.05, **P < 0.01, ***P < 0.001

HG induces EMT activation and promotes fibrosis in podocytes

To establish a pathologically relevant glucose concentration for in vitro DKD modeling, we conducted dose–response analyses in MPC5 and HPC podocytes, which revealed concentration-dependent glucose cytotoxicity. Treatment with 35 mM HG for 48 h resulted in apoptotic rates of 13.75% in HPC and 6.15% in MPC5 cells, as measured by Annexin V/PI staining, without increasing necrotic cell death (Fig. 4A–H). These findings confirm that 35 mM HG induces podocyte apoptosis and reduces viability without triggering necrosis, supporting its use for modeling sustained hyperglycemic stress in subsequent experiments.

Fig. 4.

HG promotes EMT and fibrosis progression in HPC and MPC5 cells. (A,E) Cell viability of HPC and MPC5 cells treated with different glucose concentrations for 24h, 48h, and 72h. (B, C, F, G) Representative flow cytometry plots of apoptosis measured by Annexin V-FITC/PI staining. (D, H) Quantification of total apoptotic rates (early+late apoptosis). Data are presented as the mean ±SD from three independent experiments. (I, K) Cell migration ability under HG conditions assessed by scratch wound healing and transwell assays. For wound healing, images were captured at 40× magnification with 1 field per replicate (3 biological replicates). For transwell migration, 5 images per replicate were taken at 100× magnification (3 biological replicates). (J, L) Western blot showing expression levels of p-FOS, FOS, p-Smad3, Smad3, Podocin, EMT-related markers, and fibrosis-related proteins under HG conditions. HG represents 35mmol/L glucose. **P < 0.01, ***P < 0.001

Notably, we observed upregulation of total FOS expression under sustained HG stress; however, its phosphorylation status (p-FOS) remained largely unchanged (Fig. 4J, L), suggesting that FOS may mediate HG-induced podocyte injury through a phosphorylation-independent mechanism. Extended 48 h HG exposure induced phenotypic alterations indicative of podocyte injury and fibrotic activation, as measured by a quantifiable reduction in E-cadherin with concomitant increases in N-cadherin and Vimentin [38], alongside marked elevations of Collagen I, Collagen III, and α-SMA(Fig. 4J, L).

Consistent with the loss of podocyte integrity, extended HG exposure also led to a significant downregulation of the slit diaphragm protein Podocin (Fig. 4J, L). Scratch wound healing assays and transwell assays revealed enhanced migratory capacity in HG-treated cells compared to controls (Fig. 4I, K).

This evidence suggests that HG directly triggers podocyte EMT and a concomitant pro-fibrotic response in vitro. To validate the persistence of HG-induced signaling, podocytes were exposed to HG for 48 h and subsequently returned to normal glucose (NC, 5.5mM) medium for 24 h, 48 h and 72 h. Western blot analysis revealed that total FOS and p-Smad3 protein levels remained significantly elevated throughout the 72 h recovery period compared to the constant NC control, whereas p-FOS levels returned to baseline (Fig. S3). These data indicate that the FOS/Smad3 axis exhibits sustained activation even after glucose normalization, reflecting a “metabolic memory” phenotype.

Function of FOS in DKD pathogenesis

To understand the role of FOS in mediating EMT endpoints in an in vitro model of hyperglycemic stress, we employed over-expression and gene silencing strategies in MPC5 and HPC cell lines. To delineate FOS-mediated regulation in DKD, we employed gain and loss-of-function strategies in MPC5 and HPC cell lines. Stable FOS-overexpressing cell lines exhibited marked phenotypic changes, characterized by dissolution of epithelial integrity and acquisition of mesenchymal features.

Western blot demonstrated that FOS overexpression orchestrated a mesenchymal transition. The epithelial marker, E-cadherin was reduced while N-cadherin and Vimentin increased (Fig. 5E, F). Profibrosis transformation was further evidenced by dramatic induction of collagenous matrix components (Collagen I, Collagen III) and myofibroblast marker α-SMA, surpassing levels observed in HG controls (Fig. 5E, F). In parallel, to directly assess podocyte injury under diabetic conditions, we found that FOS overexpression downregulated the slit‑diaphragm protein Podocin (Fig. 5E, F), indicating a loss of structural integrity. Migration assays confirmed enhanced cellular motility under FOS-overexpressing conditions (Fig. 5A, B). These findings collectively establish that FOS expression regulates EMT associated changes in cultured podocytes under HG conditions.

Fig. 5.

FOS functions as a "catalyst" in HG induced mobility and EMT. (A, B) FOS overexpression enhanced cell migration ability as assessed by scratch wound healing assay and transwell assay. For scratch wound healing assay, images were captured at 40× magnification with 1 field per biological replicate (n=3). For transwell assay, images were taken at 100× magnification with 5 fields per replicate (n=3). (C, D) FOS knockdown attenuated migration in HPC and MPC5 cells under HG conditions, evaluated by scratch wound healing and transwell assays, using the same imaging parameters as above. (E, F) Western blot analysis showing upregulated protein levels of EMT-related markers and fibrosis-related factors following FOS overexpression. (G, H) FOS deficiency suppressed the expression of EMT and fibrosis signatures, as demonstrated by western blot. HG represents 35mmol/L glucose. Data are presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001

Complementary siRNA-mediated FOS silencing experiments unveiled therapeutic potential. FOS knockdown cells exhibited restoration of E-cadherin expression with parallel suppression of mesenchymal markers (Fig. 5G, H). Fibrosis related changes were substantially alleviated, showing decreases in Collagen I, Collagen III and α-SMA levels (Fig. 5G, H). FOS silencing rescued the expression of structural markers like Podocin, identifying FOS as a critical upstream regulator whose inhibition restores podocyte homeostasis. Migration capacity collapsed to baseline levels, indicative of FOS-dependent motility regulation (Fig. 5C, D).

FOS functions as a molecular switch governing renal epithelial plasticity. Its overexpression drives a self-reinforcing cycle of EMT initiation→ migratory activation, while its suppression reinstates epithelial homeostasis. These findings position FOS as a central orchestrator of DKD.

Mechanistic exploration of the FOS-Smad3 axis

To elucidate the molecular basis of FOS-mediated EMT and fibrosis regulation, we focused on its interaction with the TGF-β/Smad3 signaling pathway.

Co-IP assays confirmed a direct interaction between FOS and p-Smad3 in podocytes, which was significantly strengthened under HG conditions compared to normal glucose (Fig. 6A-D). This provides direct molecular evidence that hyperglycemia potentiates their association, offering a plausible mechanism for how HG stress amplifies downstream EMT and fibrotic signaling.

Fig. 6.

FOS-Smad3 interaction drives TGF-β/Smad3-mediated renal EMT and fibrosis. (A–D) Co-IP assays confirming the direct physical interaction between FOS and p-Smad3 in HPC and MPC5 cells under normal control (NC) and HG conditions. (E, F) Smad3 phosphorylation was suppressed upon FOS knockdown. Protein levels of p-Smad3 and total Smad3 were assessed by western blot. (G, H) FOS overexpression upregulated p-Smad3 expression, as shown by western blot. (I, K) Cell migration ability evaluated by scratch wound healing assay and transwell assay across the following groups: HG, OEFOS, Vector, OEFOS + siSmad3, and OEFOS + SIS3. For scratch wound healing assay, images were acquired at 40× magnification with 1 field per replicate (n=3 biological replicates). For transwell assay, images were taken at 100× magnification with 5 fields per replicate (n=3 biological replicates). (J, L) Smad3 knockdown reversed FOS-induced EMT abnormalities and TGF-β/Smad3 pathway activation, as evaluated by western blot. HG represents 35mmol/L glucose. Data are presented as mean ± SD. ***P < 0.001

FOS knockdown attenuated HG-induced Smad3 phosphorylation, reducing p-Smad3 levels to 0.69-fold (MPC5) and 0.66-fold (HPC) of their respective HG+Vector controls (Fig. 6E, F). Conversely, FOS overexpression further enhanced phosphorylation, increasing p-Smad3 by 1.16-fold (MPC5) and 1.42-fold (HPC) (Fig. 6G, H).

Functional studies revealed FOS as a critical enhancer of Smad3 activation. Administration of the Smad3 inhibitor SIS3 or Smad3-targeted siRNA in the hyperglycemia model effectively blocked FOS-driven Smad3 activation and suppressed FOS-induced cell migration (Fig. 6I, K). Downregulation of Smad3 effectively reversed FOS-induced hyperactivation of the TGF-β/Smad3 pathway and normalized downstream fibrotic effectors (Collagen I, III, α-SMA) (Fig. 6J, L). This reversal was associated with a partial restoration of podocyte integrity, as indicated by recovered Podocin expression (Fig. 6J, L). Collectively, these data demonstrate a pathogenic axis wherein FOS amplifies TGF-β signaling via direct interaction with Smad3. The glucose withdrawal experiment further illuminated the self-sustaining nature of FOS/Smad3 signaling. The persistent elevation of total FOS and p-Smad3, despite the absence of continued hyperglycemia—suggests that HG may initiate a stable activation loop independent of upstream phosphorylation. Notably, the dissociation between sustained total FOS and transient p-FOS implies that signaling persistence relies on FOS protein stabilization and its downstream actions on Smad3, rather than persistent kinase activation. This feedforward mechanism may underlie the irreversible progression of fibrosis in DKD, providing a plausible explanation for disease progression even after glycemic control is achieved. Therefore, targeting FOS holds potential not only for preventing injury but also for halting the ongoing pathological process after glycemic control is achieved.

The reciprocal reinforcement between FOS and the TGF‑β/Smad3 axis may constitute a feedforward loop that sustains EMT and ECM deposition in vitro, processes relevant to DKD fibrosis.

Discussion

To identify key genes in DKD, we applied a multi-faceted computational strategy to two independent GEO datasets (GSE96804, GSE99339). The consensus of WGCNA, LASSO, and RF algorithms consistently pinpointed FOS as a central regulator. Crucially, this finding was biologically validated by an independent master regulator analysis using ChEA3, which predicted FOS as a top-ranked transcriptional regulator of the up-regulated genes in our dataset. This convergence of computational and evidence-based regulatory evidence significantly strengthens the proposition that FOS plays a pivotal role in DKD pathogenesis, potentially acting as a master regulator of the disease-associated gene expression signature.

The pathogenesis of DKD is intrinsically linked to dysregulated EMT and subsequent fibrosis [39, 40], yet the precise molecular drivers governing this pathology have remained elusive [41]. EMT has long been recognized as a critical gateway to renal fibrosis in DKD [42]. However, traditional paradigms attributing EMT solely to TGF-β/Smad3 signaling fail to explain why fibrosis progresses relentlessly in DKD despite systemic glucose control or TGF-β blockade in experimental models [43, 44]. This paradox suggests the existence of unrecognized amplifiers that perpetuate EMT independently of classical pathways. By establishing FOS as a podocyte amplifier of Smad3-driven EMT and fibrosis, this work redefines therapeutic targeting in DKD.

FOS dysregulation across GEO datasets and its dominance over previously proposed biomarkers in machine learning models (WGCNA+LASSO + RF) suggest that FOS induction constitutes a response to DKD. Our study not only plotted ROC curves and a nomogram to improve the accuracy and specificity of relation prediction, but also generated calibration curves for the nomogram to prove the effective clinical utility of FOS further. Secondly, DCA curves and PR curves were drawn to clarify the clinical feasibility of the risk model, and the related significance of FOS was emphasized by comparing it with known models.

Our in vivo intervention using FOS-targeted shRNA in STZ-induced diabetic rats demonstrated that FOS reduction not only reduced albuminuria (reduction in UACR and urinary microalbumin) but also improved functional renal parameters (decreased Scr and BUN), and attenuated glomerulosclerosis and interstitial fibrosis. Moreover, TUNEL staining revealed a marked reduction in renal apoptotic rates—from 25.37% in DKD rats to 11.79% following shFOS treatment—suggesting that suppression of FOS-mediated apoptosis contributes to renal protection. These in vivo results, consistent with our in vitro data demonstrating HG-induced podocyte apoptosis. Together, Our data demonstrate that shFOS administration shows improved kidney function as measure by the aforementioned endpoints.

The selection of a 35mM glucose concentration for our in vitro modeling, while higher than systemic levels observed in diabetic rodents, is justified by its capacity to recapitulate the advanced glomerulopathologic features of human DKD within an experimentally tractable timeframe. Notably, emerging evidence suggests that the local glucose concentration within the renal glomerular microenvironment may substantially exceed systemic levels due to tubular handling and tissue retention mechanism [45]. This discrepancy between circulating and intrarenal glucose levels underscores the relevance of using elevated glucose conditions to mimic the pathophysiological milieu in diabetic kidneys. In alignment with studies investigating progressive DKD, our use of 35 mM glucose effectively induced key injury phenotypes, enabling the delineation of FOS-dependent signaling in podocyte dysfunction.

We further showed that HG induces EMT in podocytes, characterized by loss of E‑cadherin and gain of N‑cadherin and vimentin, preceding collagen overproduction, supporting EMT as an early event in DKD. FOS knockdown attenuated this phenotype and restored Podocin expression, whereas FOS overexpression exacerbated EMT and fibrosis, confirming its regulatory role. These findings suggest that FOS and Smad3 inhibition could disrupt EMT-TGF-β/Smad3 axis and target FOS-mediated transcriptional amplification, a potential dual approach distinct from current anti-fibrosis paradigms. An open question remains whether FOS silencing concurrently protects against EMT and cell death; evaluating its effect on podocyte viability represents a valuable direction for future study.

Mechanistically, Co-IP revealed that FOS interacts with phosphorylated Smad3, with enhanced binding under HG, suggesting engagement with the activated Smad3 pool. This supports a model wherein FOS amplifies Smad3 signaling and may promote TGF-β expression, potentially initiating an autocrine loop that sustains Smad3 activation via TGFBR1. This idea is consistent with reports of autocrine TGF-β signaling in podocytes [46], though direct evidence, such as p-TGFBR1 detection, is still needed. We note that the absence of a normal glucose control precludes confirmation of TGF-β upregulation by HG, and the lack of p-TGFBR data limits mechanistic support for an autocrine circuit, key limitations to be addressed in future work.

Notably, this FOS–Smad3 interaction was absent in non‑renal fibrotic models, underscoring its cell‑type specificity. Although we cannot determine whether Smad3 inhibition fully restores the podocyte baseline without a normal glucose control, our data demonstrate bidirectional crosstalk: FOS silencing reduced Smad3 phosphorylation, and Smad3 inhibition counteracted FOS‑driven EMT, collagen overproduction, and Podocin loss. This reciprocity contrasts sharply with unidirectional TGF-β→Smad3 signaling in cardiac fibrosis models [47]. Subsequent studies including a normal glucose control will be essential to confirm phenotypic restoration and therapeutic potential. The FOS-driven amplification of Smad3 signaling provides a molecular explanation for the relentless progression of DKD and highlights this axis as a promising target for therapy, independent of glycemic control alone.

While our data underscore a central role for FOS in HG-induced podocyte injury, we acknowledge the complexity of AP-1 transcriptional regulation. FOS typically functions as a heterodimer with Jun family proteins (c-Jun, Jun-B, Jun-D). Interestingly, emerging evidence suggests that in certain pathological contexts, the availability of the FOS subunit itself, rather than its binding partners, can be the rate-limiting factor for AP-1 activity [48]. This may be particularly relevant in our model, where we observed a specific and pronounced upregulation of FOS in response to HG, potentially driving the formation of pathogenic AP-1 complexes. Furthermore, the FOS gene encodes several isoforms, with FOS being strongly implicated in fibrotic processes [49]. We therefore hypothesize that FOS is the primary isoform mediating the effects observed here. Future studies employing isoform-specific tools are warranted to conclusively identify the responsible FOS isoform and its key binding partners within the AP-1 complex in diabetic podocytes, which will further refine our understanding of the transcriptional drivers of DKD.

In summary, the FOS–Smad3 axis represents a disease‑initiating mechanism and a targetable vulnerability in DKD. Its amplification of fibrotic signaling provides a molecular basis for disease progression independent of glycemic control alone. Targeting this axis, particularly during the early EMT window, may shift DKD management from slowing progression to achieving disease remission.

Conclusion

FOS modulates Smad3 phosphorylation status, thereby activating the TGF-β/Smad3 signaling axis to orchestrate EMT and downstream fibrosis that exacerbate renal dysfunction in DKD. Our findings advocate for context-specific modulation of the FOS-Smad3 axis rather than broad TGF-β pathway inhibition, potentially mitigating off-target effects associated with current antifibrosis strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

DKD

Diabetic kidney disease

EMT

Epithelial-mesenchymal transition

GEO

Gene Expression Omnibus

WGCNA

Weighted gene co-expression network analysis

LASSO

Least absolute shrinkage and selection operator

RF

Random forest

STZ

Streptozotocin

Co-IP

Co-immunoprecipitation

IHC

Immunohistochemistry

HG

High glucose

Scr

Serum creatinine

BUN

Blood urea nitrogen

UA

Uric acid

eGFR

Estimated glomerular filtration rate 

UAER

Urinary albumin excretion rate

Author contributions

M. Sun designed the research, interpreted the data, analyzed the results, and was a major contributor to writing and revising the manuscript. M. Sun and W. Zhao performed the initial set of experiments. M. Sun and T. Yan were responsible for the conception, execution, and interpretation of the critical additional experiments during the revision phase. Z. Cheng designed the research, assisted with manuscript review and revision, and serves as the corresponding author. All authors read and approved the final manuscript.

Funding

Project funded by the National Key Research and Development Program of China (2022YFC2503300).

Data availability

All data generated or analyzed during this study are included in this published article.

Declarations

Ethical approval and consent to participate

The Ethics Committee of the Fourth Affiliated Hospital of Harbin Medical University approved all work.

Competing interests

The authors declare no competing interests.